Research and Measurement Issues in Gambling Studies, Volume 1

Citation preview

RESEARCH AND MEASUREMENT ISSUES IN GAMBLING STUDIES

This page intentionally left blank

RESEARCH AND MEASUREMENT ISSUES IN GAMBLING STUDIES

Garry Smith David C. Hodgins Robert J. Williams

AMSTERDAM • BOSTON • HEIDELBERG • LONDON NEW YORK • OXFORD • PARIS • SAN DIEGO SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO Academic press is an imprint of Elsevier

Academic Press is an imprint of Elsevier 30 Corporate Drive, Suite 400, Burlington, MA 01803, USA 525 B Street, Suite 1900, San Diego, California 92101-4495, USA 84 Theobald’s Road, London WC1X 8RR, UK This book is printed on acid-free paper. Copyright © 2007, Elsevier Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone: (+44) 1865 843830, fax: (+44) 1865 853333, E-mail: [email protected]. You may also complete your request on-line via the Elsevier homepage (http://elsevier.com), by selecting “Support & Contact” then “Copyright and Permission” and then “Obtaining Permissions.” Library of Congress Cataloging-in-Publication Data Research and measurement issues in gambling studies / [edited by] Garry Smith, David Hodgins, Robert Williams. p. cm. Includes bibliographical references and index. ISBN 978-0-12-370856-4 1. Gambling—Research. 2. Gambling—Social aspects. 3. Gambling—Law and legislation. I. Smith, Garry, Dr. II. Hodgins, David, Ph. D. III. Williams, Robert, Dr. HV6713.R47 2007 306.4'82072—dc22 British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. ISBN 13: 978-0-12-370856-4 ISBN 10: 0-12-370856-7 For information on all Academic Press publications visit our Web site at www.books.elsevier.com

Printed in the United States of America 07 08 09 10 9 8 7 6 5 4 3 2 1

2007013842

DEDICATION To our patient and understanding wives, Dodie, Roslyn, and Susan.

This page intentionally left blank

CONTENTS List of Contributors Preface Acknowledgments

xxiii xxvii xxxv

PART I NATURE AND SCOPE OF GAMBLING STUDIES CHAPTER 1

Situating Gambling Studies Gerda Reith

Introduction The Historical Context of Gambling The Paradigm Shift The Emergence of “Gambling Studies” Research Domains Interpretivist Approaches Sociological, Anthropological, and Psychological Research Positivist Approaches Economic and Social Cost–Benefit Analyses Biomedical Approaches Clinical Psychological Research Cognitive Psychological Research Epidemiological Research and Public Health Perspectives Conclusions: Current Trends and Future Directions

3 4 6 7 8 8 8 10 14 16 17 19 21 24

PART II MEASUREMENT ISSUES CHAPTER 2

Population Surveys Rachel A.Volberg

Introduction Purposes of Population Surveys in Gambling Studies vii

33 34

viii

Contents

Problem-Gambling Prevalence and Comorbidity Sampling Issues in Population Surveys Sample Size Sampling Frame Sampling Modality Multimodal Sampling Response Rates Weighting Population Survey Samples Constraints and Choices in Population Research

35 36 36 37 39 43 44 49 51

CHAPTER 3

Questionnaire Design: The Art of a Stylized Conversation Marianna Toce-Gerstein and Dean R. Gerstein

The Interview Context What Is an Interview? Conceptual Context Physical Context: Setting and Mode Social Context Privacy Effects Response Bias Social Desirability Bias Nonresponse Basic Communicative Principles Relevance Neutrality Ambiguity Language of Administration Translating the Questionnaire Using Interpreters Minimizing Cognitive Effects Recall Problems Limited Grasp and Computability of Quantities Scaling of Attitudes, Opinions, and Behaviors Causality Age-Graded Behavior: Special Considerations for Youth Questionnaire Construction Structuring the Questionnaire Major Questions, Sections, and Section Order Pathing Through the Interview

56 56 57 58 60 60 61 62 63 63 64 64 65 65 65 66 67 67 68 69 69 70 71 71 72 72

Contents

Question Flow and Context Effects Item Response Frames Reading Level Pretesting Informed Consent Special Considerations for Gambling Research Definition of Gambling Gambling Participation Attitudes Toward Gambling Problem Gambling Diagnosis of Pathological Gambling Correlates of Problem Gambling Problem Gambling Help and Treatment

ix

73 73 74 74 76 77 77 78 78 79 79 80 81

CHAPTER 4

Experimental Methodologies in Gambling Studies Sherry H. Stewart and Steven Jefferson

Basic Components of an Experimental Research Study Internal and External Validity Types of Experimental Designs Group Experimental Designs Control Groups Randomization Single-Case Experimental Designs Withdrawal Designs Multiple Baseline Design Sample Experimental Methodologies Behavioral Observation Explicit Cognition Implicit Cognition Think Aloud Reaction Time Tasks Conclusions

88 89 91 91 92 93 93 95 97 98 98 101 102 105 105 107

CHAPTER 5

Qualitative Methodologies Robert A. Stebbins

Grounded Theory and Exploration Exploratory Concatenation

112 113

x

Contents

Confirmatory Qualitative Research Validity and Reliability Methods of Data Collection Qualitative Research on Gambling Issues Conclusions

114 115 117 120 122

CHAPTER 6

Longitudinal Studies of Gambling Behavior Wendy S. Slutske

Introduction Existing Longitudinal Studies of Gambling Behavior Longitudinal Studies of Gambling Behavior Initiated During Preadolescence and Adolescence Montreal, Canada (Vitaro and colleagues) Minnesota (Winters and colleagues) New York (Barnes and colleagues) Quebec, Canada (Vitaro, Ladouceur, and Bujold) Longitudinal Studies of Gambling Behavior Initiated During Late Adolescence and Early Adulthood Missouri College Students (Slutske, Jackson, and Sher) New York (Barnes and colleagues) Dunedin, New Zealand (Slutske and colleagues) Longitudinal Studies of Gambling Behavior Initiated During Early to Late Adulthood New Zealand (Abbott,Williams, and Volberg) U.S. Casino Employees (Shaffer and Hall) Key Issues and Challenges in Longitudinal Gambling Research Statistical Techniques for Modeling Stability and Change Dealing with Missing Data The Low Prevalence of Pathological Gambling Disorder Important Questions and What We Know So Far Temporal Resolution of Gambling Correlates: Establishing Causality? The Stability of Gambling Behavior The Course of Gambling Behavior and Gambling Problems Sequential/Stage Theories of Gambling Involvement: Is There a “Gateway” to Problems? Developmental Changes Versus Cohort or Period Effects on Levels of Gambling Involvement “Natural Experiments” in Longitudinal Gambling Research Summary of What We Don’t Know (Yet)

128 129 129 129 131 132 132 133 133 134 134 135 135 135 136 136 137 139 140 140 142 145 146 148 149 150

xi

Contents

CHAPTER 7

Quantification and Dimensionalization of Gambling Behavior Shawn R. Currie and David M. Casey

Historical Perspectives Epidemiological Data on Gambling Expenditure, Frequency, Duration, and Type Quantification of Other Addictive Behaviors Importance of the Quantification of Gambling Variations in Sources of Data Relevant Quantitative Dimensions of Gambling Behaviors Inputs Participation Status Types of Gambling Frequency Expenditure Duration Attitudes and Cognitions Outputs Financial Legal Social and Psychological Harms Clinical Use of Quantitative Gambling Data Conclusions

156 157 159 160 161 163 164 164 168 168 169 170 170 171 171 171 172 173 173

CHAPTER 8

A Review of Screening and Assessment Instruments for Problem and Pathological Gambling Randy Stinchfield, Richard Govoni, and G. Ron Frisch

Introduction Instruments Gamblers Anonymous 20 Questions (GA-20) South Oaks Gambling Screen (SOGS) Massachusetts Gambling Screen (MAGS) DSM-IV-MR (MR = Multiple Response) Diagnostic Interview for Gambling Severity (DIGS) Gambling Treatment Outcome Monitoring System (GAMTOMS) National Opinion Research Center DSM-IV Screen for Gambling Problems (NODS) Lie/Bet Questionnaire

180 181 181 190 193 193 194 195 196 198

xii

Contents

Gambling Assessment Module (GAM) Canadian Problem Gambling Index (CPGI) Gambling Behavior Interview (GBI) Clinical Global Impression Scale (CGI) Pathological Gambling Adaptation of the Yale Brown Obsessive-Compulsive Scale (PG-YBOCS) Gambling Symptom Assessment Scale (G-SAS) Structured Clinical Interview for Pathological Gambling (SCI-PG) Conclusions and Future Research Directions

199 200 201 202 203 204 204 205

PART III EMERGING GAMBLING STUDIES RESEARCH ISSUES CHAPTER 9

The Role of Structural Characteristics in Gambling Jonathan Parke and Mark Griffiths

Background Payment Characteristics Suspension of Judgment and Cashless Gaming Smart Cards, Spending Limits, and Cashless Gaming Maximum Bet Size and Bill Acceptors Cash Versus Credit Display Playability Characteristics Feature Games and Other Specialist Play Features Stop Buttons Gamble Buttons The Near Miss The Psychology of Familiarity Speed and Frequency Characteristics Bet Frequency and Event Frequency Event Duration In-Running Betting Payout Interval Autoplay Educational Characteristics Clocks Transparency of Expenditure and Statements Warnings/Pop-Up Messages

218 223 223 224 224 225 226 226 227 229 230 231 231 232 232 233 233 234 234 235 235 235

Contents

Ambient Characteristics Light and Color Effects Sound Effects General Sound Verbal Interaction Music Reward Characteristics Multiplier Potential and Betting Lines Payout Ratios and Randomness Jackpots Reward Schedules and Reinforcers Mandatory Cashouts Issues Relating to Research and Measurement Ecological Validity: Laboratory Versus Natural Setting Experiments in the Natural Setting Ethics in Gambling Experiments Conclusions

xiii

236 236 236 237 238 238 239 239 240 241 242 242 243 243 244 245 245

CHAPTER 10

Situational Factors That Affect Gambling Behavior Max W. Abbott

Introduction Availability, Accessibility, and Exposure Dimensions of Accessibility and Exposure Availability and Participation Studies Using Official Data Surveys and Other Gambling Studies Availability, Participation, and Problem Gambling The Agent Gambling Prevalence Studies Replication Surveys Prevalence Changes in Population Sectors Other Exposure Concentrations Prospective Studies Other Location and Contextual Factors Credit Cards and ATMs Alcohol Tobacco Marketing and Advertising Conclusion

251 252 253 255 256 257 259 259 260 262 263 264 265 266 268 269 270 271 272

xiv

Contents

CHAPTER 11

Individual Characteristics and Problem Gambling Behavior Tony Toneatto and Linda Nguyen

Introduction Demographic Variables Age Gender Socioeconomic Status Marital Status Early Childhood Experiences Influence of Parental Gambling Motivation to Gamble Personality Factors Arousal and Sensation Seeking Sensation Seeking Arousal Impulsivity Drive Reduction Mood Regulation Dissociation Choice of Gambling Activity Cognitive Variables Illusion of Control Illusory Correlation: Superstitious Beliefs Interpretive Biases Attributional Biases The Gambler’s Fallacy Chasing Illusory Control over Luck Conclusion

280 280 280 281 282 282 283 283 284 286 286 286 287 288 289 289 290 290 291 292 292 293 293 294 294 294 295

CHAPTER 12

Comorbidity and Mental Illness Nancy M. Petry and Jeremiah Weinstock

Introduction Current State of Knowledge of Comorbidities with Pathological Gambling Substance Use Disorders Mood Disorders Anxiety Disorders Other Disorders

305 306 306 309 310 311

xv

Contents

Diagnostic Measurement Issues Structured Instruments for Assessing Psychiatric Disorders Assessment of Psychiatric Symptoms Research Concerns and Summary

311 312 315 318

CHAPTER 13

Research and Measurement Issues in Gambling Studies: Etiological Models Alex Blaszczynski and Lia Nower

Introduction Overview Etiological Models Psychoanalytic and Psychodynamic Public Health Social Reward Image Social Validation Behavioral Models Cognitive Conceptualizations Neurobiological, Genetic, and Biobehavioral Integrated Models General Theory of Addictions Biopsychosocial Pathways Summary

323 324 325 325 327 329 329 330 331 333 335 337 337 338 338 339

CHAPTER 14

The Neurobiology of Pathological Gambling Judson A. Brewer, Jon E. Grant, and Marc N. Potenza

Introduction Conceptualization Biochemistry Serotonin Dopamine Norepinephrine Monoamine Oxidase Stress Pathways Opioid System

345 346 347 347 348 350 351 352 353

xvi

Contents

Other Neurotransmitters Neuroimaging Structural Lesions and Decision Making Genetics Population Genetics Molecular Genetics Conclusions

353 354 356 357 357 358 359

CHAPTER 15

Treatment of Problem Gambling David C. Hodgins and Alice Holub

Introduction Major Approaches to Treatment Psychodynamic Approaches Theoretical Rationale and Therapeutic Model Psychodynamic Efficacy Research Unresolved Psychodynamic Issues Gamblers Anonymous Theoretical Rationale and Therapeutic Model Gamblers Anonymous Efficacy Research Unresolved Gamblers Anonymous Issues Behavioral Therapies Theoretical Rationale and Therapeutic Model Behavioral Efficacy Research Unresolved Behavioral Issues Cognitive and Cognitive-Behavioral Therapies Theoretical Rationale and Therapeutic Model Cognitive-Behavioral Efficacy Research Unresolved Cognitive-Behavioral Issues Brief Treatments and Self-Directed Treatments Theoretical Rationale and Therapeutic Model Brief Treatment Efficacy Research Unresolved Brief Treatment Issues Pharmacological Treatments Theoretical Rationale and Therapeutic Model Pharmacological Efficacy Research Unresolved Pharmacological Issues Alternative Approaches or Adjuncts to Therapy

372 373 373 373 373 373 374 374 374 375 375 375 377 378 379 379 380 382 382 382 383 385 385 385 386 387 388

Contents

Eye Movement Desensitization and Reprocessing Therapy Inpatient Programs Family Approaches Measurement/Evaluation Issues Summary and Conclusions

xvii

388 388 389 390 391

CHAPTER 16

Prevention of Problem Gambling Robert J.Williams, Beverly L.West, and Robert I. Simpson

Introduction Educational Initiatives to Prevent Problem Gambling Upstream Interventions Information/Awareness Campaigns More Sustained and Directed Educational Initiatives Policy Initiatives to Prevent Problem Gambling Restrictions on the General Availability of Gambling Restricting the Number of Gambling Venues Restricting More Harmful Types of Gambling Limiting Gambling Opportunities to Gambling Venues Restricting the Location of Gambling Venues Limiting Gambling Venue Hours of Operation Restrictions on Who Can Gamble Prohibition on Youth Gambling Restricting Gambling Venue Entry to Nonresidents Casino Self-Exclusion Contracts Restrictions or Alterations on How Gambling Is Provided On-Site Intervention with At-Risk Gamblers Problem Gambling Awareness Training for Employees of Gambling Venues Automated Intervention for At-Risk Gamblers at Gambling Venues On-Site Information/Counseling Centers Modifying Parameters of Electronic Gambling Machines Maximum Loss Limits Restricting Access to Money Restrictions on Concurrent Use of Alcohol and Tobacco Restricting Advertising and Promotional Activities Gambling Venue Design Independence Between Gambling Regulator and Gambling Provider Summary and Recommendations

400 401 401 401 404 406 406 406 410 411 412 413 414 414 415 415 417 417 417 418 418 420 420 421 422 423 424 424 425

xviii

Contents

CHAPTER 17

Adolescent Gambling: Current Knowledge, Myths, Assessment Strategies, and Public Policy Implications Jeffrey L. Derevensky and Rina Gupta

Introduction Adolescent Gambling Behavior Adolescent Problem Gambling Measurement Issues Related to Adolescent Problem Gambling Instruments Used to Assess Youth Problem Gambling Perspectives on the Adolescent Prevalence Data Understanding Adolescent Problem Gambling Behavior Is Pathological Gambling an Enduring Disorder? Are All Forms of Gambling Equally Dangerous? Correlates and Risk Factors Associated with Adolescent Problem Gambling Game Features, Technological Advances, and Environmental Factors Psychiatric and Mental Health Correlates Protective Factors Individual, Situational, and Environmental Factors Treatment Prevention Initiatives Concluding Remarks

438 438 439 440 440 441 442 443 443 444 447 448 448 449 450 453 456

CHAPTER 18

Cross-Cultural Comparisons Jan McMillen

Introduction Understanding the Relationship Between Gambling Research and Culture: The Importance of Theory Contending Theoretical Perspectives Normative Theories Behavioral Theories Sociological and Comparative Perspectives Problem Gambling: A Case Study of Research Cultures Theoretical Reflections The Way Forward

465 467 468 468 471 475 476 480 483

Contents

xix

CHAPTER 19

Internet Gambling: Past, Present, and Future Robert T. Wood and Robert J. Williams

Introduction History of Internet Gambling Current Situation Prevalence of Internet Gambling The Comparative Legality of Internet Gambling United Kingdom Other European Countries Australia New Zealand Canada United States Demographic Characteristics of Internet Gamblers Game-Play Patterns Why Do People Gamble on the Internet? Problems with Internet Gambling Unfair or Illegal Business Practices Unfair or Illegal Player Practices Internet Gambling by Prohibited Groups Problem Gambling Lack of Responsible Gambling Practices Future of Internet Gambling Researching Internet Gambling

492 492 494 495 496 496 497 497 497 497 498 499 500 501 501 501 502 502 503 505 506 508

CHAPTER 20

Social and Economic Impacts of Gambling Earl L. Grinols

Introduction Social Harm Economic Development The Eleven Components of Economic Development Cost–Benefit Analysis The Two-Sector Economy Revisited Measurement Summary and Conclusions

515 518 521 522 531 531 537

xx

Contents

CHAPTER 21

Gambling and Crime Colin S. Campbell and David Marshall

Introduction Categorizing Gambling and Its Relationship(s) to Crime Illegal Gambling Crimes Correlated to Problem Gambling Crimes Associated with Legal Gambling Expansion Crimes Correlated to Gambling Venues Crimes Distinct to Legal Gambling Operations Graft and Corruption Researching Illegal Gambling Researching Crime Correlated to Problem Gambling Researching the Impact of Gambling Facilities Methodological Challenges of Researching Gambling and Crime Data Problems The Social Construction of “Official Statistics” Demarcation of Gambling Concluding Observations

542 544 544 545 546 546 547 548 548 549 550 552 553 554 557 558

PART IV POLICY IMPLICATIONS OF GAMBLING RESEARCH CHAPTER 22

Values, Objectivity, and Bias in Gambling Research Jennifer Borrell and Jacques Boulet

Introduction Ontological and Epistemological Premises of Various Research Approaches Positivist (Empiricist) Approaches Interpretationist (Phenomenology) Approaches Critical-Dialectical (Structural/Structuralist) Approaches Critical-Participatory (Action-Oriented) Approaches Postmodernist (Postrelativist) Approaches Transpersonal-Ecological Approaches The Use of the Various Approaches in Existing Gambling Research Context and Corruption of Research: How Values Come to Influence Research Activities Some Contemporary Systemic Influences on Research Processes

568 568 569 570 571 571 572 573 574 575 577

Contents

Organizational and Institutional Influences Neo-Liberal Colonization of Universities and Research Institutions and the Primacy of Commercial Imperatives Government/Industry/Research Institution Collusion in Protecting Positions of Privilege Corruption of Science by Corporate Interests Within a Neo-Liberal Culture Governmental Influences Industry/Research-Institute Partnerships and Influence on Research Programs Values in Gambling Research and Their Relationship with Ideological, Cultural, and Systemic Influences Individualism Individual Pathology and Marginalization of the Problem Neo-Liberal Ideology, Individuals as Freely Choosing Consumers, and Utilitarianism Hiding Behind Putative Neutrality and Objectivity The Way Forward Multilevel Framework Understanding Harm Production Researcher Embeddedness Precautionary Principle

xxi

577 577 578 579 579 580 581 581 582 583 584 585 585 586 587 588

CHAPTER 23

Legalized Gambling: The Diffusion of a Morality Policy Patrick A. Pierce and Donald E. Miller

Introduction History of Legalized Gambling Theoretical Issues Data and Methodological Issues The Diffusion of Lotteries and Casinos The Diffusion of Innovations and Temporal Diffusion of Gambling Policies The Diffusion of Innovations and External Diffusion of Gambling Policies Lotteries Casinos Internal Diffusion of Gambling Policies The Changing Symbolic Weight of the “Sinfulness of Gambling” The Puzzle of Indian Casinos

593 594 595 596 598 598 602 603 605 606 609 609

xxii

Contents

The Future of Legalized Gambling Future Research

610 612

CHAPTER 24

Gambling and Governance Peter Collins

Introduction A Free Market in Gambling? Two Types of Arrangement for Regulating Gambling Government Objectives in Relation to Gambling Policy Vice Crime and the Gambling Industry Organized Crime and the Gambling Industry Defrauding Customers Money Laundering and Gambling Increases in Gambling Opportunities and in General Levels of Crime Problem Gambling Gambling Opportunities and Problem Gambling Numbers Economic Benefits Taxation Conclusion: Achieving Democratic Consensus About Gambling Policy

617 618 621 624 625 627 628 629 629 630 630 633 635 636

Index

641

637

CONTRIBUTORS Max W. Abbott (251), Health & Environmental Sciences, Auckland University of Technology, Auckland, 1020, New Zealand Alex Blaszczynski (323), School of Psychology, Department of Medical Psychology, University of Sydney, Sydney, New South Wales, 2150, Australia Jennifer Borrell (567), Borderlands Cooperative, Melbourne, Victoria, 3123, Australia Jacques Boulet (567), Borderlands Cooperative, Melbourne, Victoria, 3123, Australia Judson A. Brewer (345), Department of Psychiatry, Yale University School of Medicine, New Haven, Connecticut 6511 Colin S. Campbell (541), Department of Criminology, Douglas College, New Westminster, British Columbia, V3L 5B2, Canada David M. Casey (155), Addiction Centre, Foothills Medical Centre, University of Calgary, Calgary, Alberta, T2N 2T9, Canada Peter Collins (617), School of Accounting, Economics and Management Science, University of Salford, Salford, Greater Manchester, M5 4WT, United Kingdom Shawn R. Currie (155), Calgary Health Region, Calgary, Alberta, T2W 3N2, Canada Jeffrey L. Derevensky (437), International Centre for Youth Gambling Problems and High-Risk Behaviours, McGill University, Montreal, Quebec, H3A 1Y2, Canada G. Ron Frisch (179), University of Windsor, Windsor, Ontario, N9B 3P4, Canada Dean R. Gerstein (55), Claremont Graduate University, Claremont, California 91711 Richard Govani (179), University of Windsor, Windsor, Ontario, N9B 3P4, Canada Jon E. Grant (345), Department of Psychiatry, University of Minnesota, Minneapolis, Minnesota 55455 Mark Griffiths (217), Division of Psychology, Nottingham Trent University, Nottingham, NG1 4BU, United Kingdom xxiii

xxiv

Contributors

Earl Grinols (515), Department of Economics, Hankamer School of Business, Baylor University, Waco, Texas 76798 Rina Gupta (437), International Centre for Youth Gambling Problems and High-Risk Behaviours, McGill University, Montreal, Quebec, H3A 1Y2, Canada David C. Hodgins (371), Department of Psychology, University of Calgary, Calgary, Alberta, T2N 1N4, Canada Alice Holub (371), Department of Psychology, University of Calgary, Calgary, Alberta,T2N 1N4, Canada Steven Jefferson (87), Department of Psychology, Queen Elizabeth II Health Sciences Centre, Halifax, Nova Scotia, B3H 2E2, Canada David Marshall (541), Research and Community Engagement Division, Queensland Office of Gaming Regulation, Brisbane, Queensland, 4002, Australia Jan McMillen (465), Centre for Gambling Research, The Australian National University, Canberra, Australian Capital Territory, 0200, Australia Donald E. Miller (593), Department of Mathematics, Saint Mary’s College, Notre Dame, Indiana 46556 Linda Nguyen (279), Faculty of Nursing, University of Toronto, Toronto, Ontario, M5S 251, Canada Lia Nower (323), Center for Gambling Studies, Rutgers University, New Brunswick, New Jersey 08854 Jonathan Parke (217), Division of Psychology, Nottingham Trent University, Nottingham, NG1 4BU, United Kingdom Patrick A. Pierce (593), Center for Academic Innovation, Saint Mary’s College, Notre Dame, Indiana 46556 Nancy M. Petry (305), Department of Psychiatry, University of Connecticut Health Care Center, Farmington, Connecticut 6030 Marc N. Potenza (345), Department of Psychiatry, Yale University School of Medicine, New Haven, Connecticut 6511 Gerda Reith (3), Department of Sociology, Anthropology, and Applied Social Science, University of Glasgow, Glasgow, Scotland, G12 8RT, United Kingdom Robert I. Simpson (399), Ontario Problem Gambling Research Centre, Guelph, Ontario, Canada Wendy S. Slutske (127), Department of Psychological Sciences, University of Missouri–Columbia, Columbia, Missouri 65211

Contributors

xxv

Robert A. Stebbins (111), Department of Sociology, University of Calgary, Calgary, Alberta, T2N 1N4, Canada Sherry H. Stewart (87), Department of Psychiatry and Psychology, Queen Elizabeth II Health Sciences Centre, Dalhousie University, Halifax, Nova Scotia, B3H 2E2, Canada Randy Stinchfield (179), Department of Psychiatry, University of Minnesota Medical School, Minneapolis, Minnesota 55455 Marianna Toce-Gerstein (55), Georgetown University, Washington, DC 20009 Tony Toneatto (279), Clinical Research Department, Center for Addiction and Mental Health; Departments for Psychiatry and Public Health Sciences, University of Toronto, Toronto, Ontario, M5S 2S1, Canada Rachel A. Volberg (33), Gemini Research Ltd., Northampton, Massachusetts 01061-1390 Jeremiah Weinstock (305), Department of Psychiatry, University of Connecticut Health Care Center, Farmington, Connecticut 6030 Beverly L. West (399), School of Health Sciences, University of Lethbridge, Lethbridge, Alberta, T1K 3M4, Canada Robert J. Williams (399, 491), School of Health Sciences, University of Lethbridge, Lethbridge, Alberta,T1K 3M4, Canada Robert T. Wood (491), Department of Sociology, University of Lethbridge, Lethbridge, Alberta, T1K 3M4, Canada

This page intentionally left blank

PREFACE Garry Smith Alberta Gaming Research Institute University of Alberta

Humankind’s experience with gambling shows the activity to have been variously regarded as a sin, a vice, an unfortunate but tolerable weakness of human nature, an evolutionarily appropriate behavior, an adult form of play, a release from daily routines, a means of increasing arousal, an intellectual challenge, a buffer to existential anxieties caused by chance events, a prism of cultural values, and an activity that fosters preoccupation, escapism, conditioned responses, illusion of control, and addictive behavior. Gambling dates back to early recorded history, yet across societies past and present, gambling has varied considerably with respect to organization, social meanings, and how it is regarded in moral terms (Binde 2005). Moreover, contrasting degrees of tolerance, prohibition, and ambiguity have been shown toward the activity depending on the culture and historical era (Wykes 1964). For example, receptivity toward gambling in the United States has been cyclical, with the legitimacy of gambling having oscillated from prohibition to approval and back again several times. “Twice before in American history players could make legal bets in almost every state, but these waves of legal gambling came crashing down in scandal and ruin” (Rose 1991, p. 71). At the time of his writing, Rose speculated that America was in the midst of the third wave, with a reversal of gambling’s favored status expected to occur in a few more decades. While perhaps less pronounced than the American experience, public attitudes in most Western cultures toward gambling have tended to be cyclical. Despite the ubiquity and persistence of gambling through the ages, there is little evidence to suggest that any society has discovered an exemplary way to regulate it—that is, fair games offered with the proceeds going toward important societal benefits, gambling-related corruption and crime kept in check, and social and economic damages minimized. In line with this observation, David Allen (1952) posed the question: “What ought to be thought about gambling?” Referring to the role of government and commenting over a half century ago, Allen described American gambling policy as a matter of “first class importance,” ranking in gravity with “foreign policy and domestic tax issues.” While no doubt a hyperbole at the time, the exponential growth of legal gambling over the past few decades now makes Allen’s concern especially prescient. According to Allen, a consideration of best practices in gambling regulation should start by addressing three salient questions: (1) Is gambling a natural activity (i.e., rooted in the human psyche)? (2) Is gambling harmful (i.e., injurious to society or its members, and if so, under what circumstances)? and (3) Is gambling suppressible? xxvii

xxviii

Preface

In the first two-thirds of the twentieth century, the role of gambling in society was addressed in influential works by Mackenzie (1928), Asbury (1938), Devereux (1949), Ploscowe and Lukas (1950), Allen (1952), Bergler (1958), Ezell (1960), David (1962), Reid and Demaris (1963), and Turner (1965). These works served as a foundation for the emergent field of gambling studies but were written at a time when most forms of gambling were illegal and the activity still had a tainted reputation. Hence, they tended to be anti-gambling tracts, attempts to explain the appeal of gambling through the ages, efforts to fathom the minds of problem gamblers, or descriptions of particularly sordid gambling scenes. The point is that learned individuals have struggled with these questions through the ages, yet we still lack definitive answers. On a more positive note, we are moving closer to disentangling the mystery of gambling’s impact on society. Prior to a general trend that saw liberalized gambling laws and modest gambling expansion beginning in the 1970s, the study of gambling was a decidedly lowprofile academic pursuit. Gambling expansion in the 1970s occurred at a moderate pace under tightly regulated conditions and was driven by an attempt to restrict illegal markets and generate revenues for social programs such as welfare, sports, the arts, and others identified as “worthy causes” (Kingma 2004). A gambling policy paradigm shift occurred in the 1980s as a result of legislation lagging behind aggressive gambling practices, thus creating a situation whereby “politics gave in to market demands without convincing and conclusive (legal) justification” (Kingma 2004, p. 55).The upshot of this policy inversion was (1) accelerated gambling expansion, especially via electronic formats, (2) diversion of gambling proceeds away from decentralized social welfare programs and toward government treasuries, and (3) government application of corporate principles and strategies to market gambling offerings. Suddenly gambling had become a powerful economic, political, and cultural force that augmented government and gaming industry coffers but also posed significant economic and public health risks for individuals and communities. Academic interest was piqued by this rapid and radical transformation of gambling as scholars were keen to know how it was that a recently stigmatized and circumscribed activity had suddenly become a legitimate entertainment option. Why were previously outlawed activities now being sanctioned and promoted by government? And, how did this profusion of legal gambling opportunities affect citizen and community well-being, for better or worse? (Now, some twenty years later, legal gambling has gone viral and operates on a scale that was unimagined then.)1 1 While gambling is certainly big business, it is not an essential service in the same way as traditional government responsibilities, such as education, agriculture, and health care. Presumably, governments exist to make a jurisdiction a better place to live; government involvement in sanctioning and promoting gambling is thus an anomaly, because not only is gambling not an essential service, but public attitude surveys indicate that a majority of citizens feel that some forms of gambling actually detract from a community’s quality of life (Azmier 2000).

Preface

xxix

In Chapter 1 of this volume, Gerda Reith details how this rampant expansion of gambling facilitated the emergence of gambling studies as an accepted academic subfield in well-established disciplines such as medicine, law, business, psychology, economics, political science, and so forth. This nascent subfield quickly advanced through several growth phases highlighted by compulsive gambling being recognized as a “disorder of impulse control” by the American Psychiatric Association in the third edition of the Diagnostic and Statistical Manual of Mental Disorders; the development of assessment tools to identify pathological gamblers; the emergence of problem gambling prevalence surveys and national gambling impact studies; and the growing availability of research funding, which had the salutary effect of attracting new scholars to the field and, in some cases, helped to establish gambling-related institutes/centers. Current advancements in gambling studies include universities offering tenure-track faculty positions and postdoctoral fellowships in gambling studies and including gambling studies courses in the curriculum; the advent of longitudinal mega-projects featuring sophisticated research designs and interagency collaboration designed to establish broad and rich databases; and the appearance of Internet websites specializing in gambling studies topics, thus allowing scholars to easily interact with their colleagues and to access gambling reports and gambling news items from around the globe. While much has been accomplished by the dedicated scholars who have created a critical mass of knowledge in gambling studies, much remains to be done. For example, the field remains encumbered by imprecise terminology, a consensus has yet to emerge on the best way to measure out-of-control gambling in general population surveys, and little headway has been made on conducting valid and reliable cost/benefit analyses. The idea for this book originated in discussions among the editors concerning the need for more detailed information on gambling studies research. The book’s premise is that the more we research and the more precise our measuring tools, the better we will be able to comprehend gambling and its effects on individuals and society at large. Accordingly, in planning the book, we assumed two objectives: first, to expose readers to the most widely used research approaches and methodologies in gambling studies, and second, to highlight critical research issues that currently challenge scholars in the field. Research and Measurement Issues in Gambling Studies is intended as a reference text for advanced students and academics working in the field of gambling studies but should also interest problem gambling treatment providers, gambling policy makers, and laypersons interested in gambling as a cultural phenomenon. Gambling studies measurement and methodological concerns are addressed by experts who have provided state-of-the-art synopses and critiques of their specialty areas. In recruiting authors, the editors sought internationally recognized gambling studies scholars and asked them to synthesize the generally accepted knowledge in their

xxx

Preface

areas, comment on methodologies and measurement strategies used (what does and does not work), and expound on research trends, information gaps, and future research prospects. To engage this task, talented and versatile contributors from Australia, Canada, England, New Zealand, Scotland, and the United States who collectively represent a diverse array of academic disciplines (psychology, sociology, economics, cultural studies, political science, biochemistry, addiction studies, and health sciences) were enlisted. The result, we hope, is a timely and comprehensive compendium of research and measurement concerns in gambling studies. The book has four sections: (1) the nature and scope of gambling studies, (2) measurement approaches in gambling studies, (3) emergent gambling studies research issues, and (4) ethical considerations and policy implications related to gambling studies research. The first section consists of just one chapter, Gerda Reith’s “Situating Gambling Studies.” Reith’s chapter sets the stage for the remainder of the book, as she explains how gambling studies evolved, describes the research domains that formed to analyze and interpret gambling and problem gambling behavior, and speculates on how gambling studies might unfold in the future. The second section comprises seven chapters, each dealing with an important area of measurement in gambling studies research. In Chapter 2, Rachel Volberg provides a primer on the ways and means of conducting gamblingrelated population surveys. Volberg delineates why population surveys are an expedient and acceptable way to collect data on citizens’ gambling proclivities, then guides the reader through the process, indicating potential pitfalls and best practices. Chapter 3, by Marianna Toce-Gerstein and Dean Gerstein, deals with questionnaire design and development as applied to gambling studies. The authors discuss the interview context (e.g., different types of interviews, factors related to how and where interviews are conducted and response bias); communication principles pertaining to the relevance, neutrality, and ambiguity of questions; ways of minimizing cognitive effects; and how to structure questionnaires so that they flow smoothly and elicit valid responses. Sherry Stewart and Steven Jefferson review the key components of experimental research design in Chapter 4 and show how this approach is used in studies of gambling behavior.The advantages and drawbacks of this method are discussed, and the primacy of this method is noted because it is the only approach that allows the researcher to infer causality. In Chapter 5, Robert Stebbins shows how qualitative research methods can be used to improve our understanding of gambling scenes and behaviors and can ultimately lay a foundation for grounded theory. Qualitative methods are most commonly used in exploratory studies of unknown social phenomena, but Stebbins also recommends this approach for verification studies.

Preface

xxxi

In Chapter 6, Wendy Slutske explores the challenges and benefits of conducting longitudinal studies. This approach allows the researcher to resolve temporal relations between gambling behavior and its correlates and to trace gambling behavior and problem gambling development; however, it has been sparingly used in gambling studies research, mainly because of the high cost and need for experienced, academically diverse research teams. Chapter 7 focuses on the quantification and dimensionalization of gambling behavior. Shawn Currie and David Casey discuss applications to gambling from studying the quantification of other addictive behaviors and highlight the difficulties in epidemiological data collection related to gambling frequency, duration, and expenditure. The authors note that the lack of a standard unit of gambling (e.g., an equivalent to the number of alcohol drinks or number of cigarettes smoked) has hindered gambling research in this area. In Chapter 8, Randy Stinchfield, Richard Govoni, and G. Ron Frisch guide readers through the maze of assessment tools that have been used to screen for problem gambling behavior in clinical, general, and special populations. The measurement of problem gambling has been beset by controversy and a lack of precision, as evidenced by the dozen or more instruments that have been designed for this purpose. Refinements to the currently used instruments related to psychometric properties, cut scores, reliance on self-report data, and the time period being assessed are recommended. There are thirteen chapters in the third section, each dealing with a topic that has attracted scholarly attention in recent years. Chapter 9 features a discussion by Jonathan Parke and Mark Griffiths on the structural characteristics of electronic gambling machines (EGMs) and their effect on player behavior. EGMs are singled out because theirs is reputed to be the most hazardous gambling format. In addition to categorizing the structural features of EGMs, the authors call for more ecologically valid studies and note how the competing interests of gambling purveyors, gaming machine manufacturers, and gambling consumers have limited research in this area. Chapter 10, by Max Abbott, elaborates on how situational factors impact gambling behavior. These factors include the availability of various gambling formats, accessibility to gambling formats, hours of operation, pricing, alcohol and tobacco consumption while gambling, on-site automated teller machines (ATMs), and the amount and type of advertising deployed. Abbott claims there is growing evidence to link gambling availability to increased gambling participation, and in some cases, to problem gambling. Also influencing a person’s response to gambling opportunities are idiosyncratic factors such as coping styles, self-esteem, genetic makeup, faulty cognitions related to the mechanics of gambling, and demographic profile (e.g., age, gender, ethnicity, education level).The effect of these characteristics on gambling behavior is examined by Tony Toneatto and Linda Nguyen in Chapter 11. The authors

xxxii

Preface

suggest measuring the way that individuals perceive and interpret the act of gambling as a promising new research strategy. In Chapter 12, Nancy Petry and Jeremiah Weinstock probe the links between problem gambling and accompanying psychiatric conditions such as substance abuse, depressive mood states, and anxiety disorders. Diagnostic measurement issues and the assessment of psychiatric symptoms are discussed, along with implications for treatment. In Chapter 13, Alex Blaszczynski and Lia Nower describe and analyze the models used to explain the development and persistence of problem gambling. The merits and shortcomings of nine models are discussed, leaving the authors to conclude that no single empirically validated model decisively accounts for all problem gambling behavior. Researchers Judson Brewer, Jon Grant, and Marc Potenza elaborate on the neurobiological basis of problem gambling in Chapter 14. Included in their discussion is the relative influence of brain functioning and genetic factors in contributing to problem gambling behavior. Also reviewed are neuroimaging studies that examine the regions of the brain activated during gambling participation for the purpose of differentiating between problem gamblers and control subjects. Chapter 15 covers treatment issues related to problem gambling. David Hodgins and Alice Holub review the theoretical rationale and therapeutic models for seven common treatment approaches. The authors discuss the challenges related to evaluating treatment outcomes and suggest research avenues that would improve our knowledge in this area. In Chapter 16, Robert Williams, Beverly West, and Robert Simpson thoroughly review the educational and policy initiatives designed to prevent problem gambling. The authors note that some approaches are more efficacious than others, but they maintain that all efforts have some additive value and that, for maximum impact, a coordinated approach is needed. Jeff Derevensky and Rina Gupta investigate adolescent gambling in Chapter 17. Topics addressed include measuring youth problem gambling, interpreting prevalence data, and identifying correlates and risk factors associated with adolescent gambling behavior, along with treatment and prevention issues. Jan McMillen employs a sociocultural framework to review key theoretical and methodological issues pertaining to the cross-cultural study of gambling in Chapter 18. Her contention is that gambling research has been dominated by North American–based psychiatry, clinical psychology, and quantitative methodologies. The limitations of this orientation are discussed and it is suggested that gambling studies research needs a unifying rationale, which can best be achieved through a multitheoretical approach using a plurality of methods. Robert Wood and Robert Williams lead us through the murky world of Internet gambling in Chapter 19. Here, we learn about the history and prevalence of Internet gambling, its legal status in various worldwide jurisdictions, who is

Preface

xxxiii

playing the games, and their game preferences. Because Internet gambling research is a relatively recent phenomenon, the authors provide a section on the unique challenges of studying this topic. Social and economic impact studies of gambling have long been debated among gambling studies scholars; the general belief is that they are desperately needed but that there is no consensus on how to execute such studies. In Chapter 20, Earl Grinols identifies the shortcomings of previous research in the area and specifies what he believes the principles of sound empirical research to be, as well as the social cost categories that must be addressed. Chapter 21 expands on a major gambling-related social cost; namely, increased crime rates. Colin Campbell and David Marshall summarize the historical association between crime and gambling, categorize and comment on typical gambling-related crimes, and outline methodological strategies to improve our knowledge in the area. The fourth section is concerned with policy issues related to gambling studies research and includes three chapters. In Chapter 22, Jennifer Borrell and Jacques Boulet examine the underlying premises of mainstream research positions; comment on the assumed relationships between gambling studies researchers and their subjects and subject matter; and show how personal and cultural values influence the topics studied, how they are studied, and how research findings are interpreted. Chapter 23, by Patrick Pierce and Donald Miller, is a case study on the politics of gambling that details how lotteries and legal casino gambling spread from state to state in America between 1966 and 2004. Using diffusion of innovations theory, the authors identify the presence of large religious fundamentalist populations in a state as the main barrier to gambling expansion and the availability of similar gambling formats in neighboring states as the key to adoption. In Chapter 24, Peter Collins compares and contrasts the ways that governments regulate gambling and argues that commercial gambling is subject to much broader and more rigorous regulation than are most other business enterprises. Collins assesses the reasons given for this stricter regulation and concludes that sound gambling policy is a delicate matter of judging public attitudes and balancing competing interests. We wish to thank the contributors to this volume for their collegiality and enthusiasm for the project. Although the individual chapters reflect considerable variations in mode and emphasis, we appreciate the professionalism of these scholars in meeting our deadlines and adapting their writing styles and conceptualizations of the subject matter to satisfy our specific requirements. The editors realize that important gambling studies–related methodological and research issues may have been overlooked in our gathering of topics for this book. If that is the case, the blame is ours. We hope that readers of this collection will develop a better understanding of the field of gambling studies and that they will consider applying the insights generated here to their own research initiatives.

xxxiv

Preface

REFERENCES Allen, D. (1952). The Nature of Gambling. New York: Coward-McCann. Asbury, H. (1938). Sucker’s Progress: An Informal History of Gambling in America from the Colonies to Canfield. New York: Dodd, Mead and Co. (Reprinted 2003 by Thunder’s Mouth Press) Azmier, J. (2000). Canadian Gambling Behavior and Attitudes. Calgary, AB: Canada West Foundation. Bergler, E. (1958). The Psychology of Gambling. London: Bernhard Harrison Ltd. Binde, P. (2005). Gambling across cultures: Mapping worldwide occurrence and learning from ethnographic comparison. International Gambling Studies, 5, 1–27. David, F. N. (1962). Games, Gods and Gambling: A History of Probability Theory and Statistical Ideas. New York: Charles Griffin. Devereux, E., Jr. (1949). Gambling and the Social Structure: A Sociological Study of Lotteries and Horse Racing in Contemporary America. Ph.D dissertation, Harvard University. Ezell, J. (1960). Fortune’s Merry Wheel. Cambridge, MA: Harvard University Press. Kingma, S. (2004). Gambling and the risk society:The liberalization and legitimation crisis of gambling in the Netherlands. International Gambling Studies, 4, 47–67. Mackenzie,W. D. (1928). The Ethics of Gambling. Garden City, NY: Doubleday, Doran & Company. Ploscowe, M., and Lukas, E. J. (eds.). (1950, May). Gambling. Annals of the American Academy of Political and Social Science, 269. (Special issue) Reid, E., and Demaris, O. (1963). The Green Felt Jungle. New York: Trident Press. Rose, I. N. (1991). The rise and fall of the third wave: Gambling will be outlawed in forty years. In Gambling and Public Policy: International Perspectives (W. Eadington and J. Cornelius, eds.), pp. 65–86. Reno, NV: Institute for the Study of Commercial Gaming. Turner,W. (1965). Gamblers’ Money. New York: Signet Books. Wykes,A. (1964). The Complete Illustrated Guide to Gambling. Garden City, NY: Doubleday & Company.

ACKNOWLEDGMENTS We would like to acknowledge the support of the Alberta Gaming Research Institute (AGRI) for providing us with the opportunity and resources to study gambling issues and the Alberta government for its foresight in establishing the AGRI eight years ago. Finally, we owe a debt of gratitude to the encouragement and support provided by Scott Bentley and Kathleen Paoni, our Elsevier contacts; Scott, for suggesting the project and handling the administrative details, and Kathleen, for guiding us efficiently and painlessly through the process. We consider ourselves most fortunate to have partnered with them to produce this publication.

xxxv

PART I

Nature and Scope of Gambling Studies

This page intentionally left blank

CHAPTER 1

Situating Gambling Studies Gerda Reith Department of Sociology, Anthropology, and Applied Social Science University of Glasgow Glasgow, Scotland, United Kingdom

Introduction The Historical Context of Gambling The Paradigm Shift The Emergence of “Gambling Studies” Research Domains Interpretivist Approaches Sociological, Anthropological, and Psychological Research Positivist Approaches Economic and Social Cost–Benefit Analyses Biomedical Approaches Clinical Psychological Research Cognitive Psychological Research Epidemiological Research and Public Health Perspectives Conclusions: Current Trends and Future Directions

INTRODUCTION Gambling is a nearly ubiquitous activity that has been practiced throughout history and across cultures by various social groups. For hundreds of years, individuals have gambled for excitement and escapism, to win money, to gain status, to be sociable—the list is almost as diverse as the variety of games. However, in spite of this heterogeneity, one feature that is constant in all forms 3

4

Research and Measurement Issues in Gambling Studies

of gambling is the redistribution of something of value (usually money) by the operation of chance. Although the measurement of value and the degree of chance may vary substantially, the involvement of both is fundamental for a wager to take place. Explanations, criticisms, and, more recently, scientific analyses of gambling and problem gambling have consistently been embedded in their particular sociohistorical climates. Their concerns and preoccupations and even the language they are articulated in emerge out of wider cultural processes.Today,“gambling studies” is a uniquely complex, as well as contested, field. It comprises a variety of “ways of seeing” its subject, some of which are complementary, others not, and none of which have as yet fully accounted for the myriad complexities of the topic. This introductory chapter does not attempt to answer the “big questions” that still surround the study of gambling, but rather undertakes the more modest tasks of mapping, in admittedly partial form, development of the field and outlining some other major themes. The chapters in this volume then provide detailed analysis of the central research and measurement issues in gambling studies. And so, to begin, this chapter provides a brief historical introduction to criticisms and conceptualizations of gambling before moving on to examine the socioeconomic climate of the proliferation of the activity in the twentieth and twenty-first centuries. It then examines the epistemological heritage of the various research domains that have developed around gambling and problem gambling, before finally providing an overview of possible future trends.

THE HISTORICAL CONTEXT OF GAMBLING Throughout history, perhaps the only thing that has been as ubiquitous as gambling has been condemnation of the activity, a feature that has cast a persistent shadow over its popularity. Although the terminology of the debate has changed according to its sociohistorical climate, generally the tone has been one of open hostility to what has been considered a “deviant” activity—especially when the gamblers in question have been members of economically marginal groups. From the Protestant Reformation onward, games of chance have been condemned as sinful for their undermining of the work ethic, whereby diligence and effort were supposed to be rewarded with an appropriate level of wealth and success. The rewards of gambling, on the other hand, are distributed by pure chance, and entirely disconnected to effort or merit, in a way that seemingly threatens to disrupt the social hierarchy and the ideology of meritocracy and hard work that it was based on. In the more secular period of the Enlightenment,Western societies criticized the element of irrationality that appeared to be embodied in games of

Situating Gambling Studies

5

chance. Beliefs in mystical forces such as luck and fate, rather than dynamic ideas about personal agency and effort, contradicted ideals of human reason and, along with the vast amounts of waste and loss that were often involved in games, seemed to be anathema to contemporary social values. Although gambling had been widely condemned, specific explanations for gamblers’ behavior that went beyond criticisms of sin and irrationality did not appear until around the nineteenth century. At that time, the economic activities of industrializing nations such as Great Britain and the United States threw the apparently haphazard economic actions of gamblers into sharp relief and led to a renewed focus on the “wasteful” and “immoral” aspects of players themselves. With the Industrial Revolution, discussions about gambling separated slightly from broader ideological discourses of religion and reason and started to focus on individual moral defects instead. In this era of social Darwinism, biomedical “theories” sought to explain supposedly immoral or deviant behavior with reference to defects in individual physiology. Gambling was caught up in discourses about other vices, such as alcoholism, addiction, and prostitution, with concerns expressed over the possible hereditary effects of the “disease.” During this period, gambling moved into the clinic, with the psychoanalyst Sigmund Freud (1928) turning his attention to a disorder that was defined as a compulsive neurosis. Freud related gambling to the Oedipus complex, declaring it to be grounded in unresolved childhood conflict and based on guilt and masochistic self-punishment. The influence of the psychoanalytic perspective continued well into the twentieth century, with some writers, such as Edmund Bergler (1970), focusing on the dysfunctional, masochistic qualities of “degenerate gamblers,” and others, such as Robert Lindner (1976) and Ralph Greenson (1974), declaring gamblers to be sick individuals in the grip of a regressive disease. The activity fared little better in the sociological literature, dominated at that time by functionalist perspectives that attempted to explain human action in terms of its latent function within broader social structures. In these, gambling was viewed as a mechanism for releasing the tensions and frustrations of everyday life–– a cathartic role that was particularly compelling for those at the bottom of the social hierarchy, and for whom such pressures were regarded as more acute. This link between gambling and socioeconomic deprivation influenced much of the American sociology of gambling of the 1950s, 1960s, and 1970s, with researchers portraying gambling as a compensatory activity within which the frustrations of the outside world could be worked out in symbolic form (Bloch 1951; Devereaux 1949; Herman 1967; Newman 1972). These psychoanalytic and functionalist perspectives dominated explanations of gambling at a time when the activity was, although undoubtedly widespread, largely illegal throughout Europe and America. Up until the late 1960s, when not outlawed altogether, most forms of gambling were tightly regulated by restrictive legislation and were frequently dogged by scandals involving corruption and fraud, lending games of chance the air of an underground and morally dubious enter-

6

Research and Measurement Issues in Gambling Studies

prise. It is perhaps not surprising then that attempts to describe and explain what would have been a largely invisible activity were colored by the prohibitory climate in which the latter was embedded.

THE PARADIGM SHIFT In the mid-twentieth century, economic, social, political, and cultural changes transformed Western societies and encouraged a paradigm shift out of which the activity of gambling and various forms of problem gambling emerged as serious objects of study in their own right. From around the 1970s onward, neo-liberal economic policies have encouraged a reduction of state intervention in both public and private life and reluctance on the part of policymakers to regulate markets or to impose high levels of taxation. Such policies have encouraged alternative forms of revenue generation, of which taxing the profits of commercial gambling is an attractive option. At the same time, relatively increasing affluence and the spread of consumerism have begun to undermine arguments about the “immorality” of gambling and have created instead a cultural climate that if not conducive, is not exactly hostile, to the proliferation of gambling as a mainstream leisure activity. And so, since the 1980s in particular, the political economy of gambling has been increasingly deregulated, resulting in a period of dramatic expansion and the proliferation of commercial gambling as a global, multibillion-dollar enterprise. Today, between 60 and 80% of the populations of Western societies gamble regularly, including increasing numbers of younger people, women, and the middle class—the latter being the group traditionally most hostile to gambling.The normalization brought about by such a demographic shift is reflected in changing nomenclature.The euphemism “gaming,” with its connotations of play and leisure, is increasingly favored (at least by the industry) over the traditional term, “gambling,” with all its connotations of financial loss. By the start of the twenty-first century, the fortunes of gambling have been transformed, and the gambling industry has become a profitable business, selling hope, excitement, and thrills to ever-larger numbers of consumers.The industry’s task has been aided by the increased availability of credit to larger sections of the population. In particular, middle-class consumption of credit has expanded at rates similar to gambling, greatly increasing access to the activity beyond the levels imposed by “hard” money.This expansion also taps into a wider “consumerist” ethos within culture as a whole. Not only are games of chance a product of consumer culture, they also express some of its most fundamental characteristics, such as the values of instant gratification, self-fulfillment, and conspicuous consumption. Gambling can thus be described as a leisure activity with a high degree of “cultural capital” (Bourdieu 1984/1979).

Situating Gambling Studies

7

It can be seen, then, that by the end of the twentieth century a range of political, economic, and cultural factors had contributed to the dramatically changed status and expansion of commercial gambling.

THE EMERGENCE OF “GAMBLING STUDIES” Responses to the shift in the status and availability of gambling ranged from curiosity to concern and hostility as a range of groups, including health professionals, community leaders, politicians, religious organizations, and the general public, began to consider the potentially negative impacts of gambling on individual and social life. The expansion also set the stage for the study of gambling as an important and legitimate field of academic inquiry.The topic of gambling touched on many existing disciplines, including sociology, psychology, economics, law, and medicine, and as well as forming small but distinct subfields in these areas, it also established one of its own.This development is apparent in the increasing number of specialized journals, conferences, research institutes, academic departments, and funding bodies dedicated to the study of gambling that have come into being in the past two decades or so. Broadly speaking, the general area of “gambling studies” can be divided into two approaches. On the one hand, what can be loosely termed interpretivist perspectives are concerned with the interpretations of the meanings, cultures, and contexts of gambling. Such an approach is based on the premise that social meanings are created through the intentions and understandings of individuals, which in turn are embedded in culturally and historically specific conditions. Utilizing qualitative, ethnographic methods, such as interveiws, participant observation, focus groups, and case studies, researchers in disciplines such as sociology, anthropology, and sometimes psychology have attempted to understand the meanings and roles of gambling in the everyday lives of participants, as well as the attitudes and motivations of various groups of players. The focus of such approaches is frequently, although not exclusively, on recreational or nonproblematic gambling. On the other hand, what can be loosely described as positivist perspectives have developed out of the methodologies utilized by the natural sciences. In general, these methods are assumed to be value free and aim to establish causal relations between various factors that will allow the prediction and ultimately the control of behavior.These approaches are concerned with the measurement and quantification of gambling, and particularly problem gambling, in both clinical and population-based samples. Such approaches tend to utilize quantitative methodologies such as surveys, questionnaires, and laboratory-based research, and are found in psychology and sociology, as well as in disciplines such as medicine and economics. The distinction between these two approaches is by no means absolute, nor is it always clear-cut. For instance, disciplines such as psychology and sociology can

8

Research and Measurement Issues in Gambling Studies

encompass both interpretivist and positivist methodologies. The techniques and focus of qualitative and quantitative sociology are quite distinct, as are those of, for example, clinical psychology compared with social psychology, psychoanalysis, or counseling-based approaches. Moreover, many approaches to the study of gambling are interdisciplinary in nature, utilizing insights from a number of disciplines and methodologies. The remainder of this chapter now moves on to discuss developments in the field of gambling studies. For the purposes of clarity and structure, it uses the distinction between interpretivist and positivist perspectives to organize these into two research domains, although, as it is by far the most dominant approach, the discussion of the latter is considerably longer and is subdivided into a number of smaller fields. It should be noted here that the material included in this chapter is a highly selective sample. Such selectivity is justified not only by the constraints of length, but also by the fact that this essay is designed to provide an overview of broad themes and trends, rather than an exhaustive analysis of the literature on gambling. Therefore, it is not intended as a literature review, but should rather be read as a thematic overview of the development of various research domains.

RESEARCH DOMAINS INTERPRETIVIST APPROACHES Sociological, Anthropological, and Psychological Research In disciplines such as sociology and anthropology, a variety of fairly disparate studies have focused both on gambling’s structural determinants and on its meanings for participants. The subject of these types of approach tends to be gambling settings rather than gamblers as individuals, and on the sociocultural factors that influence patterns of behavior (Volberg 2001). Although relatively few in number, some studies have shed light on structural factors, such as socioeconomic class and gender, that influence patterns of gambling and preferred types of games (Chinn 1991; Clotfelter and Cook 1989; Downes et al. 1976). At the same time, the issue of individual agency has been of interest to sociologists and anthropologists, who have examined gambling as a form of leisure with its own unique experiential frame. Many writers in this tradition have been influenced by the pathbreaking studies of Erving Goffman (1969) and Clifford Geertz (1975), who utilized symbolic interactionist and interpretivist anthropological perspectives to provide “thick” accounts of the meanings and values of gambling for participants. By focusing on the role of games in the lives of individuals as well as the local cultures within which they were embedded, these authors

Situating Gambling Studies

9

highlighted the nonmaterial gains to be had from gambling and the rewards of risk taking. These included the display of character, the expression of individual and community identity, and the earning of status and prestige within one’s peer group. Insights from studies such as these have been utilized in various ways, often in research on the cultures that surround specific games and groups of players. Here, the focus is on the social contexts of gambling behavior and the social roles and rewards that make it meaningful for players. Such studies often use participant observation, ethnographic research, or various other qualitative techniques in an attempt to uncover the “social worlds” of gambling. Early studies of this type focused on the distinctive forms of sociation that made up various gambling subcultures, including horse-race betting and poker playing. Through in-depth interviews and immersion in the everyday lives of their subjects, researchers such as Henry Lesieur (1984), Marvin Scott (1968), David Hayano (1982), and John Rosecrance (1985) were able to provide firsthand accounts of the shared values, attitudes, and systems of knowledge that distinguished gambling from other forms of social behavior. More recent studies in this tradition have continued to examine the motivations, meanings, and cultures of gambling involvement, often focusing on particular social groups and sociocultural contexts and settings of play.They include studies of (mostly male) participation in horse-race betting, the social and gendered nature of bingo, the dynamics of fruit (slot) machine playing—especially among youth—and the meanings of games for lottery players (Bruce and Johnson 1992; Dixey 1996; Falk and Maenpaa 1999; Fisher 1993; Neal 1998). More recent studies include indepth ethnographic accounts of gambling in specific settings, such as the culture and kinship of the British racecourse (Cassidy 2002) and the local rituals of gambling in a Greek community (Malaby 2003). Other studies have utilized ethnographic methods to document the social and gendered meanings of gambling in traditional societies (Goodale 1987; Sexton 1987; Zimmer 1987). By outlining the meanings that games have for participants, such studies have remained within the interpretivist tradition associated with anthropologists such as Geertz and have also demonstrated how gambling is adapted to the particular needs of local communities. At the same time, many have noted the absence of “Western” concepts of problem gambling within such cultures. Finally, more theoretical accounts have used documentary analysis to describe the social and historical construction of gambling in Western societies, exploring its changing status and relation to broad social movements and global processes.These have sought to show that gambling is both historically situated and culturally specific, and also that responses to it are frequently embedded in particular political and ideological contexts.The broad sweep of such accounts has analyzed gambling on both a wide, global scale (Brenner and Brenner 1990; McMillen 1996; Reith 1999) and in terms of its development in specific Western nations, such as the United States (e.g., Abt, Smith, and Christiansen 1985), Canada

10

Research and Measurement Issues in Gambling Studies

(e.g., Morton 2003), Australia (e.g., Caldwell et al., 1985), and Great Britain (e.g., Dixon 1991).

POSITIVIST APPROACHES However, by far the greatest attention to gambling as an activity has been concerned with positivist approaches toward the measurement and quantification of gambling activity, and within that approach, has been directed largely toward gambling that is problematic. Although the potentially deviant and/or problematic nature of gambling had long been subject to a range of critical discourses, the notion of “pathological gambling” as a phenomenon in its own right did not appear until the 1980s. At that time, it was introduced into the third edition of the reference manual published by the American Psychiatric Association, the Diagnostic and Statistical Manual of Mental Disorders (DSM-III) (APA 1980), where it was described as an impulse control disorder—a compulsion characterized by an inability to resist overwhelming and irrational drives. Focus soon shifted to its addictive characteristics, and subsequent editions of the manual saw it reclassified in terms similar to those for psychoactive substance dependency, with the term “pathological gambling” consistently used to reflect its chronic, progressive character (APA 1987, 1994). Around the same time, the South Oaks Gambling Screen (SOGS) was developed—again, with loss of control as a defining category, and it became widely used for the measurement of gambling problems throughout populations (Lesieur and Blume 1987). Since then, over 20 screens have been developed for a range of purposes, including screening, diagnosis, population monitoring, and treatment planning. In general, they define “pathological” and “problem” gambling as behavior that is out of control and that has come to disrupt personal, family, social, and vocational life, with the former regarded as a more severe condition than the latter.1 With the development of a system of classification and nomenclature, these evaluations introduced a distinct “type” of individual—a pathological gambler.This individual was assessed against a checklist of formal symptoms which could be measured and compared against a norm, so distinguishing them as different in some way from other players. Commentators on this development have drawn on the theory of Michel Foucault (1976) to analyze the generation of notions of problem and pathological gambling as a distinctive social—or “discursive”—process (Castellani 2000; Collins 1 In diagnostic tests,“pathological” defines the behavior of individuals who score more than five criteria on the DSM, fourth edition (DSM-IV), and “problem” less than five. In reality, however, the terms are often used interchangeably, and it can moreover be difficult to distinguish between them because pathological gamblers will undoubtedly have been problem gamblers at some point, and both types of players can experience fluctuations in the severity of their condition.

Situating Gambling Studies

11

1996; Reith 2007). Utilizing Foucault’s notion of the “constitution of subjects,” whereby new social types are created through the process of classification, these authors note how the process of observation and measurement itself made the phenomenon of problem gambling increasingly visible and consequently “real” to scientific inquiry. It is in this way, they note, that a new area of study—problem gambling—and a new research subject—the problem gambler—came into being in the 1980s. Having noted the theoretical and discursive creation of the research subject, it is also worth pointing out that a variety of more mundane, pragmatic factors were involved in the emergence of the area of gambling studies. Since the 1950s, a handful of individuals had regarded gambling as an illness, among them the founders of Gamblers Anonymous, which was formed at that time, and a few psychiatrists. During the 1970s, insurance for psychiatric treatment expanded rapidly, which encouraged the inclusion of various mental conditions as bona fide illnesses in order to secure insurance payments for treatment. In practical terms, the formal recognition of problem gambling as a mental disorder meant that insurers could be persuaded to pay for treatment, thus increasing access to treatment for sufferers. And of course, for gamblers themselves, recognition of their problem removed some of the stigma associated with their behavior (Volberg 2001). In the 1970s, a branch of Gamblers Anonymous approached the head of an addictions treatment program, Dr. Robert Custer, for help in establishing a treatment program for gamblers. It was information from this program that contributed to the inclusion of pathological gambling in the DSM-III, as well as to the development of the criteria used to measure it (Custer and Milt 1985). So, although it is the case that the topic of “gambling studies” as an epistemological field or body of knowledge is interdependent with the wider culture in which it is embedded, it should also be noted that in many instances, the development of a new topic of inquiry is also frequently subject to more mundane pressures. Quite simply, it is often less the grand march of human knowledge than the exigencies of everyday life that provide the impetus for the development of a new branch of research. The development of the diagnostic evaluations not only helped to define the subject of problem-gambling research, but also introduced new social groups to the issue—medical, legal, academic, and treatment professionals, as well as lay groups and formal organizations. These groups brought their own distinct backgrounds and concerns to the issue and developed the field still further, stimulating debate and demands for research, which in turn encouraged the provision of funding and the creation of more specialist organizations and committees, most of which leaned in the general direction of psychology and medicine. These conditions gave rise to an increase in gambling-related research commissioned by various stakeholders, such as state or local government agencies, nongovernmental organizations (NGOs), problem-gambling and community groups, and academic bodies. In general, such policy-directed research attempted to esti-

12

Research and Measurement Issues in Gambling Studies

mate the impacts of gambling and the extent of problem gambling, analyze which factors exacerbated problems, and develop strategies to combat them. In recent years, large-scale technical reviews have been conducted on behalf of policymakers and stakeholders, with reports from the United States and Australia collating information from a wide range of sources to provide comprehensive reviews of the nature and extent of gambling and problem-gambling behavior (National Research Council 1999; Productivity Commission 1999). At this point it can be noted that the interests of specific groups—and in particular groups who have access to resources—often play a part in the shaping of emergent research fields and, through the allocation of funding, can influence which areas get the greatest attention. Research goes on in a politically charged climate where stakeholders’ interests may involve competing claims relating to, for example, expanding or limiting gambling opportunities, increasing funds for treatment and/or research, or lobbying for legislative change. Similarly, the results from research can be interpreted selectively according to prior interests, as is the case when, for example, industry-related groups emphasize the lowest estimates of prevalence rates while treatment-related groups for problem gambling emphasize higher ones. In this way, the most powerful stakeholders can (sometimes inadvertently) contribute to the establishment of particular research agendas, define the parameters of debate, and thus influence the creation of a knowledge base. As an example of the shaping of a research climate (although one that has more to do with simple expediency rather than the play of vested interests), we can look to recent developments in the United Kingdom. The passage of the 2005 Gambling Act was intended to regulate the rapidly developing gambling economy. Prior to legislative change, the government’s Department for Culture, Media and Sport was given the task of preparing the groundwork for the new law. Prior to legislative change, the state government’s Department for Culture, Media, and Sport was given the task of preparing the groundwork for the new law. As part of this, it commissioned an independent body—the Gambling Review Body—which recommended the establishment of an independent trust to oversee research, treatment, and prevention programs on problem gambling, and which would be funded by voluntary contributions from the gambling industry. In this way, the charitable, nongovernmental body, The Responsibility in Gambling Trust (RIGT), was created. Until the passage of the act and the creation of RIGT, U.K. policymakers had paid little attention to gambling as a serious object of study, and as a consequence, the area had been consistently underfunded and underresearched. It was in this context that RIGT set out to expand the slender base of gambling research in that country by explicitly advertising the funding that would be made available for it. In order to do this, they proactively targeted known researchers in the field of drug and alcohol addiction, encouraging them to apply their knowledge and skills

Situating Gambling Studies

13

in that area to what was described as a very similar field.2 Naturally, such a policy stimulated interest; but more than this, it gave a very clear indication of what “type” of thing the funders believed problem gambling to be, and what type of research they expected would be best suited to investigate it. In this way, the research agenda for a new field of inquiry was formed out of a convergence of the interests and preconceptions of a range of policymakers, governmental organizations, and NGOs, thus defining the field and outlining the parameters of the future debate. In the United Kingdom in the twenty-first century, as with Custer and the DSM-III in the 1980s, we can see that although economic, social, and cultural changes might create a climate conducive to the study of gambling, it is a combination of pragmatic concerns and expedient networks that provide the impetus for a research agenda to actually form and get off the ground. To sum up the current state of such research, we can note that today, the bulk of the study of gambling is focused on the measurement and quantification of specifically problematic behavior. Generally, when we talk about “gambling,” we are talking about problem gambling. Indeed, gambling itself has come to be defined in terms of problem gambling, with “normal” or recreational gamblers generally referred to as “nonproblem gamblers”—that is, in terms of what they are not. It could be argued that this focus on problematic or deviant aspects in gambling studies is a continuation of the emphasis on the disruptive effects of gambling, rooted in particular philosophical and religious perspectives, as traditionally expressed by the church or other guardians of public morality. In some ways, this line of thought is justifiable:The negative aspects of gambling are still being articulated today, albeit in different ways. However, where a significant break with the past has occurred is in the conception of gambling itself. The critical approaches reviewed earlier did not possess a distinctive notion of “problem” gambling as an activity separate from “recreational” gambling—instead, it viewed all gambling as inherently problematic, both as an activity in its own right and as one that could lead to further vice and disruption. Today, however, it is no longer the case that gambling per se is presented in negative terms. Indeed, after a lengthy history, it is at the precise moment when gambling is moving into the mainstream as a legitimate form of recreation that the scientific vocabulary to quantify its harmful effects is being developed. Now it is regarded as a legitimate form of consumption, albeit one that can, under certain circumstances, become problematic for some individuals. Furthermore, today the type of problem it represents is found in the domain of medicine, and formed within neutral discourses of sickness, risk, and vulnerability rather than pejorative ones of weakness, vice, and dysfunction. Within a broadly positivistic tradition, a range of explanations for the syndrome of problem or pathological gambling have been proposed, each with its own methods of research and its own conception of the subject under investigation, that 2 This is not to overlook the fact that they also addressed gambling researchers in the international community and made clear that they were eager to have input from a wide range of academic disciplines to enrich the field of gambling studies in the United Kingdom.

14

Research and Measurement Issues in Gambling Studies

is, its own idea of what “type” of thing problem gambling might be.As we shall see over the following pages, such differing conceptions have implications for the development of policy, the regulation of the gambling industry, and the treatment of problem gamblers themselves. What follows is an overview of research domains within this tradition. It should be pointed out, however, that separate presentations are undertaken simply for narrative clarity and should not be taken to imply their total independence from each other, since many domain research agendas, methodologies, and aims overlap. Furthermore, our discussion is of necessity a partial account: There are many contributions to the domain of gambling research, not all of which can be covered here, and many ways of viewing the themes which make it up, of which this outline is only one possibility. Economic and Social Cost–Benefit Analyses Within the general tradition of positivism are various analyses of the economic and social costs and benefits of gambling.These studies involve attempting to quantify, measure, and compare the effects of the activity by balancing what many argue to be its positive impacts with its negative ones, described as “negative externalities.” In the former camp are economic benefits such as commercial profits, increased employment, taxation revenues, and the stimulation of local and national economies. Meanwhile, negative impacts include the degradation of local communities and increased problem gambling, with its attendant costs of crime, bankruptcy, social disruption, and personal hardship. Although many such studies have been undertaken, findings remain generally contested and inconclusive. In many cases, studies use different criteria for measurement, measure different things, utilize different methodologies, and approach the subject from quite different perspectives. For example, not all forms of gambling have the same impacts. While widely dispersed forms of “convenience gambling” such as electronic machines and Internet games create few jobs and have little economic impact on local communities, it has been argued that other, more concentrated types such as casinos and the racing industry do stimulate economic growth by creating jobs and encouraging consumer spending. However, such claims have been contested. Some studies have found that the direct and indirect effects of, for example, casinos tend to be not so much the creation but rather the transfer of wealth away from other businesses and areas, resulting in little or no economic gain overall (Eadington 1984; Grinols and Omorov 1996; National Gambling Impact Study Commission 1999). In addition, approaches to estimating the impacts of gambling can hold different conceptions of their subject, with some conceiving it in terms of social impacts and others in terms of economic ones. For example, “gross impact stud-

Situating Gambling Studies

15

ies” tend to focus solely on calculating the economic benefits of gambling, generally quantifying basic features such as revenues, taxes, and employment gains without consideration of any corresponding costs. Such studies have frequently been funded by, or otherwise associated with, the gambling industry itself, thus attracting criticisms of research bias (Fahrenkopf 1995; see National Research Council 1999 for a review). On the other hand, some studies have attempted to measure the social costs associated with problem gambling, including costs associated with job loss, bankruptcy, divorce, ill health, arrest and incarceration, and increased uptake of unemployment and welfare benefits (Gerstein et al 1999). Studies that estimate social costs have tended to use one of two types of methodology. In one, the effect of a particular form of gambling, such as a casino, is estimated through calculating a number of variables, such as employment and crime rates. In the second, the costs generated by individual problem and pathological gamblers are calculated and combined with estimates of the prevalence of gambling problems in the general population in order to come to a figure of the total cost of gambling-related problems. A number of studies have attempted to assign financial values to the negative externalities associated with gambling in this way (Dickerson et al. 1995; Lesieur 1992; Thompson, Gazel, and Rickman 1996). One of the largest of these was conducted on behalf of the National Gambling Impact Study Commission (NGISC) (Gerstein et al. 1999). By controlling for a variety of sociodemographic factors, the study was able to estimate the financial impacts of problem gambling on individuals and to extrapolate from this economic costs to society as a whole. In this way, the researchers calculated that the annual costs of problem and pathological gambling to the United States totaled about $4 billion (Gerstein et al. 1999). This approach has also highlighted that the bulk of gambling losses come from problem and pathological gamblers, and that this group accounts for around a third of the entire gambling industry’s market (Lesieur 1998; NGISC 1999; Productivity Commission 1999). These kinds of cost-benefit analyses are examples of the application of quantitative methodologies to social phenomena—a process involving the measurement and comparison of numerical variables in order to establish “objective” information. The undertaking is based on an attempt to assign financial values to various aspects of gambling (from the profits enjoyed by industry to the social costs suffered by problem players) and rests on a conception of its subjects (i.e., gamblers and problem gamblers) as socioeconomic units whose behavior as workers, consumers, spouses, and community members is similarly quantifiable. From this premise it aims to come to some sort of conclusion as to the overall “value” of gambling. However, although economic factors such as jobs and profits are tangible and measurable, social and personal ones such as individual, familial, and community well-being are far less so and may simply not be amenable to such types of analysis at all. And so, despite the hope that methodological developments will produce more refined and therefore more accurate and more “true” calculations, the possibility that this kind

16

Research and Measurement Issues in Gambling Studies

of approach will rest on an attempt to compare phenomena which are ultimately resistant to such quantitative approaches should always be borne in mind. Indeed, two large-scale investigations into the social and economic impacts of gambling could not reach a decisive conclusion on this.The NGISC stated that an accurate cost-benefit analysis was impossible to calculate, while the report of the Australian Productivity Commission argued that since economic benefits were canceled out by economic transfers, the impact of gambling should be estimated in terms of social factors instead. The implications of cost-benefit accounts are highly significant, for if social costs were conclusively found to outweigh the benefits of gambling, then policymakers would have a responsibility to take steps to limit the spread of commercial gambling and curtail the operation of the industry. However, in a highly charged, politicized climate, the evidence base for such scenarios remains unresolved and hotly contested. Meanwhile, it appears that the daunting task of quantifying social costs presents an ongoing challenge to even the most committed positivist. The remainder of this section now turns to look at research from psychological–medical perspectives that have, since the publication of the DSM-III, dominated the field of problem gambling research. Biomedical Approaches At the most extreme end of medicalized discourses are biomedical approaches to pathological gambling.An early emphasis on its similarities with dependent substance use encouraged a view of problem gambling as a discrete pathological entity, with a material, physiological basis. As victims of a chronic condition, pathological gamblers are regarded as qualitatively different from nonproblem gamblers and subject to a lifelong medical condition. A range of biochemical, genetic, and neurological studies have attempted to explain the material bases of the disorder. For example, neurological studies have utilized magnetic resonance imaging (MRI) to attempt to identify the physiological profiles of gamblers’ brains (Breiter et al. 2001; Potenza et al. 2003), while the relationship of substances such as noradrenalin and serotonin to impulsive disorders and craving has been investigated. In addition, genetic predispositions have been implicated in pathological gambling, especially those that control neurotransmitters responsible for mood and temperament, which are also argued to be implicated in other conditions, such as drug and alcohol addiction (DeCaria, Begaz, and Hollander 1998; Comings 1998). As with discourses of addiction in general, this kind of approach can be considered to be essentialist in that it depicts pathological gamblers as distinct “types” of individuals, whose actions are reducible to primarily physiological processes. Although many researchers emphasize the role of psychological and environmental, as well as biological, factors in contributing to the development of gambling

Situating Gambling Studies

17

problems, in general the focus of such approaches is on internal, material processes within the bodies of individual subjects themselves. Such a perspective challenges ideas about free will and self-determination, since, for these individuals, behavior is often unwilled and determined by biology.These kinds of biomedical approaches have been criticized for their reductive view of human behavior and for an incorrect assumption of causation, whereby physiological factors are regarded as causes rather than simply the observable effects of behavior (Peele 1985). Such biomedical perspectives also have implications for approaches to the treatment of problem gambling. These tend to be founded on pharmacological interventions, such as the prescription of lithium and selective serotonin reuptake inhibitors (SSRIs), and/or on abstinence from all forms of gambling. Because pathological gambling is regarded as a chronic type of disorder, abstinence is frequently presented as the most effective way of preventing relapse. This is also the approach adopted by self-help groups such as Gamblers Anonymous (GA), which subscribes to a disease model of addiction, claiming that “compulsive gambling is an illness, progressive in its nature, which can never be cured, but can be arrested” (Gamblers Anonymous 2007). Support for biomedical models can arise from groups of gamblers themselves—such is the case at least of GA, whose philosophy actually converges with the epistemological foundations of some of the strictest medical accounts, and whose members may adopt the language of medicine to articulate, and in some cases lend authority to, their perceptions of their condition. The notion of pathology involved in these discourses has implications for notions of legal accountability. The loss of both reason and self-control caused by the presence of disease implies an abnegation of responsibility, which means that morally—and sometimes legally—pathological gamblers should not be held responsible for their actions, far less take charge of their future well-being (Rose 1986; Castellani 2000). At the same time, it can be interpreted in ways that would relieve the industry of at least some responsibility for gambling problems: the rationale being that it is not the product—gambling—itself that causes harm, but rather the biological makeup of a small number of unfortunate individuals which predisposes them to vulnerability. Clinical Psychological Research A range of psychological studies, often conducted in clinical settings, have examined the subjective states, motivations and cognitions of gamblers, and focus on the role of impulse, sensation, arousal and conditioning in the development and maintenance of gambling behavior.The fourth edition of the DSM (DSM-IV) classifies problem gambling as an “impulse control disorder,” along with kleptomania and pyromania, referring to an inability to resist impulsive drives and a general loss of control over behavior. It is frequently associated with the characteristics of sensation seeking—the drive for novelty and new experiences—and is related to arousal—the

18

Research and Measurement Issues in Gambling Studies

quest for excitement rather than money. In such models, uncontrolled gambling is a result of an individual’s personal inability to resist or overcome particular responses generated by games themselves. Behavior is driven by the quest to seek intense experiences in a spiraling quest for ever more exciting sensations without consideration of the long-term consequences (Blaszczynski, McConaghy, and Frankova 1990; Zuckerman 1979). Behavioral approaches have focused on the processes through which gambling is learned and reinforced, and in particular the ways in which it becomes conditioned through variable rewards, such as heightened levels of arousal and/or winning.These models of operant conditioning argue that an early gambling win provides an extremely influential reward that creates positive associations with the gambling behavior and encourages repeated play (Griffiths 1995; Orford 2001). Such approaches often highlight the similarities between the behavior of animals who can be conditioned to act in certain ways to obtain rewards (e.g., monkeys pressing levers for food) and gamblers, who repeat their behavior in the expectation of favorable results. Indeed, it has been remarked that slot machines present an almost perfect example of such operant conditioning—in other words, that players are simply responding to biopsychological cues that compel them to repeat their actions (Knapp 1997). There is some convergence between these psychological models and biomedical research, with some writers pointing to the biochemical nature of arousal and impulsivity, and some research suggesting an underlying genetic basis for a variety of impulsive-addictive-compulsive behaviors ( Jacobs 1993; Slutske et al. 2000). Again, this approach can be seen as adopting an essentialist epistemology in that its underlying assumptions are based on the idea that there exist qualitative differences between problem and nonproblem gamblers. Therefore, treatment approaches may include pharmacological interventions, as well as forms of therapy and counseling that attempt to alter particular styles of behavior and cognition. However, despite the epistemological foundations of such models, evidence for the validity of these assumptions is often ambivalent or lacking. Somewhat ironically, given the focus of much of this research (with the exception of behaviorism) on gamblers’ subjective states, many studies remain committed to the positivist paradigm of “objective” observation and measurement, an approach which has nevertheless failed to provide clear-cut findings about the internal processes of gambling behavior. The methods adopted include psychometric testing through the application of questionnaires and the use of cardiographic devices to measure physical processes such as heart rates. Some empirical studies have found pathological gamblers scoring higher than controls on sensation-seeking tests, while other studies have found no difference (Blaszczynski et al. 1990; Kuley and Jacobs 1988). Attempts to gauge arousal by measuring the cardiovascular activity of gamblers in laboratory-based experiments have failed to demonstrate raised levels of arousal, while those in genuine gambling environments have done so (Anderson and Brown

Situating Gambling Studies

19

1987). Such findings throw the “ecological validity” of artificial research settings into question, and perhaps also raise issues about the feasibility of testing, measuring, or otherwise “knowing” the subjective states of any human research subject with any degree of accuracy. At the same time that subjective states can suffer from attempts to “capture” them scientifically, the supposed objectivity of the problem-gambling assessment instruments themselves are underlined by a subjectivist bias. Despite their rigorous approach to their subject, the classifications created by the DSM-IV and the SOGS are not value free, but are formed by criteria that are socially relative and subjective. Both instruments diagnose pathological gambling on the basis of individuals’ evaluations of their motivations and moods, including states such as preoccupation, excitement, and loss of control, with the indirect effects of excess gambling inferred from negative states such as guilt, anxiety, and depression that result from disrupted personal, vocational, and financial affairs. Ultimately, then, it is gamblers’ evaluations of their own subjective feelings, rather than the measurement of more verifiable factors, that form the basis for a diagnosis of pathological gambling. In a similar vein, biomedical accounts, with which some clinical psychological perspectives overlap, can also be underlined by cultural assumptions, which turn to biological processes to explain deviation from what are essentially social norms. For example, it has been suggested that an inability to make informed decisions quickly is caused by neurobiological deficits in the prefrontal cortex and that a disregard for future consequences—“myopia”—has roots in neuroanatomical systems (Bechara 2003; Damasio 1995). It is clear that these types of investigations are heavily influenced by notions of what constitutes an “informed” decision, and how people should plan for the future. Cognitive Psychological Research It has been suggested that problem gamblers differ from nonproblem gamblers at the level of the mind: in the way that they think about gambling.This type of research has investigated the assumed irrationality of problem-gambling behavior, as revealed by the possession of a range of cognitive distortions and superstitions which the DSM-IV classifies as “disorders in thinking.” Many gamblers utilize “systems” and hold a range of superstitious beliefs which are used to attempt to control the outcomes of games and frequently justify continued, losing behavior. For example, they might look for patterns and meanings in random events when there are none and see connections in independent events which are then regarded as controllable. Clinicians and researchers have identified a range of such misperceptions and superstitions, including “biased evaluations of outcomes” (Gilovich 1983), notions of “near misses” (Reid 1986), and “illusions of control” (Langer 1975). These describe players’ tendency to overestimate their own influence in games, to

20

Research and Measurement Issues in Gambling Studies

believe that losses are attributable to external factors, to hold out unfounded optimism in their chances of winning, and to believe in mystical forces such as “luck” (Gadboury and Ladouceur 1989;Wagenaar 1988).These misperceptions have commonly been identified through the “thinking aloud” method, whereby gamblers articulate their thoughts as they are playing, and these are analyzed and interpreted by the researcher. In general, these types of accounts are based on the assumption that some type of “cognitive deficit,” whether a lack of knowledge about odds and risk or a faulty system of perception, underlies problematic behavior. Implicit in such approaches are notions of “rational economic action,” whereby individuals are assumed to be risk averse and to make informed decisions based on calculations of the benefits and risks of various forms of activity. Gambling is regarded as an activity with negative expected value (i.e., gamblers can expect to lose), which is therefore antithetical to the self-interest of rational consumers. Such a conception of problem gambling assumes the fundamental irrationality of at least the repetitive behavior that drives it, if not of the subjects themselves. The type of treatment approach that flows from this is based on rectifying irrational cognitions through the provision of accurate information about the nature of games of chance and/or various forms of therapy to modify behavior and beliefs. However, it should be pointed out that, again, the assumptions involved in these kinds of approaches rest on particular cultural values and expectations about what constitutes “rational” behavior and that some writers have pointed to other interpretations. These include the possibility that since not all gamblers are motivated purely by winning, their beliefs surrounding how best to play a game should not be regarded as necessarily irrational. Other interpretations have argued that in situations of uncertainty, belief in forces external to the self is a common and frequently functional means of asserting agency and control and, furthermore, of generating enjoyment from risk taking (Lyng 1990; Reith 1999). All the approaches in the positivist tradition examined so far embrace a conception of pathological gamblers as qualitatively different, in various ways, from social or recreational gamblers.Through their observation, measurement, and classification of subjects, often within clinical or laboratory environments, researchers subscribe to a view of pathological gambling as a mental and/or physical disorder. Their focus is on the individual subject as the locus of the problem, whether in terms of their physiological makeup, psychological characteristics, internal states, mental cognitions, or a combination of them all. Such conceptions have implications for treatment and policy. If the focus of gambling problems is the individual player, then the solution to those problems similarly lies within the individual: in various forms of treatment that rectify faulty impulses, thoughts, and even genes.To some extent, such a conception of problem gamblers can be interpreted in ways that shift the policy focus away from the gambling industry: If the roots of “the problem” lie less within the product and more with players themselves, then the issue becomes

Situating Gambling Studies

21

less a case of how to regulate or limit a potentially damaging product, and more a case of how to treat and cure a minority of vulnerable individuals. Epidemiological Research and Public Health Perspectives Recently, views of problem gambling as a discrete pathological entity have been modified by broader epidemiological approaches that examine the prevalence and incidence of gambling problems throughout the population. Contrary to broadly essentialist notions which regard problem gambling as a preexisting predisposition within the individual, such research has encouraged a focus on the interrelation of the individual with a broad range of environmental factors which together contribute to the development of gambling problems of greater or lesser severity.This type of approach expands the conception—as well as the numbers— of problem gamblers to include individuals with less severe and transient problems, with consequent implications for policy and treatment. This approach is based on the information gathered from wide-ranging surveys, which have been carried out increasingly frequently in the past two decades, to estimate the prevalence and incidence of gambling and problem gambling throughout the population. They utilize quantitative methodologies such as telephone surveys and questionnaires, which incorporate variations of such diagnostic screening tools as the DSM-IV and SOGS to produce numerical and statistical forms of knowledge. A variety of national and subnational surveys have been conducted throughout Western nations, mapping patterns of behavior among the general population as well as particular subgroups, such as adolescents, which allows rates and trends to be calculated and international comparisons to be made. Through the collection of vast amounts of demographic data on individuals’ lifestyles, relationships, incomes, and health, as well as their gambling activity, the characteristics of gamblers and problem gamblers can be compared with those of the general population. From this mass of quantifiable data, patterns of behavior, as well as relationships between various factors, can be isolated and identified, and trends in gambling behavior can be seen to emerge. Ultimately, this process of observation and identification of patterns and traits through survey research makes the subject of gambling research increasingly visible and consequently—as is revealed in increasing amounts of statistical data, charts, and tables—increasingly “real.” The information from epidemiological research provides the foundation for public health perspectives on gambling.These adopt the terminology of communicable disease, distinguishing between the agent (exposure to gambling opportunities), the host (the gambler), and the environment (the wider physical and social setting in which gambling goes on) to describe the occurrence of gambling problems throughout the population. It also utilizes statistical analyses to display the associations among various environmental, social, and physical factors, thus revealing the sociodemographic patterns of risk and vulnerability that cut across the pop-

22

Research and Measurement Issues in Gambling Studies

ulation. These relations are assessed by a continuum of harm and are expressed in terms such as “probable” and “potential” problem gambler. From this, particular features emerge as being more or less significant for the development of gambling problems. For instance, certain social groups, such as males, those under thirty years of age, those on low incomes, and those who are unmarried, are revealed to be especially vulnerable to the development of gambling problems. Situational factors such as easy availability and opportunities and early exposure via family members, as well as structural factors such as the quick, continuous action of electronic machines also put a person at risk for the development of problematic gambling behavior. At the same time, associations of all these factors with other types of problematic behaviors, such as those of mental disorders, substance abuse, and criminality, can also be revealed in what are described as “comorbid” relations. One clear emphasis that has emerged from this kind of research is on the diversity and complexity of gambling and problem gambling. Just as there is no single “type” of gambling, there is no single “type” of gambler. Rather, gamblers constitute a heterogeneous group whose behaviors are influenced by a variety of factors, including the type of game played and the psychological and social characteristics of the players themselves. In the increasingly complex relations that define it, the “nature” of problematic gambling comes to unravel beyond the individual and becomes entangled in the web of relations that tie people to the wider world. Related to this type of research has been a focus on longitudinal, dynamic models that examine the ways that gambling behavior changes over time, and the pathways through gambling behavior as “stages of change” over an individual’s life course. Such studies suggest that “regular” and “problem” players are not necessarily always distinct and separate groups, but rather can be individuals whose behavior exists on a continuum, with recreational playing at one end,“pathology” at the other, and various degrees of problematic behavior in between. It is thought that many gamblers go through cycles of behavior, when they move between extremes of playing over time: from regular to problem playing and back again; in and out of a problematic status in recurring and transient phases. A range of factors, including the influence of social networks and the availability of gambling, have been associated with gambling “pathways,” which in turn impact on individuals in different ways, making them more or less resistant to behavioral change (Abbott, Williams, and Volberg 1999; Blaszczynski and Nower 2002). Similar pathways to problematic behavior among adolescents have been investigated as stages of change (DiClemente, Story, and Murray 2000), a model originally applied to the process of change for addictive behaviors such as smoking, drinking, and drug misuse. Research has also begun to suggest that it is likely that the majority of players who experience problems, especially those of a less severe nature, recover from them on their own, without recourse to formal treatment. Such processes of “natural recovery” have been the subject of research that has not yet identified factors that predict change, but nevertheless suggests that structural factors such as

Situating Gambling Studies

23

changing social relationships and financial problems, as well as experiential factors such as life crises, may act as catalysts for change (Abbott et al. 1999; Hodgins and el-Guebaly 2000; Nathan 2003). Such a perspective challenges the view of problem gambling as a chronic, progressive disorder affecting a minority of players who for various reasons are predisposed to develop the problem. Rather, this perspective focuses attention on the wider environment in which gambling goes on and the actual gambling product itself as sources of harm that have the potential to create problems for far greater numbers of recreational players. Through both types of research, we can see a widening of the research stage: Epidemiological surveys expand it across populations, while longitudinal studies expand it over time. The result is a much broader and more fluid conception of problem gambling as a type of behavior that is integrated with and influenced by a range of factors external to the individual. However, although this focus begins to move away from deterministic models of pathology, it should be remembered that it is still located within a medicalized framework. Although stages-of-change models provide dynamic accounts of human behavior, they are borrowed from the conceptual field of “addiction research,” which rests on a clear-cut medical and normative distinction between behavior that is a sign of “sickness” and behavior that is indicative of being “well.” These differing conceptions of the nature of problem gambling have very different implications for treatment modalities.The view that problem gambling is part of a continuum of behavior and that many gamblers recover on their own or suffer from less severe problems at regular intervals in their lives supports a philosophy that is less restrictive and more responsive to the requirements of large numbers of the population than are models based on the goal of abstinence for a sick minority. Primarily throughout Australasia, parts of Europe, and Canada, public health approaches tend to be committed to the twin ideals of problem prevention and harm reduction, which are based on providing information and skills to allow informed choice and responsible play. Such a philosophy underlies strategies to raise awareness and inform players about the potential risks of gambling and the best ways to play without encountering harm.These include educational programs about the characteristics and potential hazards of games and the dissemination of information on counseling and self-exclusion programs. To a large extent, these preventative measures are based on the assumption that decisions about whether and how much to gamble should be largely left to the individual, and also that informed choice will result in rational, and therefore responsible, behavior. The tropes of “responsible gambling” found in these models reflect a continued focus on the individual as the site of gambling problems, as well as their resolution.The emphasis is on players’ responsibility to arm themselves with information, to regulate their behavior, to make appropriate decisions, and to limit how long and how much they play.At the same time, the industry is exhorted to behave

24

Research and Measurement Issues in Gambling Studies

responsibly toward its customers, by, for example, discouraging excessive play by vulnerable people who may lose more than they can afford and providing realistic estimates of the chances of losing. Most sectors of the industry pay at least lip service to such standards of “social responsibility.”

CONCLUSIONS: CURRENT TRENDS AND FUTURE DIRECTIONS It seems somewhat premature, at this point in what is only the first chapter in the collection, to begin to sum up the state of gambling research and attempt to predict future directions of study. In large part, that is part of the challenge for the readers: having read the book, to decide for themselves. However, some initial comments can nevertheless be made here. We have seen that two research domains exist—interpretivistic and positivistic—each with a different approach to its subject and different ways of studying it, the latter of which has long been dominant in the field. Cutting across these domains are very different conceptions of what gambling and problem gambling actually are, although in the main, these have been defined within a medical-psychological framework in terms of physical and/or mental disorder. In turn, such conceptions have implications for treatment and policy. So, for example, notions of individual pathology lie behind therapeutic and pharmacological treatments for relatively small numbers of individuals, while ideas about the more fluid nature of gambling problems stimulate policies and interventions to protect larger numbers of “regular” players from harm with ongoing programs of prevention and harm reduction. It is worth noting here that one commonality found in the varying conceptions of gambling and problem gambling discussed here is an ongoing discussion around the idea of responsibility. This emphasis on “responsibility”—whether in terms of the individual player or the gambling provider—rather than on, say, state regulation dovetails with wider ideologies of neo-liberalism, with its emphasis on individual freedom and choice. As the state increasingly moves away from the regulation of gambling, as laws are liberalized and the gambling industry is allowed greater freedoms, another paradigm shift seems to be coming into being. Rather than restrictive legislation, now the onus is on both individual and corporate responsibility. Gamblers are considered rational, sovereign consumers and the gambling industry a legitimate, mainstream leisure provider, and the interests of both are assumed to come together in responsible self-regulation. Debates around responsibility illuminate some of the wider issues that surround different conceptions of the nature of gambling problems, such as legal culpability, industry regulation, and the role of government, and in this debate, the stakes are high. Recently, for example, the threat of gambling to public health has been compared to that posed by alcohol or tobacco, with the director of a U.S. problem-

Situating Gambling Studies

25

gambling council warning, “The [U.S. Problem Gambling] report will act like the Surgeon General’s 1964 report on smoking and health—a wake-up call for America on the dangers of gambling.This report makes it very clear that gambling is not just another form of recreation—it is a very addictive and destructive activity. In short, gambling is the new tobacco” (Grey, in Kindt 2001). Lawsuits against gambling companies for misleading consumers on the risks of their products, akin to the class actions against tobacco, may be the result of this approach. No doubt the industry will be closely following the recent French case in which a gambler sued a casino for allowing him to lose over four million Euros in eight years. While the player claimed that the casino had a duty to provide him “information, advice, and loyalty,” the casino argued that “the idea of gambling is that one runs the risk of losing” (The Guardian, November 15, 2005, p. 25). Whichever side has the stronger case could set the parameters for legal notions of responsibility for years to come. In recent years, advances in research have substantially increased understanding of the nature of gambling and its problematic forms. In particular, the kind of information that has been produced by epidemiological and longitudinal research and formulated around policies of public health is starting to show us that gambling and problem gambling are multicausal and heterogeneous. Behaviors are fluid and shifting, encompassing different types of players who play in a variety of contexts and possess a range of motives. If there are commonalities, they appear to lie in an understanding that “gambling” is not a discrete entity but is rather part of a complex web of human behavior and that problem gamblers do not play in a vacuum, but act out behavior that is embedded in wider socioeconomic contexts. Although advances in understanding have been made, we need to acknowledge the limitations that still exist in the ways that gambling and problem gambling are studied and in the research methodologies used to examine them. Much gambling research is committed to the positivist tradition of “objective” observation, measurement, and classification of the subject by the researcher, is wedded to a medical-psychological epistemology of mental and/or physiological disorder, and is circumscribed by a highly individualistic focus. Although this tradition has informed knowledge, it provides a very particular kind of knowledge, and a very particular “way of seeing,” sometimes to the detriment of other perspectives. It is in this context that we need to bear in mind the value-laden nature of even supposedly objective measurement and to recognize that simply quantifying the subject is not always enough. We need more of the kinds of information that numbers alone cannot give us; to go beyond counting and begin to look at processes, meanings, and social contexts. At present, there is surprisingly little overlap between research into “regular,” or recreational, gambling and studies of problematic gambling, with the two tending toward separate foci of study and methods of research. Greater dialogue between problem-gambling research and investigations into normal/recreational gambling is badly needed, especially if, as some have argued, problem gambling is

26

Research and Measurement Issues in Gambling Studies

a stage within wider processes of more general gambling behavior. It is likely that many of the same individuals are, have been, or will be both problem and recreational players. Similarly, gambling researchers need to be aware of the close relationship between the two realms of study and attempt to build bridges between them in order to generate an understanding of the “big picture” of gambling as a whole. Fruitful avenues of future research point to the continued investigation of the motivations, attitudes, and beliefs of players and the social contexts of their behavior; the routes and pathways through which individuals move into gambling and problem gambling and perhaps out of them again; and a focus on which factors are associated with increased susceptibility to developing problems and which with resisting or overcoming them. All of this involves a commitment to longitudinal research and an investment in qualitative studies in order to build up a detailed picture of the meanings of gambling and the motivations of gamblers as dynamic phenomena embedded in socioeconomic contexts and subject to change over time. These are challenging areas for the future of the discipline, but, looking back at how much has been achieved in only the past 20 years, there is every reason to be optimistic for the continued progress of gambling studies in the future.

GLOSSARY Discourse forms of language and expression that construct and articulate particular worldviews and that are rooted in relations of power. Epistemology theories of knowledge and ways of knowing the world. Interpretivism a research approach that looks for culturally derived and historically situated interpretations of the social world, recognizing the role that the human creation of meaning has on this process. Positivism a research approach that rests on the assumption that the methods of the social sciences should be based on those of natural science and which thus utilizes (supposedly) value-free methods to observe and measure phenomena, whose relations can be expressed as universal laws and theories. Qualitative a research methodology that rests on the observation and interpretation of the subject by the researcher. Quantitative a research methodology that results in data being expressed in numerical form.

Situating Gambling Studies

27

REFERENCES Abbott, M. W., Williams, M., and Volberg, R. A. (1999). Seven Years On: A Follow-up Study of Frequent and Problem Gamblers Living in the Community.Wellington, NZ: Department of Internal Affairs. Abt, V., Smith, J. F., and Christiansen, E. M. (1985). The Business of Risk: Commercial Gambling in Mainstream America. Lawrence: University of Kansas Press. American Psychiatric Association. (1980). Diagnostic and Statistical Manual of Mental Disorders, third ed. Washington, DC: Author. —— . (1987). Diagnostic and Statistical Manual of Mental Disorders, 3rd ed., revised. Washington, DC: Author. —— . (1994). Diagnostic and Statistical Manual of Mental Disorders, 4th ed. Washington, DC: Author. Anderson, G., and Brown, R. I. F. (1987). Some applications of reversal theory to the explanation of gambling and gambling addictions. Journal of Gambling Behavior, 3, 179–189. Bechara,A. (2003). Risky business: Emotion, decision making and addiction. Journal of Gambling Studies, 19, 23–51. Bergler, E. (1970). The Psychology of Gambling. New York: International Universities Press. Blaszczynski, A., McConaghy, N., and Frankova, A. (1990). Boredom proneness in pathological gambling. Psychological Reports, 67, 35–42. Blaszczynski,A., and Nower, L. (2002).A pathways model of gambling and problem gambling. Addiction, 97, 487–499. Bloch, H. (1951).The sociology of gambling. American Journal of Sociology, 57, 215–221. Bourdieu, P. (1984/1979). Distinction: A Social Critique of the Judgement of Taste. London: Routledge. Breiter, H., Aharon, I., Kahneman, D., Dale, A., and Shizgal, P. (2001). Functional imaging of neural responses to expectancy and experience of monetary gains and losses. Neuron, 30, 619–639. Brenner, R., and Brenner, G. (1990). Gambling and Speculation: A Theory, a History and a Future of Some Human Decisions. Cambridge: Cambridge University Press. Bruce, A. C., and Johnson, J. E.V. (1992). Toward an explanation of betting as a leisure pursuit. Leisure Studies, 11, 201–218. Caldwell, G., Haig, B., Dickerson, M., and Sylvan, L. (eds). (1985). Gambling in Australia. Sydney: Croom Helm. Cassidy, R. (2002). The Sport of Kings: Kinship, Class and Thoroughbred Breeding in Newmarket. Cambridge, UK: Cambridge University Press. Castellani, B. (2000). Pathological Gambling:The Making of a Medical Problem. Albany: State University of New York Press. Chinn, C. (1991). Better Betting with a Decent Feller: Bookmakers, Betting and the British Working Class, 1750–1990. Hemel Hempstead, Herts, UK: Harvester Wheatsheaf. Clotfelter, C.T., and Cook, O. J. (1989). Selling Hope: State Lotteries in America. Cambridge, MA: Harvard University Press. Collins, A. F. (1996). The pathological gambler and the government of gambling. History of the Human Sciences, 9, 69–100. Comings, D. E. (1998).The molecular genetics of pathological gambling. CNS Spectrums, 3, 20–37. Custer, R., and Milt, H. (1985). When Luck Runs Out: Help for Compulsive Gamblers and Their Families. New York: Facts on File Publications. Damasio, A. (1995). On some functions of the human prefrontal cortex. Annals of the New York Academy of Sciences, 769, 241–251. DeCaria, C. M., Begaz,T., and Hollander, E. (1998). Serotonergic and noradrenergic function in pathological gambling. CNS Spectrums, 3, 38–47. Devereaux, E. (1949). Gambling and the social structure. Unpublished PhD thesis, Harvard University.

28

Research and Measurement Issues in Gambling Studies

Dickerson, M., Allcock, C., Blaszczynski, A., Nicholls, B., Williams, J., and Maddern, R. (1995). An Examination of the Socio-economic Effects of Gambling on Individuals, Families and the Community. Sydney: Australian Institute for Gambling Research. DiClemente, C., Story, M., and Murray, K. (2000). On a roll:The process of initiation and cessation of problem gambling among adolescents. Journal of Gambling Studies, 16, 289–313. Dixey, R. (1996). Bingo in Britain:An analysis of leisure and class. In Gambling Cultures: Studies in History and Interpretation (J. McMillen, ed.). London: Routledge. Dixon, D. (1991). From Prohibition to Regulation: Bookmaking, Anti-Gambling and the Law. Oxford: Clarendon Press. Downes, D., Davies, B., David, M., and Stone, P. (1976). Gambling,Work and Leisure:A Study Across Three Areas. London: Routledge and Kegan Paul. Eadington, W. (1984). The casino gaming industry. Annals of the American Academy of Political and Social Science, 474, 23–35. Fahrenkopf, F. (1995). Hearing on the Gambling Impact Study Commission Committee on Governmental Affairs. U.S. Senate.Washington DC: U.S. Government Printing Office. Falk, P., and Maenpaa, P. (1999). Hitting the Jackpot: Lives of Lottery Millionaires. Oxford: Berg. Fisher, S. (1993). The pull of the fruit machine: A sociological typology of young players. Sociological Review, 41, 446–474. Foucault, M. (1976). The History of Sexuality, vol. 1 (R. Hurley, trans.). Harmondsworth: Penguin. Freud, S. (1928). Dostoevsky and parricide. In Collected Papers, vol. 5 (J. Strachey, ed.). London: Hogarth Press. Gadboury, A., and Ladouceur, R. (1989). Erroneous perceptions and gambling. Journal of Social Behavior and Personality, 4, 411–420. Gamblers Anonymous. (2007). What is compulsive gambling? “Q & A” page, website, www.gamblersanonymous.org Geertz, C. (1975).The Interpretation of Cultures. London: Hutchinson. Gerstein, D., Hoffmann, J., Larison, C., Engelman, L., Murphy, S., Palmer, A., Chuchro, L., Toce, M., Johnson, R., Buie, T., and Hill, M. A. (1999). Gambling Impact and Behavior Study: Report to the National Gambling Impact Study Commission. Chicago: National Opinion Research Center. Gilovich, T. (1983). Biased evaluation and persistence in gambling. Journal of Personality and Social Psychology, 44, 1110–1126. Goffman, E. (1969). Where the Action Is:Three Essays. London: Allen Lane. Goodale, J. (1987). Gambling is hard work: Card playing in Tiwi society. Oceania, 58, 6–21. Greenson, R. (1974). On gambling. In The Psychology of Gambling (J. Halliday and P. Fuller, eds.). London: Allen Lane. Griffiths, M. (1995). Adolescent Gambling. London: Routledge. Grinols, E., and Omorov, J. (1996). Development or dreamfield delusions? Assessing casino gambling’s costs and benefits. Journal of Law and Commerce, 16, 49–87. Hayano, D. (1982). Poker Faces:The Life and Work of Professional Card Players. Berkeley and Los Angeles: University of California Press. Herman, R. (ed.). (1967). Gambling. London: Harper and Row. Hodgins, D. C., and el-Guebaly, N. (2000). Natural and treatment-assisted recovery from gambling problems: A comparison of resolved and active gamblers. Addiction, 95, 777–789. Jacobs, D. (1993). Evidence supporting a general theory of addiction. In Gambling Behavior and Problem Gambling (W. Eadington and J. Cornelius, eds.). Reno, NV: Institute for the Study of Gambling and Commercial Gaming. Kindt, J. (2001). The costs of addicted gamblers: Should the states initiate mega-lawsuits similar to the tobacco cases? Managerial and Decision Economics, 22, 17–63. Knapp,T. (1997). Behaviorism and public policy: B. F. Skinner’s views on gambling. Behavior and Social Issues, 7, 129–139.

Situating Gambling Studies

29

Kuley, N., and Jacobs, D. (1988). The relationship between dissociative-like experiences and sensation seeking among social and problem gamblers. Journal of Gambling Behavior, 4 197–207. Langer, E. (1975).The illusion of control. Journal of Personality and Social Psychology, 32, 311–328. Leiseur, H. (1984). The Chase: Career of the Compulsive Gambler. Cambridge, MA: Schenkman Books. —— . (1992). Compulsive gambling. Society, 29, 43–50. —— . (1998). Costs and treatment of pathological gambling. Annals of the American Academy of Political and Social Science, 556, 153–171. Leiseur, H., and Blume, S. (1987).The South Oaks Gambling Screen (SOGS):A new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, 1184–1188. Lindner, R. (1976).The psychodynamics of gambling. In The Psychology of Gambling ( J. Halliday and P. Fuller, eds.). London: Allen Lane. Lyng, S. (1990). Edgework: A social psychological analysis of voluntary risk taking. American Journal of Sociology, 95, 851–886. Malaby, T. (2003). Gambling Life: Dealing in Contingency in a Greek City. Urbana: University of Illinois Press. McMillen, J. (ed.). (1996). Gambling Cultures: Studies in History and Interpretation. London: Routledge. Morton, S. (2003). At Odds: Gambling and Canadians, 1919–1969. University of Toronto Press. Nathan, P. (2003). The role of natural recovery in alcoholism and pathological gambling. Journal of Gambling Studies, 19, 279–286. National Gambling Impact Study Commission. (1999). Final Report.Washington DC: U.S. Government Printing Office. National Research Council. (1999). Pathological Gambling: A Critical Review.Washington, DC: National Academy Press. Neal, M. (1998). “You lucky punters!” A study of gambling in betting shops. Sociology, 32, 581–600. Newman, O. (1972). Gambling: Hazard and Reward. London: Athlone Press. Orford, J. (2001). Excessive Appetites: A Psychological View of Addictions. Chichester, UK:Wiley. Peele, S. (1985). The Meaning of Addiction. San Francisco: Jossey-Bass. Potenza, M., Steinberg, M., Skudlarsky, P., Fulbright, R., Lacadie, C.,Wilbur, C., Rounsaville, B., Gore, J., and Wexler, B. (2003). Gambling urges in pathological gambling:A functional magnetic resonance imaging study. Archives of General Psychiatry, 60, 828–836. Productivity Commission. (1999). Australia’s Gambling Industries. Canberra. Reid, R. (1986).The psychology of the near miss. Journal of Gambling Behavior, 2, 32–39. Reith, G. (1999). The Age of Chance: Gambling in Western Culture. London: Routledge. —— . (2007, in press.). Gambling and the contradictions of consumption: A geneaology of the “pathological” subject. American Behavioral Scientist. Rose, I. N. (1986). Gambling and the Law. Hollywood, CA: Gambling Times/Lyle Stuart. Rosecrance, J. (1985). The Degenerates of Lake Tahoe:A Study in Persistence in the Social World of Horse Race Gambling. New York: Peter Lang. Sexton, L. (1987).The social construction of card playing among the Daulo. Oceania, 58, 38–46. Scott, M. (1968). The Racing Game. Chicago: Aldine. Slutske,W. S., Eisen, S.,True,W. R., Lyons, M. J., Goldberg, J., and Tsuang, M. (2000). Common genetic vulnerability for pathological gambling and alcohol dependence in men, Archives of General Psychiatry, 57, 666–673. Thompson,W., Gazel, R., and Rickman, D. (1996).The social costs of gambling in Wisconsin. Wisconsin Policy Research Institute Report, 9, 1–44. Volberg, R. A. (2001). When the Chips Are Down: Problem Gambling in America. New York:The Century Foundation Press. Wagenaar,W. (1988). Paradoxes of Gambling Behaviour. London: Lawrence Erlbaum. Zimmer, L. (1987). Playing at being men. Oceania, 58, 22–37. Zuckerman, M. (1979). Sensation Seeking: Beyond the Optimal Level of Arousal. Hillsdale, NJ: Erlbaum.

This page intentionally left blank

PART II

Measurement Issues

This page intentionally left blank

CHAPTER 2

Population Surveys Rachel A.Volberg Gemini Research, Ltd. Northampton, Massachusetts

Introduction Purposes of Population Surveys in Gambling Studies Problem-Gambling Prevalence and Comorbidity Sampling Issues in Population Surveys Sample Size Sampling Frame Sampling Modality Multimodal Sampling Response Rates Weighting Population Survey Samples Constraints and Choices in Population Research

INTRODUCTION Population surveys of gambling participation and problem-gambling prevalence play an important role in monitoring the impacts of legal gambling. In this chapter, we consider some of the critical decisions that researchers must make in planning population research on gambling and problem gambling. Some of these decisions relate to sample size, sampling frame, and sampling modality. Other decisions relate to the challenges of achieving an acceptable response rate, one important measure of the representativeness of the sample, and of properly weighting the results of the survey to yield unbiased information about gambling and problem gambling in the population. Ultimately, population surveys are always constrained 33

34

Research and Measurement Issues in Gambling Studies

by the available resources, and it is up to the researcher to decide how to most effectively deploy those resources. The material presented here is intended to assist researchers in making these decisions and in understanding possible choices.

PURPOSES OF POPULATION SURVEYS IN GAMBLING STUDIES In the 1980s, following rapid expansion in the availability of legal gambling in North America, state and provincial governments began to establish services for individuals with gambling problems.Almost immediately, questions arose about the number of problem gamblers in the general population who might seek help for their gambling difficulties. These questions required population research to identify the number (or “cases”) of problem gamblers in a jurisdiction, ascertain the demographic characteristics of these individuals, and determine the likelihood that they would utilize treatment services if these became available. In the 1990s, population surveys of gambling and problem gambling became an essential component in the monitoring of legal gambling in many countries (Abbott and Volberg 1999; Abbott,Volberg, Bellringer, et al. 2004). Population surveys of gambling participation and problem-gambling prevalence play an important role in monitoring the impacts of legal gambling. From a policy and planning perspective, population surveys are useful in tracking changes in attitudes toward gambling and gambling participation over time. Population surveys can also be helpful in assessing the proportion of gambling revenues derived from problem gamblers, an important factor in the rational calculus of public gambling policy (Volberg, Gerstein, et al. 2001). From a public health perspective, population surveys are valuable for identifying which sectors of the population contain the highest concentrations of problem gamblers and are thus important in the development and refinement of problem-gambling prevention and treatment services. By comparing problemgambler profiles from population surveys with client records and records of calls to helplines, it is possible to ascertain how well services are reaching those most in need and to introduce measures to enhance outreach. From a basic research perspective, population research is important in identifying risk factors that problem and pathological gambling have in common with other disorders, as well as those specific to the disorder. Population research is needed to improve our understanding of the relative contribution of different risk factors in the development of problem and pathological gambling. Finally, population research is needed to improve our understanding of the impact that prevention and treatment efforts may have on different types of problem gamblers.

35

Population Surveys

PROBLEM-GAMBLING PREVALENCE

AND

COMORBIDITY

Some forms of gambling have a particularly strong association with problem gambling, most notably those that are continuous in nature and involve an element of skill or perceived skill (e.g., electronic gambling machines, casino table games) (Abbott,Volberg, Bellringer, et al. 2004). Population surveys in a number of countries have found that people with preferences for, frequent involvement in, and substantial expenditures on these forms of gambling have a high probability of being problem gamblers. For example, while it is generally estimated that between 2% and 5% of the adult population are problem or pathological gamblers in jurisdictions with “mature” gambling markets, prevalence rates among regular machine players and track bettors can be as high as 25% (Abbott and Volberg 2000; Gerstein et al. 1999; Productivity Commission 1999; Schrans, Schellinck, and Walsh 2000; Smith and Wynne 2004). Early adult population surveys conducted in the United States, Canada, Australia, Spain, and New Zealand found that male gender, age under 30 years, low income, and single marital status were almost universally risk factors for problem gambling. Low occupational status, less formal education, non-Caucasian ethnicity, and residence in a large city were additional risk factors in a number of studies.Youth surveys in North America found that people in their mid- to late teenage years had higher prevalence rates than adults (Abbott,Volberg, Bellringer, et al. 2004). In recent years, some jurisdictions have seen a marked increase in the proportion of women problem gamblers, while in other jurisdictions the proportion of men has expanded. In some jurisdictions where tribal casinos have become operational, there have also been increases in the proportion of problem gamblers who are non-Caucasian. From these studies, it appears that change in the availability of particular types of gambling is instrumental in altering the sociodemographic characteristics of problem gamblers (Volberg 2004). While research generally supports the notion that problem-gambling prevalence is associated with greater exposure to high-risk gambling activities, there are some groups in the population with interesting “bimodal”gambling patterns. In comparison with other groups, they contain large proportions of people who do not gamble or gamble infrequently, as well as moderate to large proportions of frequent, high-spending gamblers. Groups in this category include some ethnic minorities and recent immigrant groups (e.g.,African Americans in the United States, Pacific Islanders in New Zealand, and eastern European immigrants in Sweden). These appear to be sectors of the population in the early stages of introduction to high-risk forms of gambling, and some of these groups have exceedingly high levels of problem gambling (Abbott 2001;Abbott,Volberg, Bellringer, et al. 2004). Like others with addictive disorders, pathological gamblers have much higher rates of co-occurring psychiatric conditions and substance abuse than are found in

36

Research and Measurement Issues in Gambling Studies

the general population.The two most recent national surveys in the United States found rates of alcohol and substance dependence among problem and pathological gamblers in the general population that were approximately ten times higher than among low-risk gamblers and nongamblers (Gerstein et al. 1999;Welte et al. 2001). There is also evidence that mood disorders, primarily major depression and anxiety, frequently co-occur with problem and pathological gambling (Gerstein et al. 1999; Specker et al. 1996). The co-occurrence of pathological gambling and attention deficit/hyperactivity disorder (ADHD) parallels research reporting ADHD in people with other addictions (Ozga and Brown 2000; Rounsaville et al. 1991; Rugle and Melamed 1993). Finally, high rates of comorbidity between pathological gambling and alcohol, nicotine and drug-use disorders, mood disorders, anxiety disorders, and personality disorders were identified in the recent National Epidemiologic Survey of Alcohol and Related Conditions (Petry, Stinson, and Grant 2005).

SAMPLING ISSUES IN POPULATION SURVEYS In the most general terms, the aim of a sampling design is to recruit a representative set of the eligible population into the study. A representative sample is necessary to enable the findings of a study to be generalized beyond just the people who are included in the study.There are many decisions that researchers must make in planning a population survey. Some of the most important questions relate to sample size,sampling frame, and sampling modality.

SAMPLE SIZE With regard to sample size, the low base rate of problem gambling in the general population poses a particular challenge. Sample sizes in population surveys of gambling and problem gambling have typically been too small to detect differences between subgroups in the population that are at highest risk for gambling problems. Given small sample sizes, the margins of error associated with problemgambling prevalence estimates tend to be quite large. In the case of many subgroups within these studies, error terms may be so large that little confidence can be placed in findings pertaining to them. An important consideration in deciding on the number of completed cases required for a survey is the ability to detect differences in prevalence rates between groups.While it is generally desirable for sample sizes to be as large as possible, statistical power calculations can be helpful in determining whether a simple random sampling strategy is adequate or a more complex sampling strategy, which trades overall power for adequate representation of specific subgroups, is needed. Statistical power is the probability of rejecting the null hypothesis that the

Population Surveys

37

prevalence rates of two groups are the same. An important point is that the appropriate size of samples for subgroups is substantially affected by differences in prevalence rates for different groups. It is easier to establish whether differences in prevalence rates are statistically significant if these rates differ substantially. Even if the final sample includes fewer respondents from specific subgroups than desired, it is still possible to assess the significance of differences in prevalence rates, albeit with a lower degree of statistical confidence (e.g., 90% rather than 95%). There are numerous texts that provide assessments of statistical power (e.g., Cohen 1988), as well as publicly available computer programs. In practice, it is helpful to consult with a sampling statistician who can examine scenarios for difference tests by subgroups of particular interest and determine the most efficient sample size for a survey—that is, a sample size that permits the most testing without greatly exceeding thresholds of statistical adequacy.

SAMPLING FRAME Fundamentally, there are two methods for sampling from populations: probability and nonprobability. Probability sampling is any method that uses some form of random selection—a process that ensures that each member of the population of interest has an equal probability of being chosen. Nonprobability sampling does not involve random selection, which means that the probability that the population is well represented cannot be calculated and that confidence intervals for statistical tests on the resulting data are difficult to estimate. In general, researchers prefer probabilistic or random sampling methods. However, there may be circumstances where random sampling methods are not feasible, practical, or theoretically sensible (Trochim 2000). Where the objective of the study is to be able to generalize the results to the general population, the best approach is to utilize random sampling. Random sampling is generally done by assigning a random number to each member of the population and then selecting the requisite number of respondents or by systematically selecting every nth member of the population. Use of random sampling makes it reasonable to generalize the results from the sample to the population. However, random sampling may not provide adequate representation when certain small subgroups in the population—such as ethnic minorities or problem gamblers—are of particular interest. Gambling researchers have used a variety of alternate sampling approaches to overcome the challenge of low base rates of problem gambling in the population. These approaches include stratified sampling and quota sampling. Stratified random sampling involves dividing the population into homogeneous subgroups and taking a random sample from within each.The use of stratified random sampling means that there is adequate representation of both the overall population and key subgroups within the population.As long as the groups

38

Research and Measurement Issues in Gambling Studies

that are sampled are homogeneous, stratified random sampling will generally have more statistical precision than simple random sampling. Finally, stratified random sampling means that the study will include enough cases from each subgroup to allow for meaningful subgroup inferences (Trochim 2000). Stratified sampling is appropriate in situations where the population is heterogeneous, with subgroups that vary considerably in behaviors of interest (e.g., gambling participation or problem-gambling prevalence). Proportional allocation is generally used to determine the size of the sample in each stratum based on the relative size of the strata in the population.When stratified sampling is employed, poststratification weighting can be used to adjust the achieved sample to reflect the population as a whole. Stratified sampling has been used in a number of studies of gambling and problem gambling, including prevalence surveys in Australia, Nevada, and North Dakota and in a study funded by the National Institute of Drug Abuse (NIDA) one of prevalence and predictors of pathological gambling (CunninghamWilliams et al. 2005; Productivity Commission 1999; Volberg 2001, 2002). Stratified sampling was also used in the patron survey component of the National Gambling Impact and Behavior Study in the United States (Gerstein et al. 1999). Another approach that is sometimes used in population surveys of gambling is quota sampling. In quota sampling, the population is segmented into mutually exclusive subgroups, as for stratified sampling. However, quota sampling does not involve random selection procedures within strata. Instead, researchers use their judgment to establish ahead of time the number of respondents they wish to interview within each stratum. Once the desired number of interviews is achieved, interviewers stop trying to recruit individuals in that stratum. The use of quotas means that the sample can no longer be considered representative of the population, since the probability of selection differs for individuals or households with different characteristics. In the case of quota sampling, poststratification weighting can be much more problematic and may not fully adjust for the differential selection probabilities associated with the original sample. It is worth noting the use of one other sampling strategy in population surveys of gambling and problem gambling: convenience sampling. Selection of respondents in this approach is based on their availability and willingness to participate in the study. Examples of convenience samples in the gambling studies field include college students in introductory psychology courses, clients in a clinical practice, and people who respond to an invitation from the researcher for volunteers (Blackman, Simone, and Thoms 1989; Hodgins and el-Guebaly 2000; May et al. 2003). The problem with all of these types of samples is that there is no evidence that the participants are representative of the populations to which the research is interested in generalizing (Trochim 2000). Convenience sampling trades ease of recruiting for representativeness of the sample. However, convenience sampling can be a good choice in situations where researchers are conducting exploratory research rather than attempting to represent an entire population.

Population Surveys

39

SAMPLING MODALITY The most expensive sampling method is area probability sampling. Area probability sampling methods were developed to permit random sampling of a population distributed across a large geographic area. Area probability sampling methods are not a concern in postal or telephone surveys, where geographic distances have little impact on the cost of recruiting participants. Area probability studies typically involve in-person interviews, with respondents residing at randomly selected residences in a geographically representative selection of neighborhoods. A variety of approaches can be used to collect data in face-to-face interviews, including computer-aided personal interviews (CAPI), paper-and-pencil interviewer-completed questionnaires, and self-completion questionnaires that the respondent fills out in the interviewer’s presence or on his/her own. Area probability sampling is generally employed in situations where it is necessary to administer lengthy, complicated questionnaires or where no other adequate means of sampling the target population exists. While area probability surveys are generally superior to telephone surveys in terms of coverage of subgroups in the population and higher response rates, such studies are costly, generally requiring at least five times the budget per completed case. Only a few area probability surveys have been completed in the gambling studies field (Cunningham-Williams et al. 1998; Orford et al. 2003;Volberg and Vales 2002). The most widely used approach to obtaining a random sample of the population is a random-digit-dial (RDD) telephone survey. Although RDD sampling can achieve cost-effective probability samples of households with telephones, households without telephones will not be covered. Coverage rates also tend to be lower for rural areas, large households, households with unemployed persons, households with young heads, African Americans, Hispanics, single persons, and persons with low income (Groves et al. 1988). However, given the constraints on time and resources generally available for population surveys of gambling and problem gambling, telephone sampling designs are the approach most commonly used in surveys of gambling and problem gambling worldwide. Telephone surveys typically use the household as the unit of analysis. The Scandinavian countries are an exception to this rule; because of the availability of official registers in Sweden and Norway, population surveys of gambling in these countries are able to use individuals as the unit of analysis (Lund and Nordlund 2003;Volberg, Abbott, et al. 2001). An emerging challenge to conducting population surveys by telephone relates to the growing number of cell-phone–only households in developed countries. As of 2004, approximately 6% of U.S. households were not covered by RDD sampling methods because they did not have landline services (Tucker et al. 2004). Individuals with mobile telephone service are generally considered outside the

40

Research and Measurement Issues in Gambling Studies

scope of such studies unless they also have access to landline telephone service.The primary challenge of mobile phone usage to survey sampling is that mobile phones tend to be associated with individual persons, while landline telephones tend to be associated with households.This means that samples based on differential access to telephone service in fact represent different units of analysis. Another difficulty in considering sampling frames based on mobile phone versus landline telephone usage relates to multiple opportunities for recruitment among individuals with access to both types of service. To date, only one population survey of gambling and problem gambling has included a sample of mobile phone users (Kavli and Berntsen 2005). Another significant recent development in the telecommunications industry relates to telephone number portability.While wireless-to-wireless and wireless-towireline provisions do not impact RDD sampling strategies, wireline-to-wireless does because it creates the possibility of inadvertently including wireless telephone numbers in the RDD sample. Broadly, evidence suggests that the results of surveys conducted by telephone compared with self-completion questionnaires (whether completed in the presence of an interviewer or mailed in later) can sometimes be incompatible. Responses to self-completion surveys tend to be skewed toward those who have a particularly strong opinion on the subject of the survey. Another challenge posed by self-completion of questionnaires is that respondents can view all of the questions as well as all of the response options prior to completing the questionnaire and may change how they answer specific items based on this information. Most significantly, the quality of data from self-completion surveys is often poor, with questions left blank and with indecipherable or “out-of-range” responses as well as limited responses to open-ended questions. Conversely, there are some advantages to self-administration (whether by postal questionnaire or using computerized aids in face-to-face interviews), particularly in surveys of sensitive behaviors.The most significant advantage is improved validity, since self-administered surveys consistently elicit higher and more accurate reports of socially sensitive behavior than do other methods (Aquilino 1997; Fendrich et al. 1999; Tourangeau and Smith 1996; van der Heijden et al. 2000). Other advantages of self-administration include the ability for respondents to proceed at their own pace, elimination of variability due to inter-interviewer administration, and minimization of interaction effects between the interviewer and the respondent (Fendrich et al. 1999; Johnson et al. 2000). Given that so many of the population surveys of gambling and problem gambling have been conducted by telephone, there is little information in the gambling studies field about the possible impact of interview modality on the results of such surveys. Only a few gambling surveys have used a self-completion approach; these include national surveys in Great Britain, Norway, and Sweden (Lund and Nordlund 2003; Orford et al. 2003; Rönnberg et al. 1999; Volberg, Abbott, et al. 2001).

Population Surveys

41

Other large health surveys that have included gambling modules in recent years include the Canadian Community Health Survey (CCHS) and the U.S. National Epidemiologic Survey of Alcoholism and Related Conditions (NESARC) (Petry, Stinson, and Grant 2005; Statistics Canada 2002). The gambling and problemgambling questions in these two surveys were administered in face-to-face interviews but did not involve computerized self-administration. Williams and Wood (2004) argue that the method of survey administration can significantly affect reporting of problem-gambling prevalence rates because respondents are less likely to acknowledge problems related to their gambling in a face-to-face interview than in a more anonymous telephone interview or in a face-to-face interview that includes computerized self-administration of questions about socially sensitive topics. They note that prevalence rates for problem gambling attained in the CCHS were less than half the rates obtained in several provincial surveys in the same time period. Similarly, substantially lower rates of problem and pathological gambling were identified in the NESARC (completed in 2001) than in the most recent national survey of gambling and problem gambling carried out by telephone in 1999 and 2000 (Welte et al. 2001). While self-administration may produce more valid estimates of sensitive behaviors because it offers greater anonymity, it is also possible that the much higher response rates in both the Canadian and U.S. health surveys contributed to lower prevalence rates through the inclusion of much larger numbers of nongamblers and infrequent gamblers than is usual in surveys with lower response rates (see the section Response Rates below). In the British Gambling Prevalence Survey—in which respondents were recruited face-to-face but the questionnaires were self-completed and collected later—4% of the questionnaires could not be used in determining the prevalence of problem gambling because more than half of the responses were missing (Orford et al. 2003). This supports the view that the data from self-completion questionnaires tend to be of lower quality than data from telephone surveys, in which automatic routing through the questionnaire and controls for inappropriate responses can be employed to ensure that respondents answer all of the questions within an established format. The first Swedish Prevalence Survey provides additional information on the comparability of self-completion versus telephone administration in problemgambling prevalence surveys (Rönnberg et al. 1999). Prior to the main survey, a pilot study was carried out with a randomly selected sample of 3,000 weekly gamblers to assess the impact of interview modality on problem-gambling prevalence estimates. Half of the respondents were interviewed by telephone and half were interviewed via a postal questionnaire. Overall, response rates for both the telephone interview (80%) and the postal questionnaire (70%) were quite high, and there was no significant difference in the problem-gambling prevalence rates in the two samples.

42

Research and Measurement Issues in Gambling Studies

Rather different results were found in the main part of the Swedish survey. Out of the 9,917 individuals selected for the survey, telephone contact was made with 8,845 (89% of the total sample). The remaining eligible respondents (1,071, or 11% of the total sample) could not be reached by telephone and were sent a postal questionnaire. In contrast to the pilot study, the response rate in the main survey was considerably lower with the postal questionnaire (31%) than with the telephone interview (77%). Also in contrast to the pilot study, the prevalence rate of problem gambling among respondents surveyed by mail was significantly higher than among those surveyed by telephone. Furthermore, the respondents to the postal questionnaire were significantly younger and less likely to have been born in Sweden than the telephone sample. A national prevalence survey conducted by the Norwegian Institute for Alcohol and Drug Research provides further information on the comparability of self-completion versus telephone administration in problem gambling prevalence surveys (Lund and Nordlund 2003). In the Norwegian survey, out of the 9,529 individuals selected, telephone contact was made with 5,484 individuals (58% of the total sample) and interviews were completed with 3,581 individuals (65% of those contacted). The remaining eligible respondents (4,045, or 42% of the total sample) received the postal questionnaire, and 1,651 questionnaires (41%) were received (S. Nordlund, personal communication, January 12, 2006). As in the Swedish survey, the response rate for the telephone sample was significantly higher than that for the postal questionnaire, and the prevalence of problem gambling was significantly lower. As in Sweden, the respondents in the postal sample in Norway were significantly younger and less likely to have been born in Norway than the telephone sample (I. Lund, personal communication, January 13–19, 2006). Finally, a recent survey of Norwegian youth aged 12 to 18 provides additional information on the comparability of self-completion versus telephone administration in problem-gambling prevalence surveys ( Johansson and Götestam, 2003). Both telephone and postal methods were used in the survey of adolescents in Norway. The majority of the sample of 3,237 adolescents (59%, n = 1,913) was interviewed by telephone, and the remainder (41%, n = 1,324) completed a postal questionnaire. In contrast to the adult surveys in Norway and Sweden, the rate of problem gambling was significantly higher among adolescent respondents interviewed by telephone in Norway compared with those who returned postal questionnaires ( Johansson 2006). As yet, Internet and email approaches have received little attention from gambling researchers, although increased exploration of these methods is likely in the near future. Online survey methods offer several advantages, including the ability to invite large numbers of potential respondents to participate at a much lower cost than those associated with mail, telephone, or face-to-face recruitment; automated routing of respondents through the questionnaire with concomitant improvements in data quality and reductions in interview length; and instant data capture and rapid availability of results.

Population Surveys

43

The critical challenge to Internet and email approaches in population research lies in ensuring the representativeness of the sample. For the most part, Internet and email surveys rely on panels of volunteers who are willing to complete surveys because of their interest in a topic or for various incentives, including merchandise. Internet surveys are advantageous where the target population is known and available through email or where the target population is difficult to reach using more traditional survey methods. It is important to note that although Internet surveys incur virtually no coding or data-entry costs, the labor costs for design and programming can be high (Schonlau, Fricker, and Elliott 2002). One recent study examined the potential of online methods for gambling surveys among adolescents (Lahn, Delfabbro, and Grabosky 2006). Comparing results from a school-based survey with the same questionnaire administered in an online format, these researchers found that the principal advantages of the online approach were greater flexibility in the timing of the survey and reductions in the amount of teacher time required for administration. Disadvantages of the online method included difficulties in obtaining adequate response rates, lack of control over the administration context, and difficulties obtaining detailed open-ended responses.

MULTIMODAL SAMPLING Finally, it is worth considering the use of multimodal sampling strategies in conducting population surveys of gambling and problem gambling. As noted above, the relative infrequency of problem and pathological gambling in the general population means that surveys must either recruit and screen very large numbers of respondents to identify adequate numbers of problem gamblers for subsequent analysis or use one of a variety of strategies to prescreen or filter respondents to reduce the number of people assessed. A variety of measures have been used to filter respondents, including regular gambling or reported gambling expenditures. “Dual frame” sampling has been used in one survey in the United States, and multiple interviews per household has been used in one survey in the United Kingdom (Gerstein et al. 1999; Orford et al. 2003). As described above, a strategy used in surveys in Norway and Sweden involves obtaining selfcompleted postal questionnaires from respondents who cannot or will not complete a telephone interview (Lund and Nordlund 2003;Volberg, Abbott, et al. 2001). Dual frame sampling was used in the U.S. national survey to capture large numbers of frequent (hence more likely to be problematic) gamblers efficiently relative to their prevalence in the household population. This approach was taken because the overarching goal of the study was to examine the socioeconomic impacts of problem gambling rather than problem-gambling prevalence alone.This survey used two separate sampling frames and interview modalities. One component consisted of a national random sample of residential telephone numbers designed to

44

Research and Measurement Issues in Gambling Studies

proportionately represent adult household residents; these respondents were interviewed by telephone.The other component consisted of a random stratified sample of individuals sampled proportionally to their frequency of entering commercial gambling venues; these respondents were interviewed in person at the venues. Since the questionnaires for the two samples were nearly identical, it was possible to use statistical procedures to combine the data from the two different frames. A statistical approach that took into account the differential opportunities for respondents to be included in the final sample was used to combine the samples and reweight the resulting file. In other words, the intercepted patrons shared the sample weights assigned initially to the telephone cases whom they most resembled in terms of gender, age, and past-year gambling behavior. In the British Gambling Prevalence Survey, a large, random, nationally representative sample of 7,680 British residents aged 16 and older was recruited from 4,619 households, using postal addresses as the sampling frame. Interviewers visited each of the randomly selected household addresses, completed a brief interview with the highest-income householder, and then asked each member of the household aged 16 and over to complete a self-administered questionnaire (Orford et al. 2003).While the data were weighted to correct for nonresponse and for differing probabilities of household selection, no adjustments were made for possible differential coverage of the population due to factors besides age and gender. Nor does it appear that adjustments were made to account for the clustering of interviews within households. Despite the use of more elaborate sampling procedures, researchers rarely make subsequent adjustments to the data for factors such as differential likelihood of inclusion in the sample or for clustering of interviews within households. Furthermore, few researchers take into account the impact of such sampling strategies on the effective sample size, despite the fact that design effects due to clustering and multimodal approaches can be substantial.

RESPONSE RATES There are benefits and drawbacks to any research approach. Postal surveys are inexpensive but generally have low response rates and long completion times. Online surveys take relatively little time to complete but, like postal surveys, tend to have low response rates and are difficult to assess with regard to representativeness. Face-to-face surveys typically achieve high response rates but may miss people who are infrequently at home and are far more costly than other sampling strategies.Telephone surveys are less expensive than face-to-face surveys, generally obtain comparable results, and, for some sensitive topics, provide a higher degree of anonymity. The great majority of population surveys in the gambling studies field have used this method.

Population Surveys

45

Although response rates alone should not be the sole measure of survey data quality, they are a crucial indicator of potential nonresponse bias and, hence, the representativeness of the sample. Given the importance of achieving a representative sample in most population surveys, response rates are also an important measure of the reliability of the estimates of gambling participation and/or problem-gambling prevalence identified in general population samples. Response rates, particularly for telephone surveys, have declined rapidly in recent years as people become increasingly reluctant to participate in this type of research and as technological barriers proliferate. Another challenge is that RDD samples of telephone numbers generally include substantial proportions of numbers that ring when dialed but are potentially out-of-scope for the survey. These numbers include businesses and public telephones as well as banks of numbers activated by telephone companies but not yet in service (as in geographic areas experiencing rapid increases in population) (Volberg 2002). While there is great uncertainty about the characteristics of individuals who choose not to participate in gambling surveys, it has generally been assumed that people who are not contacted or decline to be interviewed in gambling surveys include disproportionate numbers of problem gamblers (Lesieur 1994). However, it has been suggested that both people with little involvement and/or interest in gambling and problem gamblers may be overrepresented among nonrespondents in surveys with low to medium response rates (Abbott and Volberg 1991). If this is the case, the effects of their omission may partially or totally cancel each other out. It is possible that surveys that attain relatively high response rates pick up disproportionately more people with low involvement and/or interest in gambling with concomitantly lower prevalence estimates (Abbott, Volberg, and Rönnberg 2004). Abbott (2001) examined this possibility by comparing the most recent New Zealand problem gambling prevalence estimates with those obtained from the national Australian survey conducted at about the same time (Productivity Commission 1999). Data collection for the New Zealand Prevalence Survey was carried out in 1999 by Statistics New Zealand. The nationally representative random sample included 6,452 adults aged 18 years and older interviewed by telephone.The response rate for the New Zealand survey, conservatively defined, was 75%.This is probably because of the involvement of the country’s official statistics agency (Abbott and Volberg 2000; Abbott,Volberg, and Rönnberg 2004). In contrast, like most previous gambling prevalence surveys, the Australian study was undertaken by a private research company.The Australian survey used a two-phase sampling strategy with a brief “screening” questionnaire administered to identify broad patterns of gambling participation and a more detailed questionnaire completed by respondents using a selective strategy based on the intensity of gambling involvement. Although the final study included a stratified sample of 10,500 respondents selected by geographic area, age, and gender, the full

46

Research and Measurement Issues in Gambling Studies

questionnaire was administered to only 3,498 respondents, and the survey had a response rate below 50%. The data were subsequently weighted to generate nationally representative results (Productivity Commission 1999). The New Zealand prevalence estimate was very similar to those obtained by the two Australian states that had similar expenditure on continuous forms of gambling and markedly lower than those from Australian states and territories with higher expenditure. In other words, the prevalence estimates from these surveys were consistent with expectations based on known associations between expenditures on continuous forms of gambling and problem-gambling prevalence. On the basis of this analysis, Abbott (2001) concluded, like Shaffer, Hall, and Vander Bilt (1997), that problem gambling is a “robust phenomenon” largely impervious to differences in researcher and research methodology and quality. One consequence of the decline in response rates for telephone surveys has been that these rates are now calculated in a variety of ways. Although all of these approaches involve dividing the number of respondents by the number of contacts believed to be eligible, there can be substantial differences in response rates that result from different ways of calculating the denominator—that is, the number of individuals deemed eligible to respond. The most liberal approach to calculating response rates includes in the denominator only the total valid sample (e.g., households known to be eligible for inclusion in the sample).This approach is probably based on the response rate calculation long accepted as the standard for face-toface surveys. Using this approach—more properly called the completion rate in telephone surveys—the response rate is calculated by dividing the number of completed interviews by the sum of completes, refusals, and terminations. A more conservative approach is the method adopted by the Council of American Survey Research Organizations (CASRO). The CASRO method for calculating response rates entails determining the resolution rate, the screening rate, and the completion rate and multiplying these together. The resolution rate includes the status of the entire released sample of telephone numbers, including nonworking, nonresidential, and known and likely households that were nevertheless not screened.The screening rate involves calculating the proportion of eligible and ineligible households compared with the proportion of the sample determined to be known and/or likely households that were not screened. The completion rate is the number of completed interviews divided by the sum of completes and known eligibles that did not result in a completed interview. Researchers seldom include enough information in published reports to allow others to assess the accuracy of reported response rates. However, detailed information about the final disposition of the entire sample is valuable and should be included in methodological reports on population surveys of gambling and problem gambling to enable readers to assess the quality of the data for themselves. Table 2.1 presents information about what is needed to calculate response rates using CASRO and other methods.

Population Surveys

47

Overall, given uncertainty about the characteristics of individuals who choose not to participate in surveys, it is important to attain the highest possible response rates (within the constraints of available resources) in population surveys of gambling and problem gambling. There are a variety of steps that survey researchers can take to achieve better response rates. One important measure for improving response rates is budgeting for and completing as many attempts as possible to recruit respondents. In telephone surveys, it is important that calls be made at different times of the day and evening and on different days of the week, including weekends. Another related measure is to space these attempts out over the full period of data collection to allow for contact with respondents who may be away from their place of residence for an extended period. However, the most important measure for improving response rates in gambling surveys is to plan for an extended period of data collection in order to complete a sufficient number of call attempts to telephone numbers that are believed or determined to be eligible. Williams and Wood (2004) report on the results of an investigation into the appropriate number of call attempts to make in RDD surveys of problem gambling. Contrary to expectation, problem gamblers were no more difficult to reach than nonproblem gamblers. However, the investigators found that 95% of the contactable and cooperative sample was obtained within 15 to 16 attempts (with the majority of attempts made in the evenings/weekends and spread out over a minimum of 6 to 8 weeks) (Williams and Wood 2004).

Table 2.1 Information Required to Calculate Population Survey Response Rates. Total Sample Preresolved Nonworking Noncontact Nonresidential Answering machine Known household (unscreened) Likely household (unscreened) Ineligible (known household) Eligible (not complete) Complete Resolution rate (RR) Screening rate (SR) Completion rate (CR) CASRO response rate

D NC NR I U1 U2 J Ks Kc (D+NR+U1+U2+J+Ks+Kc)/total sample ( J+Ks+Kc)/( J+Ks+Kc+U1+U2) Kc/(Kc+Ks) RR*SR*CR

CASRO; Council of American Survey Research Organizations.

48

Research and Measurement Issues in Gambling Studies

Another helpful approach to improving response rates is the use of computerized survey management systems. Computerized survey management systems can be updated daily throughout the data collection period, allowing for tight control over the sample and efficient monitoring of response rates. These systems are capable of producing a variety of detailed reports that allow researchers to monitor progress in the field and reallocate resources, if necessary, to ensure that specific targets are met. Survey management systems are also useful in managing “reissues” of telephone numbers, where interviewers have previously recorded no contact with a household or a soft refusal that might be converted. Interviewer training is an important component in ensuring the quality of the data collection effort in any population survey. It is important that new interviewers go though general training to become acquainted with the principles and practice of survey research interviewing. General training typically consists of an overview of telephone interviewing procedures, including the use of computeraided telephone interviewing (CATI) technology and the role of the interviewer in the research process.Training should cover such issues as probing for clarity and quality of response, gaining cooperation, and avoiding refusals. Project-specific training is also important, for both new and experienced interviewers. Projectspecific training includes orientation to the questionnaire as well as mock interviews before interviewers are ready to begin data collection in earnest. Experienced interviewers are an especially important resource for ensuring the highest possible response rates in population surveys. Experienced interviewers are adept at averting refusals and at converting refusals into completed interviews. In large population surveys, it can be helpful to maintain a small team of experienced interviewers who are specially trained in techniques for converting soft refusals into interviews and thereby improving the overall response rate. Survey researchers have found that sending advance letters to potential respondents or households prior to the beginning of data collection can help boost response rates. Advance letters explain the purpose and importance of the survey and reassure respondents about anonymity and confidentiality. In general, it is best to send out advance letters in waves to minimize the gap between receiving the letter and an interviewer calling or coming to the respondent’s house. It is important to craft advance letters with care so as not leave respondents in any doubt about what the survey involves or how the data will be used. In North America, commercial services using multisource databases are able to link business and residential telephone numbers (both listed and unlisted) with current names and addresses. These links can be updated on a regular basis and provide a cost-effective way to contact potential respondents and solicit their participation. Another helpful measure for enhancing response rates is mailing conversion letters midway through the data collection period. Using survey management

Population Surveys

49

systems and well-trained interviewers means that over the course of data collection, researchers are able to analyze the reasons that potential respondents who are reluctant to participate have given to interviewers. Using interviewer call notes and address matching, it is possible to tailor refusal conversion letters to fit respondents’ reasons for refusing to participate and increase the likelihood that additional efforts will result in a completed interview. Along with scrupulous planning and monitoring of the sample, the measures discussed here can be expected to result in higher response rates. However, survey researchers generally agree that the most effective steps for increasing response rates are (1) to keep the average interview to an acceptable length (e.g., no more than 20 minutes for a telephone survey and no more than an hour for face-to-face interviews) and (2) to establish as long a data collection period as possible to provide the greatest opportunity to space out attempts on each piece of sample and maximize the chances of reaching each potential respondent.And regardless of the uncertainties, the available data suggest that achieving a high response rate must be balanced against other important objectives in conducting surveys of gambling and problem gambling in the population.

WEIGHTING POPULATION SURVEY SAMPLES The ultimate goal of a survey is to generate unbiased estimates of behaviors in the target population. Before analyzing the results of population surveys, it is important to ensure that the profile of the sample mirrors the profile of the population it is meant to represent. Otherwise, the results would be biased or skewed toward subgroups that are overrepresented in the sample. Sample weighting can quickly become very complex, and it is wise to involve a sampling statistician in the design of a population survey at the earliest possible stage. In general, survey data must be weighted prior to analysis to account for differential probabilities associated with selection, response rates, and population coverage rates. The latter includes an allowance for noncoverage of the eligible population in nontelephone households and underreporting of the eligible population in telephone households. A comprehensive weighting scheme for a telephone survey will include adjustments for nonresolution of telephone numbers, screener nonresponse, multiple telephone lines in a single household, withinhousehold selection probability, interview nonresponse, and poststratification to align the achieved sample with the known characteristics of the population. Until quite recently, only large, national population surveys of gambling and problem gambling included such comprehensive weighting schemes. Poststratification weighting (sometimes called calibration) is a far more common feature of population surveys in the gambling studies field.

50

Research and Measurement Issues in Gambling Studies

Poststratification weighting corrects, as far as possible, any observable bias in the data by examining key characteristics of the sample and comparing these characteristics to reference data such as the census. If the difference is considered significant, the survey data are weighted to reflect the reference data. The most common form of corrective weighting is to compare the gender, age, and ethnicity profile of respondents with the population profile and weight the sample to match the population. However, poststratification weighting cannot account for differential nonresponse within subgroups in the population (Abbott and Volberg 2000). There is a price to pay for weighting, since these procedures can increase the statistical error of the survey results—for example, by widening confidence intervals. Each type of weighting will contribute separately to the effect on the survey results, and the extent to which this occurs is measured as the design effect, or DEFF. Design effect is more commonly considered in the development phase of a survey when researchers consider the question of sample size. In general, sample size calculations assume the use of simple random samples and the impact of alternative sample designs must be taken into account when determining the sample size for a study. Complex sampling designs are used more often than simple random sampling in population research, mainly as a means of saving money. However, complex sampling strategies result in less variability in the final sample than if random sampling is used, with the result that the effective sample size is reduced.This loss of power is measured as the ratio of the actual variance, under the sampling method actually used, to the variance under the assumption of simple random sampling.The design effect is a direct way of assessing the impact of the sampling design on sampling variability. In general, the smaller the design effect, the more reliable the results of the survey are considered to be. However, estimates of design effect are usually based on past experience, since the statistical literature provides little guidance (Rosander 1977). A separate challenge from the corrections required to account for complex sampling design relates to the appropriate calculation of confidence intervals around very low or high prevalence estimates. Statistical inference, used to examine the strength and significance of relationships between different variables or change over time on the same measure, requires the construction of accurate confidence intervals.When sample sizes are large and the proportion not near zero or one, confidence intervals can be determined using conventional approximations. However, if proportions are very small or very large, orthodox methods of calculating errors of measurement and confidence intervals may be inappropriate. This is because estimators are often not distributed normally at these extremes. In this situation, alternative approaches are needed to provide more accurate and stable confidence intervals.

Population Surveys

51

CONSTRAINTS AND CHOICES IN POPULATION RESEARCH Population surveys are always constrained by the available resources, and researchers must make numerous decisions about how to most effectively deploy them (Groves 1989). Over the course of a study, researchers must continuously balance the conflicting demands of achieving an adequate sample size, achieving adequate coverage of subgroups in the population, obtaining reliable answers from respondents to a reasonable number of questions, and attaining an acceptable response rate. The general message is that costs and sources of error in surveys are related. Cutting corners in any one area will always affect some other important property of the study. Resources that are spent on recruiting a large sample into the study cannot be spent on increasing the length of the questionnaire to obtain information on a topic that may be of great interest but is tangential to the main goals of the study. Reducing the number of attempts to contact eligible respondents may permit a researcher to increase the sample size but only at the cost of increasing nonresponse. If the emphasis is on attaining the highest possible response rate, the researcher will have to minimize the length of the questionnaire or accept a smaller sample size. Population research will continue to be an important tool in efforts to monitor gambling and problem gambling in the coming decades. Given the limited resources that have been available for conducting population research on gambling and problem gambling, we need (1) ways to pool resources for gambling research so that larger and more complex studies can be conducted and (2) ways to pool expertise so that gambling studies can benefit from developments in social science, survey methodology, and statistics. Despite the challenges of conducting population research on gambling and problem gambling, the pragmatic demands imposed by the rapid expansion of legal gambling have led researchers to carry out surveys that are often limited in terms of sample size, coverage of the population, response rate, and reliable data. While much of the research on gambling and problem gambling can be criticized, this does not mean that the findings are without validity and practical utility. Such studies are generally an advance on anecdotal evidence or nonscientific approaches to gather information and reach decisions in a highly politicized arena. In practice, all research falls along a spectrum in terms of quality and fallibility. Too little of it, in gambling studies as in many other disciplines, is found near the quality end of the continuum. Probably no single study will ever excel in all respects. However, few studies are likely to be totally devoid of information, and reviewers as well as researchers must use reasoned, but ultimately personal, judgment to decide whether the methodological weaknesses of a study entirely undermine its results.

52

Research and Measurement Issues in Gambling Studies

GLOSSARY Design effect this is the ratio of the true variance of a statistic (taking the complexity of the sample design into account) to the variance of the statistic for a simple random sample of the same size. Design effects differ for different subgroups and different statistics but can affect confidence in the results of the study. Population research studies that use questionnaires and quantitative techniques to collect information from large groups of people. Such studies are generally used to describe populations or to test hypotheses and examine relationships between variables of interest. Response rate the proportion of individuals invited to participate in a study who actually do so.The method of collecting data has an important impact on the response rate, and response rates can be calculated in a variety of ways. Sampling frame the list of members of the population of interest from which a sample for study can be drawn or the procedure used to ensure that each member of the population has an equal probability of being chosen for study. Sampling modality the method used to collect information in a survey. Sampling modalities vary widely in cost but can also affect the reliability and validity of the information collected. Weighting statistical procedures used to adjust the achieved sample to reflect the population.While weighting can correct for nonresponse and differing probabilities of selection, such procedures cannot account for differential nonresponse within subgroups in the population.

REFERENCES Abbott, M.W. (2001). What Do We Know About Gambling and Problem Gambling in New Zealand? Report Number Seven of the New Zealand Gaming Survey.Wellington: Department of Internal Affairs. Abbott, M.W., and Volberg, R.A. (1991). Gambling and problem gambling in New Zealand: Report on Phase One of the National Survey. Research Series No. 12.Wellington: Department of Internal Affairs. ———. (1999). Gambling and Problem Gambling in the Community:An International Overview and Critique. Report Number One of the New Zealand Gaming Survey.Wellington: Department of Internal Affairs. ———. (2000). Taking the Pulse on Gambling and Problem Gambling in New Zealand: Phase One of the 1999 National Prevalence Survey. Report Number Three of the New Zealand Gaming Survey. Wellington: Department of Internal Affairs. Abbott, M.W.,Volberg, R. A., Bellringer, M., and Reith, G. (2004). A review of research on aspects of problem gambling. London: Responsibility in Gambling Trust. Abbott, M. W., Volberg, R. A., and Rönnberg, S. (2004). Comparing the New Zealand and Swedish National Surveys of gambling and problem gambling. Journal of Gambling Studies, 20, 237–258. Aquilino, W. S. (1997). Privacy effects on self-reported drug use: Interactions with survey mode and respondent characteristics. NIDA Research Monograph, 167, 383–415.

Population Surveys

53

Blackman, S., Simone, R. V., Thoms, D. R. (1989). The Gamblers Treatment Clinic of St. Vincent’s North Richmond Community Mental Health Center: Characteristics of the clients and outcome of treatment. International Journal of the Addictions, 24, 29–37. Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences, 2nd ed. Hillsdale, NJ: Lawrence Erlbaum Associates. Cunningham-Williams, R. M., Cottler, L. B., Compton, W. M., and Spitznagel, E. L. (1998). Taking chance: Problem gamblers and mental disorders—results from the St. Louis Epidemiological Catchment Area (ECA) study. American Journal of Public Health, 88, 1093–1096. Cunningham-Williams, R. M., Grucza, R.A., Cottler, L. B.,Womack, S. B., Books, S. J., Przybeck,T. R., Spitznagel, E. L., and Cloninger, C. R. (2005). Prevalence and predictors of pathological gambling: Results from the St. Louis Personality, Health and Lifestyle (SLPHL) study. Journal of Psychiatric Research, 39, 377–390. Fendrich, M., Johnson, T., Shaligram, C., and Wislar, J. (1999). The impact of interview characteristics on drug use reporting by male juvenile arrestees. Journal of Drug Issues, 29, 37–58. Gerstein, D. R.,Volberg, R.A.,Toce, M.T., Harwood, H., Palmer,A., Johnson, R., Larison, C., Chuchro, L., Buie, T., Engelman, L., and Hill, M. A. (1999). Gambling Impact and Behavior Study: Report to the National Gambling Impact Study Commission. Chicago: National Opinion Research Center at the University of Chicago. Groves, R. M. (1989). Survey Errors and Survey Costs. New York:Wiley & Sons. Groves, R. M., Biemer, P. P., Lyberg, L. E., Massey, J.T., Nicholls,W. L., and Waksberg, J. (1988). Telephone Survey Methodology. New York:Wiley & Sons. Hodgins, D. C., and el-Guebaly, N. (2000). Natural and treatment-assisted recovery from gambling problems: A comparison of resolved and active gamblers. Addiction, 95, 777–789. Johansson, A. (2006). General Risk Factors for Gambling Problems and the Prevalence of Pathological Gambling in Norway. Doctoral dissertation, Norwegian University of Science and Technology,Trondheim. Johansson, A., and Götestam, K. G. (2003). Gambling and problematic gambling with money among Norwegian youth (12–18 years). Nordic Journal of Psychiatry, 57, 317–321. Johnson,T. P., Fendrich, M., Shaligram, C., Garcy,A., and Gillespie, S. (2000).An evaluation of the effects of interviewer characteristics in an RDD telephone survey of drug use. Journal of Drug Issues, 30, 77–102. Kavli, H., and Berntsen, W. (2005). Gambling Habits and Gambling Problems in the Population. Report to Norsk Tipping. Oslo: MMI. Lahn, J., Delfabbro, P., and Grabosky, P. (2006). Classroom or cyberspace: Ethical and methodological challenges of online gambling surveys for adolescents. Journal of Gambling Issues, 16. Available at http://www.camh.net/egambling. Lesieur, H. R. (1994). Epidemiological surveys of pathological gambling: Critique and suggestions for modification. Journal of Gambling Studies, 10, 385–398. Lund, I., and Nordlund, S. (2003). Pengespill og Pengespillproblemer i Norge [Gambling and problem gambling in Norway]. Oslo: Norwegian Institute for Alcohol and Drug Research. May, R. K., Whelan, J. P., Steenbergh, T. A., and Meyers, A. W. (2003). The Gambling Self-Efficacy Questionnaire: An initial psychometric evaluation. Journal of Gambling Studies, 19, 339–357. Orford, J., Sproston, K., Erens, B.,White, C., and Mitchell, L. (2003). Gambling and Problem Gambling in Britain. Hove, UK: Brunner-Routledge. Ozga, D., and Brown, H. J. (2000). ADHD Screening of Adult VLT/Slot Machine Pathological Gamblers. Presented at the 11th International Conference on Gambling and Risk Taking, Las Vegas, NV. Petry, N. M., Stinson, F. S., and Grant, B. F. (2005). Comorbidity of DSM-IV pathological gambling and other psychiatric disorders: Results from the National Epidemiologic Survey on Alcohol and Related Conditions. Journal of Clinical Psychiatry, 66, 564–574. Productivity Commission. (1999). Australia’s gambling industries, Report No. 10. Canberra: AusInfo. Rönnberg, S.,Volberg, R. A., Abbott, M. W., Moore, W. L., Andrén, A., Munck, I., Jonsson, J., Nilsson, T., and Svensson, O. (1999). Gambling and Problem Gambling in Sweden. Stockholm: National Institute of Public Health.

54

Research and Measurement Issues in Gambling Studies

Rosander, A. C. (1977). Case Studies in Sample Design. New York: M. Dekker. Rosenthal, R. J., and Fong,T. (2004). The Etiology of Pathological Gambling. Report to the California Office of Problem Gambling. Sacramento: Department of Alcohol and Drug Programs. Rounsaville, B. J., Anton, S. F., Carroll, K., Budde, D., Prusoff, B. A., and Gawin, F. (1991). Psychiatric diagnoses of treatment-seeking cocaine abusers. Archives of General Psychiatry, 48, 43–51. Rugle, L. J., and Melamed, L. (1993). Neuropsychological assessment of attention problems in pathological gamblers. Journal of Nervous and Mental Disease, 181, 107–112. Schonlau, M., Fricker, R. D., and Elliott, M. N. (2002). Conducting Research Surveys via Email and the Web. Rand Monograph MR-1480-RC. Santa Monica, CA: Rand Corporation. Schrans, T., Schellinck, T., and Walsh, G. (2000). Technical Report: 2000 Regular VL Players Followup: A Comparative Analysis of Problem Development and Resolution. Halifax, Nova Scotia: Focal Research Consultants Ltd. Shaffer, H. J., Hall, M. N., and Vander Bilt, J. (1997). Estimating the Prevalence of Disordered Gambling Behavior in the United States and Canada: A Meta-Analysis. Boston: Harvard Medical School Division on Addictions. Smith, G. J., and Wynne, H. J. (2004). VLT Gambling in Alberta:A preliminary analysis. Lethbridge, Canada: Alberta Gaming Research Institute. Specker, S. M., Carlson, G. A., Edmonson, K. M., Johnson, P. E., and Marcotte, M. (1996). Psychopathology in pathological gamblers seeking treatment. Journal of Gambling Studies, 12, 67–81. Statistics Canada. (2002). Canadian Community Health Survey, Mental Health and Well-being Section (CCHS 1.2). Tourangeau, R., and Smith,T.W. (1996).Asking sensitive questions:The impact of data collection mode, question format, and question context. Public Opinion Quarterly, 60, 275–304. Trochim, W. M. (2000). The Research Methods Knowledge Base, 2nd ed. Cincinnati, OH: Atomic Dog Publishing. Tucker, C., Brick, J. M., Meekins, B., and Morganstein, D. (2004). Household Telephone Service and Usage Patterns in the U.S. in 2004. Unpublished monograph. van der Heijden, P.,Van Gils, G., Bouts, J., and Hox, J. (2000). A comparison of randomized response, computer assisted interview, and face-to-face direct questioning: Eliciting sensitive information in the context of welfare and unemployment benefit. Sociological Methods and Research, 28, 505–537. Volberg, R.A. (2001). Gambling and Problem Gambling in North Dakota:A Replication Study, 1992 to 2000. Bismarck: Office of the Governor. ———. (2002). Gambling and Problem Gambling in Nevada. Carson City: Department of Human Resources. ———. (2004). Fifteen years of problem gambling research: What do we know? Where do we go? Electronic Journal of Gambling Issues, eGambling Issue 10. Volberg, R. A., Abbott, M. W., Rönnberg, S., and Munck, I. M. (2001). Prevalence and risks of pathological gambling in Sweden. Acta Psychiatrica Scandinavica, 104, 250–256. Volberg, R. A., Gerstein, D. R., Christiansen, E. M., and Baldridge, J. (2001). Assessing self-reported expenditures on gambling. Managerial and Decision Economics, 22, 77–96. Volberg, R. A., and Vales, P. (2002). Estimados de prevalencia sobre el juego patológico en Puerto Rico [Prevalence estimates of pathological gambling in Puerto Rico]. Revista Puertorriqueña de Psicología, 13, 71–98. Welte, J., Barnes, G.,Wieczorek,W.,Tidwell, M.-C., and Parker, J. (2001). Alcohol and gambling among U.S. adults: Prevalence, demographic patterns and comorbidity. Journal of Studies on Alcohol, 62, 706–712. Williams, R., and Wood, R. (2004). Final Report: The Demographic Sources of Ontario Gaming Revenue. Toronto: Ontario Problem Gambling Research Centre. Available at http://www.gamblingresearch.org.

CHAPTER 3

Questionnaire Design:The Art of a Stylized Conversation Marianna Toce-Gerstein

Dean R. Gerstein

Georgetown University Washington, DC

Claremont Graduate University Claremont, California

The Interview Context What Is an Interview? Conceptual Context Physical Context: Setting and Mode Social Context Privacy Effects Response Bias Social Desirability Bias Nonresponse Basic Communicative Principles Relevance Neutrality Ambiguity Language of Administration Translating the Questionnaire Using Interpreters Minimizing Cognitive Effects Recall Problems Limited Grasp and Computability of Quantities Scaling of Attitudes, Opinions, and Behaviors Causality Age-Graded Behavior: Special Considerations for Youth Questionnaire Construction Structuring the Questionnaire Major Questions, Sections, and Section Order Pathing Through the Interview Question Flow and Context Effects Item Response Frames 55

56

Research and Measurement Issues in Gambling Studies

Reading Level Pretesting Informed Consent Special Considerations for Gambling Research Definition of Gambling Gambling Participation Attitudes Toward Gambling Problem Gambling Diagnosis of Pathological Gambling Correlates of Problem Gambling Problem Gambling Help and Treatment

The principal issue in developing a research questionnaire is how to minimize the extent to which the information that is collected is invalid or inaccurate—that is, to minimize the extent of missing, imprecise, biased, or unreliable data.There are not many published studies on the accuracy of measurement of gambling, in terms of the specific motives, attitudes, knowledge, experiences, and behaviors of those who engage in it. However, a substantial literature exists on survey1 methodology in general, based on experimental and other systematic studies, and it has yielded firmly proven results as well as expert-judgmental “best advice” for many aspects of questionnaire design dealing with similar subjects, such as substance use, sexual and reproductive behavior, and income and wealth.We recommend that anyone interested in developing a questionnaire first review one or more of the many excellent texts on this topic, which cover these issues in much more detail than we are able to here (see, e.g., DeVellis 2003; Groves et al. 2004; Schwarz 1996; Sudman, Bradburn, and Wansink 2004; Tourangeau, Rips, and Rasinski 2000). In order to help both producers and consumers of gambling data, this chapter first outlines major design principles and then identifies and recommends key practices for developing questionnaires on personal gambling behaviors and attitudes, their correlates, and their individual and social impacts.

THE INTERVIEW CONTEXT WHAT IS

AN INTERVIEW?

The research interview—and for simplicity, we will henceforth use the term interview to refer to all individual data collection episodes, including the 1 By “survey,” we are referring to a method of collecting data from a sample of individuals, with the intent of generalizing the findings to a population of interest.

The Art of a Stylized Conversation

57

completion of self-administered questionnaires by the respondent—is a complex social-psychological event or sequence of communicative behaviors. Although in most cases the interview involves direct communication between only the interviewer/instrument and the respondent, there are actually three other types of implicit participants who need to be taken into account in designing the questionnaire: the researchers, the intended audience, and unintended auditors. The question-and-answer format of interviews is deceptively simple. For design purposes, each interview is best viewed as a brief, stylized, dramatic conversation, partly scripted and partly improvised, in which a set of interconnected story lines are rapidly sketched out and recorded. Developing these story lines and making them as clear and accurate as possible can require a lot of work—mental concentration and sometimes emotional control—on the part of the respondent as well as the interviewer.

CONCEPTUAL CONTEXT Surveys come in two polar types, what we might call “portrait” and “landscape.” The portrait survey focuses on a single subject matter against a selective background of its most important contexts. The landscape survey incorporates a broad sweep of subjects, with limited details or contexts discernible for any one subject. Landscape surveys are often composites in which the question modules are sponsored by different clients. Surveys of either type may be strung together as longitudinal panels, meaning that the same questions are given to the same persons repeatedly over time. This permits one to look at change over time in persons or cohorts, and such longitudinal studies can provide the strongest available evidence about sequence, cause, and effect. In general, putting a small number of key gambling questions into a landscape survey is less expensive than devoting an entire survey to gambling issues, and that approach can be cost-effective if the right questions are asked and interest is limited to certain domains in which not many questions need be asked to get useful information. For example, this approach works well for measuring general attitudes toward gambling, awareness of gambling-related public messages or services, or the recency/frequency of participation in the most popular games, such as lotteries, casino slot machines, poker, and betting on major sports events. Portrait surveys are much more expensive to mount but permit far more questions to be asked about gambling, which is necessary for acquiring useful information about gambling-related problems or gambling expenditures. Perhaps the most problematic surveys attempt to be both portrait and landscape—to cover quite a number of topical areas, each in great depth.This type of survey is generally considered “too long” and can badly compromise the extent and quality of information obtained, especially for questions in the latter parts of the questionnaire.

58

Research and Measurement Issues in Gambling Studies

Researchers should be aware that the way in which gambling-related questions are contextualized will influence how respondents perceive the research and their role. The philosopher Gilbert Ryle made the observation (in Geertz 1973) that in the absence of context, we are unable to interpret the meaning of a wink.The winker might be attracted to us, she might be trying to communicate something secretly, or something else might be going on. As the context changes, the meaning of the wink changes. And so it is with questions; asking about gambling participation in a survey focused on alcohol use will send a different message than would the same questions in a casino customer satisfaction survey. This is a neutral and unavoidable fact of any communication—the way talk is framed influences participants’ understanding of its meaning.

PHYSICAL CONTEXT: SETTING

AND

MODE

Surveys can be administered—that is, the survey questions given to and answers received from the respondent—through various modes of communication. The classical mode is face-to-face, one-to-one, paper-and-pencil interviewing (PAPI), in which the interviewer reads each question in person to a single respondent from a printed questionnaire script and writes each spoken answer, or a categorical code representing it, in a designated answer blank, usually on the same printed page as the question.Variations of PAPI are nearly endless, and can include the following, in almost any combination: ●

●

●

●

●

●

The interviewer, instead of reading questions from a paper questionnaire, may read questions from a computer screen. The interviewer, instead of filling in answer blanks on paper, may key the verbal or precoded answers into a computer, which may in turn be a laptop, desktop, or handheld device. The respondent, rather than listening to questions spoken by a live interviewer, may read the questions on a printed page or computer screen (that is, self-administer the questions), or the questions may be audio or video recorded and administered to the respondent through headphones, a telephone, or a computer. The respondent, instead of speaking or writing down her answers, may enter them into a computing device using a keypad, keyboard, or pointing device (such as a mouse). Contact, instead of being face-to-face, may be via telephone, Internet connection, fax machine, or mail. Contact, instead of being one-on-one, may be in a group session.

The selection of a survey’s mode (or mixture of modes) is typically governed by balancing survey cost—usually calculated as average cost per completed case—against

The Art of a Stylized Conversation

59

(1) data quality, that is, the extent of missing items, errors, and invalid answers, (2) sample quality, or how well the surveyed sample represents the targeted, inferential population, and (3) sample size, that is, the number of completed cases. A review of the role of mode in measurement error can be found in Lyberg and Kasprzyk (1991). Usually the mode(s) is(are) chosen before the detailed questionnaire is developed. In most respects, there is little interaction between mode choice and questionnaire content—that is, the mode choices do not make any difference in designing the questions or in the answers that would be obtained. (Note that mode choices can dramatically affect sample coverage and response rates.) There are, however, a few specific implications for questionnaire development. First, it is much easier to successfully implement complicated skip patterns or other inter-item contingencies (such as wording variations) when using computers instead of paper to generate questions and record answers. As a rule, selfadministered paper questionnaires are the least desirable mode choice in this respect. Second, all other things being equal, different modes of data collection have different psychological effects (Tourangeau et al. 2000). For example, sensitive information tends to be disclosed more fully and accurately under intrinsically more impersonal circumstances, such as computer-assisted selfadministered interviews (Epstein, Barker, and Kroutil 2001;Tourangeau and Smith 1996). Finally, information that involves many categories or compartments, including information about changes over time, tends to be more readily obtained in person, where one can display and use visual aids such as calendars, lists, or graphics. Of course, there are benefits and drawbacks to each possible mode. For example, postal surveys are inexpensive but usually have relatively high rates of missing items and very low response rates; Internet surveys may fare even worse (Kwak and Radler 2002). However, with careful multiple-wave approaches that involve additions but not multiples to the cost, postal and Web surveys may produce much better than usual response rates (see Dillman 1999 and Couper; Traugott, and Lamias 2001). On-site surveys—for example, of patrons at gambling venues—cannot produce a representative sample of all gamblers or of the general population without extensive statistical adjustments that may not be possible to properly calibrate.Area probability surveys, although generally the “gold standard” for representing the general household population, may seriously underrepresent people who are infrequently at home, and even area probability surveys can have high refusal rates if insufficient persistence and persuaders are deployed. For example, Dickerson and colleagues (1996) reported that a remarkable 49% of the householders that they attempted to recruit in a “door knock” survey in New South Wales refused to participate. Telephone surveys, in which the interviewer reads questions from a computer screen and keys in the answers, are widely used in studies of health and social

60

Research and Measurement Issues in Gambling Studies

behavior and are the most common mode used today for gambling surveys. This mode is generally called computer-assisted telephone interviewing (CATI). It is less expensive than an area probability survey, generally obtains comparable results in general population studies, and for some sensitive topics provides a higher degree of anonymity, which, as mentioned above, can obtain more accurate data. However, survey research professionals in the United States and Canada report that response rates for telephone surveys in the general population have declined rapidly since their introduction in the 1970s.

SOCIAL CONTEXT Generally, respondents to a survey do their best to be cooperative partners, attempting to infer what the researcher means by a question and even the kind of response that is expected (Schwarz 1996). Nevertheless, there are predispositions and characteristics of respondents—some stable and others more transient—that influence not only the way in which one participates in a given interview, but whether one participates at all. Privacy Effects A privacy effect is a bias that results from participants’ desire to keep certain aspects of their lives private from persons either directly or tangentially involved in the interview. Typically the term refers to bias created by third-party observers, although bias may also result with regard to the interviewer.Third-party observers might include other members of the respondent’s household, strangers in proximity to where the interview is taking place, and even interpreters and interviewer supervisors. In fact, observers may not even be present during the interview; for example, cognitive interviews are often videotaped for later analysis. Finally, third parties may be nonexistent; the respondent may simply provide inaccurate information because he believes that someone might try to obtain information about him in an unethical fashion. Obviously, privacy effects will vary depending on a variety of factors, including the respondents’ personalities, their status in the community, personal events in their lives, and even events in the news. Studies of the role that third-party observers play with regard to sensitive issues such as drug use and sexual activity have had mixed results, some showing that the respondent is likely to provide less accurate information, others showing no effects, and a small number indicating that more accurate information is obtained by having a third party present (when data are collected from cohabitating couples). The one exceptional group is adolescents, for whom the evidence repeatedly shows that the presence of a parent results in the underreporting of sensitive behaviors (Aquilino 1997). In fact, even the mere possibility

The Art of a Stylized Conversation

61

of a known third party viewing the information has resulted in underreporting, such as that created by the practice of asking youth to write their names on questionnaires (ibid.). A key issue, obviously, is anonymity. As mentioned above in the section Physical Context, the more impersonal the mode of data collection, the more accurate responses tend to be (Epstein et al. 2001; Tourangeau and Smith 1996). A recent Dutch survey of adolescents found that they were more likely to report on certain sensitive behaviors such as alcohol use when the survey was anonymous than when it was simply confidential (van de Looij-Jansen, Goldschmeding, and Jan de Wilde 2006). Similarly, a study of adolescent smoking found that the survey mode that produced the most accurate estimates was in fact the most impersonal— that where participants used a telephone keypad to indicate their responses to computer-generated questions (Currivan et al. 2004). The important lesson to take from this body of work is that an individual’s responses may differ depending on the perceived audiences for those responses. The underlying motivators appear to be fear of reprisal and social desirability bias (see the next section). In practical terms, since gambling behaviors often create tension between family members, we recommend that interviews take place where the respondent cannot be overheard; (it may be worth rescheduling a time when the respondent will be alone). If interviews are conducted via telephone, care should be taken so that answers which might be spoken aloud do not reveal personal information. Response Bias Two response styles are of particular concern in surveys: acquiescent responding and extreme responding. Extreme responding is the tendency to choose one or another extreme when offered any ordinal or polar scale of values. Acquiescent responding is the tendency to be agreeable and positive to whatever the researcher asks, that is, to “yea-say.” Acquiescent respondents, when asked to evaluate persons or events, tend to downplay negative feelings or attitudes and accentuate the positive in their responses. In effect, acquiescent respondents adopt a higher midpoint. Research has suggested that this tendency may vary by demographic characteristics such as age, education, culture, and gender (Knäuper 1999; O’Muircheartaigh et al. 2000; Smith 2003). Acquiescent and extreme responding can coexist; for example, consider the case where, on a customer satisfaction survey, a person rates every service without exception as “excellent.” Jon Krosnick’s theory of survey satisficing builds on the notion of acquiescent responding; it states that some respondents may shortcut the cognitive process of responding to questions in two ways.The first way he calls weak satisficing, in which the respondent executes all cognitive steps involved in formulating responses, but incompletely and with bias. The second form is strong satisficing, in which the

62

Research and Measurement Issues in Gambling Studies

respondent offers responses that will seem reasonable to the interviewer without any memory search or information integration. Satisficing will result in the respondent typically selecting a no-opinion response option when it is offered, choosing socially desirable responses, failing to differentiate when a battery of questions asks for ratings of multiple objects on the same response scale, and/or tending to agree with any assertion (Krosnick, Narayan, and Smith 1996). Research on response effects has most typically focused on how they influence rating scales; such scales are typically used to measure attitudes and opinions but can also be used to measure such concepts as general health status and relative frequency of certain behaviors.They typically have an odd number of categories, with strong ones at either end (e.g.,“very favorably” to “very unfavorably”) and a neutral category in the middle. To avoid response bias from satisficing, some have argued that researchers should abandon the general practice of offering middle or neutral alternatives (Converse and Presser 1986). But this comes at a risk, particularly in attitude and opinion questions, since this response frame communicates to the respondent that she should have an opinion; the consequence (seen repeatedly across studies) is that very few people will volunteer that they do not (e.g., Schuman and Presser 1981)2; instead, we have seen that respondents who would have put themselves in the middle of the scale randomly select from one of the alternative categories (O’Muircheartaigh et al. 2000). (For more information on response scales, see the section Scaling of Attitudes, Opinions, and Behaviors.)

Social Desirability Bias Studies on survey response have identified another major type of response bias: social desirability bias. Social desirability bias is the motive to appear socially desirable, even if means bending or denying the facts in order to achieve or appear to meet socially desirable goals. Social desirability bias is a statistical tendency, not a feature of every person, but it colors enough responses in surveys to require attention in the design, operational, and analytical phases of research. Most respondents want to appear socially desirable, and many to some extent treat the survey experience like a job interview or a test on which they will be graded. Hence, many respondents tend to minimize or evade giving information about themselves that falls short of normative standards that they think the interviewer or survey team holds, or standards they hold for themselves (Tourangeau et al. 2000).The classic epigram in consumer research is that people drink half the amount of alcohol sold and use twice the amount of soap. 2 Attitude questions that do not have a middle alternative will usually give the option to the interviewer of coding “No Opinion” if this response is volunteered by the respondent.

The Art of a Stylized Conversation

63

Social desirability bias cannot be eliminated, but its influence on results can and should be rigorously minimized. Interviewers should be trained out of conventional conversational cueing behavior, such as nodding or smiling (or headshaking or frowning) in response to answers (see Fowler and Mangione 1990; van der Zouwen, Dijkstra, and Smit 1991). Wordings should be sifted carefully to find the most value-neutral and nonpresumptive alternatives—for example, “Next I have some questions about the people you live with” instead of “Next I would like to know about your family.” The selection of mode should also take issues of respondent “image management” into consideration (see Tourangeau et al. 2000).

NONRESPONSE Participation in an ethically conducted survey is completely voluntary, meaning that the respondent is free to decline outright, decline to answer individual questions or groups of questions, or stop the interview early. Participants are also free to be lazy or inconsistent in their answers, to invent stories or opinions, and to shade or deny the truth. A well-designed questionnaire is not proof against such outcomes, but a poorly designed questionnaire will increase their frequency, aside from the problems it may create in analyzing the data. Several factors can affect the response rates of a survey. A well-designed questionnaire lets the respondent know why her participation is important and that her privacy and safety are not at risk. A thoughtfully crafted introduction that addresses these issues is essential in averting initial refusals. Response rates can also be negatively affected if the initial questions lose the respondent’s interest or the instrument is too lengthy. Questions should make sense to the respondent both individually and in the way they are ordered, should not dwell at length on matters that are irrelevant to his life or interests, should not create discomfort—or if they must, should cover unpleasant topics discreetly, quickly, and without reproach—and should not seek any information that is not needed for the stated research purposes.

BASIC COMMUNICATIVE PRINCIPLES Within the questionnaire itself, a great many factors can influence how individual questions are approached by the respondent, including item wording, sequencing of questions, and complexity.The questionnaire designers also needs to take into account various cognitive, linguistic, and emotional barriers to effective and accurate communication, and the statistical tools and interpretive strategies that the survey’s analysts will likely use.We can only touch briefly on some of these

64

Research and Measurement Issues in Gambling Studies

issues here; for more detail and breadth on writing survey questions, we recommend Fowler (1995), Schaeffer and Presser (2003), and Sheatsley (1983).

RELEVANCE Human communication is the exchange and interpretation of information, and such communication processes are remarkably intricate and subject to many kinds and degrees of success and failure. Substantial fields of study have focused on the way that survey questions are understood and addressed by respondents, especially how comprehension of and response to questions are influenced by prior questions. Perhaps the cardinal principle of communication that pertains to survey response is relevance—what experiences, persons, norms, attributes, and/or subjective states the respondent is mentally focusing on when providing the descriptive or evaluative information that a particular question evokes. Respondents constantly search for and interpret cues about what should be considered relevant for each item of a questionnaire. Effective questionnaire design is the mastery of this cueing process, so that the respondent considers relevant to each question only and exactly what the researcher does. One key aspect of relevance is that for any particular item, the personal or partisan desires or preferences of the researcher (or interviewer) are not made relevant to the respondent. The only relevant preference that should be cued by the questions is that the researcher desires the respondent’s unblinkered, unedited opinions, beliefs, perceptions, accounts, or estimates.

NEUTRALITY Neutrality of perspective is a second critical principle in designing questions— that is, neither questions nor response frames should be phrased or sequenced so as to “lead” the respondent to select answers that the respondent would consider forced or not quite accurate. Non-neutral questions provide less valid data in themselves and annoy respondents, making them less cooperative with the survey. Neutrality is partly a matter of avoiding invidious or normatively loaded wording (such as “Are you married, not yet married, or did your marriage break up?”), but also a matter of acknowledging the nearly infinite variability and continuing evolution of social arrangements and attitudes.This often means that the designers should develop a logical series of branching questions, each with very simple choices, rather than trying to pack numerous categorical possibilities into a single question. For further discussion on how question wording and response frames can intentionally or unintentionally bias results, see Knäuper (1998) and Schwarz and Hippler (1991).

The Art of a Stylized Conversation

65

AMBIGUITY Words are very flexible instruments for conveying meaning, and that flexibility includes the fact that many words and phrases can convey different meanings depending on the context, the listener’s previous experience with the word or phrase, or the intonation as spoken or read. It is critically important that questions convey the same meaning to each respondent and that it is the same meaning that the researcher intends. Researchers cannot always be there to make sure their respondents understand them, or to answer questions when they do not. Studies have found that terms one might consider equivalent can, in fact, produce different outcomes (Smith 1987;Tourangeau et al. 2000). More problematic, though, is when a respondent asks the interviewer what she means by a particular concept—for example, betting on a game of skill (“Does it have to be on a game that I was playing, or could it also be on a game my friends were playing?”); the typical response of “Whatever it means to you” would come up short here, resulting in not only frustration on the part of the respondent, but an over- or underestimation of what the researcher is seeking to measure (Fowler 1992; Schober, Conrad, and Fricker 2004). A related issue concerns what are known as vague quantifiers—words and phrases like “hardly” and “very.” Vague quantifiers have been the subject of numerous investigations in experimental psychology and psycholinguistics; not only has this research revealed differences in how individuals understand these terms, but some work has shown that these differences may vary by race, education, and age. For more information, we refer the reader to Bradburn and Miles (1979), Schaeffer (1991), and Wallsten et al. (1986). Researchers typically attempt to avoid ambiguity by using questions whose clarity and stability of meaning have already been tested successfully in prior studies and through careful pretesting of the instrument (see the section Pretesting). The importance of this phase of questionnaire development cannot be overstated.

LANGUAGE

OF

ADMINISTRATION

Monolingual respondents must be interviewed in their own language, and bilingual or multilingual speakers in their language of greatest fluency and/or situational comfort. Study investigators can choose to tailor a given questionnaire for multiple linguistic groups through translation of the instrument, live interpretation, or both. Translating the Questionnaire Very few studies have been conducted on best practices for translating questionnaires for survey research. Nevertheless, many general-population surveys are developed in the official or most commonly spoken language first (the “source”

66

Research and Measurement Issues in Gambling Studies

language), with the written questionnaire then being translated into one or more additional languages (the “target” languages). This helps ensure comparability among instruments to the extent that this is possible. In addition to requiring accuracy and semantic equivalence in the target language, the instrument will also typically require some degree of tailoring to the cultural context; this will require consideration of the kinds of question-and-answer formats that respondents will be familiar with, culture-specific norms of politeness, and norms regarding the sharing of sensitive information (Harkness, van de Vijver, and Johnson 2003). When administering a questionnaire in two or more languages, one can only speculate for any given question the extent to which between-group differences actually exist or are artifacts of the translation. A number of techniques have been developed for translating questionnaires. The most basic is a simple direct translation that is performed by a single bilingual individual who translates the questionnaire from the source language into the target language. Another method of translation is called back translation; this iterative technique uses direct translation but then retranslates the translated text back into the source language to check for semantic equivalence. If necessary, changes are made to the source items and the process is repeated. While this method increases the likelihood that the meaning of the original question is more accurately conveyed, it unfortunately does not respect the cultural translations that should also take place. One preference is for the committee approach to instrument translation (Behling and Law 2000; Harkness et al. 2003; Pan and de la Puente 2005). In this method, the committee ideally comprises at least one arbitrator, several translators, translation reviewers, subject matter specialists, the questionnaire designers, and the person who will be in charge of conducting the pretests. Several translators work simultaneously and independently of each other to translate the full instrument. When finished, the team convenes to compare and discuss the translations. A new version is then created, which is submitted to the arbitrator to review and for any final decisions. Pretesting then commences for the translated version. Once the pretest is completed, the team reconvenes to discuss the findings and make any additional changes before main data collection begins. Using Interpreters Nonstandardized, “on-the-fly” interpretation of interviews by bilingual interviewers and third-party interpreters are problematic and should be avoided whenever translation is possible. In fact, it is better, prima facie, to explicitly omit a linguistic group from a survey than to incorporate and report on data that may or may not be valid. Despite this, we are aware that circumstances (and funding agencies) sometimes require that interpretation be used; therefore, we will describe three methods for improving the quality of data from these interviews.

The Art of a Stylized Conversation

●

●

●

67

First, interviewers must be fluent in the respondents’ target language, and preferably also the dialect. Second, training is essential for both interpreters and the interviewers who will work with them. Interpreters require training similar to that which the interviewers receive, including information about specific requirements of the data collection and a copy of the questionnaire. Additionally, interviewers should be trained in how to work with interpreters—for example, in maintaining a moderate pace (Gerver 1969) and in communicating adequately to ensure that the interpreter does not get overtired. Third, we recommend that telephone interviews be monitored in real time to determine whether the interpreter is following good survey protocol and doing a satisfactory job of interpreting the questions.

MINIMIZING COGNITIVE EFFECTS RECALL PROBLEMS The human mind is a remarkable instrument, but it did not evolve for the express purpose of responding to research surveys.The most common recall problem is certainly outright forgetting, but a second frequent problem, which may be just as important, is backward or forward telescoping, in which a particular event is misplaced in time—thought by the respondent to have occurred more recently or less recently than it actually happened. A third problem is confuting one event with another; that is, taking features of two or more different events or points in time and ascribing them to the same event or point in time. Of course, there is a counterpart called splitting, in which features of one event or point in time are ascribed to separate occasions. All of these issues pose challenges to questionnaire designers and call for investigators to know the relevant research literature and technologies in order to maximize the validity and usefulness of the data. Questions about time frames need be sensitive to findings on the average trajectories of forgetting and the directions of telescoping bias for different types of events. A very important class of tools includes calendar aids, which help anchor events in time by first establishing salient markers such as birthdays and major holidays or, for longer time frames, major life events such as births, weddings, graduations, and so forth.When asking respondents to recall and date events in the distant past, interval approximations should be offered. One cannot get blood from a stone, but at least a stone will not mind the effort; however, the human respondent becomes frustrated and resentful of repeated questions that are too demanding and cannot be answered.

68

Research and Measurement Issues in Gambling Studies

LIMITED GRASP

AND

COMPUTABILITY

OF

QUANTITIES

Just as people are not digital imaging devices, they also are not digital computing devices. In general, for example, people perform poorly when asked to produce frequency reports (e.g., Burton and Blair 1991).Training and aptitude in basic arithmetic varies, and very few people of any aptitude keep running accounts and dated logs of everything they do. So when asked during an interview to sum up the number of instances of a behavior or to compute ratios of behavior over time (two very common forms of survey measurement), the results will nearly always be a rough estimate. Further, these estimates will become rougher—that is, more approximate and more subject to recall errors and biases—the longer the period covered. In methodological studies, participants have been shown primarily to use two methods to calculate the frequency of events: (1) episode enumeration, in which interviewees attempt to recall specific events of the behavior in question and sum across time, and (2) rule-based estimation, in which they calculate how often they usually perform the activity in question in a given time frame and then multiply this to determine the frequency for the time period that has been requested. Burton and Blair (1991) suggest that rule-based estimation is superior for events that are numerous, similar, and regular in frequency, whereas episode enumeration will likely produce the most accurate estimates for events that are less frequent and more distinctive and occur with greater irregularity.While all gamblers play differently, this recommendation does suggest that we may achieve better results by encouraging respondents to use rule-based estimation for lottery play (e.g., how many tickets are purchased in a typical week), and episode enumeration for less routinized activities. Questions that ask respondents to count detailed frequencies should never extend more than 12 months into the past, and even this will likely still produce underestimation of nonregular events.Although research has not yet pinpointed an ideal time frame, it likely would lie somewhere between the past month and past six months (assuming that seasonal variation were not an issue). For regular events, it can be expected that respondents will rely on memory of recent typical behavior and extrapolate to longer time frames, with recent habits overlaid on earlier ones; therefore, there would likely be no loss of accuracy were a shorter time frame used. Other well-established methods for obtaining estimates of behavioral frequency and expenditures go beyond the question itself. The best known and most effective of these is the use of prospective diaries, in which respondents make a daily record of the behavior in question and then provide these records to the researcher at predetermined intervals. To our knowledge, there have been no general population studies of gambling that have used this method. However, researchers have achieved some success using a second method, known as the

The Art of a Stylized Conversation

69

timeline follow-back, which uses calendar aids (described above) to enhance recall. Recently this method was found to achieve good to excellent reliability and fair to good validity for report of gambling behaviors in a small treatment outcomes study of problem gamblers (Hodgins and Makarchuk 2003).

SCALING

OF

ATTITUDES, OPINIONS, AND BEHAVIORS

A wide variety of scales have been used in the history of studies to measure degrees of experience. For example, one might be asked by an interviewer, “On a scale of one to five, how would you describe your general health? ‘One’ is poor and ‘five’ is excellent.” Scales can include as few as two points to as many as the researcher would like. However, studies have shown that cross-sectional reliability, test–retest reliability, and criterion validity are optimal for scales of between five and eight points (Smith 1994). Likewise, interrater agreement is best, and susceptibility to context effects is lowest, when rating scales have between five and eight points (Wedell, Parducci, and Lane 1990).We recommend using the shortest scale within this range that is appropriate, keeping in mind that the length of time required to respond will increase with the number of items in the scale. If the survey mode permits visual aids, one can represent the scale along a horizontal axis. In telephone interviews, the questionnaire designers can include the endpoints as part of the question, while providing wording for other points on the scale for the interviewer to use as prompts when necessary.

CAUSALITY The challenge of finding and proving causal relationships is a serious one for the social and behavioral sciences. This is attributable mainly to the complicated objective/subjective nature of nearly all psychological and social phenomena and the many ways that people can get into any particular kind of trouble, whether financial indebtedness, family discord, criminal culpability, deep mental distress, or something else. The major diagnostic screens for pathological gambling tap a variety of gambling-specific behaviors, feelings, and motives, as well as adverse consequences of gambling for work, school, finances, and family relationships, in order to determine whether—and with what overall severity—gambling is a personal problem for the respondent. A different line of questioning in many surveys asks respondents whether certain specific current or past adverse events or circumstances may apply to them, such as bankruptcy, job loss, arrest, or the like, and then, for any items affirmed, asks whether their own gambling contributed causally to this result.

70

Research and Measurement Issues in Gambling Studies

In our own work (Gerstein et al. 1999), we found some disjunction between results for the last form of attribution, on the one hand, and statistical correlations between overall diagnostic status and the occurrence of adverse events and circumstances, on the other.That is, we found that respondents whose self-reports of gambling-related signs and symptoms classified them as problem and pathological gamblers were also much likelier—by multiples of between two and five to one—to report bankruptcy, arrest, and other problems. Yet when asked directly, few of the same respondents directly attributed these adverse events to their gambling. Of course, statistical correlation and cause-and-effect are not the same thing. The correlational strength in such findings may be due in whole or in part to a reverse causal path (other problems leading to gambling disorders) or to gambling disorders and other problems sharing one or more underlying causes. Also, the apparent attributional deficit may be due to systematic underreporting of adverse linkages, similar to the minimization of gambling losses. Sorting out these causal pathways is very difficult in a cross-sectional study.We recommend that researchers continue to use multiple approaches, asking about specific diagnostic markers of gambling and suspected adverse consequences of gambling, with follow-up questions of attribution.

AGE-GRADED BEHAVIOR: SPECIAL CONSIDERATIONS FOR YOUTH As a rule, studies of minors pose special difficulties for questionnaire designers. Minor children are subject in many jurisdictions to extra protections and requirements, such as the need for parental consent either implicitly (the parent is offered a fair and clear chance to opt the child out of participation, and does not do so) or explicitly (by signing a form or giving an interviewer direct verbal approval). However, consent is only the first hurdle. Children’s responsibilities, routines, language, and levels and modes of comprehension and emotional control are very different from those of adults and change substantially as children age. By and large, researchers do not sample children under the age of 9 or 10—that is, younger children are not asked to provide self-reports on their own or others’ behavior or to express their attitudes or opinions for statistical analysis. This is not due to lack of interest in child behavior, but distrust that younger children have the cognitive skills, emotional maturity, and social dexterity needed to provide useful and valid data. Nevertheless, some methodological studies with children have achieved some success. For example, Amato and Ochiltree (1987) compared the data quality obtained from children aged 8–9 and adolescents aged 15–16 and concluded that while the data from the younger children were inferior, the differences were minimal, with data

The Art of a Stylized Conversation

71

quality from both groups being high in absolute terms. The researchers recommended that studies with younger children focus (1) on the present, as past events are difficult for younger children to recall, and (2) on subjects that are within the children’s direct realm of experience—for example, what they do at school, rather than what a parent does at work.The article also suggested a number of strategies for interacting with young respondents to improve participation and data quality. A study by Backett and Alexander (1991) confirmed some of these findings and additionally suggested that researchers avoid items regarding opinions and evaluations. By and large, adolescents do have the skills needed to complete surveys. However, many questions that are framed for adult respondents change meaning or lack relevance when asked of youngsters, and so are often incommensurable or of doubtful comparability. For example, few adolescents provide all or an appreciable fraction of the income that supports them, but most adults do. Most adolescents are enrolled in school, but their routines change substantially during summer months when school is out of session. Adolescents change their behavior much more rapidly than adults. Finally, adolescents appear to be more cautious in their responses when the data are not anonymous or the possibility exists that a parent may be listening in (see the section Privacy Effects). In short, although the same basic principles of communication and survey technique apply to studies of youngsters and adults, the applications can be quite different, and all questions and procedures need separate age-specific testing. We recommend that readers interested in this topic review Fontana (2002).

QUESTIONNAIRE CONSTRUCTION STRUCTURING THE QUESTIONNAIRE It is natural to surveys—as opposed to detailed case studies, for example— that the same overall set of questions be asked of each participant, although some subset of questions may be selected or omitted based on individual response patterns (see the section Pathing Through the Interview). The ways and degrees to which questions and answers are constrained toward uniformity is usually called the structuring of the interview. Less structured interviews—with less uniform, more improvisational questioning and more open response frames—lend themselves much more readily to qualitative analysis, while more structured interviews are better suited to quantitative (statistical) analysis, particularly when collected from large samples of respondents. Neither approach is intrinsically superior to the other, but they have different strengths and weaknesses, and limitations and

72

Research and Measurement Issues in Gambling Studies

potentials.While the present authors are versed in both qualitative and quantitative interviewing, our focus in this chapter is on the development of highly structured interviews characteristic of modern quantitative survey research. However, many of the design principles discussed here also apply to other types of interviews.

MAJOR QUESTIONS, SECTIONS, AND SECTION ORDER With limited exceptions for experimental or other special purposes, individual questions should not be distributed randomly across a survey, but grouped into topical sections.The most general rule to apply in deciding how to organize questions by topic is conventionality—the topics should accord with the commonsense ways that most respondents conceive the world, such as job, home, health, and recreation, and each section should begin with an introduction of the topic, such as “Next I have some questions about your sources of income.” In other words, topics should not be organized by the technical concepts of the researcher. The ordering of sections should follow four principles. First, items most essential for survey interpretation or analysis should come early in the sequence; second, survey order, when possible, should be arranged so as to minimize the need to repeat items; third, the designers should take into consideration the difficulty and interest of different topics and try to alternate between tedious/interesting and hard/easy sections; and fourth, the most sensitive questions should be placed as close to the end of the interview as possible.

PATHING THROUGH THE INTERVIEW Pathing refers to the sequence of specific questions that a given respondent is offered during an interview, taking into account the structure of logical contingencies that may be present in the questionnaire—such as branches, gates, filters, fills, and skip patterns. A questionnaire with no contingencies, in which every respondent is asked the identical set of questions, is usually said to have no pathing (strictly speaking, it has a single path). Pathing is generally based on very straightforward logic—a respondent’s answer to one question (“Have you ever been married? ‘No’”) may logically provide the answer to one or more additional questions, which can therefore be skipped without losing any information. Sometimes path options are simple, such as skipping a single question depending on the answer to the one immediately before it. They can be highly complex, making the pathing choice contingent on the answers to multiple, widely spaced items in complicated Boolean formulas. Pathing permits the tailoring of questions more precisely to respondent characteristics and can minimize the total number of questions or average length of interviews needed to obtain the desired information

The Art of a Stylized Conversation

73

(pathing is the complement of adaptive testing in psychometrics). However, the more elaborate the pathing, the greater the challenge in completing the interview correctly and the greater the challenge to analysts in correctly interpreting the resulting pattern of answers and nonanswers. Elaborate pathing is greatly facilitated by computer-assisted administration, which can automate nearly all pathing decisions, and by the postsurvey step of data imputation—that is, replacing answer codes that mean “question was not administered” with the logically imputed answer.

QUESTION FLOW

AND

CONTEXT EFFECTS

There is a substantial literature in survey methodology devoted to understanding how one question affects the questions that follow, or more exactly, how the answer to one question can depend on not only the contents of that question but also the contents of questions and related survey information that preceded it (this group of phenomena is also known as context effects). This has to do both with communicative principles such as relevance and its determination and with other workings of human cognition, especially with how people learn to pay attention, retrieve memories, and calculate and estimate numerical values. Putting questions in the right order can make it easier for respondents to understand correctly what a question is about, to remember correctly things that they have done or observed, and to compute quantities more accurately. Study practices well informed by these methodological studies will yield more valid and useful results than practices that ignore them. For more information on this important subject, we recommend Moore (2002), Tourangeau et al. (2000), and Sudman, Bradburn, and Schwarz (1996).

ITEM RESPONSE FRAMES The possible ways in which an answer to a question can be recorded are referred to as response frames. The major types of response frames are fixed, open, and partially open. In a fixed frame, only a small number of all possible responses may be recorded, usually corresponding to exclusive logical possibilities that are either directly presented to the respondent for selection or else interpreted and confirmed by the interviewer from what the respondent offers (“yes/no,” “zero/one/two/three or more,” etc.). An open frame permits any response to be recorded that is clearly relevant to the question. Partially open frames permit any relevant response that falls within certain limits, with a fixed procedure for encoding responses outside those limits (e.g., in a computer-administered survey, the question “How many natural or adopted children of your own have you raised?” may have an upper limit, beyond which the user would be prompted to revise the response).

74

Research and Measurement Issues in Gambling Studies

Generally, fixed and partially open response frames are supplemented by standardized reserve codes, including “Don’t Know,” “Refused,” “Legitimate Skip,” and sometimes an all-purpose miscellaneous category such as “Other—Specify: [verbatim].”“Legitimate Skip” would be used as a response when the respondent is not asked a question due to his response to an earlier question (as discussed in the section Pathing Through the Interview).

READING LEVEL Another key element to be considered when constructing the questionnaire is achieving maximum clarity for the target audience.A prime consideration is the level of education that will be assumed for respondents; this should be determined in advance, and the questionnaire prepared accordingly. Next, the questionnaire should be tested for readability according to grade-level targets.The questionnaire designers should continue to be on the lookout for issues throughout the pretest, so that they can be resolved prior to the main data collection. Probably the most important aspects of readability pertain to syntax, grammar, and vocabulary. Even though two sentences that convey the same information may have the same number of words with the same number of syllables (hence, the same reading level), one may still be much more difficult to understand than the other. Questions are more comprehensible if they do not use technical terms and have a conversational tone. One should avoid unnecessary and uncommon words and make questions as simple as possible. In addition, keep in mind that even though a word may count as “big” because it has more than two syllables, if a word is encountered regularly in day-to-day life, it is likely that a literate adult will grasp it quickly (Tefki 1987). While there is clearly more to the readability of a document than measuring the frequency of big words or long sentences, nearly all of the readability formulas that are available (more than 40) use algorithms based on these qualities to provide a reasonably accurate estimate of grade level. The four most commonly used indices are the Flesch Reading Ease Score, Flesch-Kincaid Index, Fog Index, and SMOG (Simplified Measure of Gobbledygook). Even though different formulas tend to result in different scores, they have been found to correlate highly with one another (Meade and Smith 1991). As a practical note, Microsoft Word uses the Flesch-Kincaid Index to determine the grade level of a text, while the Fog and the SMOG are calculated by hand.

PRETESTING Any questionnaire, no matter how brief, simple, or apparently foolproof, should be pretested prior to going into the field. The pretest should examine,

The Art of a Stylized Conversation

75

where relevant, any training materials, calling respondent selection procedures, and the overall design, flow, and respondent comprehension of the questionnaire. Before the pretest, a number of steps need to be taken. First, if the questionnaire has been programmed into a software system, the programming needs to be carefully tested so that interviewers can use the pretest to familiarize themselves with the correct instrument. Testing the questionairre pathing is critical to ensuring that respondents are not skipped over questions they should be asked, as well as for preventing respondents from being asked inappropriate questions.In many studies we have worked on, we have seen numerous crises due to insufficient skip-pattern testing, where valuable data were lost due to incorrect programming and rigorous quality control procedures not having been in place.We strongly recommend that the questionnaire designer take part in this phase, since no one will know the instrument as thoroughly. Second, it is important for the design team to record and distribute notes for interviewers on specific questionnaire items; such notes typically include clarifications about specific words or concepts within a question that some respondents may not understand.These are usually introduced during the interviewer training. Lastly, interviewer training should take place prior to the pretest. In our experience, the questionnaire designers should work as part of a team with the lead interviewers for this training, since it is during this time that the most egregious oversights are usually identified; these should be handled before the main data collection begins. One of the additional ways researchers pretest specific questionnaire items is through cognitive testing. This method is used to gather detailed information from individuals who have not previously encountered the interview and to examine how they formulate their responses. Extensive probing usually takes place following the interview or while the respondent is answering the questions.A number of styles of cognitive testing are available (see Willis [2004] for an overview). For example, the concurrent think-aloud method asks individuals to discuss their thought processes during and immediately following each question. They may be asked to provide feedback on their reactions to each question’s structure and wording and, as part of this process, repeat back each item for the interviewer in their own words. Respondents may also describe the experiences or behaviors which they believe qualify or do not qualify them to offer a particular response. The interviewer then probes as needed for additional information and for sources of confusion or misunderstanding. Research has shown that cognitive interviewing of only a few respondents— even fewer than ten—can detect major ambiguities in question wording (Fowler 1992). However, if cognitive interviews are not to be used, we recommend that as part of the pretest, interviewers ask respondents to comment on the questionnaire introduction, question wording, and the content and flow of the instrument. Interviewers should then fill out a “thumbnail sketch” summarizing the respondent’s comments during the interview, identifying problem questions, and indicating any instrument programming issues (where relevant).

76

Research and Measurement Issues in Gambling Studies

Once the pretest has been completed, the questionnaire designers will need to incorporate what has been learned from the pretest into the instrument. For this reason, the questionnaire designers should participate along with the interviewers and other team members in a pretest debriefing in order to review major findings and discuss any additional observations.Taken together, interviewers’ thoughts and observations will provide important insight in determining whether a question or answer option is working as designed, and if not, why. An excellent and thorough work on testing questionnaires is Presser and colleagues (2004). This useful guide documents a wide variety of methods for evaluating and improving instrument design.

INFORMED CONSENT Any given study typically presents both benefits and risks to participants. Benefits might include exploring an interesting topic, enjoying the social exchange of the interview, and knowing that one is making an important contribution to the body of knowledge in a particular area. However, certain information could cause harm to the respondent were it to be revealed, such as past suicide attempts and stealing to pay gambling debts. One needs to inform participants of such risks and maximize protection of their interests. Researchers need to adopt as part of every data collection effort a set of measures for informing prospective participants about possible and likely risks and benefits and obtaining clear, unforced, verifiable agreement to participate. Outlines of such measures have been codified in professional customs and in legislation in many jurisdictions. These measures differ depending on the status of the research group (university-, government-, or other-affiliated), as well as the legal status of the respondent; for example, additional consent may need to be gained from the parents of minors (Coyne [1998] provides an interesting discussion of informed consent issues with children) or from institutions in the case of studies of prisoners, patients, and the like. As soon as a respondent is contacted, and prior to the administration of any part of the interview, the interviewer should read a consent script and obtain a positive expression of informed consent.The script should identify the sponsoring institution and/or survey organization, the purpose and nature of the survey, and whether the data are anonymous and how confidentiality will be protected. In the event there will be any later contact with the respondent or her information, such as a request to participate in further data collection, this possibility should also be mentioned. Respondents should be explicitly told that they have the right to discontinue the interview at any time.They may also be provided with the name and telephone number of the principal investigator or equivalent in case they have any questions, concerns, or complaints about the study.

The Art of a Stylized Conversation

77

SPECIAL CONSIDERATIONS FOR GAMBLING RESEARCH This section discusses specific issues the researcher may want to explore with regard to gambling.The discussion here is meant to provide guidance on some of the more common variables, based on the authors’ experience conducting their own gambling studies, as well as working with data from other studies in the field. The topical areas described are not meant to be exhaustive, nor will all of them be relevant to all of our readers. They illustrate many of the questionnaire design principles discussed in earlier sections.

DEFINITION

OF

GAMBLING

Folk definitions of gambling vary widely. For example, many people would consider buying stocks as gambling, and many people would not consider buying lottery tickets as gambling—although these are not the majority’s ideas. Due to these divergent definitions, any questionnaire items about gambling, no matter how few or how many, must be preceded by a very clear definition, followed by specification of exactly what kinds of activities the questionnaire designers want the respondent to cover when answering questions. Although there is no single standard, an example of current practice in definitions is as follows: I would like to begin by asking about your experience with various kinds of wagering or betting, including what kinds of gambling facilities are located near you. By “betting,” I mean placing a bet on the outcome of a game of skill or chance, or playing a game in which you might win or lose your money.

Then the interviewer should name specific games or venues as each one is queried; or, if general questions are being asked about gambling, such as global positive or negative attitudes, the interviewer should provide at the end of the definition a list of four or five specific examples, preferably ones that nearly all people in the places being sampled will understand and recognize—for example,“betting on slot machines or table games in a casino, betting on horse races off-track or at the track, buying lottery tickets, playing in a bingo hall.” If the interest is strictly in publicly licensed games and not private betting, or only betting within certain geographic locations (a region or province), that needs to be specified. One should avoid examples that involve proper or trade names, specialized jargon, less common games, or ambiguity about whether the activity actually involved gambling—for example, “betting at a simulcast . . . betting on a class II device . . . going to Las Vegas.” We also recommend that researchers use multiple words to describe gambling for clarity.The word “gambling” itself carries negative connotations based on

78

Research and Measurement Issues in Gambling Studies

its historical ties with organized crime.“Betting” and “wagering” are two substitutes we have used to ask about gambling, with no apparent loss of understanding. We do not recommend “gaming,” as this term is often used to describe videogame playing and other types of fantasy play (e.g., Dungeons and Dragons) that typically do not involve financial risk.

GAMBLING PARTICIPATION The most accurate data about participation comes when it is framed by venue, with respondents being asked about their wagering and their playing of specific games by venue.We find it helps respondents if they can visualize a specific place as a recall aid, before describing their activities within it. In contrast, gambling participation has been found to be substantially underreported when researchers have asked global questions (e.g., “Have you ever gambled?”) (Volberg and Banks 1990). New types of gambling that cross standard venue boundaries are created frequently. Furthermore, different jurisdictions will have different types of venues and games. Therefore, the venues and game descriptions should be tailored to reflect local contexts. Decisions should be made about how to characterize relatively rare types of gambling (e.g., purchasing lottery tickets over the Internet). Where necessary in a written questionnaire, this information may be conveyed as a parenthetical note at the end of the question. It is desirable to keep spoken questions concise, so unusual permutations may be described in materials provided to the interviewers discussed in interviewer training, then provided to respondents when clarification is needed.

ATTITUDES TOWARD GAMBLING While a number of variables can be used to learn about public opinion with regard to gambling, researchers should be warned that looking at gambling participation rates and patterns alone tells us little besides how common a behavior is. In the United States, opinion polls and surveys have demonstrated since the 1970s both an increasingly positive view toward gambling as well as increases in the number of individuals who gamble (Gallup Organization 1999; Kallick et al. 1976; Scripps Howard News Service 2006; Volberg, Toce, and Gerstein 1999). However, it is critical to look beneath the surface of general approval data. Though most Americans view gambling’s effects on society as “neutral” or a balance of positive and negative, it is still the case that far more Americans believe that gambling’s effects on society are negative rather than positive (Gerstein et al. 1999). A 1996 Gallup poll found that even though 70% of respondents disagreed that gambling

The Art of a Stylized Conversation

79

was immoral, 67% agreed that it encouraged people who could least afford it to “squander” their money, and 61% agreed that it could make a “compulsive gambler” out of someone who would never gamble illegally (Heubusch 1997). Another important attitudinal variable is problem typology—gambling attitudes vary significantly depending on the degree of problems an individual has experienced.Volberg and coauthors (1999) found that individuals who reported a modest to moderate problem with gambling were much more positive about gambling’s overall effects on society than current gamblers who reported no problems. However, nearly half of pathological gamblers reported that the overall effect of legalized gambling on society was “bad” or “very bad.” Another common way that polls and surveys have explored attitudes toward gambling is by asking individuals why they have or have not gambled recently. The first national survey we are aware of that did this was the 1975 U.S. National Commission survey (Kallick et al. 1976). Unfortunately, while polls tend to ask why people approve or disapprove of gambling, they seldom inquire about why people do or do not gamble. However, an understanding of why people do and do not gamble can play an important role in the prevention and treatment of gambling problems. In this regard, understanding such differences on a group level can contribute to our knowledge about vulnerabilities and protective factors (see Gerstein et al. [1999] for further discussion).

PROBLEM GAMBLING Diagnosis of Pathological Gambling Administering any of the existing screens for problem and pathological gambling takes at least several minutes of interview time, which is costly in large-scale general population surveys. Researchers therefore typically use filtering mechanisms to skip low-frequency gamblers over diagnostic screening items. For example, surveys have skipped respondents who never lost more than $100 gambling in a single year of their lives (Toce-Gerstein, Gerstein, and Volberg 2003) or never gambled more than five times in any given year of their lives (Petry, Stinson, and Grant 2005). Alternative brief screens to filter out unlikely problem gamblers and reduce survey costs include the two-item Lie/Bet ( Johnson et al. 1997) and three-item NODS-CLiP (NORC Diagnostic Survey—Control, Lies, and Preoccupation) (Toce-Gerstein and Volberg 2003). Another way that questionnaire designers attempt to save time is by reducing the time frame covered by screening questions. The Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV), puts the entire lifetime in play (“Have you ever. . .”), but many researchers ask only about more recent periods (e.g., the past 6 or 12 months). Shaffer, Hall, and Vander Bilt (1997) argue

80

Research and Measurement Issues in Gambling Studies

that the lifetime measure inflates estimates of problem gambling, and they recommend that researchers rely instead on past-year (or other “current”) time frames “as the most accurate measure of the existence of clustered indicators of a gambling disorder” (p. 64). However, it is equally plausible and consistent with the DSM-IV that an active case be defined as anyone with a history of pathological gambling who exhibits one or more criteria in the past year, as was true of 83% of pathological gamblers in the NORC Gambling Impact and Behavior Study sample (Toce-Gerstein and Gerstein 2004). We recommend that researchers continue to estimate lifetime prevalence in the diagnosis of pathological gambling, since no data exist from which we can judge how many “current” DSM-IV criteria are needed for a “current” diagnosis. We also, however, recommend that researchers continue to assess which criteria respondents have experienced in a recent time frame for the following reasons: (1) Recall is best for more recent events; (2) the criteria can be linked to each other as well as to other characteristics in a fixed period of time; and (3) eventually research may suggest that a certain number of criteria, or specific criteria, are indicative of acute episodes of pathological gambling. Finally, researchers should look at the clustering of problems across longer periods of time, to further our understanding of precursors to problem gambling and aids to recovery. Correlates of Problem Gambling Gambling behaviors, including problems, can vary by a wide variety of respondent characteristics.The most basic of these are demographic characteristics that are collected as part of virtually all studies, such as gender, age, and ethnicity/ culture. Variables that have been shown to interact with gender include venue most often frequented, favorite game, age first gambled, age first had gambling problems, and age first sought treatment (Abbott,Volberg, Bellringer, et al. 2004; Grant and Kim 2002; Ladd and Petry 2002; Potenza et al. 2001;Tavares et al. 2001). Ethnicity and culture items should be defined or categorized comparably to census data in order to assess sample representativeness. Further detail may be added to reflect local contexts or specific study questions. Ethnicity and culture have been shown to be related to problem gambling in that disproportionate numbers of disadvantaged group members tend to experience gambling problems (Abbott, Volberg, Bellringer, et al. 2004; Volberg 2001; Volberg and Abbott 1997; Welte et al. 2001; Zitzow 1996). Recent immigration to a country may also play a role in gambling problems (Abbott,Volberg, and Rönnberg 2004; Petry et al. 2003). Furthermore, game preferences and help-seeking behavior can vary by group (Raylu and Oei 2002). Depending on which of these areas are of interest, one may ask a series of demographic items, including whether the respondent was born in the region of study; if not, how long ago he moved there; whether his parents were born there; and his primary language.

The Art of a Stylized Conversation

81

Employment status and income are also typically collected as part of demographic data; this information is especially important in studies of economic impacts of gambling. Again, these questions should be asked in such a way that the data are comparable to census standards. Respondent’s job title is important for constructing more nuanced social class variables in conjunction with education and household income variables (see, e.g., Nakao and Treas 1994), but this requires specialized coding skills. Job title will also identify respondents in gaming occupations, which has been shown to be an important risk factor (Shaffer,Vander Bilt, and Hall 1999). Income questions should seek precision and include nonwage sources such as alimony, welfare, unemployment compensation, disability pay, pensions, annuities, and so forth. It is essential to determine household composition—minimally, the number of adults and minor children who live in the household, and their relationships to one another. Ideally, one should collect both respondent’s own income and all household income; however, there will be appreciable missing data on the latter unless all income-generating household members are interviewed. It is advisable to place income questions late in the questionnaire, after the interviewer has had time to establish rapport with and a pattern of responsiveness from the respondent. Another important issue is access; the evidence that increased gambling availability leads to an increase in problem gambling is appreciable (Abbott 2001; Abbott and Volberg 2000; Gambling Review Body 2001; National Research Council 1999; Productivity Commission 1999; Shaffer, LaBrie, and LaPlante 2004; Volberg 2002). Increased gambling opportunities sometimes create more problem gamblers by increasing the risk of exposure; as more people gamble, the risks that individuals with vulnerabilities will gamble and develop problems increase. Capturing the influence of gambling facilities requires correlating residence and venue data with map coordinates. In most countries, locations and revenues of venues such as casinos, racetracks, cardrooms, and lottery outlets can be obtained or derived from public sources; distances, densities, and gross revenue flows can be associated with respondent locations using residential locators such as self-reported street addresses or postal codes. Finally, it is well known that depression and problem gambling co-occur often in individuals (Petry et al. 2005). A number of depression screens are in use with excellent psychometric properties; researchers can choose the one that best suits their needs. Problem Gambling Help and Treatment General population surveys generally capture very few individuals who have sought treatment for gambling problems—in fact, only a tiny percentage of the individuals who indicate serious gambling problems in their lives report having

82

Research and Measurement Issues in Gambling Studies

sought out treatment (Gerstein et al. 1999). Report of help-seeking behavior should enumerate the main sources of help, including seeing a doctor or other health or counseling professional, enrolling in a residential treatment program, visiting an online group or website for problem gamblers, attending a Gamblers Anonymous meeting, and seeking gambling-related help from another mutual support group. Barriers to treatment are an important element for inquiry (Hodgins and el-Guebaly 2000; Rockloff and Schofield 2004). Specific questionnaire items about treatment barriers might ask whether there was ever a time when the respondent thought she should see a doctor, counselor, or other health professional or seek any other help for her wagering but did not go, and then follow up affirmative responses with an item probing why. Response categories, which should be tailored to the local context and ideally be pretested in a treatment population, might include that health insurance did not cover treatment, respondent was afraid she would have to stop gambling, respondent did not have transportation, respondent was too embarrassed to discuss the problem, the hours were inconvenient, respondent could not find a program in her preferred language, and/or respondent stopped gambling on her own. Finally, for those who have received treatment, its perceived effectiveness should be of paramount importance. A single question whether (or the extent to which) the treatment helped the person stop or cut back on gambling behavior is a good solution if survey time is at a premium.

GLOSSARY Questionnaire a written script used to ask questions and record responses.The goal of a questionnaire is to obtain information that is valid. Response effects partially or wholly invalid data resulting from issues related to questionnaire construction or contexts; examples include the desire to present a good image, fatigue or annoyance with the questions, and cultural or linguistic inconsistencies in understanding the meaning of questions.

REFERENCES Abbott, M. W. (2001). What Do We Know About Gambling and Problem Gambling in New Zealand? Report Number Seven of the New Zealand Gaming Survey.Wellington: Department of Internal Affairs. Abbott, M. W., and Volberg, R. (2000). Taking the Pulse on Gambling and Problem Gambling in New Zealand: Phase I of the 1999 National Prevalence Survey, Report #3. Wellington: Department of Internal Affairs.

The Art of a Stylized Conversation

83

Abbott, M. W., Volberg, R., Bellringer, M., and Reith, G. (2004). A Review of Research on Aspects of Problem Gambling. Report to the Responsibility in Gambling Trust. Auckland, NZ: Aukward University of Technology. Abbott, M. W., Volberg, R., and Rönnberg, S. (2004). Comparing the New Zealand and Swedish national surveys of gambling and problem gambling. Journal of Gambling Studies, 20, 237–258. Amato, G., and Ochiltree, G. (1987). Interviewing children about their families: A note on data quality. Journal of Marriage and the Family, 49, 669–675. Aquilino,W. S. (1997). Privacy effects on self-reported drug use. In The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates (L. Harrison and A. Hughes, eds.) (NIH Publication No. 97-4147), pp. 383–415. Rockville, MD: National Institute on Drug Abuse. Backett, K., and Alexander, H. (1991). Talking to young children about health: Methods and findings. Health Education Journal, 50, 34–38. Behling, O., and Law, K. S. (2000). Translating Questionnaires and Other Research Instruments: Problems and Solutions. London: Sage Publications, Inc. Bradburn, N., and Miles, C. (1979).Vague quantifiers. Public Opinion Quarterly, 43, 92–101. Burton, S., and Blair, E. (1991).Task conditions, response formulation processes, and response accuracy for behavioral frequency questions in surveys. Public Opinion Quarterly, 55, 50–79. Converse, J. M., and Presser, S. (1986). Survey Questions: Handcrafting the Standardized Questionnaire. Beverly Hills, CA: Sage. Couper, M.,Traugott, M., and Lamias, M. (2001).Web survey design and administration. Public Opinion Quarterly, 65, 230–253. Coyne, I. T. (1998). Researching children: Some methodological and ethical considerations. Journal of Clinical Nursing, 7, 409–416. Currivan, D. B., Nyman, A. L., Turner, C. F., and Biener, L. (2004). Does telephone audio computerassisted self-interviewing improve the accuracy of prevalence estimates of youth smoking? Public Opinion Quarterly, 68, 542–564. DeVellis, R. F. (2003). Scale Development:Theory and Applications, 2nd ed.Thousand Oaks, CA: Sage. Dickerson, M. G., Baron, E., Hong, S. M., and Cottroll, D. (1996). Estimating the extent and degree of gambling related problems in the Australian population. Journal of Gambling Studies, 12, 161–178. Dillman, D. (1999). Mail and Internet Surveys:The Tailored Design Method. New York:Wiley. Epstein, J. F., Barker, P. R., and Kroutil, L. A. (2001). Mode effects in self-reported mental health data. Public Opinion Quarterly, 65, 529–549. Fontana,A. (2002). Interviewing children and adolescents. In Handbook of Interview Research: Context and Method ( J. F. Gubrium and J. A. Holstein, eds.).Thousand Oaks, CA: Sage Publications. Fowler, F. (1992). How unclear terms affect survey data. Public Opinion Quarterly, 56, 218–231. —— . (1995). Improving Survey Questions: Design and Evaluation. Applied Social Research Methods Series, vol. 38.Thousand Oaks, CA: Sage Publications. Fowler, F., and Mangione,T. (1990). Standardized Survey Interviewing: Minimizing Interviewer-Related Error. Newbury Park, CA: Sage. Gallup Organization. (1999). Gambling in America 1999: A Comparison of Adults and Teenagers. Available at: http://poll.gallup.com/content/default.aspx?ci=9889. (Accessed June 6, 2006.) Gambling Review Body. (2001). Gambling Review Report. Department for Culture, Media and Sport. Norwich, UK: Her Majesty’s Stationery Office. Geertz, C. (1973). The Interpretation of Cultures: Selected Essays. New York: Basic Books. Gerstein, D., Volberg, R., Toce, M., Harwood, H., Johnson, R., Buie, T., Christiansen, E., Chuchro, L., Cummings,W., Engelman, L., Hill, M., Hoffmann, J., Larison, C., Murphy, S., Palmer,A., Sinclair, S., and Tucker, A. (1999). Gambling Impact and Behavior Study: Report to the National Gambling Impact Study Commission. Chicago: National Opinion Research Center and partners. Gerver, D. (1969).The effects of source language presentation rate on the performance of simultaneous conference interpreters. In The Interpreting Studies Reader (F. Pöchhacker and M. Shlesinger, eds.). New York: Routledge.

84

Research and Measurement Issues in Gambling Studies

Grant, J. E., and Kim, S. W. (2002). Gender differences in pathological gamblers seeking medication treatment. Comprehensive Psychiatry, 43, 56–62. Groves, R. M., Fowler, F. J., Couper, M. P., Lepkowski, J. M., Singer, E., and Tourangeau, R. (2004). Survey Methodology. New York:Wiley. Harkness, J., van de Vijver, F., and Johnson,T. (2003). Questionnaire design in comparative research. In Cross-Cultural Survey Methods (J. Harkness, F. van de Vijver, and P. Mohler, eds). New York:Wiley. Heubusch, K. (1997, May).Taking chances on casinos. American Demographics. Hodgins, D., and el-Guebaly, N. (2000). Natural and treatment-assisted recovery from gambling problems: A comparison of resolved and active gamblers. Addiction, 95, 777–789. Hodgins, D., and Makarchuk, K. (2003). Trusting problem gamblers: Reliability and validity of selfreported gambling behavior. Brief report. Psychology of Addictive Behaviors, 17, 244–248. Johnson, E. E., Hamer, R., Nora, R. M.,Tan, B., Eisenstein, N., and Engelhart, C. (1997).The Lie/Bet Questionnaire for screening pathological gamblers. Psychological Reports, 80, 83–88. Kallick, M., Suits, D., Dielman, T., and Hybels, J. (1976). Survey of American Gambling Attitudes and Behaviors. USGPO Stock No. 052-003-00254.Washington, DC: U.S. Government Printing Office. Knäuper, B. (1998). Filter questions and question interpretation. Public Opinion Quarterly, 62, 70–78. —— . (1999). The impact of age and education on response order effects in attitude measurement. Public Opinion Quarterly, 63, 347–70. Krosnick, J.A., Narayan, S. S., and Smith,W. R. (1996). Satisficing in surveys: Initial evidence. In Advances in Survey Research (M.T. Braverman and J. K. Slater, eds.), pp. 29–44. San Francisco: Jossey-Bass,. Kwak, N., and Radler, B. (2002). A comparison between mail and Web surveys: Response pattern, respondent profile, and data quality. Journal of Official Statistics, 18, 257–273. Ladd, G.T., and Petry, N. M. (2002).A comparison of pathological gamblers with and without substance abuse treatment histories. Experimental and Clinical Psychopharmacology, 11, 202–209. Lyberg, L. E., and Kasprzyk, D. (1991). Data collection methods and measurement error. In Measurement Errors in Surveys (P. P. Biemer, R. M. Groves, L. E. Lyberg, N. A. Mathiowetz, and S. Sudman, eds.). New York:Wiley. Meade, C. D., and Smith, C. F. (1991). Readability formulas. Patient Education and Counseling, 17, 153–158. Moore, D. (2002). Measuring new types of question-order effects. Public Opinion Quarterly, 66, 80–92. Nakao, K., and Treas, J. (1994). Updating occupational prestige and socioeconomic scores. In Sociological Methodology (P. Marsden, ed.), pp. 1–72.Washington, DC: American Sociological Association. National Research Council. (1999). Pathological Gambling: A Critical Review.Washington, DC: National Academy Press. O’Muircheartaigh, C., Krosnick, J. A., and Helic, A. (2000). Middle Alternatives, Acquiescence, and the Quality of Questionnaire Data. Irving B. Harris Graduate School of Public Policy Studies. Available at http://harrisschool.uchicago.edu/about/publications/working-papers/ pdf/wp_01_3.pdf. (Accessed January 9, 2007.) Pan, Y., and de la Puente, M. (2005). Census Bureau Guideline for the Translation of Data Collection Instruments and Supporting Materials. Research Report Series #2005-06. Washington, DC: U.S. Bureau of the Census. Petry, N. M., Armentano, C., Kuoch,T., Mova, S., Norinth,T., and Smith, L. (2003). Gambling participation and problems among South East Asian refugees to the US. Psychiatric Services, 54, 1142–1148. Petry, N. M., Stinson, F. S., and Grant, B. F. (2005). Comorbidity of DSM-IV pathological gambling and other psychiatric disorders: Results from the NESARC. Journal of Clinical Psychiatry, 66, 564–574. Potenza, M. N., Steinberg, M. A., McLaughlin, S. D., Wu, R., Rounsaville, B. J., and O’Malley, S. S. (2001). Gender-related differences in the characteristics of problem gamblers using a gambling helpline. American Journal of Psychiatry, 158, 1500–1505. Presser, S., Rothgeb, J. M., Couper, M. P., Lessler, J.T., Martin, E., Martin, J., and Singer, E. (eds.). (2004). Methods for Testing and Evaluating Survey Questionnaires. Hoboken, NJ:Wiley.

The Art of a Stylized Conversation

85

Productivity Commission. (1999). Australia’s Gambling Industries, Report No. 10. Canberra: AusInfo. Rasinski, K.A., and Wright, D. (2000). Practical aspects of disclosure analysis. Of Significance, 2, 35–41. Raylu, N., and Oei, T. P. S. (2002). Pathological gambling: A comprehensive review. Clinical Psychology Review, 22, 1009–1061. Rockloff, M. J., and Schofield, G. (2004). Factor analysis of barriers to treatment for problem gambling. Journal of Gambling Studies, 20, 121–126. Schaeffer, N. (1991). Hardly or constantly? Group comparisons using vague quantifiers. Public Opinion Quarterly, 55, 395–423. Schaeffer, N., and Presser, S. (2003).The science of asking questions. Annual Review of Sociology, 29, 65–88. Schober, M. F., Conrad, F. G., and Fricker, S. S. (2004). Misunderstanding standardized language. Applied Cognitive Psychology, 18, 169–188. Schuman, H., and Presser, S. (1981). Questions and Answers in Attitude Surveys. New York:Academic Press. Schwarz, N. (1996). Cognition and Communication. Hillsdale, NJ: Erlbaum. Schwarz, N., and Hippler, H.-J. (1991). Response alternatives: The impact of their choice and presentation order. In Measurement Errors in Surveys (P. Biemer, R. Groves, L. Lyberg, N. Mathiowetz, and S. Sudman, eds.), pp. 41–56. New York:Wiley. Scripps Howard News Service. (2006). Gambling Poll Results. Available at: http://www.shns.com/shns/ g_index2.cfm?action=detail&pk=GAMBLESIDE-03-15-06. (Accessed June 6, 2006.) Shaffer, H. J., Hall, M. N., and Vander Bilt, J. (1997). Estimating the Prevalence of Disordered Gambling Behavior in the United States and Canada:A Meta-Analysis. Cambridge, MA: Harvard Medical School. Shaffer, H. J., LaBrie, R. A., and LaPlante, D. (2004). Laying the foundation for quantifying regional exposure to social phenomena: Considering the case of legalized gambling as a public health toxin. Psychology of Addictive Behaviors, 18, 40–48. Shaffer, H. J.,Vander Bilt, J., and Hall, M. N. (1999). Gambling, drinking, smoking and other health risk activities among casino employees. American Journal of Industrial Medicine, 36, 365–378. Sheatsley, P. B. (1983). Questionnaire construction and item writing. In Handbook of Survey Research (P. H. Rossi, J. D.Wright, and A. B. Anderson, eds.), pp. 195–230. New York: Academic Press. Smith T. (1987). That which we call “welfare” by any other name would smell sweeter: An analysis of the impact of question wording on response patterns. Public Opinion Quarterly, 51, 75–83. —— . (1994). A Comparison of Two Confidence Scales. GSS Methodological Report 80. Chicago: NORC. —— . (2003). Developing comparable questions in cross-national surveys. In Cross-Cultural Survey Methods ( J. Harkness, F. van de Vijver, and P. Mohler, eds.). New York:Wiley. Sudman, S., Bradburn, N., and Schwarz, N. (1996). Thinking About Answers:The Application of Cognitive Processes to Survey Methodology. San Francisco: Jossey-Bass. Sudman, S., Bradburn, N., and Wansink, B. (2004). Asking Questions:The Definitive Guide to Questionnaire Design for Market Research, Political Polls, and Health Questionnaires. New York:Wiley. Tavares, H., Zilberman, M. L., Beites, F. J., and Gentil,V. (2001). Gender differences in gambling. Journal of Gambling Studies, 17, 151–159. Tekfi, C. (1987). Readability formulas: An overview. Journal of Documentation, 43, 261–273. Toce-Gerstein, M., and Gerstein, D. (2004). Of time and the chase: Lifetime versus past-year measures of pathological gambling. eGambling: The Electronic Journal of Gambling Issues, 10. Available at: http://www.camh.net/egambling/issue10/ejgi_10_gerstein.html. (Accessed January 13, 2007.) Toce-Gerstein, M., Gerstein, D., and Volberg, R. (2003). A hierarchy of gambling disorders in the community. Addiction, 98, 1661–1672. Toce-Gerstein, M., and Volberg, R. (2003). The NODS CLiP:A New Brief Screen for Pathological Gambling. Paper presented at the 17th National Conference on Problem Gambling, Louisville, KY, June 19. Tourangeau, R., Rips, L. J., and Rasinski, K. (2000). The Psychology of Survey Response. Cambridge: Cambridge University Press. Tourangeau, R., and Smith,T.W. (1996).Asking sensitive questions:The impact of data collection mode, question format, and question context. Public Opinion Quarterly, 60, 275–304.

86

Research and Measurement Issues in Gambling Studies

van de Looij-Jansen, P., Goldschmeding, J., and Jan de Wilde, E. (2006). Comparison of anonymous versus confidential survey procedures. Journal of Youth and Adolescence, 35, 652–658. van der Zouwen, J., Dijkstra, W., and Smit, J. (1991). Studying respondent–interviewer interaction: The relationship between interviewing style, interviewer behavior, and response behavior. In Measurement Errors in Surveys (P. Biemer, R. Groves, L. Lyberg, N. Mathiowetz, and S. Sudman, eds.). New York: Wiley. Volberg, R. (2001). When the Chips Are Down: Problem Gambling in America. New York: Century Foundation. —— . (2002). Gambling and Problem Gambling in Nevada. Carson City: Report to the Nevada Department of Human Resources. Volberg, R., and Abbott, M. W. (1997). Gambling and problem gambling among indigenous peoples. Substance Use and Misuse, 32, 1525–1538. Volberg, R., and Banks, S. (1990). A review of two measures of pathological gambling in the United States. Journal of Gambling Behavior, 6, 153–163. Volberg, R., Toce, M., and Gerstein, D. (1999). From back room to living room: Changing attitudes toward gambling. Public Perspective, 10, 8–13. Wallsten, T., Budescu, D., Rapoport, A., Zwick, R., and Forsythe, B. (1986). Measuring the vague meaning of probability terms. Journal of Experimental Psychology: General, 115, 348–365. Wedell, D. H., Parducci,A., and Lane, M. (1990). Reducing the dependence of clinical judgment on the immediate context: Effects of number of categories and type of anchors. Journal of Personality and Social Psychology, 58, 319–329. Welte, J., Barnes, G.,Wieczorek,W.,Tidwell, M.-C., and Parker, J. (2001). Alcohol and gambling among US adults. Journal of Studies on Alcohol, 62, 706–712. Willis, G. B. (2004). Cognitive Interviewing:A Tool for Improving Questionnaire Design.Thousand Oaks, CA: Sage. Zitzow, D. (1996). Comparative study of problematic gambling behaviors between American Indian and non-Indian adults in a northern plains reservation. American Indian and Alaska Native Mental Health Research, 7, 27–41.

CHAPTER 4

Experimental Methodologies in Gambling Studies Sherry H. Stewart

Steven Jefferson

Departments of Psychiatry and Psychology, Queen Elizabeth II Health Sciences Centre, Dalhousie University Halifax, Nova Scotia, Canada

Department of Psychology Queen Elizabeth II Health Sciences Centre Halifax, Nova Scotia, Canada

Basic Components of an Experimental Research Study Internal and External Validity Types of Experimental Designs Group Experimental Designs Control Groups Randomization Single-Case Experimental Designs Withdrawal Designs Multiple Baseline Design Sample Experimental Methodologies Behavioral Observation Explicit Cognition Implicit Cognition Think Aloud Reaction Time Tasks Conclusions

Behavioral scientists can study gambling behavior and gambling disorders in the same way that scientists study other phenomena, such as the development of viral infections or novel methods of generating power—namely through the scientific or experimental method.Through experimental methods, behavioral scientists can help answer such questions as “Why do some people develop gambling problems?” and “What is the best way to treat someone suffering from pathological gambling?” Many ingenious methods have been developed for studying the behaviors that constitute problem gambling, why people develop gambling problems, and 87

88

Research and Measurement Issues in Gambling Studies

how to treat gambling disorders. In this chapter, we focus on one such set of methods for studying gambling behaviors and disorders: scientific or experimental research methods. The experimental method has the major advantage of being the only type of research design that can be used to draw conclusions about causation.

BASIC COMPONENTS OF AN EXPERIMENTAL RESEARCH STUDY We begin by examining three basic components of an experimental research study: the research hypothesis, the independent variable, and the dependent variable.When designing a research study, scientists begin with a research hypothesis, or an educated guess about the outcome of the study that will be tested empirically (i.e., through the collection of data). For example, you might hypothesize that availability of gambling establishments causes increases in gambling disorders. Or you might hypothesize that the best way to treat a problem gambler is by challenging his or her beliefs about the benefits of gambling. Once a researcher has decided what to study, the next step is to state the hypothesis clearly and in a form that is testable. For example, consider a study of the effects of gambling on drinking behavior (Stewart et al. 2002).The researchers observed a sample of 30 regular gamblers in a simulated bar in a laboratory environment which contained two video lottery terminals (VLTs). The researchers posed the hypothesis that engagement in VLT play, relative to engagement in a control activity (i.e., movie watching) in the same environment, would lead to an increase in purchases of alcoholic beverages from the “bar.” Phrasing the expected outcome in this manner made it testable—a characteristic that is important for the advancement of science. For example, it was possible that involvement in VLT play would have no impact whatsoever on purchase of alcoholic beverages; this alternative statement is called the null hypothesis in that it predicts no relationship between gambling and drinking behaviors. The researchers stated the hypothesis in a manner such that the results would lead to one of two possible conclusions. Either (1) alcohol consumption behavior is adversely affected by VLT gambling, so let’s study this more and examine the policy implications (e.g., the advisability of the common practice of housing VLTs in bars), or (2) alcohol consumption behavior is not affected by VLT gambling, so let’s look elsewhere for risk behaviors that might be affected by gambling. The next two important components of a research study are the independent and dependent variables. When a behavioral scientist develops a research hypothesis, he or she also specifies the independent and dependent variables. The dependent variable is the outcome variable or factor that the researcher expects to be influenced or to change in the study. In the area of gambling and gambling disorders, the dependent variable might be overt gambling behaviors, gamblers’ thoughts or feelings, physiological variables (e.g., heart rate while gambling;

Experimental Methodologies in Gambling Studies

89

see Stewart et al. 2005), or symptoms of a gambling disorder. For example, in a recent study by Doiron and Nicki (in press) which investigated the effectiveness of a new problem gambling prevention program, the main dependent measures included levels of symptoms of pathological gambling, self-reported gambling behaviors, and gambling-related cognitions. Independent variables are those factors thought to affect the dependent variables. The independent variable is the factor that is manipulated by the experimenter. In the Doiron and Nicki study, the independent variable was the new intervention (participants either did or did not receive the new intervention). Statistical tests are then conducted to determine if the independent variable (e.g., exposure to the intervention) does indeed have an effect on the dependent variable (e.g., levels of gambling problems). If statistical tests do support an effect of the independent variable on the dependent variable, then the researcher must reject the null hypothesis of no relationship between these variables.

INTERNAL AND EXTERNAL VALIDITY A challenge faced by researchers conducting experimental research in the area of gambling behavior and gambling disorders is balancing ecological validity against the need for strict experimental control. In designing any research study, the researcher must balance concerns for internal validity against concerns that the findings are externally valid as well. Internal validity refers to the degree of experimental control within the study design and consequently the degree to which one can be confident that the manipulation of the independent variable is responsible for causing the outcome on the dependent variable(s). External validity, on the other hand, is the degree to which the findings can be generalized beyond the particular experiment—to individuals who were not engaged in the research study in question, and to settings beyond the laboratory (i.e., to the real-world gambling context).The more stringent the controls exerted within the research design, the greater the internal validity of the study. But internal validity is often achieved at the expense of internal validity, and vice versa. To illustrate the importance of external validity in gambling study design, research by Leary and Dickerson (1985) showed that gambling studies conducted in sterile laboratory environments that used mock gambling machines and participants who played for nonmonetary incentives did not generalize to gambling behavior in the real world. The findings of Leary and Dickerson suggest that it is important to set up the gambling situation within experimental gambling studies to be as close to the real-world gambling situation (i.e., as naturalistic) as possible. For example, one might choose to use real-world games such as commercially available gambling machines in the research, as opposed to simulated gambling tasks. And one might design the study to have participants playing for real money as opposed to using nonmonetary incentives. As another example, environmental cues of the testing

90

Research and Measurement Issues in Gambling Studies

environment can be matched as closely as possible to the real-world gambling context (e.g., testing in a bar-lab type of situation for VLT gambling research or mocking up a casino environment for a study of casino betting). These types of design features enhance the ecological validity of the study and increase its external validity, or the chances that the results of the research will generalize to the real-world gambling situation (see Stewart, Blackburn, and Klein 2000 for a review). A study by Ladouceur et al. (1991) confirms the importance of attending to issues of ecological validity of the testing environment. These authors found few significant differences on gambling cognitions and behaviors between nonpathological gamblers (video poker players) examined in a laboratory versus natural settings. However, the authors carefully simulated many aspects of the real-world gambling situation within their laboratory environment. For example, in both environments, participants played with real money, the amount of which was determined by each participant’s regular weekly bet; participants were allowed to keep all winnings from the study gambling situation; and participants played on commercially available video poker machines. On the other hand, one must attend to internal validity in experimental gambling research design. In experimental research, it is important to exert control over extraneous, potentially confounding variables.Without a reasonable degree of control of irrelevant variables, it is impossible to draw confident conclusions about the important factors underlying an interesting behavioral finding. Let us look at a sample experimental gambling study to illustrate the issue of balancing concerns about external validity against the need for internal validity within the study design. In a study by Ellery, Stewart, and Loba (2005), on the effects of alcohol intake on risk taking during VLT play, regular VLT players were randomly assigned to one of two beverage conditions. In the experimental condition, the beverage was a fixed, mildly intoxicating dose of vodka mixed with orange juice. A second condition was included where participants consumed only the orange juice, to control for the effects of drinking per se. Drinking only orange juice while playing VLTs may not be a typical experience for many VLT players, so the decision to include this control group may have compromised ecological validity of the study and reduced external validity. However, this control was necessary to determine whether risk taking observed among VLT players who drank vodka and orange juice was specifically due to alcohol intake. Another control exerted in this study was alcohol dose. All of those in the experimental group were administered a fixed dose of vodka that was based on their body weight and was designed to target a specific blood alcohol concentration (i.e., 0.06%, which is equivalent to mild intoxication).This particular dose was chosen because it is the usual dose of alcohol reportedly consumed by regular VLT players when they are gambling at the machines (Focal Research 1998). In other words, this dose was chosen to maximize ecological validity, and it was standardized across participants to enhance internal validity (i.e., so the researchers could make confident statements about the effects of this particular mildly intoxicating dose of alcohol on risk taking during

Experimental Methodologies in Gambling Studies

91

VLT gambling). However, the decision to standardize dose across participants may have enhanced internal validity at the expense of external validity; presumably, this particular chosen dose is not the dose of alcohol that would normally be consumed by some of the regular-gambler participants while playing VLTs in the real world (i.e., some drink more heavily and some drink less than the chosen dose). Another threat to external validity pertains to alcohol restrictions in simulated experiments. For example, Diskin and Hodgins (2001) conducted a study designed to assess the presence of dissociative-like states in VLT players. Participants, who were community-based VLT players, were required to respond to a flash of light emitted in their periphery while they were using the machines.The time required to respond to the light was one of the dependent variables in this experiment. One participant was not permitted to take part in the study because he was under the influence of alcohol. From the perspective of internal validity, this seems reasonable, given that reaction time is considerably influenced by alcohol (e.g., Holloway 1995). In terms of external validity, however, the decision to exclude this participant might seem inappropriate when one considers that the majority of VLT players drink while playing the VLT machines (Focal Research 1998). Upon first consideration, the task of enhancing the ecological validity of the experimental testing environment may not seem difficult; lighting, music, and other nuances of gambling settings can generally be emulated rather easily in laboratories. However, particularly for regular gamblers who tend to gamble in one location (e.g., a local bar), there are likely to be myriad stimuli that have come to be associated with gambling (e.g., through higher-order conditioning processes), and such stimuli may be difficult or impossible to reproduce in an experimental laboratory setting. One such stimulus, for example, might be particular staff members of an individual’s regular gambling setting. Thus, it is important to recognize that even though one can go to great lengths to enhance the ecological validity of the experimental laboratory testing environment in gambling research, there will always be limits that can impact the external validity of the research findings.

TYPES OF EXPERIMENTAL DESIGNS GROUP EXPERIMENTAL DESIGNS The most common type of experimental research design used in gambling studies is the group experimental design, which involves comparisons of groups of individuals. To enhance the internal validity of the study (i.e., the degree to which we can be certain that changes in the dependent variable are actually due to changes in the independent variable), participants in an experimental group (who receive one level of the independent variable) are compared with another group of individuals, referred to as the control group, or comparison group, who receive the other level of the independent variable. In the Ellery et al. (2005) study on the effects of alcohol

92

Research and Measurement Issues in Gambling Studies

on VLT play, those in the experimental group received a moderate dose of alcohol, while those in the comparison group received a control (i.e., nonalcoholic) beverage. The inclusion of a control group for comparison helps the researcher rule out alternative explanations for the change in the dependent variable. Control Groups There are several types of control groups that are available to gambling researchers depending on the purposes of the study. For example, in the area of gambling interventions research, there are at least four types of control groups that can be employed. First, there is the no-treatment control, where participants randomized to the experimental intervention are compared against a control group of individuals who receive no intervention. A no-treatment group controls for any influences of the passage of time, since individuals in both groups complete the dependent variables at pretreatment and again at posttreatment (and/or follow-up) after the intervention is completed.The major drawback of this type of design is an ethical one: Should treatment be withheld from an individual who needs it for a gambling problem? The remaining three types of control groups do not pose this type of ethical concern. Another control group for the passage of time is the wait-list control group (or delayed treatment control group). In this type of design, control group participants have treatment withheld but only temporarily, until the two sets of dependent measures have been completed. In the Doiron and Nicki (in press) intervention study, those in the experimental group were exposed to the novel preventive intervention, while those in the control/comparison group were not. Rather, the control participants were assigned to what is referred to as a wait list, where they completed the same dependent measures as those in the experimental group and then were exposed to the intervention at a later time. By including a wait-list control group in their study design, Doiron and Nicki (in press) were able to conclude that changes in levels of gambling problems, gambling behaviors, and gambling cognitions were due to their new intervention itself rather than just the passage of time per se. Otherwise, the researchers would have seen similar changes on the various dependent variables among those assigned to the wait-list control group. This type of design takes advantage of the unfortunate fact that many addiction treatment centers have long wait lists, where potential patients/clients have to wait for services until a counselor becomes available. A third type of comparison group controls not only for the passage of time, but for expectations that clients may hold about the treatment. For example, although it was clear in the Doiron and Nicki (in press) study that some aspect of the intervention other than simply the passage of time was responsible for the observed changes in gambling behaviors, cognitions, and problems, it was not clear whether the observed improvements were caused by the contents of the intervention itself, by patient expectations for changes, or by a combination of the two. A placebo control treatment is designed to control for patient expectancies about treatment effects. A placebo (from Latin, “I shall please”) such as a harmless sugar cube is typically given to members of control groups in drug studies to make them believe that they are

Experimental Methodologies in Gambling Studies

93

getting the real medication being given the experimental group (e.g., MacDonald et al. 2001). Although this control for expectancy is relatively easy to achieve in medication studies (e.g., Kim et al. 2001), it is not always so easy to “pull off ” in studies examining psychological treatments for gambling problems because it is not always easy to devise a placebo “treatment” that problem gamblers believe would help them but that does not include the component the researcher thinks will be effective (see Barlow, Durand, and Stewart 2006). Most often, placebo psychological treatments involve therapist attention, time, and concern but do not involve any active provision of the components of therapy the researchers believe are important for change in gambling behaviors (e.g., cognitive or behavioral techniques). Finally, a fourth type of control group involves comparing a new treatment (experimental group) with an existing, established treatment (control group) in a type of study called comparative treatment research, also commonly referred to as a “race horse” study, because it will determine which treatment comes in first. For example, the results on various dependent measures in a cognitive-behavioral treatment for problem gambling might be compared against those in treatments with a medication like naltrexone. This type of research study answers the clinically important question of whether a new treatment performs as well or better than other, established treatments. Randomization Another method used to increase internal validity in group experimental designs is random assignment to the experimental or the control group (or randomization). Random assignment, where membership in the experimental or control group is determined by chance (e.g., by flipping a coin or by using a random number table), means that each participant has an equal chance of being placed in any group within the research design. This method improves internal validity by eliminating any systematic bias in assignment (Asmundson, Norton, and Stein 2002). For example, if a researcher examining a new treatment for pathological gamblers at a gambling clinic did not assign participants to the experimental and control groups randomly, but instead assigned the first 50 who presented to the clinic to the control group and the second 50 to her new treatment, she might introduce bias if referral patterns changed over time (e.g., if the gamblers referred to treatment had more severe problems over time). This type of confound might seriously affect the internal validity of her study; thus, the process of randomization is extremely important in group experimental designs.

SINGLE-CASE EXPERIMENTAL DESIGNS When large groups of patients are not available, or when a researcher has first developed a new treatment strategy, the single-case experimental design can be considered as an alternative to the traditional group design. B. F. Skinner formalized

94

Research and Measurement Issues in Gambling Studies

the scientific research method of single-case experimental designs (see Barlow et al. 2006 for more information).This methodology involves the systematic study of individuals under various experimental conditions. Single-case experimental designs differ from case studies (i.e., the detailed description of individual cases) in that the researchers use a number of different strategies to improve internal validity, thus reducing the influence of potential confounds. Single-case experimental designs have both advantages and disadvantages relative to traditional group experimental designs, described in the previous section. Single-case designs allow us to learn a lot about the behavior of one individual, whereas traditional group designs involve making only a few observations of a large group, allowing us to make conclusions about the “average” response.The single-case experimental methodology has greatly helped us understand the factors involved in individual psychopathology (Barlow et al. 2006). In terms of applications to the study of gambling, single-case experimental designs can help us explain why individual people engage in gambling behavior, as well as how to treat those with problem or pathological gambling. One of the most important features of single-case experimental designs is repeated measurement. In contrast to the traditional group design, where the behavior of interest is measured only once before and once after the independent variable is applied, in single-case experimental designs the behavior is measured several times. More specifically, the researcher takes the same measurements over and over to learn how variable the behavior is (How does it change day to day?) and whether it shows any obvious trends (Is it getting better or worse?). Suppose a young man, Devin, comes into the therapist’s office complaining about a preoccupation with gambling. When we ask him to rate the level of his gambling preoccupation on a scale from 1 to 10, he gives it a 9, indicating a severe level of preoccupation. After several weeks of treatment, Devin rates his gambling preoccupation at 6, indicating a moderate level of preoccupation. Can we say that the treatment reduced his preoccupation with gambling? Not necessarily. Suppose we had measured Devin’s preoccupation with gambling each day during the weeks before his visit to the office (repeated measurement) and observed that it differed greatly. On particularly good days, he rated his preoccupation from 5 to 7. On bad days, it was up between 8 and 10. Suppose further that even after treatment, his daily ratings continued to range from 5 to 10.The rating of 9 before treatment and 6 after treatment may only have been part of the daily variations he experienced normally. In fact, Devin could just as easily have had a good day and reported a 6 before treatment, and then had a bad day and reported a 9 after treatment, which would imply that the treatment made him worse! Repeated measurement is part of each single-case experimental design. It helps identify how a person is doing before and after intervention and whether the treatment accounted for any changes. Consider yet another possibility with respect to Devin’s preoccupation with gambling. Maybe Devin’s preoccupation

Experimental Methodologies in Gambling Studies

95

with gambling was on its way down before the treatment.This alternative pattern would also have been obscured with just a single before-and-after treatment measurement. Maybe he was getting better on his own, and the treatment didn’t have much, if any, effect on his gambling preoccupation. Although the first scenario shows how the variability from day to day could be important in an interpretation of the effect of treatment in Devin’s case, the second scenario shows how the trend itself can also be important in determining the cause of any change in his gambling preoccupation. There are thus three important parts of repeated measurements in singlecase experimental designs (see Barlow et al. 2006): (1) the level or degree of behavior change with different interventions, (2) the variability or degree of change over time, and (3) the trend or direction of change. Again, before-and-after scores alone do not necessarily show what is responsible for behavioral changes. For example, the researcher would feel much more confident that the treatment was responsible for Devin’s change in gambling preoccupation if (1) there were consistently high scores over repeated assessments at baseline (prior to treatment) and consistently lower scores over repeated assessments after the treatment was applied (the issue of variability); (2) the trend in the data suggested no change at baseline, versus his gambling preoccupation showing a downward trend only after the treatment was applied (the issue of trend); and (3) the degree of change in gambling preoccupation from pre- to post-treatment were fairly substantial when collapsed across the repeated measurements (the issue of level of change in the variable of interest). Now we will examine two of the most common types of single-case experimental designs and illustrate how they might be applied to the study of gambling behavior or gambling problems.The two types we will examine here are the withdrawal design and the multiple baseline design (see Barlow et al. 2006 for further detail). Withdrawal Designs One of the more common strategies used in single-case experimental research is a withdrawal design, also known as an ABAB design. In this design, the researcher tries to determine whether the independent variable is responsible for changes in behavior.The effect of Devin’s treatment could be tested by stopping it for a period of time to see whether his gambling preoccupation increased.A simple withdrawal design has four parts. First, a person’s condition is evaluated before treatment, to establish a baseline, which is the first “A” in the ABAB design.Then comes the first change in the independent variable—in Devin’s case, the beginning of treatment, or the first “B” in the ABAB design.Third, treatment is withdrawn (the return-tobaseline phase, or the second “A” in the ABAB design) and the researcher assesses whether Devin’s level of preoccupation with gambling changes again (i.e., increases)

96

Research and Measurement Issues in Gambling Studies

as a function of this step. Finally, the independent variable is manipulated once again; in Devin’s case, the treatment is applied once more: the second “B” in the ABAB design. If with the first application of the treatment, Devin’s gambling preoccupation lessens in comparison to baseline, then worsens again after treatment is withdrawn, and then improves once more when the treatment is reapplied, the researchers can confidently conclude that the treatment has reduced Devin’s gambling preoccupation. An important difference between this design and that of a case study is that the change in treatment is designed specifically to show whether treatment caused the changes in behavior. Although case studies often involve treatment, they don’t include any effort to learn whether the person would have improved without the treatment.A withdrawal, or ABAB, design gives researchers a better sense of whether or not the treatment itself caused behavior change. In spite of their advantages, withdrawal designs are not always appropriate. Two main arguments have been presented against withdrawal designs: one ethical and the other practical. First, the researcher is required to remove what might be an effective treatment, a decision that is sometimes difficult to justify for ethical reasons. In Devin’s case, a researcher would have to decide that there was a sufficient reason to deliberately produce his gambling preoccupation again. Suppose we knew that high levels of gambling preoccupation have often been accompanied by marked depression and suicidal thinking in Devin’s case. Would it be ethical to remove what might be an effective treatment for his gambling preoccupation at the risk of inducing suicidal thoughts, just to learn whether it was really the treatment that was responsible for Devin’s decreased gambling preoccupation? Barlow et al. (2006) have presented several counterarguments to this particular concern about the use of withdrawal designs. First, they note that treatment is routinely withdrawn when medications are involved so clinicians can determine whether the medication is responsible for the treatment effects and thus avoid any unnecessary medications. Second, they note that the withdrawal phase does not have to be prolonged; a very brief withdrawal may still clarify the role of the treatment. The second problem identified with withdrawal designs is a practical one. A withdrawal design is unsuitable when the treatment can’t be removed by the researcher. Suppose Devin’s treatment involved having him visualize himself by a pool at a southern resort. It would be very difficult—if not impossible—to stop Devin from imagining this situation, making it very difficult to institute the treatment withdrawal phase. Similarly, some treatments involve teaching clients skills which might be impossible to unlearn. If, through treatment, Devin learned skills that helped him be less preoccupied with gambling, it would be difficult to reverse this even if the treatment itself were withdrawn.Another single-case experimental design, the multiple baseline design, which we will examine next, addresses this particular limitation.

Experimental Methodologies in Gambling Studies

97

Multiple Baseline Design Another commonly employed single-case experimental design strategy that doesn’t have some of the drawbacks of a withdrawal design is the multiple baseline design. Rather than stopping the intervention to see whether it is effective, the researcher starts treatment at different times across settings (e.g., bar vs casino), behaviors (e.g.,VLT gambling vs. racetrack betting), and/or participants (e.g., different people with a gambling disorder). For example, after waiting a period of time and taking repeated measures of Devin’s gambling preoccupation both at a bar and at the casino (the baseline), the researcher/clinician could treat him first in the bar setting. When the treatment begins to be effective in the bar setting, intervention could then begin in the casino setting. If he improves in the bar setting only after beginning treatment and improves in the casino setting only after treatment is also used there, we can confidently conclude that the treatment was effective. This is an example of using multiple baselines across settings. Internal validity improves with a multiple baseline design because other explanations for the results can be ruled out. Devin’s gambling preoccupation improved only in the settings where it was treated, which rules out competing explanations. For example, if Devin got married to a very supportive spouse at the same time that treatment started and his gambling preoccupation decreased in all gambling-related situations, we couldn’t conclude that his condition was affected by treatment. Suppose a researcher wanted to assess the effectiveness of a treatment for a problem gambler who appeared to be addicted to multiple forms of gambling behavior.Treatment could first focus on his VLT gambling, then on a second problem, such as racetrack betting, and then a third problem, such as casino blackjack. If the treatment was first effective only in reducing VLT gambling behavior, then was effective for his racetrack betting only after the second intervention, and then was effective for his blackjack gambling only after the third intervention, the researcher could conclude that the treatment, not something else, accounted for the improvements.This is an example of a multiple baseline design conducted across behaviors. Single-case experimental designs are sometimes criticized because they tend to involve only a small number of cases, leaving their external validity in doubt. In other words, we can’t say that the results we saw with a few people would be the same for everyone. However, although they are called single-case designs, researchers can and often do use them with several people at once, in part to address the issue of external validity. Symes and Nicki (1997) recently studied the effectiveness of a behavioral treatment called exposure and response prevention (ERP) in the treatment of problem gambling. Briefly, the theory behind ERP, drawn from the treatment of obsessive-compulsive disorder, is that exposing a problem gambler to the stimuli associated with gambling (e.g., the VLTs and the associated setting) while preventing the response of gambling will allow the urge to gamble to extinguish. Using multiple baselines across participants, the researchers

98

Research and Measurement Issues in Gambling Studies

introduced this treatment to a set of two pathological gamblers—a 23-year-old female and a 20-year-old male. Their dependent variables were self-reports of gambling behavior and urges to gamble. Only when treatment began did each pathological gambler’s gambling urges and behaviors improve. This design let the researchers rule out coincidence or some other change in the problem gamblers’ lives as an explanation for the improvements. Among the advantages of the multiple baseline design in evaluating treatments is that it does not require withdrawal of treatment, since, as we’ve seen, withdrawing treatment is sometimes difficult or impossible. Furthermore, the multiple baseline design typically resembles the way treatment would naturally be implemented. A clinician can’t help a client with numerous problems simultaneously (e.g., one who is addicted to multiple forms of gambling or one who is addicted to both gambling and drugs/alcohol) but can take repeated measures of the relevant behaviors and observe when they change.A clinician who sees predictable and orderly changes related to where and when the treatment is used can conclude that it is the treatment that is causing the change.

SAMPLE EXPERIMENTAL METHODOLOGIES In the next section of the chapter, we turn our attention to several types of research methodologies that researchers can make use of in experimental gambling studies. Specifically, we review techniques involving behavioral observation and those involving the measurement of cognition (both explicit and implicit cognitive measures, and both “offline” and “online” cognitive measures).

BEHAVIORAL OBSERVATION Generally speaking, scientific knowledge is accrued through observation. In many cases, it is possible to observe directly the variable of interest. For example, an animal researcher may believe that overcrowding leads to aggression in rats. In order to test this hypothesis, he or she may place rats in densely populated quarters and then observe their behavior. Instances of aggression (the variable of interest) can be directly observed, provided that they are appropriately operationally defined. Gambling behaviors that can be observed directly include engaging in superstitious rituals and “chasing” losses. In other cases, the variable in question cannot be observed directly. Its presence and/or strength, however, can be inferred by observing its purported effects. For example, astronomers have discovered dozens of planets outside of our solar system by observing the manner in which their gravitational pull causes nearby stars to “wobble” as they rotate on their axes. Detection of these planets via direct

Experimental Methodologies in Gambling Studies

99

observation is not currently possible because they are too small and too far away. In the field of gambling studies, irrational beliefs concerning the outcome of gambling events cannot be observed directly.Their existence, however, may be inferred by observing gamblers’ behavior during a gambling session. For instance, Henslin (1967) observed that experienced dice players modified the force with which they rolled the dice in order to obtain desired numbers. Specifically, they rolled the dice hard for high numbers and softly for low ones. From this behavior, one might infer that dice players believe that there is an association between the force of the throw and the outcome of the roll. In gambling studies, researchers can utilize two different types of observation: Naturalistic observation and simulated observation. In naturalistic observation, participants are observed in their usual gambling setting, such as in a bar or casino. As an example of this method, King (1990) observed superstitious behaviors exhibited by bingo players while they played in a community bingo parlor. Simulated observation consists of observing gamblers in an “artificial” setting, typically in the researcher’s laboratory. Quite often, efforts are made by the researcher to ensure that the setting does not differ markedly from “natural” gambling venues, as alluded to at the beginning of the chapter.Thus, a simulated gambling setting may include soft lighting or music playing in the background. One of the main problems with naturalistic observation is reactivity, which refers to changes in behavior that occur as a result of being monitored. In gambling research, participants may feel the need to “sanitize” some of their gambling-related behaviors in an effort to avoid embarrassment and/or maintain social desirability. For example, individuals who frequently swear at VLTs may be reluctant to do so when they know that they are being watched. Reactivity may attenuate over time, as participants become increasingly accustomed to the observer’s presence (Spiegler and Guevremont 1998). An alternative is for the experimenter to be as unobtrusive as possible to reduce reactivity. A rather extreme example is a naturalistic observation study by Parke and Griffiths (2004) in which they monitored a sample of 303 slot machine players who were engaged in gambling. The site chosen for this study was a gambling arcade that housed 54 slot machines. The principal investigator obtained employment as an arcade supervisor at this establishment to minimize the experimenter’s obtrusiveness. In this role, he observed gamblers over four 6-hour periods. The “participants” in this study were unaware that their behavior was being monitored for research purposes.This lack of awareness, combined with the fact that the principal investigator actually appeared to be an arcade employee, maximized the ecological validity of the study and prevented reactivity. Nonetheless, in making decisions about how to best minimize experimenter obtrusiveness in naturalistic observation studies, one must balance these concerns against the participants’ ethical right to informed consent. Another problem with naturalistic observation is that it is sometimes difficult to define operationally a construct of interest. For example, consider dissociation

100

Research and Measurement Issues in Gambling Studies

during VLT gambling. It has been suggested that VLT play is particularly appealing to gamblers who seek escape from their difficulties (Dickerson 1993; Wynne 1994) and that VLT players in particular become highly engrossed while using the machines. However, while observing VLT gamblers in their usual setting, what behaviors would lead one to suspect that they were experiencing dissociative-like states? One might propose that obliviousness toward surrounding stimuli (e.g., music, others’ attempts to make conversation with the gambler) suggests the presence of dissociation.This may be valid, but how could one operationally define such behaviors? Observational studies in which target behaviors are vaguely defined (or not amenable to clear operational definitions) are likely to result in low levels of interobserver reliability (i.e., agreement between observers) and therefore poor replicability. Finally, as a research method, naturalistic observation is descriptive, not explanatory. Controlled conditions are not present, which makes it difficult or impossible to discover cause-and-effect relationships. For example, in the naturalistic observation study by Parke and Griffiths (2004), slot machine players were monitored for instances of aggression while gambling. A total of 165 instances of aggression were noted over the observation periods, including verbal aggression aimed at arcade staff, slot machines, and other gamblers. Physical aggression against the machines (e.g., kicking) was also noted.The authors speculated that introjected anger was the cause of aggression in this study (particularly the verbal aggression meted out against the arcade staff).Although this may be a tenable statement (especially among proponents of Freudian theory!), the observational nature of this study did not allow the researchers to draw conclusions about the underlying causes of the “aggressive behavior” with confidence. Simulated observation in gambling research is more consistent with the experimental method in that it carries the notable advantage of allowing the experimenter to control important variables. Examples of variables that have been controlled in simulated gambling settings include alcohol intake (Stewart et al. 2006), payout rate of gambling activities (Bechara et al. 1994), and structural characteristics of VLTs (Loba et al. 2001). Although experimental control is generally desirable, it may pose serious threats to external validity, as discussed earlier in this chapter. A difficulty with both naturalistic and simulated observation is that behavioral data gleaned from these methods may occur relatively infrequently—that is at a low base rate. For example, in a bar-based study of the effects of VLT machine manipulations (e.g., of the apparent speed at which the reels spin in a spinningreels game; see Loba et al. 2001), some of the behaviors of interest occurred at a very low rate (e.g.,“cashing out” occurred on average substantially less than once per hour). Thus, researchers need to plan appropriately long observation intervals to capture the phenomena of interest when conducting naturalistic observation studies.A more practical alternative can be the use of simulated observation where the situation is designed to elicit the phenomenon of interest. If one is interested in examining players’ responses to wins, and winning is a low-base-rate phenomenon

Experimental Methodologies in Gambling Studies

101

(which it often is in gambling!!), a researcher may decide to manipulate the win rate (see Ladouceur et al. 1988) to ensure that the player experiences a win during the period of observation. In fact, means exist of adjusting win frequency experimentally. For example, MacLin, Dixon, and Hayes (1999) have developed a computerized slot machine simulation that was designed to examine many of the potential variables involved in gambling behavior. This simulation program is designed to run on a computer and allows the experimenter to manipulate a number of variables, including probabilities of payoffs. It also allows the researcher to program specific sequences of losses and wins. Data are recorded on a trial-by-trial basis and can be easily imported into many statistical analysis packages.

EXPLICIT COGNITION Individuals with gambling problems often exhibit distorted beliefs related to the outcome of gambling events. One of these beliefs, for example, is the gambler’s fallacy, which is the notion that the longer one has gambled without winning, the greater become the chances of placing a winning bet. This line of reasoning is incorrect in any gambling activity in which there is independence among the outcomes. In order to assess the magnitude of gambling cognitive distortions, a number of methods have been employed. These include self-report inventories, think-aloud procedures, and behavioral observation. In behavioral observation, an individual’s conviction in a particular gambling belief is inferred from his or her behavior, as we discussed in the behavioral observation section above. Self-report inventories, on the other hand, require the examinee to reflect on a particular gambling belief and then indicate whether and/or the extent to which they hold it to be true (this is typically accomplished through true-false questionnaires or inventories involving Likert scale ratings). For instance, the Informational Biases Scale (IBS) ( Jefferson and Nicki 2003) requires examinees to rate their conviction regarding 25 statements related to VLT gambling. Sample items include the following: “After a long string of wins on a VLT, the chances of losing become greater” and “There are certain strategies (e.g., betting all of your credits at once) that one can use with VLTs to help him or her win.” An implicit assumption often made by researchers is that beliefs influence behavior. For example, an individual who wholeheartedly endorses the gambler’s fallacy would be expected to continue gambling in spite of mounting losses. However, it appears that behaviors and beliefs are not always congruent. For example, Steenbergh et al. (2004) examined the impact of warning and informational messages on irrational beliefs and gambling behavior. A total of 101 undergraduate students, all of whom had gambled at least once, were matched on self-efficacy and level of gambling distorted beliefs and then randomly assigned to receive a warning

102

Research and Measurement Issues in Gambling Studies

message, a warning message plus information about limit setting and gambling cognitive distortions, or a 10-minute film about the history of gambling. Participants then played an electronic version of roulette. It was found that the individuals in the two message conditions exhibited greater knowledge about the risks of gambling than did those who watched the video only. Also, participants who received the warning message plus the additional information described above exhibited significant reductions in their levels of gambling cognitive distortions. However, there were no differences in gambling behavior among participants across the three groups.Another example of the discrepancy between cognitions and behavior was obtained in a recent study by Ellery, Stewart, and Collins (2006).They examined the effects of alcohol versus a placebo beverage on the gambling cognitions and behaviors of high- and low-risk VLT gamblers. Consistent with the original hypotheses, the authors found that alcohol consumption increased risk-taking behaviors during VLT play, particularly among the high-risk gamblers. However, although alcohol increased irrational gambling cognitions (on the IBS), this occurred only in the low-risk gamblers! These two studies show that self-reported gambling cognitions as reported on questionnaires are not always congruent with actual gambling behaviors. Conversely, although individuals may behave as though they hold a certain belief, it does not necessarily follow that they actually endorse that belief. For example, in a study conducted by Langer (1975), participants played a card game known as “war” against either an opponent who presented as shy and awkward or one who appeared confident and debonair.War is a very simple card game—each player is dealt a card and the one with the higher card wins. Langer reported that participants in this experiment behaved as though they were more likely to beat the shy opponent than the confident one. Presumably, however, it was clear to the participants (who were university students) that luck was the only factor in this game that determined who would win and who would lose. Thus, it is highly doubtful that any of them would have responded affirmatively if asked whether personality characteristics conferred an advantage (or disadvantage) in the game of war. Given these examples of the sometimes divergent results between measures of explicit gambling-related cognitions and gambling behavior, it is advisable to include measurements of both in research studies examining explicit cognition. It might also be advisable to study another aspect of gambling-related cognition that is theoretically less susceptible to the many forms of bias involved in self-report cognitive measures—namely, measures of implicit cognition.

IMPLICIT COGNITION Implicit cognitive processes are mental activities that transpire outside of our conscious awareness and control. It is believed that such processes influence our perception, decision making, and behavior. Several branches of psychology

Experimental Methodologies in Gambling Studies

103

have utilized implicit cognition research designs in order to obtain a deeper understanding of phenomena of interest. For example, social psychologists frequently use the Implicit Association Test (IAT), which provides information concerning the automatic associations that people make between concepts (e.g., science and the arts) and attributes about themselves or others. Investigators who study various psychological disorders have also been known to explore implicit processes in their research. To date, little research has been conducted concerning implicit processes in gamblers. However, several researchers have acknowledged the potential importance of such factors in addictive behaviors. For example, Stacy, Leigh, and Weingardt (1994) purport that implicit associations in memory mediate addictive behavior (see also Hills and Dickerson 2002). Implicit cognition paradigms that have been utilized in gambling research include Stroop interference tasks, IATs, and Lexical Salience Tasks (LSTs) (see below and review by Zack and Poulos 2006). In the original Stroop (1935) task, various stimuli (e.g., geometric shapes, a series of Xs, commonly used words) were printed in colored ink. Participants were required to name, as quickly as possible, the color of ink in which each stimulus was printed. It was consistently found that participants took longer to name the color of ink for color words (e.g., to say “red” in response the word blue printed in red ink) than to name the ink color of shapes, a string of Xs, or non–color words. Stroop’s interpretation of this phenomenon was that the automatic (i.e., unconscious) processing of the color word’s meaning interferes with the task of rapid ink-color naming. McCusker and Gettings (1997) used a Stroop-like task to explore interference with gambling-related words in problem gamblers who either used VLTs or bet on horse races. Participants were presented with colored words that were neutral in content and drug and gambling related. It was found that overall, the problem gamblers took longer to color-name gambling-related than neutral or drug-related words. A specificity effect was also noted: VLT players exhibited interference effects for VLT-related words only, and horse race gamblers displayed interference only to track-related words. The IAT requires the test-taker to rapidly sort various stimuli into four categories using two response keys, where two of the categories (e.g., alcoholic beverages and reward outcomes) are assigned to the first response key and the other two categories (e.g., nonalcoholic beverages and relief outcomes) to a second response key. It is assumed that rapid categorization when two concepts are assigned the same key is indicative of strong implicit association between the two concepts (e.g., a strong semantic association between alcoholic beverages and reward outcomes). Comparatively slow sorting indicates a lack of implicit association or a semantic dissociation between the two concepts assigned to a given key. Using an IAT paradigm, Zack et al. (2005) assessed responses to alcoholic or nonalcoholic beverage words paired with words related to gambling win outcomes

104

Research and Measurement Issues in Gambling Studies

and words related to gambling losses. All participants in the study met the South Oaks Gambling Screen (SOGS) (Lesieur and Blume 1987) criteria for problem gambling. It was found that gamblers who drank when they won exhibited faster response times on trials in which alcohol-related words were paired on the same response key with gambling win words (e.g., jackpot) than when they were paired on the same response key with gambling loss words (e.g., forfeit).The authors suggested that drinking in response to gambling wins creates associations between gambling wins and alcohol in memory and that such implicit associations may promote drinking while gambling (e.g., to celebrate winning or the anticipation of winning). Implicit cognition tasks frequently make use of the fact that reading is a highly automatic activity. In an LST, participants are asked to read a list of words as quickly as possible.The time required to read a word is taken to be an index of the word’s importance to the reader. The assumption is that faster response times for the reading of schema-relevant words are indicative of schema activation engendered by exposure to those words. In LSTs, words are often “degraded” by intersticing arbitrary symbols among the letters (e.g., j*a*c*k*p*o*t) (see Stanovich and West 1983). An example of an LST in gambling research comes from Zack and Poulos (2004), who investigated the effects of amphetamine administration on implicit cognition.They employed a factorial design that included two groups of gamblers (problem vs. nonproblem). The amphetamine administered was a 30-mg dose of d-amphetamine, a psychomotor stimulant.This drug was chosen because previous research indicated that problem gamblers tend to describe an imagined gambling session in much the same manner that psychostimulant users describe the psychoactive effects of their drug (Hickey, Haertzen, and Henningfield 1986).The idea behind examining the effects of the amphetamine on implicit gambling cognitions was that the amphetamine might induce a state similar to that experienced by gamblers when gambling and thus might automatically activate their gambling schema. The investigators used five categories of words in their LST: gambling, alcohol, positive affect, negative affect, and neutral (e.g., window). It was found that amphetamine produced different effects in problem gamblers versus nonproblem gamblers. Nonproblem gamblers exhibited faster latency times for all classes of words—a nonspecific stimulation effect. In the problem gamblers, however, it appeared that amphetamine led to rapid responding to gambling words but inhibited responses to neutral words.Amphetamine had no effect on problem gamblers’ response time to alcohol words. Thus, this study made successful use of the LST implicit cognition task to demonstrate that, consistent with hypothesis, amphetamine automatically and selectively activates problem gamblers’ gambling schema. We now turn to methodologies that allow for examination of gambling-related cognitions “online,” while gamblers are actually engaged in gambling behavior.

Experimental Methodologies in Gambling Studies

105

The first of these is a measure that would be considered to tap explicit gamblingrelated cognitions online (i.e.,Think Aloud), and the second would be considered to tap online implicit cognition (i.e., a reaction time task).

THINK ALOUD In the Think Aloud (TA) method (Ericsson and Simon 1980), gamblers verbalize their thoughts while they are engaged in a gambling activity.These verbalizations are typically recorded by the researcher, transcribed, and then examined for the presence of irrational statements. It is assumed that irrational statements are manifestations of underlying gambling cognitive distortions. One problem with the TA method is reactivity; once voiced, a certain statement may sound dubious or surprising to the participant and may therefore influence his or her subsequent thoughts and/or actions. An illustration of this comes from a study by Griffiths (1993), who analyzed 30 VLT gamblers via the TA method. After each participant was tested, he or she was given the option of listening to the audio recording of the session. Only four wished to do so, and, according to Griffiths, all were quite surprised by what they had said during the session. Several months after the study, Griffiths accidentally encountered one of the participants, a 19-year-old male. This young man had been diagnosed as a problem gambler at the time of the study. Griffiths described the content of their brief conversation: [S]ince taking part in my study, his gambling behavior had declined and subsequently ceased. He claimed that a large factor in the cessation of his gambling was hearing the playback of his recording. He claimed he could still remember some of the things he heard on the tape but reiterated his disbelief at what he had verbalized. He further claimed it was his disbelief that prompted him to examine and monitor his [gambling] behavior more closely.Through this self-introspective process, he realized the futility of his gambling and eventually stopped playing. (p. 296)

REACTION TIME TASKS An advantage of the TA method is that it provides a means of assessing cognition in participants while they are involved in a gambling task. However, in addition to the problem of reactivity, other problems associated with the assessment of explicit cognition (e.g., assumption that gamblers are aware of their cognitive content and can report on it accurately) plague the TA method. Other methods of assessing gambling-related cognition online are not subject to these problems and fall more under the category of implicit cognitive measures, as described earlier. One such alternative method, involving reaction time (RT) was provided by

106

Research and Measurement Issues in Gambling Studies

Diskin and Hodgins (1999), who explored the tendency to become highly absorbed during gambling sessions as a function of problem gambling status. Pathological and occasional VLT players were required to respond to a light that appeared intermittently in the periphery of their vision while they were playing VLTs in a simulated laboratory setting.The investigators hypothesized that pathological gamblers would focus their attention more intensely on VLT play than would the occasional gamblers and, as a result, would take longer to respond to the light. As predicted, the problem VLT gamblers took significantly longer than the occasional gamblers to respond to the peripheral light. Additionally, pathological gamblers reported more symptoms of general dissociation than did the occasional gamblers. In a later study, Diskin and Hodgins (2001) again compared pathological and occasional VLT gamblers on their response times to the light stimulus. However, in this study, the experimental design was slightly modified. Participants were randomly assigned to one of two conditions: baseline first and VLT first.The baseline-first condition consisted of a five-minute RT test to the light stimulus while the VLT screen was covered. Immediately after the baseline-first task, the VLT was uncovered and participants were asked to play the machine in their normal manner while simultaneously responding as quickly as possible to the light. Individuals who were assigned to the VLT-first condition played the VLT first and simultaneously responded to the light, and then completed the RT test with the VLT screen covered. Although it was predicted that the pathological gamblers would take generally longer to respond to the light stimulus while playing the VLT, this hypothesis was not confirmed when data from both conditions were combined. However, when the effect of task order was analyzed, some interesting findings emerged. For instance, among participants who played the VLT first, the pathological gamblers took almost twice as long as the occasional gamblers to respond to the light while playing the VLT. However, in the baseline-first task, in which participants had the opportunity to complete the RT test before playing the VLT, the pathological gamblers responded to the light stimulus almost seven times faster than the occasional gamblers during subsequent VLT play. This suggests that even though they were playing the VLT at the time, they were able to focus intensely on the light stimulus. It appeared, then, that although the pathological gamblers exhibited a tendency to become highly absorbed in VLT play under certain circumstances, they were capable of shifting their focus away from VLT play onto other stimuli. This set of studies illustrates the utility of using online cognitive assessment methods, as they produced findings that would not likely have been uncovered by offline methods such as a self-report cognitions inventory. Moreover, these findings have important treatment implications, as they suggest that there are ways that we can modify problem gamblers’ tendencies to become highly absorbed in VLT play.

Experimental Methodologies in Gambling Studies

107

CONCLUSIONS We hope that this chapter has provided a useful glimpse into the various ways in which experimental methodologies have been, and can be, used in the study of gambling behaviors and gambling disorders. Experimental methods are the only set of research methods that can definitively determine causality because the experiment is high in internal validity, or control of potentially confounding variables. Thus, experimental methodologies can help definitively determine the causes of gambling problems and effective interventions for pathological gamblers. However, there are limits to experimental methods, including the external validity of the findings, or their generalizability to the real-world setting, and the obvious ethical limitations to conducting experimental research on the causes of gambling disorders. For example, it clearly would not be ethical to subject participants to factors such as prolonged exposure to gambling in order to determine whether such exposure caused a gambling disorder! For these reasons, it is important that converging research methods are used in gambling research to address important gambling issues, where the experimental methodologies described in this chapter are supplemented with the more externally valid research methods discussed in other chapters of this book.

GLOSSARY Dependent variable the outcome variable or factor that the researcher expects to be influenced or to change in the study. In the area of gambling and gambling disorders, the dependent variable might be overt gambling behaviors, gamblers’ thoughts or feelings, physiological variables, or symptoms of a gambling disorder. External validity the degree to which the findings can be generalized beyond the particular experiment—to individuals who were not engaged in the research study in question and to settings beyond the laboratory (i.e., to the real-world gambling context). External validity can be enhanced in experimental studies in the gambling area by ensuring that the testing environment closely resembles the real-world gambling setting. Hypothesis an educated guess about the outcome of the study that will be tested through the collection of data. The hypothesis must be stated clearly in a form that is testable, where the null hypothesis (predicting no relationship between the independent and dependent variables) is contrasted with the experimental hypothesis (which predicts a relationship between the independent and dependent variables). Independent variable the factor that is manipulated by the experimenter and that is expected to have an effect on the dependent, or outcome, variable.

108

Research and Measurement Issues in Gambling Studies

In a pathological gambling treatment outcome study, for example, the independent variable is the application of the treatment (e.g., treatment vs no treatment) that is expected to reduce gambling problems. Internal validity refers to the degree of experimental control within the study design and consequently the degree to which one can be confident that the manipulation of the independent variable is responsible for causing the outcome on the dependent variable. Experimental methodologies have the advantage of being high in internal validity relative to other research methods.

ACKNOWLEDGMENTS The authors would like to extend their thanks to Pamela Collins, Laboratory Manager at the Dalhousie Gambling Laboratory, for her research and administrative assistance in preparing this chapter. Dr. Stewart would like to acknowledge the support she and her colleagues have received from the Nova Scotia Gaming Foundation, the Ontario Problem Gambling Research Centre, and the Canadian Institutes of Health Research; Dr. Jefferson would like to acknowledge the support he and his colleagues have received from the Atlantic Lottery Corporation, New Brunswick, and the Hospital Corporation, Region 3, New Brunswick.

REFERENCES Asmundson, G. J. G., Norton, G. R., and Stein, M. B. (2002). Clinical Research in Mental Health: A Practical Guide.Thousand Oaks, CA: Sage. Barlow, D., Durand, M., and Stewart, S. H. (2006). Abnormal Psychology, 1st Canadian ed. Toronto: Nelson-Thompson. Bechara, A., Damasio, A. R., Damasio, H., and Anderson, S. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50, 7–15. Dickerson, M. (1993). Internal and external determinants of persistent gambling: Problems in generalizing from one form of gambling to another. Journal of Gambling Studies, 15, 17–28. Diskin, K., and Hodgins, D. (1999). Narrowing of attention and dissociation in pathological video lottery gamblers. Journal of Gambling Studies, 15, 17–28. —— . (2001). Narrowed focus and dissociative experiences in a community sample of experienced video lottery gamblers. Canadian Journal of Behavioural Science, 33, 58–64. Doiron, J., and Nicki, R. (in press). Prevention of pathological gambling:A randomized controlled trial. Cognitive Behaviour Therapy. Ellery, M., Stewart, S. H., and Collins, P. (2006). An evaluation of irrational beliefs as possible mediators of the behavioral effects of alcohol consumption on video lottery terminal (VLT) play among probable pathological and non-pathological gamblers [Summary]. Alcoholism: Clinical and Experimental Research, 30, 96a. Ellery, M., Stewart, S. H., and Loba, P. (2005). Alcohol’s effects on risk-taking during video lottery terminal (VLT) play among probable pathological and non-pathological gamblers. Journal of Gambling Studies, 21, 299–324. Ericsson, K. A., and Simon, H. A. (1980).Verbal reports as data. Psychological Review, 87, 215–251.

Experimental Methodologies in Gambling Studies

109

Focal Research. (1998). Nova Scotia Video Lottery Players’ Survey 1997/98. Halifax, Canada: Nova Scotia Department of Health, Problem Gambling Services. Griffiths, M. D. (1993). Pathological gambling: Possible treatment using an audio playback technique. Journal of Gambling Studies, 9, 295–297. Henslin, J. M. (1967). Craps and magic. American Journal of Sociology, 73, 316–360. Hickey, J. E., Haertzen, C. A., and Henningfield, J. E. (1986). Simulation of gambling responses on the Addiction Research Center Inventory. Addictive Behaviors, 11, 345–349. Hills, A. M., and Dickerson, M. (2002). Emotion, implicit decision making, and persistence at gaming. Addiction, 97, 598–599. Holloway, F. A. (1995). Low-dose alcohol effects on human behavior and performance. Alcohol, Drugs, and Driving, 11, 39–56. Jefferson, S., and Nicki, R. (2003). A new instrument to measure cognitive distortions in video lottery terminal users: The Informational Biases Scale (IBS). Journal of Gambling Studies, 19, 387–401. Kim, S. W., Grant, J. E., Adson, D. E., and Shin, Y. (2001). Double-blind naltrexone and placebo comparison study in the treatment of pathological gambling. Biological Psychiatry, 49, 914–921. King, K. M. (1990). Neutralizing marginally deviant behavior: Bingo players and superstition. Journal of Gambling Studies, 6, 43–61. Ladouceur, R., Gaboury, A., Bujold, A., LaChance, N., and Tremblay, S. (1991). Ecological validity of laboratory studies of videopoker gaming. Journal of Gambling Studies, 7, 109–116. Ladouceur, R., Gaboury, A., Dumont, M., and Rochette, P. (1988). Gambling: Relationship between the frequency of wins and irrational thinking. Journal of Psychology: Interdisciplinary and Applied, 122, 409–414. Langer, E. J. (1975).The illusion of control. Journal of Personality and Social Psychology, 32, 311–328. Leary, K., and Dickerson, M. (1985). Levels of arousal in high- and low-frequency gamblers. Behaviour Research and Therapy, 23, 635–640. Lesieur, H. R., and Blume, S. B. (1987).The South Oaks Gambling Screen (SOGS): A new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, 1184–1188. Loba, P., Stewart, S. H., Klein, R. M., and Blackburn, J. R. (2001). Manipulations of the features of standard video lottery terminal (VLT) games: Effects in pathological and non-pathological gamblers. Journal of Gambling Studies, 17, 297–320. MacDonald, A. B., Stewart, S. H., Hutson, R., Rhyno, E., and Loughlin, H. L. (2001). The roles of alcohol and alcohol expectancy in the dampening of responses to hyperventilation among high anxiety sensitive young adults. Addictive Behaviors, 26, 841–867. MacLin, O. H., Dixon, M. R., and Hayes, L. J. (1999). A computerized slot machine simulation to investigate the variables involved in gambling behavior. Behavior Research Methods, Instruments, and Computers, 31, 731–734. McCusker, C. G., and Gettings, B. (1997). Automaticity of cognitive biases in addictive behaviours: Further evidence with gamblers. British Journal of Clinical Psychology, 36, 543–554. Parke, A., and Griffiths, M. (2004). Aggressive behaviour in slot machine gamblers: A preliminary observational study. Psychological Reports, 95, 109–114. Spiegler, M. D., and Guevremont, D. C. (1998). Contemporary Behavior Therapy, 3rd ed. Pacific Grove, CA: Brooks/Cole Publishing Company. Stacy, A.W., Leigh, B. C., and Weingardt, K. R. (1994). Memory accessibility and association of alcohol use and its positive outcomes. Experimental and Clinical Psychopharmacology, 2, 269–282. Stanovich, K. E., and West, R. F. (1983). On priming by a sentence context. Journal of Experimental Psychology: General, 112, 1–36. Steenbergh,T. A.,Whelan, J. P., Meyers, A.W., May, R. K., and Floyd, K. (2004). Impact of warning and brief intervention messages on knowledge of gambling risk, irrational beliefs, and behaviour. International Gambling Studies, 4, 3–16.

110

Research and Measurement Issues in Gambling Studies

Stewart, S. H., Blackburn, J. R., and Klein, R. M. (2000, Spring). Against the odds: Establishment of a video lottery terminal research laboratory in a naturalistic setting. Nova Scotia Psychologist, 3–6. Stewart, S. H., Collins, P., Blackburn, J. R., Ellery, M., and Klein, R. M. (2005). Heart rate increase to alcohol administration and video lottery terminal (VLT) play among regular VLT players. Psychology of Addictive Behaviors, 19, 94–98. Stewart, S. H., McWilliams, L. A., Blackburn, J. R., and Klein, R. M. (2002). A laboratory-based investigation of relations among video lottery terminal (VLT) play, negative mood, and alcohol consumption in regular VLT players. Addictive Behaviors, 27, 819–835. Stewart, S. H., Peterson, J. B., Collins, P., Eisnor, S., and Ellery, M. (2006). Heart rate increase to alcohol administration and video lottery terminal (VLT) play among probable pathological gamblers and nonpathological gamblers. Psychology of Addictive Behaviors, 20, 53–61. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643–662. Symes, B.A., and Nicki, R. M. (1997).A preliminary consideration of cue-exposure, response-prevention treatment for pathological gambling behaviour: Two case studies. Journal of Gambling Studies, 13, 145–157. Wynne, H. J. (1994). A description of problem gamblers in Alberta: A secondary analysis of the Gambling and Problem Gambling in Alberta study. Edmonton, Canada: Alberta Alcohol and Drug Abuse Commission. Zack, M., and Poulos, C. X. (2004). Amphetamine primes motivation to gamble and gambling-related semantic networks in problem gamblers. Neuropsychopharmocology, 29, 195–207. —— . (2006). Implicit cognition in problem gambling. In Handbook of Implicit Cognition and Addiction (R.W.Wiers and A.W. Stacy, eds.), pp. 379–391.Thousand Oaks, CA: Sage. Zack, M., Stewart, S. H., Klein, R. M., Loba, P., and Fragopoulos, F. (2005). Contingent gamblingdrinking patterns and problem drinker status moderate implicit gambling–alcohol associations in problem gamblers. Journal of Gambling Studies, 21, 325–354.

CHAPTER 5

Qualitative Methodologies Robert A. Stebbins Department of Sociology University of Calgary Calgary, Alberta, Canada

Grounded Theory and Exploration Exploratory Concatenation Confirmatory Qualitative Research Validity and Reliability Methods of Data Collection Qualitative Research on Gambling Issues Conclusions

Let us be clear from the outset what we mean by some of the key terms routinely used in the field of qualitative methodologies. The broadest concept in this area is that of qualitative research, defined by Norman K. Denzin and Yvonna S. Lincoln (1994, p. 2) as data collection that employs one or more interpretive and naturalistic methods to study its subject matter.Theirs is a generic definition, however, meant to embrace the multitude of more specific definitions of the concept that abound in this field. In addition, and in harmony with the preceding definition, I see qualitative research as composed of three elements: (1) the various data collection techniques (e.g., participant observation, semi-structured interviews), (2) the process of collecting data (e.g., the use of these techniques in a particular study), and (3) the development of grounded theory from the collected data.The order of these elements, as just presented, is chronological in that once a research problem is determined, one or more appropriate techniques of data collection are then selected, and the study gets under way by using those techniques. Next, through analysis of the data collected, grounded theory is developed. These three will be discussed here in reverse order, however, for by doing so we can accent the primordial goal of the first two elements, which is to develop grounded theory. 111

112

Research and Measurement Issues in Gambling Studies

GROUNDED THEORY AND EXPLORATION The second element may be more precisely understood as exploration, the aim of which is to generate new ideas (using the techniques in the first element) and weave them together to form what Glaser and Strauss (1967) call grounded theory.This is the set of generalizations derived inductively from data collected directly through a third element, exploratory research. Robert A. Stebbins (2001, p. 3) defines and describes social scientific exploration as follows: a broad-ranging, purposive, systematic, prearranged undertaking designed to maximize the discovery of generalizations leading to description and understanding of an area of social or psychological life. Such exploration is, depending on the standpoint taken, a distinctive way of conducting science—a scientific process—a special methodological approach (as contrasted with confirmation), and a pervasive personal orientation of the explorer.The emergent generalizations are many and varied; they include the descriptive facts, folk concepts, cultural artefacts, structural arrangements, social processes, and beliefs and belief systems normally found there.

The various generalizations about the group, process, activity, or situation under study discovered while exploring it, when logically integrated with one another, become the components of a grounded theory of that area. Hayano’s (1984, pp. 158–161) work on types of professional poker and blackjack players illustrates well the development and integration of generalizations (or types) by way of qualitative exploration. He identifies four types of professional poker and blackjack players. The “worker-professional” is a man or woman who holds a nondeviant job but is dedicated to a career in gambling and spends a good deal of time pursuing it.The nondeviant job, however, provides the steady income needed to live from day to day. For this type, playing poker or blackjack is, thus, leisure. The “outside-supported professional” has an even better extragambling income, derived from savings, investments, retirement funds, or a similar source. In this category we find retired people, housewives, and the independently wealthy. Hayano (1984) includes here students who, presumably, use their gambling winnings as a source of pin money. These players can also be classified as leisure participants. The “subsistence professional” is a consistent winner, but of only relatively small bets. He or she uses the winnings to live on, to pay the bills of everyday living.There is little interest here in a career in gambling or in moving on to higher stakes and stiffer competition. This, however, is precisely the orientation of the “career professional.” Such a person is much more likely to be a man than a woman, compared with the first three types, and is most highly committed to gambling as an occupation. Even when broke, he prefers to borrow money in order to return to gambling. Hayano (1977) found in his study of poker professionals in Gardena, California (one of the

Qualitative Methodologies

113

few places in North America where commercial poker gambling is legal), that they see their occupation as arduous. They emphasize a work ethic, including a commitment to poker, honesty, and integrity in borrowing and lending money to pay their own or others’ gambling debts.They also pride themselves on “class” personal demeanor, whether winning or losing. Exploration in the social sciences may be further understood by discussing what it is not. It is decisively not serendipity. Serendipity is the quintessential form of informal experimentation, accidental discovery, and spontaneous invention (Stebbins 2001, p. 4) contrasts sharply with exploration, just described as a broadranging, purposive, systematic, prearranged undertaking.The first is highly democratic—at least in principle anyone can experience serendipity—whereas the second is more narrowly select, the province of those creative people who must routinely produce new ideas. In certain fields of serious leisure and professional work, artists, scientists, and entertainers, for example, routinely explore, while, in some forms of casual leisure, people at play (both children and adults), sociable conversationalists, and seekers of sensory stimulation never do this (Stebbins 1997), an observation that holds equally well for many nonprofessional kinds of work. For the second group, new ideas and other discoveries can come only by way of serendipity, compared with the first group, where discovery, though occasionally serendipitous, is nonetheless far more likely to flow from exploration. Exploration and the grounded theory it generates constitute the early steps of the scientific process leading to verificational, or confirmatory, research, wherein hypotheses are tested in controlled circumstances, the experiment being the archetypical example.Where hypothesis-driven research has been preceded by exploration, as it ought to be ideally, the hypotheses tested are often the generalizations that make up the grounded theory in the area under study. So confirmation has a different goal from exploration. It relies on control of variables and prediction of outcomes using hypotheses. By contrast, exploration requires flexibility in looking for data and open-mindedness about where to find them (Stebbins 2001, pp. 6, 9–10). In general, exploration is the preferred methodological approach under at least three conditions: when a group, process, activity, or situation (1) has received little or no systematic empirical scrutiny, (2) has been largely examined using prediction and control rather than flexibility and open-mindedness, or (3) has grown to maturity but has changed so much along the way that it begs to be explored anew.

EXPLORATORY CONCATENATION In social science research exploratory concatenation refers, at once, to a longitudinal research process and the resulting set of open-ended field studies,

114

Research and Measurement Issues in Gambling Studies

which are linked together, as it were, in a chain leading to cumulative, often formal, grounded theory (Stebbins 1992, 2001). Studies near the beginning of the “chain” are wholly or predominantly exploratory in scope. Each study, or link, in the chain examines or, at times, reexamines a related group, activity, or social process or aspect of a broader category of groups, activities, and so on. Where this metaphor of a chain of studies becomes inadequate is in its failure to suggest the accretive nature of properly executed, concatenated exploration. In the metaphor of the chain, each link is equally important, whereas in scientific concatenation the studies in the chain are not only linked, but are also predicated on one another; that is, later studies are guided, in significant measure, by what was found in earlier research in the same area as well as by the methods used and the samples examined there. Thus each link plays a somewhat different part in the growing body of research and in the emerging grounded theory. Furthermore, note that the earlier studies only guide later exploration; they do not control it to the point where discovery is thwarted by preconceptions. Exploration describes the nature of the overall approach to data collection that is followed especially at the beginning of the chain and, to a significant degree, all along it as well. Glaser and Strauss (1967) observe that exploration may be qualitative or quantitative, even though most researchers in this area seem to favor mixing the two, with qualitativeness being primary and quantitativeness secondary. Still, as the chain of studies lengthens, quantitative data may grow in proportion and importance vis-à-vis qualitative data. Consequently, the terms “exploration” and “exploratory research” subsume both forms of data, whatever their ratio and significance in any particular study in the chain of studies or in the entire chain itself.

CONFIRMATORY QUALITATIVE RESEARCH Although qualitative methods are used mostly in exploratory research, they may, quite logically, be employed to test hypotheses generated in earlier exploration or deduced from established theory.The open-ended nature of these methods (see next section) discourages their use in large-scale surveys, where fixed-answer questionnaires and scales are far more efficient. Nevertheless, some research problems lend themselves poorly to quantification, opening up thereby a place for the openended methods of qualitative data collection. For example, hypotheses about highly emotional matters, such as the anguish of repeated, costly losses at gaming or fractious relations with a spouse over gambling habits, are often better tested (have greater validity) using open-ended, qualitative procedures than ones that are quantitative and answered using fixed responses. One of the widely acknowledged strengths of quantitative methods is their ease of application, and hence their capacity for studying large samples of people.

Qualitative Methodologies

115

This feature is obviously lost when qualitative methods must be used. But the latter can still be used to test hypotheses, even if the question of the broader distribution of the object of verification remains a problem. Confirmatory qualitative research is relatively uncommon and appears to be next to nonexistent in gambling studies.Thus the remainder of this chapter is devoted to gambling research carried out at the exploratory level.

VALIDITY AND RELIABILITY The question of validity in exploratory research, which goes at times by the name of “credibility,” refers to whether a researcher can gain an accurate or true impression of the group, process, or activity under study, and if so, how this can be accomplished. According to McCall and Simmons (1969, p. 78), validity is problematic in this realm of social scientific inquiry in at least three ways: 1. reactive effects of the observer’s presence or activities on the phenomenon being observed, 2. distorting effects of selective perception and interpretation on the observer’s part, and 3. limitations on the observer’s ability to witness all relevant aspects of the phenomena in question. These three problems worry exploratory and, even more so, confirmatory researchers alike.Viewed from the angle of the latter, exploration appears to evoke the greatest suspicion with reference to the second problem, primarily because of the heavy subjective element involved when a lone researcher (the usual party in an exploration) observes and interviews, employing an open-ended design. Exploratory researchers try to enhance the validity of their studies in various ways. Many of them discuss their emergent generalizations with the people they are investigating to determine whether these ideas have a familiar ring and whether, in the eyes of these people, the generalizations seem plausible. Some of these researchers may even be given written text to read. Additionally, aware that personal bias can distort perception and interpretation of observed events, competent explorers look assiduously for evidence that might contradict their observations.This approach is successful to the extent that the researcher is aware of his or her biases—and it is likely that no one is fully aware of them all—and that they are not held with great, unbending conviction.Third, these researchers constantly ask themselves whether they have observed a sufficient number of occurrences of an event, process, or activity to constitute grounds for a valid generalization. There are, furthermore, three central points to be made about validity in qualitative/exploratory research:

116

Research and Measurement Issues in Gambling Studies

1. Validity in qualitative/exploratory research is, in at least one way, substantially different from validity in confirmation. In the former, the focus of concern is on the explorer’s capacity to acquire directly an accurate impression of a group, process, activity, or situation, whereas in the confirmatory validity, the focus is on the investigator’s capacity to find measures and indices that indirectly convey an accurate impression of these phenomena. In fact, the validity issue is more easily resolved in exploration—accomplished, for example, by using different methods to examine the same group or activity (known as triangulation), asking key informants to comment on the familiarity and reasonableness of observations, and finding recurrent evidence for each generalization. Unfortunately, confirmatory researchers sometimes have to evaluate exploratory work, even though they are accustomed to working only with indirect measures and often know little about how validity is achieved and assessed in the realm of discovery. In this situation, they are wont to try to force the exploratory study into the Procrustean bed of verificational validation. 2. Confirmatory researchers are inclined to use what might be called a “one-shot” approach to assessing both validity and reliability. Each study undertaken must meet established criteria on both accounts. By contrast, exploratory researchers, to the extent that they concatenate their research, take a more global approach, arguing that judgments about validity and reliability are to be made with reference to a set of studies, which together demonstrate most convincingly how these two conditions have been realized. Although validity and reliability are also important to exploratory researchers in each study they execute, they recognize that the most authoritative statement about them both can be made only down the road, in the wake of several open-ended investigations. 3. Validity in exploration has a lot to do with representativeness of the sample of groups, processes, or activities being examined. Validity is strongest when hypothetical generalizations emerge from direct, empirical study of a set of representative instances. That is, validity rests on the number of times a regularity of thought or behavior is observed in talk or action, which must be often enough to seem general to all or a main segment of the people in the group, process, or activity being examined. If, for example, a study purports to be an exploration of routine life in casinos in a particular city and every casino there is observed, then representativeness of that sample is assured by means of this 100% coverage. But should casinos in socioeconomically disadvantaged areas of town be omitted from the study, the claim of representativeness in the city being examined would be false and the validity of the project, for this reason, called into question. Reliability refers to the replicability of a researcher’s observations; it turns on the question of whether another researcher with similar methodological

Qualitative Methodologies

117

training, understanding of the field setting, and rapport with its subjects can make similar observations. The same three problems identified earlier by McCall and Simmons (1969) also affect reliability in exploratory research. Here, too, after all the fretting about reliability in a particular study, experience remains the best teacher: Sufficient concatenation with different researchers participating in the process is needed to demonstrate most convincingly that the researchers can make similar or compatible observations on the same or related groups, processes, or activities. Moreover, as with validity, great concern with reliability of the study at hand is most appropriate toward the end of the research chain, where confirmation is the rule, compared with the beginning of that chain, where exploration dominates. More extensive treatments of validity and reliability in qualitative research are available from Kirk and Miller (1986) and Stewart (1998).

METHODS OF DATA COLLECTION Before discussing some of the key methods used to collect exploratory data, let us consider what researchers actually look for when exploring an area of social life. Note first that many exploratory researchers are reluctant to be very prescriptive about what to look for. They plump for something no more constraining than the old research formula of searching for field data according to the five Ws: who, what, whom, when, and where. (I consider, as data generating devices, the five in Stebbins 2001, p. 23. See also Denzin 1970, pp. 269–284.) In exploration the researcher wants to learn who is doing (thinking, feeling) what to (with, for, about) whom and when and where. Open-ended procedures generate data on these five questions, data that in turn become the basis for generalizations in the form of concepts and their interrelationship in propositions.What the old formula neglected (and, consequently, what I neglected in Stebbins 2006) is that there is also a most important additional question of how. How do the people being observed do what they do? This is not so much a conceptual interest, however, as a descriptive one. The answer to this question gives the descriptive, ethnographic foundation on which the explorer constructs more abstract grounded theory revolving around the five Ws. Furthermore there is also another W, which however, is theoretic; it is why people do what they have been observed to do. Still, answering this question does not steer data collection. Rather it greatly aids interpretation of the data that have been gathered. In sum, the five Ws and the H combine to guide data collection, leading the exploratory researcher to ask why the data collected are as they are.

118

Research and Measurement Issues in Gambling Studies

The watchword for every method of data collection in qualitative research is open-endedness, which serves as a way of implementing the characteristics of flexibility and open-mindedness of exploration. That is, each method is designed to reveal a wide range of behavior, activity, attitude, or other object of investigation. Five widely used open-ended methods are covered here: participant observation, semi-structured interviewing, focus group, narrative, and content analysis. All five have several variants that relate to special research situations (which, because of spatial limitations, will not be discussed here). These five methods may be used to generate, or in some instances help generate, either an ethnography or what I will call here a set of problem-centered data. An ethnography is a systematic, broad-ranging description of all or a major segment of the social or cultural life of a category or group of people. As the preceding definition stated, ethnographies are developed using one or more openended methods of data collection. Problem-centered data, which also may be collected using one or more open-ended methods, bear on a smaller, more specialized domain of knowledge within an ethnographic area, such as, in gambling, perceived risk, relations with spouses, and financing debt. Participant observation is the practice of directly watching and recording (even if only in memory) what is happening in the group, activity, or situation being examined. Adler and Adler (1987) point out that such observation may be conducted by one of three types of observer: a “complete member,” or bona fide member of the group (e.g., a gambler studying gamblers); an “active member,” or outsider to the group who nevertheless has access to all group involvements; or a “peripheral member,” or someone who is known to group members but lacks access to the group’s core activities.The latter two types are, in the case of gambling, commonly nongambling social scientists who have obtained different degrees of involvement with the group. Additionally, the research role of the observer may be known to the people being studied (overt observation) or unknown to them (covert observation). The adjective “participant” is not to be taken literally, to mean that “participant observers” are actually doing the activity in question. Rather, it signifies that those observers are physically present, and thus are able to see what is going on. This method is open, because the observer is in a physical position to view much, if not all, that there is to see in naturally occurring behavior and activity. Participant observation is the classic method used to develop an ethnography of an area. All science is based on some kind of description of the phenomenon being examined. Accordingly the ethnographies, which undergird exploration in an area, are generated by way of participation observation and are necessarily highly descriptive.Were we to carry out an ethnography of a Las Vegas casino, we would start by looking at who (e.g., the gamblers, card dealers, servers of food and drink) is doing what (e.g., placing bets, dealing cards, bringing food and drink) to or for whom (e.g., with the dealer, to the gamblers, for the gamblers), when

Qualitative Methodologies

119

(e.g., evenings), and where (e.g., casino rooms in large commercial establishments in a special geographic area of Las Vegas). The next four methods provide data that help explain why ethnographic pictures like the one just sketched exist. Semi-structured interviewing is conducted using an “interview guide.” This guide consists of a set of broad questions that bear on the research problem. Each item (question) in the guide is open-ended in that no preconceived categories are available into which interviewers must place their respondents’ answers. Still the interview has some minimal structure, for the items direct interviewees to talk about particular broad areas of their lives. McCracken (1988) argues that semistructured interviews are especially well adapted for gathering personal information that only the respondent is in a position to give, which includes personal opinion, attitude, belief, accomplishments, and major life events, as well as personal background data and that person’s interpretation of all this. Focus groups consist of small numbers of participants who are assembled by the researcher for the purpose of discussing one or a few common problems, issues, situations, and the like. Morgan (1997) says that such groups are good for determining collective positions on the problems, issues, and situations, though the problem, issue, or situation must be something everyone who is present is willing and able to talk about with the others. During the interactive session, participants may be reminded by the remarks of the others of their own thoughts on the subject at hand and possibly even crystallize, then and there, their own position on it.The researcher typically directs discussion, trying to ensure that everyone present has his or her say and that, through the open-endedness of the procedure, all leads bearing on the collective position are followed up. As for the size of the focus group, Morgan (1997, pp. 42–43) recommends in general, after weighing a variety of practical and substantive considerations, no less than 6 and no more than 10 participants. Although there is scant research in gambling studies that has used focus groups, such research can be mounted to good effect. Thus, we might organize focus groups among members of several local chapters of Gamblers Anonymous (GA), with the aim of considering the question of why some problem gamblers fail to join GA. Or a focus group could be assembled to discuss how GA might better realize its goals. Recreational gamblers could be assembled into such groups to talk about the differential appeals of the many forms of gaming. Narratives are constructed by people from their own biographic material. Riessman (1993, p. 3) used, as one definition of narrative in her book on narrative analysis “talk organized around consequential events.” A narrative may be constructed, for example, about a single event (e.g., a traffic accident, a theatrical performance), about a period in a person’ life,or even about an entire life (i.e., a life story). Narratives are commonly obtained through interviewing (often in two or three sessions, if about a person’s entire life or large part of it), which has even less structure than its semi-structured cousin.

120

Research and Measurement Issues in Gambling Studies

A narrative study of gamblers might begin with a question or prompt from the interviewer addressed to each respondent, such as: “Tell me how you got into gambling.” The interviewer might then ask about when gambling became problematic and about the nature of the circumstances leading to this condition. And then a third question might be posed bearing on how the respondent came to grips with the gambling problem. Note that narratives differ from biographies, which usually take in much bigger segments of the person’s life and are not intended to be generalized or used to develop grounded theory. According to Krippendorff (2004), content analysis is the search for generalizations carried out on any kind of previously unanalyzed material, be it reports, musical recordings, books, articles, paintings, diaries, and so on. This method is open-ended in that the material selected for study has not yet been analyzed and principal themes pulled out. Unlike participant observation, neither the narrative nor content analysis is well suited as a sole method for working up an ethnography. By contrast, all five methods can be fruitfully used to generate data on a particular problem. And, to repeat, the last four are well suited for explaining (through emerging generalizations) the data collected. Just as with focus groups, research in gambling studies that employs content analysis is rare. Still the method can be used to advantage as an exploratory device, even if it must be joined with other open-ended methods to generate a full ethnography. For instance, books or newspaper articles on gambling could be analyzed for the themes that run through them. Or, on a more detailed level, printed material on how to win at different kinds of games could be so analyzed. Finally, the goal in using these five methods is to produce generalized conclusions. That is, science is a generalizing (in philosophic terms, “law-giving”) enterprise, consisting as it does of propositions (or hypotheses, general statements) about the area under study.This property of science is stressed here, because of the tendency in exploration for some researchers to get caught up with the details of, for example, the narrative, turning this method into a biography of the person interviewed. As such, no generalization is attempted, and consequently the development of ground theory, and hence of science, is also arrested. (Of course, if a biography is what the interviewer wants, as, say, a scholar in the humanities, this goal is perfectly acceptable [see Stebbins 2001, pp. 11–12].)

QUALITATIVE RESEARCH ON GAMBLING ISSUES Qualitative research on gambling has tended to center on the process of gambling, the gamblers themselves, and the social and cultural circumstances in which these people pursue gambling as leisure activity or are driven to gambling by compulsion. Note, however, that by no means is all social scientific research on these subjects qualitative; a good deal of quantitative, hypothesis-driven work has

Qualitative Methodologies

121

also been conducted on them. In this section we concentrate exclusively on some representative qualitative studies in the field of gambling.The favored methods here have been participant observation and semi-structured interviewing. Both of these methods were employed by Livingston (1974) in his classic ethnography of compulsive gamblers in a city in the United States. Livingston observed several meetings of a GA chapter and later interviewed a number of its members. His study constitutes an ethnography of the problem gambler, his personality (all research participants were male), his deviant career in gambling as the activity became more compulsive and less recreational, his career in abstaining from gambling, and the role of the organization (i.e., GA) in helping facilitate the career in abstinence. Livingston also provides a rich demographic portrait of his sample, including the variables of age, education, income, religion, and marital status. The qualitative work of Lesieur (1977)—another classic in gambling studies—revolving specifically, as it did, around the deviant career of the compulsive gambler, might lead us to classify his work as problem-centered qualitative research. Indeed, he used semi-structured interviews to discern the stages of this type of career as experienced among a sample of GA members and non-GA gamblers (including even some in prison). But his research interests also spread to gamblers’ families, the gamblers’ occupations, and, in some cases, gamblers’ criminal activities (to get money to gamble), as well as bookmakers and their business and the lending practices of certain financial institutions. In short, Lesieur’s study was sufficiently broad to be qualified as an ethnography. Turning to gambling as leisure, Hallebone (1999) used semi-structured interviews in a problem-centered study of spokespersons and counselors for non–English-speaking women from four ethnic subcultures in Victoria, Australia. She found that these women—Indochinese, Chinese, Italian, and Greek—used their beliefs in karma and predestination to justify their gambling interests. Downsizing in Australian manufacturing and retailing was found to be another factor explaining those interests. Nonetheless, for many of these women, gambling pursued in this economic situation failed to last as leisure, primarily because they came to be dependent on this activity. A companion study conducted by Hallebone in Melbourne using a similar sample showed that those women became problem gamblers because they were using the activity in hope of regaining a positive sense of identity while also trying to escape isolation, loneliness, boredom, and domestic abuse. She observed that addictive gambling worked to undermine the desires of these women, in that they lost control to an addicting activity and a burgeoning personal debt. Neal (2005) studied gambling on horse races as leisure, using participant observation to gather data in a betting shop and at a racetrack in the United Kingdom. He learned, while observing and talking informally with gamblers at these two locations, that their betting was an integral part of their routine lives.

122

Research and Measurement Issues in Gambling Studies

It was not seen by them as irrational or pathological, but as something interesting and enjoyable to do during their free time. Neal’s interest in the rationality/irrationality question makes his work too specific to be judged an ethnography. It was, rather, a problem-centered study.That having been said, he moved from his findings to a discussion of gambling as leisure, thereby expanding his scope to “an ethnography of leisure gambling.” He observed that there was an overwhelming preoccupation in the United Kingdom with the compulsive side of gaming (which represented only 0.8 to 1.2% of the gambling population there), while looking on it as leisure was far more representative of all people who gambled. In Britain the betting shops and race courses are two principal axes of the leisure gaming scene. Rosecrance’s (1985) study of the horse-racing scene at Lake Tahoe, Nevada, is similar in some ways to Neal’s work. Both are concerned with the social world of the gambler, considering it an important part of his lifestyle as well as a source of motivation, knowledge, and sympathetic understanding. Both authors also observed problem and leisure gamblers. Rosecrance’s study, however, also took in professional gamblers who bet on horse races. Both he and Neal used a combination of participation observation and semi-structured interviews to generate ethnographic accounts of the gambling scenes they studied. In a rare study in the field of gambling behavior, Pratt and colleagues (2005) organized 34 focus groups, each with four to nine members, from a sample of Ontario and Quebec adolescents aged 12 to 18 years. The object was to identify what adolescents understood as “gambling,” to explore their awareness and participation in gaming activity, to identify the benefits they saw as accruing from gambling, and to identify the risks they believed were associated with such activity. Pratt and her team found, among other things, that the sample had a broad understanding of what constituted gambling, ranging from scratch tickets and personal bets to the gaming activities found in casinos and the Las Vegas gambling establishments. Turning to content analysis—another unusual method in gambling studies— McMullan and Mullen (2001) carried out such an analysis of media coverage of casino and electronic gambling between 1992 and 1997 in Nova Scotia. The authors analyzed 234 gambling stories to learn that pro-gambling corporate and political newspaper sources had conducted a successful media campaign in support of new gaming products, services, and institutions.The media gave both form and visibility to these structured messages, thereby helping create expectations about gambling and economics as well as gambling and government.

CONCLUSIONS The principal focus of this chapter has been exploratory research leading to grounded theory, with but brief mention of qualitative research methods as a

Qualitative Methodologies

123

verificational approach.The most profound argument to be mounted in favor of exploration in the social sciences is that there may be unknown phenomena in social life (gambling included) that warrant open-ended examination to discover what they are and what significance they may hold for science and community life. The field of gambling studies has been blessed with a number of fine qualitative/exploratory investigations, some of which have been described in this chapter. Still, plenty of exploration remains to be done. For instance, earlier, Neal emphasized the need to put more effort into exploring gambling as leisure and as leisure lifestyle. Only a small proportion of all gamblers have psychological problems with an activity that, to be sure, started for them as a controllable and agreeable leisure pastime. Research is also urgently needed the matter of the stigma of gambling, whether the gambling in question is for fun or is done compulsively. Qualitative methods are well suited for casting light on this condition. The meaning to men vis-à-vis women of the different kinds of gaming is still another area begging study, as an area of human psychology that is inherently qualitative. Nevertheless, such qualitative research is time consuming. And the explorer is often the only person interested in the object of study. In the typical case, he or she is alone in the field to arrange and conduct all the necessary observations, interviews, and focus groups (see Stebbins 2001, pp. 52–55, for an account of the lifestyle of the social scientific explorer).There is no one to farm out this work to, while interviews can run from 2 to 3 hours, and a researcher can spend days or nights (or both) observing action at the casinos, racetracks, and bingo halls. It is no wonder, then, that there is a shortage of research of this kind, even while the importance of gambling as leisure and as pathology dictates that we social scientists not fail to thoroughly explore this area.

GLOSSARY Ethnography a systematic, broad-ranging description of all or a major segment of the social or cultural life of a category or group of people. Exploratory concatenation both a kind of longitudinal research (process) and the resulting set of open-ended field studies that are linked together, as it were, in a chain, leading to cumulative and often formal grounded theory (product). Exploratory research a broad-ranging, purposive, systematic, prearranged undertaking designed to maximize the discovery of generalizations leading to description and understanding of an area of social or psychological life. Grounded theory a set of generalizations derived inductively from data collected directly through exploratory research.

124

Research and Measurement Issues in Gambling Studies

Qualitative research data collection that uses one or more interpretive and naturalistic methods to study its subject matter. Reliability the replicability of a researcher’s observations. It turns on the question of whether another researcher with similar methodological training, understanding of the field setting, and rapport with its subjects can make similar observations. Serendipity the quintessential form of informal experimentation, accidental discovery, and spontaneous invention. Validity in exploratory research, refers to whether a researcher can gain an accurate or true impression of the group, process, or activity under study, and if so, how this can be accomplished. Verificational (confirmatory) research data collection guided by hypotheses derived from established theory and conducted to test, in controlled circumstances, those hypotheses.

REFERENCES Adler, P. A., and Adler, P. (1987). Membership Roles in Field Research. Beverly Hills, CA: Sage. Denzin, N. K. (1970). The Research Act: A Theoretical Introduction to Sociological Methods. Chicago: Aldine. Denzin, N. K., and Lincoln, Y. S. (1994). Introduction: Entering the field of qualitative research. In Handbook of Qualitative Research (N. K. Denzin and Y. S. Lincoln, eds.), pp. 1–18. Thousand Oaks, CA: Sage. Glaser, B. G., and Strauss,A. L. (1967). The Discovery of Grounded Theory: Strategies for Qualitative Research. Chicago: Aldine Atherton. Hallebone, E. (1999). Women and the new gambling culture in Australia. Loisir et Société/Leisure and Society, 22, 101–125. Hayano, D. M. (1977). The professional poker player: Career contingencies and the problem of respectability. Social Problems, 24, 556–564. —— . (1984, July).The professional gambler. The Annals, 474, 157–167. Kirk, J., and Miller, M. L. (1986). Reliability and Validity in Qualitative Research (Qualitative Research Methods Series 1). Newbury Park, CA: Sage. Krippendorff, K. (2004). Content Analysis:An Introduction to Its Methodology, 2nd ed.Thousand Oaks, CA: Sage. Lesieur, H. R. (1977). The Chase: Career of the Compulsive Gambler. Garden City, NY: Doubleday Anchor. Livingston, J. (1974). Compulsive Gamblers: Observations on Action and Abstinence. New York: Harper & Row. McCall, G. J., and Simmons, J. L. (eds.). (1969). Issues in Participant Observation: A Text and Reader. Reading, MA: Addison-Wesley. McCracken, G. (1988). The Long Interview.Thousand Oaks, CA: Sage. McMullan, J. L., and Mullen, J. (2001). What makes gambling news? Journal of Gambling Studies, 17, 321–352. Morgan, D. L. (1997). Focus Groups as Qualitative Research, 2nd ed.Thousand Oaks, CA: Sage. Neal, M. (2005). “I lost, but that’s not the point”: Situated economic and social rationalities in horserace gambling. Leisure Studies, 24, 291–310.

Qualitative Methodologies

125

Pratt, L., Derevensky, J. L., Gillespie, M., and Gupta, R. (2005, August). The Development of an Instrument to Assess the Role of Gambling Outcome Expectancies for Adolescents: A Qualitative Analysis of Perceived Risks and Benefits of Adolescent Gambling. Final report to the Ontario Problem Gambling Centre, International Centre for Youth Gambling Problems and High-Risk Behaviors, McGill University, Montreal. Riessman, C. K. (1993). Narrative Analysis. Newbury Park, CA: Sage. Rosecrance, J. D. (1985). The Degenerates of Lake Tahoe: A Study of Persistence in the Social World of Horse Race Gambling. New York: Peter Lang. Stebbins, R. A. (1992). Concatenated exploration: Notes on a neglected type of longitudinal research. Quality and Quantity, 26, 435–442. —— . (1997). Casual leisure: A conceptual statement. Leisure Studies, 16, 17–25. —— . (2001). Exploratory Research in the Social Sciences (Qualitative Research Methods Series 48). Thousand Oaks, CA: Sage. —— . (2006). Concatenated exploration: Aiding theoretic memory by planning well for the future. Journal of Contemporary Ethnography, 35, 483–494. Stewart, A. (1998). The Ethnographer’s Method (Qualitative Research Methods Series 46). Thousand Oaks, CA: Sage.

This page intentionally left blank

CHAPTER 6

Longitudinal Studies of Gambling Behavior Wendy S. Slutske Department of Psychological Sciences University of Missouri–Columbia Columbia, Missouri

Introduction Existing Longitudinal Studies of Gambling Behavior Longitudinal Studies of Gambling Behavior Initiated During Preadolescence and Adolescence Montreal, Canada (Vitaro and colleagues) Minnesota (Winters and colleagues) New York (Barnes and colleagues) Quebec, Canada (Vitaro, Ladouceur, and Bujold) Longitudinal Studies of Gambling Behavior Initiated During Late Adolescence and Early Adulthood Missouri College Students (Slutske, Jackson, and Sher) New York (Barnes and colleagues) Dunedin, New Zealand (Slutske and colleagues) Longitudinal Studies of Gambling Behavior Initiated During Early to Late Adulthood New Zealand (Abbott,Williams, and Volberg) U.S. Casino Employees (Shaffer and Hall) Key Issues and Challenges in Longitudinal Gambling Research Statistical Techniques for Modeling Stability and Change Dealing with Missing Data The Low Prevalence of Pathological Gambling Disorder 127

128

Research and Measurement Issues in Gambling Studies

Important Questions and What We Know So Far Temporal Resolution of Gambling Correlates: Establishing Causality? The Stability of Gambling Behavior The Course of Gambling Behavior and Gambling Problems Sequential/Stage Theories of Gambling Involvement: Is There a “Gateway” to Problems? Developmental Changes Versus Cohort or Period Effects on Levels of Gambling Involvement “Natural Experiments” in Longitudinal Gambling Research Summary of What We Don’t Know (Yet)

INTRODUCTION Longitudinal research on gambling behavior is in its infancy—the earliest study based on longitudinal data reviewed in this chapter is the paper by Winters, Stinchfield, and Kim (1995), published little more than a decade ago. As this field continues to develop, it will be important to be mindful of the wisdom of more seasoned veterans from outside of the gambling research community who have spent many decades in the longitudinal research trenches. For example, Rutter (1981, p. 334) presents his balanced perspective on the role of longitudinal research in understanding behavior: There are numerous scientific and policy questions which can only be answered effectively through the availability of longitudinal or follow-up data. In addition, there are others for which longitudinal data, although not essential, greatly improve the strength of hypothesis-testing. Nevertheless, it would be quite wrong to assume that longitudinal studies are necessarily the best means of answering psychiatric questions, or even developmental questions. All too often, longitudinal studies have been planned without any clear aims or hypotheses in mind, have involved a mindless collection of large amounts of data which are never adequately analyzed, and which continue over so many years that by the end the original measures are hopelessly inappropriate for the purposes for which they are being used.

In this chapter, I review the existing database of longitudinal studies of gambling behavior, including studies of gambling participation and gambling-related problems (problem and pathological gambling).Any study of gambling participation or gambling-related problems in which individuals from a systematically ascertained

Longitudinal Studies of Gambling Behavior

129

or representative community-based sample have been assessed on at least two separate occasions across an interval of at least 1 year have been included. I have limited this review to observational/correlational longitudinal investigations and have not covered studies in which there is some type of experimental manipulation, such as an intervention study.These are covered in Chapter 15 of this volume. I will first provide a summary of the methods used in these studies, then briefly discuss three key issues and challenges in conducting longitudinal research on gambling behavior, and then provide a review of the results of longitudinal gambling research, focusing primarily on those questions that can be answered only with longitudinal or follow-up data.

EXISTING LONGITUDINAL STUDIES OF GAMBLING BEHAVIOR Table 6.1 presents a summary of published longitudinal studies of gambling behavior, grouped according to the developmental period of the participants at the initiation of the study. Of the nine longitudinal studies listed, four were initiated in preadolescence or adolescence, three were initiated in late adolescence or early adulthood, and the remaining two studies were initiated when the participants were in early to late adulthood. My review of the existing studies will be organized around these three broad developmental groupings. The existing longitudinal studies of gambling behavior can be categorized into two basic types. The purpose of the first type of study is to examine the prospective associations between one or more predictors and later gambling behavior. The other type of study is one in which repeated assessments of gambling behavior are conducted in an effort to characterize the stability, change, and patterning of gambling behavior over time.

LONGITUDINAL STUDIES OF GAMBLING BEHAVIOR INITIATED DURING PREADOLESCENCE AND ADOLESCENCE Montreal, Canada (Vitaro and colleagues) The gambling behavior research program of Vitaro and colleagues (Vitaro, Arsenault, and Tremblay 1997, 1999;Vitaro et al. 2001, 2004) was part of a longitudinal study of 1034 boys living in disadvantaged neighborhoods in Montreal, Canada (Tremblay et al. 1994). Boys were first studied when they were in kindergarten in 1984. Given that the primary focus of the study was antisocial behavior development, the participants were recruited from the 53 schools whose students came from households with the lowest socioeconomic status in order to enrich the sample for boys at risk for the eventual development of delinquency.

Table 6.1

Setting and Investigators

Time Span of Study (years)

Number of Waves of Data Collection Included in Published Reports Total

Gambling Pathology

Measure of Gambling Pathology

Number of Subjects

% Female

Mean Age of Subjects at First Wave

903

0

11

532/305

49

16

522

57

14.5

130

Summary of Existing Longitudinal Studies of Gambling Behavior.

Gambling

Preadolescence and Adolescence 12

10

8

3

8

3

3

3

7

6

2

0

SOGS-RA/ SOGS SOGS-RA/ SOGS —

3

3

1

0

—

631

0

10

DSM-III/ III-R/IV —

468

54

18.5

597

0

18

Modified SOGS

939

49

18.0

Late Adolescence and Early Adulthood Missouri college students Slutske, Jackson, and Sher New York Barnes,Welte, Hoffman, and Dintcheff Dunedin, New Zealand Slutske, Caspi, Moffitt, and Poulton

11

4

1

4

5

2

2

0

3

2

0

1

Early to Late Adulthood New Zealand Abbott and Volberg U.S. casino employees Shaffer and Hall

7

2

2

2

SOGS-R

143

46

18–65+

2

3

3

3

SOGS

639

58

37

SOGS, South Oaks Gambling Screen; SOGS-RA, South Oaks Gambling Screen–Revised for Adolescents; SOGS-RA, South Oaks Gambling Screen–Revised; DSM, Diagnostic and Statistical Manual of Mental Disorders; III, third edition, R, revised, IV, fourth edition.

Research and Measurement Issues in Gambling Studies

Monteal, Canada Vitaro, Arsenault,Tremblay, et al. Minnesota Winters, Stinchfield, et al. New York Barnes,Welte, Hoffman, and Dintcheff Quebec, Canada Vitaro, Ladouceur, and Bujold

Longitudinal Studies of Gambling Behavior

131

Prospective predictors of gambling outcomes in this study included self-rated impulsivity (Vitaro et al. 1997, 1999, 2001, 2004), risk taking (Vitaro et al. 2004), friends’ deviancy (Vitaro et al. 2001), parental supervision (Vitaro et al. 2001), and teacher-rated impulsivity (Vitaro et al. 1997, 1999), all assessed at ages 13 and 14 years of age; self-rated delinquency, drug and alcohol use, gambling frequency, and gambling problems assessed at age 16 years (Vitaro et al. 2001); and a composite inhibition measure based on teacher ratings obtained at ages 6, 10, 13, and 14 years (Vitaro et al. 2004). In addition, at ages 13 and 14, 333 boys who were persistently high or low in teacher-rated aggressiveness or high or low in teacher-rated anxiousness were invited to participate in a laboratory-based study that included tasks designed to measure response perseveration and inability to delay gratification (Vitaro et al. 1999). Gambling outcomes in this study included problem gambling assessed at ages 17 (Vitaro et al. 1997, 2001) and 23 (Wanner et al. 2006) and an empirically derived latent gambling trajectory class membership derived from the dichotomized responses to a single-item assessment of past-year gambling obtained every year from age 11 to 16 (Vitaro et al. 2004;Wanner et al. 2006). Minnesota (Winters and colleagues) Winters and colleagues (Winters, Stinchfield, and Kim 1995; Winters et al. 2002, 2005) conducted a three-wave longitudinal study of a community-based sample of Minnesota adolescents. The study extended across 8 years from adolescence to early adulthood. Participants were recruited from across the state of Minnesota, and were nearly evenly split between those residing in urban and rural areas.They were randomly selected from a list generated by a market research firm of households expected to include an adolescent (based on high school information, state driver and voter registration records, and previous market research). At the first wave of the study, conducted in 1990, telephone interviews were conducted with 702 adolescents aged 15–18 years (Winters, Stinchfield, and Fulkerson 1993a, b). At the second wave of the study, conducted in 1992, telephone interviews were conducted with 532 adolescents and young adults 16–20 years of age. At the third wave of the study, conducted in 1997–1998, a decision was made to identify and select 160 high-risk and 190 low-risk participants for follow-up based on the frequency of their gambling and their scores on the South Oaks Gambling Screen–Revised for Adolescents (SOGS-RA) obtained at the earlier two waves of the study. At the third wave of the study, telephone interviews were conducted with 305 young adults 21–26 years of age. This study has examined a rich array of gambling outcomes, including the frequency of participating in 11 different types of gambling in the past year (and a variety of indicators derived from this assessment), and past-year at-risk and problem gambling based on the SOGS-RA (Winters et al. 1995, 2003) assessed at all

132

Research and Measurement Issues in Gambling Studies

three study waves. Participants were also classified into one of five different developmental gambling groups (“resistor,”“persistor,”“desistor,”“new incidence cases,” and “other”) based on their past-year problem-gambling status (no problem gambling, at-risk gambling, or problem gambling) at the three waves of the study (Winters et al. 2005). Prospective predictors of gambling outcomes at wave 3 in this study were early onset of gambling, parental gambling history, delinquency, regular substance use, psychological distress, poor school performance, and at-risk and problem gambling assessed at wave 1 or 2. In addition to spanning an important developmental transition of the participants—adolescence to early adulthood—this study also obtained measures before and after many of the participants had reached the legal age to gamble in the state of Minnesota (18 years) and before and after the introduction of the Minnesota state lottery in 1990. New York (Barnes and colleagues) Barnes and colleagues (1999, 2002, 2005) included measures of past-year gambling involvement at waves 5 and 6 of their six-wave, 7-year longitudinal study of the development of alcohol misuse. The baseline wave of the study was conducted in 1989, when the participants were 13–16 years of age, and subsequent waves were conducted at yearly intervals. Participants were a representative household sample of adolescents (with an oversampling of blacks) from the Buffalo, New York, metropolitan area who were identified through random digit dialing. Gambling assessments were conducted in 1994–95 and 1996, when the participants were aged 17–21 and 18–22. Gambling outcomes in this study included a composite measure of past-year gambling frequency based on the frequency of participating in 11 different types of gambling assessed at wave 5 (Barnes et al. 1999), a latent gambling construct based on a confirmatory factor analysis of the same 11 gambling frequency items assessed at wave 5 (Barnes et al. 2005), and a rationally derived gambling pattern variable (“flat-low,”“increasing,”“flat-medium,” “flat-high,” and “decreasing”) based on the frequency of gambling at waves 5 and 6 (Barnes et al. 2002). Prospective predictors of gambling involvement at wave 5 (ages 17–21) were self-reported impulsivity and peer delinquency (assessed at ages 15–18), moral disengagement (assessed at ages 16–19), a composite cross-wave selfreport measure of parental monitoring, and sociodemographic characteristics (Barnes et al. 2005). Self-reported impulsivity was also examined as a prospective predictor of the rationally derived gambling pattern (Barnes et al. 2002). Quebec, Canada (Vitaro, Ladouceur, and Bujold) Participants in the prospective gambling study of Vitaro, Ladouceur, and Bujold (1996) were recruited from schools throughout the province of Quebec,

Longitudinal Studies of Gambling Behavior

133

Canada. All of the participants were male, were primarily white (94%), and represented the full range of socioeconomic status. At age 13, 631 boys completed a questionnaire in which they self-reported their past-year frequency of participating in six different categories of gambling activities and the greatest amount of money that they bet in a single day in the past year.These seven items were combined into a single scale for the gambling outcome measure and were also used to identify groups of frequent gamblers (n = 33) and nongamblers (n = 108).Teacher and mother ratings of hyperactivity, impulsivity, aggressivity, and anxiety/withdrawal obtained at ages 10 and 11 were available for 441 of the 631 boys who completed the self-report questionnaire at age 13.These teacher and mother behavior ratings were used as prospective predictors of gambling involvement and gambling group membership at age 13.

LONGITUDINAL STUDIES OF GAMBLING BEHAVIOR INITIATED DURING LATE ADOLESCENCE AND EARLY ADULTHOOD Including the three longitudinal studies reviewed above that continued into late adolescence and early adulthood, and the three studies reviewed below, there are actually six longitudinal studies altogether on gambling behaviors across the critical ages of 18–25. During a critical developmental period in which one expects a great deal of change (such as the ages that span the years when it becomes legal to gamble, or when there are several important milestones), it is especially valuable to be able to study individuals at each specific age, rather than combining them into broader age bands at the analysis stage.Two of the three studies of gambling behavior during late adolescence and early adulthood, the Missouri (Slutske, Jackson, and Sher 2003) and Dunedin (Slutske et al. 2005) studies, were “age homogeneous,” that is, all of the subjects were born within the same year; this is also true of the Montreal study of Vitaro and colleagues.Thus, these three studies are especially sensitive to developmental change. (Note that an age-homogeneous study is not the only available option for conducting sensitive analyses of developmental change. In principle, any large sample can be subdivided into agehomogeneous cohorts.) Missouri College Students (Slutske, Jackson, and Sher) Participants in the gambling study by Slutske, Jackson, and Sher (2003) were 468 first-time freshmen aged 18–19 years who were part of a longitudinal study of the development of alcohol use patterns and associated problems (Sher et al. 1991). Subjects were selected from among all incoming freshmen at the University of Missouri–Columbia based on the presence or absence of alcoholism and associated psychopathology in their first- and second-degree biological relatives.

134

Research and Measurement Issues in Gambling Studies

The subjects were predominantly white (94%), and 213 of the subjects were male (46%). Data from years 1, 4, 7, and 11 were used in the gambling study, which corresponded to ages 18–19, 21–22, 24–25, and 28–29 of the participants. All individuals from the first year of the study were targeted for contact at later years, regardless of their current college enrollment status or place of residence. Gambling outcomes in this study included lifetime problem gambling assessed at all 4 years of the study, past-year problem gambling assessed at years 4, 7, and 11, and lifetime involvement in 10 different gambling activities assessed at year 11. The measures of problem gambling at all 4 years were combined to form developmental trajectories of problem gambling. New York (Barnes and colleagues) Barnes and colleagues (1999, 2002, 2005) included measures of past-year gambling involvement at every wave of their three-wave, 5-year longitudinal study of delinquency in young men. The gambling component of this study was designed to be similar to the other New York study by Barnes and colleagues reviewed above, and the results of the two studies have been published together in the same reports in tandem in order to establish the replicability of the findings. The baseline wave of the study was conducted in 1992, when the participants were 16–19 years of age, and the two subsequent waves were conducted at 1.5-year intervals. Participants were identified through random-digit dialing of households in the Buffalo, New York, metropolitan area with an oversampling of telephone districts with high-crime rates.Two-thirds of the sample was selected because the adolescent male was identified at telephone screening as being at high risk for developing delinquency, and the remaining one-third of the sample was randomly selected. Although gambling involvement was assessed at all three waves, published reports to date have utilized the gambling measures from waves 2 and 3, which were conducted in 1994–1995 and 1996–1997, when the participants were aged 17–21 and 19–22 years. Gambling outcomes and prospective predictors in this study were the same as in the other New York study by Barnes and colleagues reviewed above. Dunedin, New Zealand (Slutske and colleagues) The prospective problem gambling study of Slutske and colleagues (2005) was part of a much larger longitudinal study based in Dunedin, New Zealand.The Dunedin Multidisciplinary Health and Development Study is an ongoing longitudinal investigation of the health and behavior of a complete cohort of consecutive births between April 1, 1972, and March 31, 1973, in Dunedin (Silva and Stanton 1996; Moffitt et al. 2001).The birth cohort was constituted when the participants were 3 years of age. The investigators were able to successfully enroll 91% of the

Longitudinal Studies of Gambling Behavior

135

eligible births, which yielded a cohort size of 1037 children (52% male).The cohort families represent the full range of socioeconomic status in the general population of New Zealand’s South Island and are primarily (~93%) of white European ancestry. Scales from a comprehensive assessment of self-reported personality traits conducted at age 18 were used to predict structured interview–based diagnoses of past-year problem gambling, as well as alcohol, cannabis, and nicotine dependence at age 21.

LONGITUDINAL STUDIES OF GAMBLING BEHAVIOR INITIATED DURING EARLY TO LATE ADULTHOOD Six of the seven longitudinal studies of gambling behavior reviewed above continued into early adulthood. Only the two published longitudinal studies of gambling behavior reviewed below have included participants who were in or beyond their fourth decade of life. New Zealand (Abbott, Williams, and Volberg) Abbott, Williams, and Volberg (1999, 2004) conducted a 7-year follow-up in 1998 of 143 adults who were selected from 4053 participants in the 1991 New Zealand National Prevalence Survey. Individuals in four different subgroups were selected for the follow-up: regular noncontinuous gamblers, regular continuous gamblers, lifetime problem gamblers, and lifetime probable pathological gamblers. Outcome measures included the SOGS-Revised; a 10-item scale based on the criteria from the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV), for pathological gambling; and the frequency of participating in 21 different gambling activities. Prospective predictors of gambling outcomes at the follow-up were the frequency of participation in gambling activities, gambling subgroup classification, gambling problems, self-reported current psychological distress and past-year alcohol problems, and demographic characteristics, all assessed at baseline. U.S. Casino Employees (Shaffer and Hall) The longitudinal gambling study of Shaffer and Hall (2002) represents the only study reviewed in this chapter that focused on a high-risk population—in this case, casino employees. As part of a health survey for an employee assistance and health improvement program, the researchers assessed gambling and alcoholrelated problems via self-report questionnaires. Employees at six different casinos were invited to participate in three surveys conducted at intervals of approximately 1 year. Unfortunately, participation and retention rates in this longitudinal study were very low, in part owing to the large job turnover at the casinos and the

136

Research and Measurement Issues in Gambling Studies

inability to track individuals who were no longer employees.The response rate at the first year was 67% (n = 6067), of which 52% (n = 3174) and 19% (n = 1176) participated at years 2 and 3, respectively; there were 639 individuals who provided problem gambling information at all three years. Gambling outcomes were scores and gambling problem severity categories (i.e., levels 1, 2, and 3) derived from the SOGS, assessed at all 3 years (Lesieur and Blume 1987). Prospective predictors of gambling outcomes were a four-item measure of alcohol-related problems, a oneitem measure of depression, and a variety of demographic, health, risky-behavior, and biomedical variables culled from the larger health survey.

KEY ISSUES AND CHALLENGES IN LONGITUDINAL GAMBLING RESEARCH Given space constraints, here I will briefly draw attention to three important topics for longitudinal gambling research: (1) statistical techniques for modeling stability and change, (2) missing data, and (3) approaches for coping with the low prevalence of pathological gambling disorder.

STATISTICAL TECHNIQUES FOR MODELING STABILITY AND CHANGE Many recent advances have been made in the development and refinement of growth modeling, a general family of techniques for analyzing longitudinal data. With but one exception (Vitaro et al. 2004), growth modeling has not yet made its way into the longitudinal gambling research literature. Growth modeling does not require the same restrictive (and often implausible) assumptions as, and affords a number of advantages over, more traditional analytic approaches for analyzing longitudinal data such as repeated-measures analysis of variance (Gibbons et al, 1993;Tomarken and Waller, 2005). Hierarchical linear modeling (HLM) is a regression-based framework (see Singer and Willett 2003) for simultaneously modeling change over time at both the group and individual levels. HLM analysis of longitudinal data consists of two levels—a level 1 model that estimates (from each individual’s repeated measurements) the initial mean level (intercept) and direction and rate of change over time (slope) for each individual, and a level 2 model that estimates the group level differences in mean levels and change over time. One can think about a two-level longitudinal HLM as a systematic way of conducting individual regressions for each participant based on his repeated measurements and then using the resulting parameters obtained from these individual regressions as the input data for a regression analysis for the entire group. HLM goes beyond this simplified scenario by “borrowing strength” from the

Longitudinal Studies of Gambling Behavior

137

information provided by other participants with similar characteristics to derive a better estimate of the trend for the individual (Gibbons et al. 1993, p. 743). Latent variable growth curve modeling (LGM) is a structural equation modeling–based framework (see Duncan et al. 1999) for simultaneously modeling change over time at both the group and individual levels. LGM is similar to confirmatory factor analysis (Curran and Hussong 2003). The repeated measurements obtained over the course of a longitudinal study are treated as indicators of two correlated latent factors—a mean-level factor (intercept) and a growth structure factor (slope), with fixed factor loadings of the repeated measurements on the latent factors representing the passage of time. In LGM, the latent factor means represent the mean level and growth at the group level, and the latent factor variances represent individual differences around these factor means. HLM and LGM are flexible data-analytic approaches that could potentially be used to test a number of different hypotheses about gambling behavior over time. For example, one could examine (1) whether increases or decreases in gambling involvement with age are linear or nonlinear, (2) whether there is an acceleration in growth in gambling involvement when people reach the legal age to gamble, (3) whether changes in gambling-related problems correspond to changes in alcoholrelated problems, and (4) the predictors, correlates, and sequelae of various aspects of changes in gambling behavior. (See Curran and Hussong [2003] and Muthén and Muthén [2000] for some applications and extensions of growth modeling.)

DEALING WITH MISSING DATA Missing data are inevitable in longitudinal research. Over the course of a longitudinal study, not all of those who participate at earlier waves will participate at later waves (and some of those who participate at later waves may have been unavailable at earlier waves).There are two potential consequences of missing data in a longitudinal study. At the very least, even when the data are missing completely at random, there will be a reduction in sample size and a corresponding loss of statistical power. This can be especially consequential when accumulated over many waves of a longitudinal study. However, data are usually not missing completely at random, and this may be especially true in longitudinal studies of gambling behavior. Individuals with gambling pathology may be more likely to thwart follow-up efforts or to drop out of a longitudinal study compared with individuals without gambling pathology; they may be more likely to be geographically mobile—to move or disconnect their telephones to escape from creditors—or to be incarcerated. For example, Abbott et al. (1999) found that those who felt that they had a gambling problem were less likely to participate in the 7-year follow-up of the New Zealand study than those who did not, and Slutske et al. (2003) showed that problem gambling at the first wave of the Missouri study was a significant predictor

138

Research and Measurement Issues in Gambling Studies

of nonparticipation at subsequent waves conducted 3, 6, and 10 years later. In addition to a reduction in sample size and a loss of statistical power, nonrandom missing data can also lead to biased results and incorrect conclusions (Acock 2005; Raghunathan 2004). Experts argue strongly against the common practice of ignoring missing data at the analysis stage by including only individuals with complete data (i.e., “listwise deletion”). However, it is important to recognize that even the best missing-data techniques are not panaceas and cannot completely correct all of the biases that can result from missing data. The appropriate handling of missing data is one area of statistics in which there appears to be a general consensus among experts about the relative merits of different strategies (e.g., Acock 2005; Raghunathan 2004; Schafer and Graham 2002; Tomarken and Waller 2005). Three approaches are especially promising for dealing with missing data in the context of a longitudinal study: (1) weighting, (2) multiple imputation, and (3) full information maximum likelihood. Weighting (Raghunathan 2004) is commonly used in large-scale crosssectional epidemiologic surveys when the characteristics of the target population are known. One can compensate for the underrepresentation of certain groups in the obtained sample by assigning a sampling weight to each observation that reflects the probability of the groups’ inclusion in the sample—observations from undersampled groups will be weighted more heavily and those from oversampled groups will be weighted less heavily.The same logic can be applied to account for individuals who drop out of a longitudinal study. Those individuals from groups who are underrepresented at follow-up based on characteristics at the baseline assessment can be weighted more heavily. Weighting can also be used when data are missing by design, such as when a two-stage sampling design is used. Multiple imputation (Acock 2005; Schafer and Graham 2002;Tomarken and Waller 2005) is an extension of single imputation in which missing values in a data set are imputed based on the available nonmissing data. In multiple imputation, this process is repeated in order to generate 5–20 imputed data sets. Data analyses are conducted on each of these imputed data sets, and then pooled estimates of the parameters and standard errors are derived from the set of solutions. Multiple imputation is preferred over single imputation because the degree of uncertainty in estimating the imputed data points is taken into account in the statistical analyses. Full information maximum likelihood (Little and Rubin 2002) is generally considered the preferred approach to handling missing data.This approach does not fill in missing data, but rather uses all available data to derive proper parameter estimates.This missing-data technique is not available for all software packages and statistical routines, but when it is available, its application requires very little effort on the part of the end user. Full information maximum likelihood estimation can be found in all major structural equation modeling and HLM software packages (Acock 2005).

Longitudinal Studies of Gambling Behavior

139

THE LOW PREVALENCE OF PATHOLOGICAL GAMBLING DISORDER In a recent U.S. national survey of 43,093 adults, the number of individuals identified with past-year and lifetime DSM-IV diagnoses of pathological gambling disorder was only 74 and 185, respectively (Slutske 2006).As it is currently defined, pathological gambling disorder is no more prevalent than schizophrenia or anorexia nervosa. One of the greatest challenges to conducting systematically ascertained community-based longitudinal studies of pathological gambling disorder is the fact that it is not very common. Here I summarize two approaches to dealing with this problem. One approach for dealing with the low base rate of pathological gambling disorder is to employ strategies for developing enriched samples. High-risk prospective studies were originally proposed in response to the challenge of prospectively predicting the onset of another rare disorder, schizophrenia (Mednick and McNeil 1968). Typically, unaffected individuals have been deemed at high risk based on a history of the same disorder in their parents. Although this approach has not yet been used in studies of pathological gambling, a related approach was used in the recently completed follow-up to the Vietnam Era Twin study (Eisen et al. 1998; Slutske et al. 2000). In this study, unaffected identical and fraternal twins of men with pathological gambling disorder were selected for the 10-year follow-up with the assumption that they were at high risk for the development of pathological gambling. A strategy for maximizing the number of individuals in a sample who already are experiencing problems is to use a two-stage sampling design. The first stage involves the initial screening of a larger pool of potential participants, and the second stage involves more intensive study of the subsample consisting of those who were identified at the first stage as already having experienced problems.With the parallel collection of data from a randomly selected unaffected control sample, one can use missing-data techniques such as weighting to obtain estimates based on the larger representative stage 1 sample.A possible solution to the low base rate problem in systematically ascertained community-based research on pathological gambling disorder might be to pool data across several research sites. Another approach for dealing with the low base rate of pathological gambling disorder is to focus on continuous measures of gambling pathology and on subclinical problem gambling.There has been a recent general trend in psychopathology research to use continuous, rather than categorical, measures of pathology.This has two clear benefits for longitudinal studies of gambling pathology—a gain in statistical power and provision of a more sensitive metric of intra-individual change. Although typically used to derive categorical outcomes, all of the commonly used measures of problem and pathological gambling, such as the SOGS, can also be used to derive continuous outcome measures (Lesieur and Blume 1987).

140

Research and Measurement Issues in Gambling Studies

There are a number of reasons to focus on a broad range of gambling pathology, including subclinical problem gambling, rather than restricting the focus of research to diagnosable pathological gambling disorder: (1) Subclinical gambling pathology is still clinically significant (Slutske et al. 2003); (2) there is no solid empirical evidence supporting the current diagnostic threshold, and studies that include the entire range of gambling pathology, including subclinical problem gambling, will provide the evidence needed to determine where the diagnostic threshold should be drawn; (3) problem gambling may be an important developmental precursor in the downward spiral leading to pathological gambling disorder; and (4) if problem and pathological gambling truly represent different points on an underlying continuum of liability, then studies that identify risk factors for problem gambling will also further our understanding of the risk factors for pathological gambling disorder because the only difference between the two outcomes would be that more (but not different) risk factors would be required to develop pathological gambling disorder than problem gambling. Although we may never know whether problem and pathological gambling truly represent different points on an underlying continuum of liability, it is likely that the relevant risk factors for problem and pathological gambling substantially overlap.

IMPORTANT QUESTIONS AND WHAT WE KNOW SO FAR In this section I focus on the following five important issues that can be resolved effectively only with longitudinal or follow-up data: (1) resolving the temporal relation between gambling behavior and its correlates, (2) establishing the stability of gambling behavior, (3) characterizing the course of gambling behavior, (4) identifying sequences and stages in the development and progression of gambling behavior, and (5) understanding age differences in levels and rates of gambling behavior. To date, the main focus of nearly all of the existing longitudinal studies of gambling behavior has been on the first and third questions. I will also discuss how longitudinal data can improve the strength of hypothesis testing in the context of natural experiments.

TEMPORAL RESOLUTION OF GAMBLING CORRELATES: ESTABLISHING CAUSALITY? Longitudinal data are essential for establishing the temporal relation that is necessary for inferring a cause-and-effect association. For example, are crime-prone individuals more likely to become involved in gambling activities, or do gambling activities lead to criminal behavior in some individuals?

Longitudinal Studies of Gambling Behavior

141

Is socioeconomic disadvantage a cause or a consequence of problem gambling? When considered in isolation, correlational longitudinal data are insufficient for answering such causal questions (even when one is conducting “causal modeling” analyses) (Cliff 1983) and are best thought of as one piece of a puzzle. Although correlational longitudinal data cannot definitively prove causal relationships, they can rule out some possibilities based on a temporal relation that is inconsistent with a particular direction of causation. There are essentially two ways to establish temporal precedence of one construct over another. The most intuitively appealing is to study individuals early enough so that one or both behaviors of interest have not yet developed. In the Montreal study by Vitaro and colleagues (2004), a substantial fraction of boys were already gambling by age 11, so one would need to initiate a study even earlier than this to prospectively predict gambling involvement using this strategy. Most existing longitudinal studies of gambling behavior have not studied individuals prior to the potential development of the gambling outcomes of interest. For example, in the Dunedin birth cohort study (Slutske et al. 2005), the higher-order personality dimensions of negative emotionality (including lowerorder scale indicators of nervousness or worry, anger or aggressiveness, and feeling mistreated or victimized) and (low) behavioral constraint (including lower-order scale indicators of risk taking, impulsivity, and rebelliousness) assessed at age 18 were prospectively associated with problem gambling at age 21. Even though pastyear problem gambling was assessed as the outcome of interest at age 21, some individuals may have had problems at or before age 18, and some aspects of personality, particularly those related to negative emotionality, could conceivably be a consequence of these earlier gambling problems, rather than a potential cause of problems at age 21. This possibility could not be empirically evaluated because problem gambling was not assessed at age 18. Although such prospective associations add an important piece to the “causality puzzle,” they can be interpreted in several different ways. For instance, an association between personality at baseline and problem gambling at follow-up might be partially due to the cross-sectional association of personality with (unmeasured) problem gambling at baseline, and the temporal stability of problem gambling from baseline to follow-up. Therefore, a second way that researchers might attempt to establish the temporal precedence of one construct over another is through the use of statistical controls. Measures of the outcome variable (in this example, problem gambling) would also be included at baseline so that the effect of problem gambling at baseline on the association between personality at baseline and problem gambling at follow-up can be statistically controlled. A related technique that has been used to disentangle the direction of causality of two correlated constructs is cross-lagged panel correlation (CLPC) analysis (Kenny 1979). In CLPC analysis, both constructs of interest are each measured over two or more occasions.The premise of CLPC analysis is to compare the strength

142

Research and Measurement Issues in Gambling Studies

of the association between, for example, personality at time 1 and problem gambling at time 2 with problem gambling at time 1 and personality at time 2. The rationale is that if the former association is stronger than the latter (i.e., personality → problem gambling > problem gambling → personality), then the correlation between personality and problem gambling can be predominantly attributed to the prospective association between personality and problem gambling. Conversely, if the reverse is true (i.e., problem gambling → personality > personality → problem gambling), then the correlation between personality and problem gambling can be attributed predominantly to the prospective association between problem gambling and personality. If the associations do not differ from each other, then the correlation between personality and problem gambling is “spurious” (i.e., due to the effect of one or more common variables that causes them both) (Campbell and Kenny 1999). CLPC analysis can be a potentially useful exploratory data-analytic technique (Campbell and Kenny 1999; Kenny 1979; Rutter 1981). For example, once replicable cross-sectional and longitudinal relationships between a putative cause and effect are established, one might then examine the difference in the crosslagged associations to see if they are consistent with a causal hypothesis before proceeding to more sophisticated analyses (Campbell and Kenny 1999). There are a number of problems that limit the interpretability of CLPC analyses of manifest (rather than latent) variables (Campbell and Kenny 1999; Rogosa 1980). For example, different implementations of CLPC assume that, for example, problem gambling and personality are measured with equal reliability at both time points, or that individual differences in problem gambling and personality are equally stable across time. Many of these problems can be circumvented by examining the cross-lagged associations between latent variables estimated within the context of a structural equation model, and this is now the preferred approach to conducting such analyses (e.g., see Sher et al. 1996; Sher and Wood 1997; for a nontechnical general review of structural equation modeling, see Loehlin 1998).

THE STABILITY

OF

GAMBLING BEHAVIOR

Developmental psychologists and statisticians draw distinctions between at least three different types of trait stability—mean level stability, stability in individual differences (“rank-order” or “inter-individual stability”), and intra-individual stability (see e.g., Caspi and Roberts 2001; Roberts and DelVecchio 2000; Rogosa 1995)—that are relevant to understanding gambling behavior. Mean level stability refers to the aggregate stability over time in the average level of a trait (in the case of continuous outcomes) or the percentage of individuals possessing a trait (in the case of dichotomous categorical outcomes) and is assessed by examining group means or prevalences over time. In the context of a longitudinal study of a

Longitudinal Studies of Gambling Behavior

143

relatively age-homogeneous sample, mean level stability provides insight into developmental changes in a gambling outcome, and in the context of a natural experiment, mean level stability can reflect the effects of an important historical change. Three existing longitudinal studies present evidence relevant to the issue of mean level stability of gambling behavior across development. In the Montreal study (Vitaro et al. 2004), the prevalence of gambling at least once in the past year steadily increased among boys from age 11 until age 15, but then appeared to level off from age 15 to 16 (actual prevalences by age were not reported). In the Minnesota study (Winters et al. 2002), the prevalences of any gambling (80–88%) and regular gambling (13–18%) did not significantly differ across the three waves of the study conducted at ages 15–18, 16–20, and 21–26. This apparent mean level stability, however, masked considerable variability when individual gambling activities were considered. Involvement in unregulated forms of gambling (i.e., cards, betting on games of personal skill or on sports) significantly declined, whereas involvement in regulated forms of gambling (i.e., scratch tabs, gambling machines, and lottery) significantly increased over this developmental (and historical) period. Although there were no differences in the prevalences of problem gambling (2–4%), there was a significant increase in the prevalence of at-risk gambling from waves 1 and 2 to wave 3 of the study (12–15% vs 21%). In the Missouri study (Slutske et al. 2003), the past-year prevalences of problem gambling did not significantly differ from ages 21–22 (3%), 24–25 (3%), or 28–29 (2%). Stability in individual differences refers to the retention of an individual’s rank within a group over time and is assessed by correlating the same repeated measures of a trait obtained from a group of individuals on two separate occasions. As mentioned in the previous section, a prospective association between a correlated baseline characteristic and a gambling outcome actually may be (at least partially) due to the inter-individual stability of the gambling outcome itself. Inter-individual stability inevitably diminishes over time for psychological traits, and based on the scant evidence available, this also appears to be true for problem gambling. Only two longitudinal studies have reported the inter-individual stabilities of gambling behavior (Slutske et al. 2003;Vitaro et al. 2001), but stabilities were also calculable from the published data of another study (Winters et al. 2005). The inter-individual stabilities of problem gambling across 2 to 10 years from the Missouri and Minnesota studies (Slutske et al. 2003;Winters et al. 2005) are shown in Figure 6.1. Note that, as expected, there is a substantial correlation of −0.50 between the length of the test–retest interval and the inter-individual stability (estimated by the tetrachoric correlation). Overall, there appears to be substantial stability of individual differences in problem gambling, with an average test–retest correlation across 2 to 10 years of 0.53. (Interestingly, results from the Montreal study [Vitaro et al. 2001] are inconsistent with the Missouri and Minnesota studies [Slutske et al. 2003;Winters et al. 2005]. In the Montreal study,

144

Research and Measurement Issues in Gambling Studies

the inter-individual stability of problem gambling across 1 year from age 16 to 17 was only 0.20, but this may have been due to the fact that an abbreviated threeitem version of the SOGS-RA was used at age 16 and the full twelve-item scale was used at age 17 [Vitaro et al. 2001].) In the context of a longitudinal study of a relatively age-homogeneous sample, the patterns of inter-individual stabilities over time can provide clues to the causes of individual differences in gambling outcomes at different ages or phases of life (Cole 2006; Fraley and Roberts 2005). For example, Cole (2006, p. 22) argued that the “fact that stability estimates tend to diminish as the time interval increases suggests the influence of processes that are somewhat transitory.”The fact that the stabilities for problem gambling are greater than zero also suggests the influence of “something enduring or traitlike” (ibid.), at least over the short time span of 2 to 10 years covered in these two studies.What we don’t know yet are the longerterm inter-individual stability of problem gambling and whether the curve in Figure 6.1 will eventually reach an asymptote greater than zero or the stabilities will continue to diminish to zero.The former curve (i.e., one that asymptotes at > 0) would lead to the conclusion that individual differences in problem gambling are enduring over the life course and that problem gambling later in life can be predicted from problem gambling earlier in life. The latter curve (i.e., one that approaches 0) would lead to the conclusion that individual differences in problem gambling do not endure over the life course and that problem gambling later in life cannot be predicted very well from problem gambling earlier in life.

1

Test−retest correlation

0.9

MN 0.82

0.8

r = −50

MN 0.74

0.7

MO 0.66

0.6

MO 0.51

0.5

MO 0.57

MN 0.54

0.4

MO 0.37

0.3

MO 0.35

MO 0.2

0.2 0.1 0 0

2

4

6

8

10

12

Years between assessments

Figure 6.1. Stability of individual differences in problem gambling as a function of time from two longitudinal studies. (Data points from Slutske, Jackson, and Sher [2003] are denoted by a square; data points calculated from Winters et al. [2005] are denoted by a diamond.)

145

Longitudinal Studies of Gambling Behavior

Intra-individual stability refers to the consistency within an individual of the amount or level of a trait (for continuous outcomes) or of belonging in a particular category (in the case of categorical outcomes). Thus, studies that focus on the course of gambling behaviors and gambling problems are studies of intra-individual stability.

THE COURSE

OF

GAMBLING BEHAVIOR PROBLEMS

AND

GAMBLING

There is now consistent evidence emerging from longitudinal research suggesting that the course of problem gambling is variable—for some individuals the gambling problems are relatively transient, whereas for others the gambling problems are persisting and chronic (Slutske 2006).This is an important distinction that has been overlooked until recently because it was assumed that disordered gambling behavior was invariably progressive and chronic. The first compelling evidence came from the 7-year follow-up of the 1991 New Zealand National Prevalence Survey (Abbott et al. 1999). The New Zealand researchers found that of the 13 individuals who had a past-6-month diagnosis of probable pathological gambling in 1991, only 3 (23%) had a similar diagnosis in 1998; of the 22 individuals with a past-6-month diagnosis of problem gambling in 1991, only 3 (14%) had escalated to probable pathological gambling, and only 2 (9%) had a problem gambling diagnosis in 1998. (Somewhat surprisingly, this study also obtained mean level differences, that is, an unexpected decrease in the prevalence of current disordered gambling behavior over time in this age-heterogeneous cohort.) The Minnesota (Winters et al. 2005) and Missouri (Slutske et al. 2003) studies tracked the course of problem gambling over 8 and 11 years, respectively, among individuals ranging in age from 15 to 29 years. Both studies were consistent in demonstrating high mean level stability, moderate inter-individual stability, but low intra-individual stability of problem gambling across this age range. The intra-individual stability of problem gambling from these two studies is depicted in Figure 6.2.The figure illustrates that the persistence of past-year problem gambling across all three time points in these two studies was exceedingly rare, persistence across two consecutive time points was also uncommon, and at each time point the numbers of resistant cases and incident cases often equaled or outnumbered the number of persistent cases. Several studies have examined the course of nonpathological gambling behavior. In the Montreal study of Vitaro and colleagues (2004; Wanner et al. 2006), participants were categorized according to their pattern of gambling at least once in the past year based on assessments conducted annually from age 11 to 16. Three trajectory groups were empirically identified using discrete mixture modeling (Nagin 1999). Sixty-two percent of the boys were assigned to a

146

(a) Winters et al.,

Research and Measurement Issues in Gambling Studies

2005

T1 in 1992 mean age = 16.0 years

PG 7

T2 in 1994 mean age = 17.6 years T3 in 1997−1998 mean age = 23.8 years

No PG 298

PG 4 PG 2

No PG 3

No PG 2

PG 0

persistent

No PG 3

PG 12 PG 5

No PG 286 PG 5

No PG 281

incident

resistant

No PG 7

desistant

(b) Slutske et al., 2003 Year 4 in 1990−1991 age = 21−22 years

PG 11

Year 7 in 1993−1994 age = 24−25 years Year 11 in 1998−1999 age = 28−29 years

PG 4 PG 1 persistent

No PG 3

No PG 377 No PG 7 PG 0

No PG 7 desistant

PG 8 PG 1

No PG 369

No PG 7

PG 5

incident

resistant

No PG 364

Figure 6.2. Course of past-year problem gambling across three time points from two longitudinal studies. (Note: PG, problem gambling; data from Winters et al. [2005] and Slutske, Jackson, and Sher [2003].)

group that was unlikely to gamble at any given age, 22% of the boys were assigned to a group that was likely to gamble at any given age, and 16% of the boys were assigned to a group that was unlikely to gamble at ages 11, 12, and 13, but likely to gamble at ages 14, 15, and 16; there were no trajectory groups identified in which the probability of gambling decreased from age 11 to 16. In the two parallel New York studies of Barnes and colleagues (Barnes et al. 2002), participants were categorized according to their pattern of gambling frequency across two study occasions that occurred 12 to 18 months apart when participants were 17–21 and 18–22 years of age, respectively (essentially picking up at the age where the Montreal study left off). The New York studies (Barnes et al. 2002) identified a subgroup of young adults (16–26%) who decreased the frequency of their gambling.

SEQUENTIAL/STAGE THEORIES OF GAMBLING INVOLVEMENT: IS THERE A “GATEWAY” TO PROBLEMS? Longitudinal studies of gambling behavior can provide interesting insights into the typical developmental sequence of participating in different gambling activities in much the same way that longitudinal substance use research has yielded important insights into the typical sequence of the use of different

Longitudinal Studies of Gambling Behavior

147

substances. An influential paper (widely cited as evidence that marijuana is the “gateway” to the use of other illicit drugs) showed that there was a typical sequence of stages of substance involvement among adolescents, starting with no use, to the use of beer or wine, to the use of hard liquor or cigarettes, to the use of marijuana, to the use of other illicit drugs (Kandel 1975).The identification of these stages was based on the observation that there were very few individuals who used a substance at a later stage who hadn’t previously used substances from the earlier stage(s). The cross-sectional 1999 U.K. gambling prevalence survey of 7680 adolescents and adults (Sproston, Erens, and Orford 2000) showed a pattern of involvement in various gambling activities that suggested a similar stage-like process that might be tested in a longitudinal study by examining the initiation of new activities over time. Participants from the U.K. study were classified into groups according to whether they had participated in one, two, three, four, five, or six or more different gambling activities in the past year. With each additional activity, all of the activities of the less involved groups were endorsed by the majority of the individuals in the more involved groups.That is, gambling activities appeared to conform to a Guttman scale progressing from participation in the national lottery, to scratch cards, to gambling machines, to private betting with friends, to horse/dog races. In the United Kingdom, lottery playing was less strongly associated with problem and pathological gambling than was betting on dog races because it was an activity that was ubiquitous across all levels of gambling involvement. Betting on dog races was an activity that was associated with an overall greater level of gambling involvement in the United Kingdom, and so was also more indicative of problem and pathological gambling. One could also examine the typical sequence in the development of symptoms of problem and pathological gambling in much the same way that symptoms of alcohol dependence have been examined. For example, Toce-Gerstein, Gerstein, and Volberg (2003) presented results from a cross-sectional 1999 U.S. prevalence survey in which they examined the frequency of endorsement of each pathological gambling symptom among those who varied in the severity of their gambling pathology based on the total numbers of lifetime symptoms they endorsed. Toce-Gerstein et al. (2003) identified the following four sets of symptoms characteristic of increasing levels of severity of gambling pathology (from least to most severe): (1) chasing losses; (2) preoccupation, lying about gambling, and gambling to escape; (3) withdrawal, loss of control, tolerance, bailout, and risked relationships; and (4) illegal acts. This ordering of symptoms might also reflect a temporal sequence that could be tested in a future longitudinal study by examining the development of new symptoms over time. In the absence of longitudinal data, a cross-sectional study with retrospectively reported ages of onset of pathological gambling symptoms might be a valuable first step (e.g., see Nelson et al. 1996).

148

Research and Measurement Issues in Gambling Studies

DEVELOPMENTAL CHANGES VERSUS COHORT OR PERIOD EFFECTS ON LEVELS OF GAMBLING INVOLVEMENT In cross-sectional epidemiologic surveys, the prevalence of problem gambling varies with age, with higher prevalence estimates typically obtained for younger than older individuals. For example, in the 1999 U.K. gambling prevalence survey (Sproston et al. 2000), there was a strong negative association (r = −0.98) between age group and past-year problem gambling, with estimates based on the SOGS ranging from 1.7%, 1.2%, 0.8%, 0.7%, 0.5%, and 0.1% among 16–24 year-olds, 25–34 year-olds, 35–44 year-olds, 45–54 year-olds, 55–64 year-olds, and those aged 65 years and older, respectively. Similarly, in a meta-analysis of 119 North American cross-sectional prevalence surveys of problem and pathological gambling, both lifetime and past-year estimates were substantially higher for adolescents than for adults (Shaffer, Hall, and Vander Bilt, 1999).What is unclear from these cross-sectional data is whether this reflects a developmental trend or an effect of different rates of problem gambling among individuals from different birth cohorts. More recently born individuals may be at greater risk for developing gambling-related problems; alternatively, this prevalence difference may be due to a developmental effect, such that gambling-related problems are at their peak during adolescence and young adulthood, with a gradual maturing-out of problems with age. There are limited longitudinal data on mean level stability to answer this important question, and the existing studies do not span the entire life course. In the Minnesota longitudinal study of Winters and colleagues (2002), the prevalence of problem gambling was unchanged from adolescence to young adulthood, although there was a significant increase during the young adult years of at-risk gambling that appeared to coincide with an increase in participation in regulated gambling activities. In the Missouri study of Slutske and colleagues (2003), the lifetime and past-year rates of problem gambling did not differ from age 18 to 29. Thus, the results from longitudinal research appears to contradict the results of the cross-sectional research by suggesting that the prevalence of problem gambling actually may be relatively stable from adolescence to early adulthood or may actually be higher among young adults than among adolescents. Taken together, the evidence from the cross-sectional and longitudinal research, coupled with the results of a “cross-temporal” meta-analysis of crosssectional prevalence surveys conducted between 1977 and 1997 indicating higher prevalences of disordered gambling behavior obtained in more recent studies than in earlier studies (Shaffer et al. 1999), suggests that age-related differences in prevalences of problem and pathological gambling may be due, at least in part, to a cohort effect or an interaction between a developmental and a period effect (i.e., effects of historical changes only influencing those at certain critical ages; see Menard 2002; Rice, Moldin, and Neuman 1991), rather than simply reflecting developmental changes.

Longitudinal Studies of Gambling Behavior

149

“NATURAL EXPERIMENTS” IN LONGITUDINAL GAMBLING RESEARCH A longitudinal study can potentially provide a sensitive test (i.e., using a within-subjects design that is unconfounded by selection factors, rather than a between-subjects design) of the effect of a new gambling-related legislation or policy or other macro-environmental event or change. In effect, one has a natural experiment comparing the same individuals before and after such an event or change has occurred. Such events or changes cannot always be anticipated, but ongoing longitudinal studies may be well positioned to seize a unique opportunity in assessing their impact.A nice example examining the possible causal connection between socioeconomic disadvantage and psychiatric symptomatology comes from an ongoing 8-year community-based longitudinal study of mental illness among urban and rural youth recruited from eleven counties in western North Carolina (“the Great Smoky Mountains Study,” Costello et al. 2003). Some of the youth were American Indians living on a federal reservation that extended into this region. In year 4 of the longitudinal study, tribal members began to receive income from a gambling casino that opened on the reservation, and there was also an increase in the number of jobs available in the newly opened casino and surrounding businesses.This new source of income raised 14% of the American Indian families from below to above the poverty line and corresponded to a reduction in externalizing psychiatric symptoms of these previously poor youth. Another ingenious use of this same longitudinal study comes from a scheduled interview that was planned to occur during the year 2001.Two-thirds of the sample were interviewed prior to the terrorist attacks on the United States on September 11, and the remaining one-third were interviewed after 9/11.The investigators examined (this time, using a between-subjects design) the effect of the September 11 attacks on reports of current substance use and psychiatric symptomatology among 19–21 year-olds (Costello et al. 2004). The only longitudinal study that has examined the effect of a change in legislation or policy on subsequent gambling behavior is the Minnesota study of Winters and colleagues (1995), who examined the effect of the introduction of lottery games into the state of Minnesota. They were introduced in two stages— instant scratch tabs first and then, 4 months later, online or number selection lottery games. The baseline survey of the Minnesota study, conducted when the participants were 15–18 years of age, occurred prior to the second stage, but due to administrative delays, after the first stage; the follow-up was conducted 1.5 years later—well after all of the stages of the introduction of the lottery were complete. Mean level wave 1/wave 2 comparisons in gambling outcomes were made for two groups: those who were of legal age (18+ years) for at least half of the interval between wave 1 and wave 2 and those who were of legal age for less than half of this interval.The legal age subsample reported a greater frequency of participating

150

Research and Measurement Issues in Gambling Studies

in regulated forms of gambling (casino machines, scratch tabs, and lottery) at wave 2 than at wave 1, whereas there was no such increase in the underage subsample. However, there were no mean-level increases in regular, at-risk, or problem gambling from wave 1 to wave 2 for either the legal-age or underage subsamples.These results were “reassuring to public health officials who were concerned that the onset of the state’s high-stakes and heavily promoted lottery would trigger a significant increase in the rate of problem gambling among youth” (Winters et al. 1995, p. 178).

SUMMARY OF WHAT WE DON’T KNOW (YET) Although the first decade of community-based longitudinal gambling research has led to a number of important insights about gambling behavior, a review of the research conducted to date highlights a number of areas in which there are gaps in our knowledge. ●

●

●

●

●

●

Longitudinal studies are just beginning to disentangle the temporal and potential causal relations between gambling behavior and important correlates. Establishing prospective prediction is an important first step but must eventually be followed by more probing analyses that can rule out alternative hypotheses. We don’t yet know the longer-term mean level, inter-individual, and intra-individual stability of gambling behavior. There are major spans of the life course and segments of the population that have not yet been the subjects of developmentally sensitive longitudinal gambling research.We know virtually nothing about the predictors, stability, and course of gambling behaviors beyond the third decade of life. We also know virtually nothing about the predictors, stability, and course of gambling behaviors during childhood and early adolescence among girls. We don’t yet know the factors that predict individual differences in the desistance from or maintenance or escalation of gambling behavior. Such questions may require the use of more contemporary statistical techniques such as growth modeling. Longitudinal research has not yet tackled important issues related to the typical sequence in which different gambling activities are initiated and whether individual differences in this pattern are related to gambling outcomes. Nor has longitudinal research examined the typical progression of problem and pathological gambling symptom development. We have not yet capitalized on the potential power of longitudinal research for examining the effect on gambling behavior of changes in gambling legislation or policy or other macro-environmental events.

Longitudinal Studies of Gambling Behavior

●

151

Finally, nearly all of the longitudinal gambling research has been focused on nonpathological gambling involvement or subclinical problem gambling.There is very little information currently available from communitybased longitudinal research on the predictors, stability, and course of diagnosable pathological gambling disorder.

GLOSSARY Cohort effects an interaction between a developmental and time-period effect in which the effects of historical events uniquely influence those at certain critical ages rather than simply reflecting developmental changes. Critical developmental period a period in which one expects a great deal of change (such as the ages that span the years when it becomes legal to gamble, or when there are several important milestones). Intra-individual stability the consistency within an individual of the amount or level of a trait (for continuous outcomes) or of belonging in a particular category (in the case of categorical outcomes). Longitudinal study study of individuals from a systematically ascertained or representative community-based sample who are assessed on at least two separate occasions across an interval of at least 1 year. Mean level stability the aggregate stability over time in the average level of a trait (in the case of continuous outcomes) or the percentage of individuals possessing a trait (in the case of dichotomous categorical outcomes). It is assessed by examining group means or prevalences over time. Natural experiment a study of the effect of a new gambling-related legislation or policy or other macro-environmental event or change by comparing the same individuals before and after such an event or change has occurred. Stability in individual differences the retention of an individual’s rank within a group over time assessed by correlating the same repeated measures of a trait obtained from a group of individuals on two separate occasions.

REFERENCES Abbott, M.W.,Williams, M., and Volberg, R.A. (1999). Seven Years On:A Follow-Up Study of Frequent and Problem Gamblers Living in the Community.Wellington, NZ: Department of Internal Affairs. —— . (2004). A prospective study of problem and regular nonproblem gamblers living in the community. Substance Use and Misuse, 39, 855–884. Acock, A. C. (2005).Working with missing values. Journal of Marriage and Family, 67, 1012–1028.

152

Research and Measurement Issues in Gambling Studies

Barnes, G. M.,Welte, J.W., Hoffman, J. H., and Dintcheff, B.A. (1999). Gambling and alcohol use among youth: Influences of demographic, socialization, and individual factors. Addictive Behaviors, 24, 749–767. —— . (2002). Effects of alcohol misuse on gambling patterns in youth. Journal of Studies on Alcohol, 63, 767–775. —— . (2005). Shared predictors of youthful gambling, substance use, and delinquency. Psychology of Addictive Behaviors, 19, 165–174. Campbell, D.T., and Kenny, D. A. (1999). A Primer on Regression Artifacts. New York: Guilford Press. Caspi, A., and Roberts, B.W. (2001). Personality development across the life course:The argument for change and continuity. Psychological Inquiry, 12, 49–66. Cliff, N. (1983). Some cautions concerning the application of causal modeling methods. Multivariate Behavioral Research, 18, 115–126. Cole, D. A. (2006). Coping with longitudinal data in research on developmental psychopathology. International Journal of Behavioral Development, 30, 20–25. Costello, E. J., Compton, S. N., Keeler, G., and Angold, A. (2003). Relationships between poverty and psychopathology: A natural experiment. Journal of the American Medical Association, 290, 2023–2029. Costello, E. J., Erkanli, A., Keeler, G., and Angold, A. (2004). Distant trauma: A prospective study of September 11th on young adults in North Carolina. Applied Developmental Science, 8, 211–220. Curran, P. J., and Hussong, A. M. (2003). The use of latent trajectory models in psychopathology research. Journal of Abnormal Psychology, 112, 526–544. Duncan, T. E., Duncan, S. C., Strycker, L. A., Li, F., and Alpert, A. (1999). An Introduction to Latent Variable Growth Curve Modeling: Concepts, Issues, and Applications. Mahwah, NJ: Lawrence Erlbaum Associates. Eisen, S.A., Lin, N., Lyons, M. J., Scherrer, J., Griffith, K.,True,W. R., Goldberg, J.,Tsuang, M.T. (1998). Familial influences on gambling behavior:An analysis of 3,359 twin pairs. Addiction, 93, 1375–1384. Fraley, R. C., and Roberts, B. W. (2005). Patterns of continuity: A dynamic model for conceptualizing the stability of individual differences in psychological constructs across the life course. Psychological Review, 112, 60–74. Gibbons, R. D., Hedeker, D., Elkin, I.,Waternaux, C., Kraemer, H. C., Greenhouse, J. B., Shea,T., Imber, S. D., Sotsky, S. M., and Watkins, J.T. (1993). Some conceptual and statistical issues in analysis of longitudinal psychiatric data: Application to the NIMH Treatment of Depression Collaborative Research Program dataset. Archives of General Psychiatry, 50, 739–750. Kandel, D. (1975). Stages in adolescent involvement in drug use. Science, 190, 912–914. Kenny, D. A. (1979). Correlation and Causality. New York: John Wiley & Sons. Lesieur, H. R., and Blume, S. B. (1987). The South Oaks Gambling Screen (SOGS): A new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, 1184–1188. Little, J. R., and Rubin, D. (2002). Statistical Analysis with Missing Data, 2nd ed. New York:Wiley. Loehlin, J. C. (1998). Latent Variable Models: An Introduction to Factor, Path, and Structural Analysis, 3rd ed. Mahwah, NJ: Lawrence Erlbaum. Mednick, S. A., and McNeil,T. F. (1968). Current methodology in research on the etiology of schizophrenia: Serious difficulties which suggest the use of the high-risk group method. Psychological Bulletin, 70, 681–693. Menard, S. (2002). Longitudinal Research.Thousand Oaks, CA: Sage. Moffitt, T. E., Caspi, A., Rutter, M., and Silva, P. A. (2001). Sex Differences in Antisocial Behaviour: Conduct Disorder, Delinquency, and Violence in the Dunedin Longitudinal Study. Cambridge, UK: Cambridge University Press. Muthén, B. O., and Muthén, L. K. (2000). Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcoholism: Clinical and Experimental Research, 24, 882–891.

Longitudinal Studies of Gambling Behavior

153

Nagin, D. S. (1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139–157. Nelson, C. B., Little, R. J. A., Heath, A. C., and Kessler, R. C. (1996). Patterns of DSM-III-R alcohol dependence symptom progression in a general population survey. Psychological Medicine, 26, 449–460. Raghunathan,T. E. (2004).What do we do with missing data? Some options for the analysis of incomplete data. Annual Review of Public Health, 25, 99–117. Rice, J. P., Moldin, S. O., and Neuman, R. (1991). Age, period, and cohort effects on rates of mental disorders. In Genetic Issues in Psychosocial Epidemiology (M.T.Tsuang, K. S. Kendler, and M. J. Lyons, eds.). New Brunswick, NJ: Rutgers University Press. Roberts, B.W., and DelVecchio,W. F. (2000).The rank-order consistency of personality traits from childhood to old age: A quantitative review of longitudinal studies. Psychological Bulletin, 126, 3–25. Rogosa, D. (1980). A critique of cross-lagged correlation. Psychological Bulletin, 88, 245–258. —— . (1995). Myths and methods: “Myths about longitudinal research” plus supplemental questions. In The Analysis of Change (J. M. Gottman, ed.). Mahwah, NJ: Lawrence Erlbaum Associates. Rutter, M. (1981). Longitudinal studies: a psychiatric perspective. In Prospective Longitudinal Research:An Empirical Basis for the Primary Prevention of Psychosocial Disorders (S. A. Mednick, A. E. Baert, and B. P. Bachmann, eds.). Oxford, UK: Oxford University Press. Schafer, J. L., and Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147–177. Shaffer, H. J., Hall, M. N., and Vander Bilt, J. (1999). Estimating the prevalence of disordered gambling behavior in the United States and Canada:A research synthesis. American Journal of Public Health, 89, 1369–1376. Shaffer, H. J., and Hall, M. N. (2002). The natural history of gambling and drinking problems among casino employees. Journal of Social Psychology, 142, 405–424. Sher, K. J.,Walitzer, K. S.,Wood, P. K., and Brent, E. E. (1991). Characteristics of children of alcoholics: Putative risk factors, substance use and abuse, and psychopathology. Journal of Abnormal Psychology, 100, 427–448. Sher, K. J., and Wood, P. K. (1997). Methodological issues in conducting prospective research on alcohol-related behavior:A report from the field. In The Science of Prevention: Methodological Advances from Alcohol and Substance Abuse Research (K. J. Bryant, M.Windle, and S. G.West, eds.).Washington, DC: American Psychological Association Press. Sher, K. J.,Wood, M. D.,Wood, P. K., and Raskin, G. (1996). Alcohol outcome expectancies and alcohol use: A latent variable cross-lagged panel study. Journal of Abnormal Psychology, 105, 561–574. Silva, P. A., and Stanton, W. R. (1996). From Child to Adult: The Dunedin Multidisciplinary Health and Development Study. Auckland, NZ: Oxford University Press. Singer, J. D., and Willett, J. B. (2003). Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press. Slutske,W. S. (2006). Natural recovery and treatment-seeking in pathological gambling: Results of two US national surveys. American Journal of Psychiatry, 163, 297–302. Slutske, W. S., Caspi, A., Moffitt, T. E., and Poulton, R. (2005). Personality and problem gambling: A prospective study of a birth cohort of young adults. Archives of General Psychiatry, 62, 769–775. Slutske, W. S., Eisen, S. A., True, W. R., Lyons, M. J., Goldberg, J., and Tsuang, M. T. (2000). Common genetic vulnerability for pathological gambling and alcohol dependence in men. Archives of General Psychiatry, 57, 666–673. Slutske,W. S., Jackson, K. M., and Sher, K. J. (2003).The natural history of problem gambling from age 18 to 29. Journal of Abnormal Psychology, 112, 263–274. Sproston, K., Erens, R., and Orford, J. (2000). Gambling Behaviour in Britain: Results from the British Gambling Prevalence Survey. London: National Centre for Social Research. Toce-Gerstein, M., Gerstein, D. R., and Volberg, R. A. (2003). A hierarchy of gambling disorders in the community. Addiction, 98, 1661–1672.

154

Research and Measurement Issues in Gambling Studies

Tomarken, A. J., and Waller, N. G. (2005). Structural equation modeling: Strengths, limitations, and misconceptions. Annual Review of Clinical Psychology, 1, 31–65. Tremblay, R. E., Pihl, R. O.,Vitaro, F., and Dobkin, P. L. (1994). Predicting early onset of male antisocial behavior from preschool behavior. Archives of General Psychiatry, 51, 732–739. Vitaro, F., Arseneault, L., and Tremblay, R. E. (1997). Dispositional predictors of problem gambling in male adolescents. American Journal of Psychiatry, 154, 1769–1770. —— . (1999). Impulsivity predicts problem gambling in low SES adolescent males. Addiction, 94, 565–575. Vitaro, F., Brendgen, M., Ladouceur, R., and Tremblay, R. E. (2001). Gambling, delinquency, and drug use during adolescence: Mutual influences and common risk factors. Journal of Gambling Studies, 17, 171–190. Vitaro, F., Ladouceur, R., and Bujold, A. (1996). Predictive and concurrent correlates of gambling in early adolescent boys. Journal of Early Adolescence, 16, 211–228. Vitaro, F.,Wanner, B., Ladouceur, R., Brendgen, M., and Tremblay, R. E. (2004).Trajectories of gambling during adolescence. Journal of Gambling Studies, 20, 47–69. Wanner, B., Vitaro, F., Ladouceur, R., Brendgen, M., and Tremblay, R. E. (2006). Joint trajectories of gambling, alcohol, and marijuana use during adolescence: A person- and variable-centered approach. Addictive Behaviors, 31, 566–580. Winters, K. C., Stinchfield, R. D., Botzet, A., and Anderson, N. (2002). A prospective study of youth gambling behaviors. Psychology of Addictive Behaviors, 16, 3–9. Winters, K. C., Stinchfield, R. D., Botzet, A., and Slutske, W. S. (2005). Pathways of youth gambling problem severity. Psychology of Addictive Behaviors, 19, 104–107. Winters, K. C., Stinchfield, R. D., and Fulkerson, J. (1993a). Patterns and characteristics of adolescent gambling. Journal of Gambling Studies, 9, 371–386. —— . (1993b). Toward the development of an adolescent gambling problem severity scale. Journal of Gambling Studies, 9, 63–84. Winters, K. C., Stinchfield, R. D., and Kim, L. G. (1995). Monitoring adolescent gambling in Minnesota. Journal of Gambling Studies, 11, 165–183.

CHAPTER 7

Quantification and Dimensionalization of Gambling Behavior Shawn R. Currie

David M. Casey

Calgary Health Region Calgary, Alberta, Canada

Addiction Centre Foothills Medical Centre University of Calgary Calgary, Alberta, Canada

Historical Perspectives Epidemiological Data on Gambling Expenditure, Frequency, Duration, and Type Quantification of Other Addictive Behaviors Importance of the Quantification of Gambling Variations in Sources of Data Relevant Quantitative Dimensions of Gambling Behaviors Inputs Participation Status Types of Gambling Frequency Expenditure Duration Attitudes and Cognitions Outputs Financial Legal Social and Psychological Harms Clinical Use of Quantitative Gambling Data Conclusions

155

156

Research and Measurement Issues in Gambling Studies

HISTORICAL PERSPECTIVES Gambling has been present in our society for many centuries. Nonetheless, attempts to monitor and control the amount of gambling in the general population are relatively recent. The onset of government regulation of gambling in Western countries, which occurred roughly in the late 1960s in both the United States and Canada (Campbell and Smith 1998; Petry 2005), can be reasonably viewed as the first organized effort to quantify gambling behavior.The motives at the time were largely financial. Governments were concerned about the vast amounts of untaxed income being generated by casinos, horse tracks, and other gambling venues. Regulation by the government provided the means for provincial and state participation in this new and growing economy. The onset of government regulation also signaled the first official recognition that gambling, like alcohol, drugs, and tobacco, required external controls to protect the public from potential harm. During the first 20 years or so of government regulation and monitoring, gambling failed to attract the attention of researchers (Campbell and Smith 2003). The first large-scale surveys of gambling behavior on the individual level were not conducted until the late 1970s (Smith and Wynne 2000). Systemic data on the extent of gambling in the population has been collected primarily by government regulatory bodies and the industry itself. In both cases, the focus of data collection continues to be on the economic impact of gambling. The industry, for example, monitors variables such as patron origin (e.g., locals vs tourists), frequency of visitation, the average expenditure per visit, and types of games played.This information, typically not disclosed to the general public or government, provides the industry with a primary source of data for marketing and growth initiatives. Government regulators collect similar data for monitoring purposes. Furthermore, in some countries such as Canada, the government is also the main provider of gambling opportunities, putting it in the unique position of using data for both regulatory and marketing purposes. In terms of the economic impact of gambling, the government closely monitors the number of gambling outlets (e.g., number of electronic gaming machines, lottery outlets, casinos, and racetracks), per capita revenue generated from games of chance, general populace attitudes toward gambling, and jobs created directly by gambling and by spin-off economies (tourism, hospitality, and construction). A significant historical trend in the last 20 years has been the emphasis by government- and industry-run gambling to attract local consumers rather than tourists (MacDonald, McMullen, and Perrier 2004). The expansion of gambling venues in Canada and the United States has been largely driven by a “convenience gambling” model, providing opportunities to gamble as close to consumers as possible to prevent the migration of persons across provincial, state, and national borders to the larger centers of gambling.This market trend is not limited to casino

Quantification and Dimensionalization of Gambling Behavior

157

gambling. It has influenced all types of gambling products. For example, lottery gambling saw a rapid expansion of available products in the late 1980s, including the introduction of instant-win tickets and sports betting (MacDonald et al. 2004; Smith and Wynne 2000).The intent is to put more opportunities to gamble within reach of the average consumer. These trends have important implications for the collection and interpretation of data on gambling behaviors. Apart from known gambling hot spots (Las Vegas, Atlantic City, etc.), information gathered on individual gambling habits via population surveys, government revenues, and industry records in specific jurisdictions are now used locally to monitor socioeconomic trends. Hence, data collected on resident gambling habits can be used for policymaking purposes at the local level. Furthermore, the impact of responsible gambling initiatives can be more effectively measured when the bulk of gambling is performed by local residents.

EPIDEMIOLOGICAL DATA ON GAMBLING EXPENDITURE, FREQUENCY, DURATION, AND TYPE There have been numerous large population surveys conducted in the last 10 years across North America to measure individual gambling habits. Every province in Canada has conducted its own problem gambling prevalence survey, with sample sizes ranging from 1500 to 4000 (Williams and Wood 2004). In addition, national data on gambling habits were collected as part of a large household survey on mental health completed in 2002 (Statistics Canada 2002a). Similar surveys have been conducted throughout the United States and other countries (Shaffer, Hall, and Vander Bilt 1997). In contrast to the paucity of gambling prevalence data that existed 15 years ago, researchers are now faced with a surplus of data. Unfortunately, a variety of instruments and questions have been employed on surveys, making comparisons across regions extremely difficult. All provincial surveys in Canada conducted over the last 5 years have used a common instrument, the Canadian Problem Gambling Index (CPGI) (Ferris and Wynne 2001). Nevertheless, subtle differences in the wording of questions related to the type of gambling, amount of money spent, and frequency of play exist across the surveys. Due to space limitations, it is not possible to review the prevalence data from each population survey conducted on gambling in the last 10 years. To provide a national snapshot, Canadian data from the gambling module collected in the 2002 Canadian Community Health Survey (CCHS)–Mental Health and Well-being cycle (Statistics Canada 2002a) are presented in Table 7.1. Briefly, the CCHS-1.2 was a cross-sectional survey of a nationally representative sample of over 36,000 individuals aged 15 and older in all provinces and territories (Cox et al. 2005; Statistics Canada 2002a). Gambling was assessed using a short version of the CPGI (Ferris and Wynne 2001). Unfortunately, the CCHS-1.2 version did not ask participants about

158

Research and Measurement Issues in Gambling Studies

time invested in gambling activities, so national prevalence data on gambling duration were unavailable. In addition, because frequency varied by type of gambling, a composite index of gambling frequency was calculated based on the type occurring most often (e.g., an individual who plays the lottery once per week and electronic slot machines every day was classified as a daily gambler). Nonetheless, the frequency data are likely underestimates given that the frequency questions for each type of gambling were administered independently (e.g., an individual who engages in Table 7.1 Canadian Population Statistics for Frequency, Expenditure, Percent Income, and Type of Gambling. Dimension

% or mean

Gambling Participation % population ≥ 15 years of age who are gamblers

76

Frequency of Any Gambling Daily

1

2–6 times/week

11

About once/week

17

2–3 times/month

9

About once/month

9

6–11 times per year

7

1–5 times per year

22

Never

24 Expenditure

Self-reported Government-reported revenue % Gross income

$272 $1080 0.9

Type of Gambling1 Lotteries

65

Instant-win

36

Casinos

22

Bingo

8

Video lottery terminals outside of casinos

6

Horse racing Other

4 21

Data sources: Currie, Hodgins, Wang, el-Guebaly, and Wynne 2006a; Canadian Community Health Survey: Mental Health and Well-being (Marshall and Wynne 2003); Survey of Household Spending (MacDonald, McMullan, and Perrier 2004); and provincial government financial reports (Azmier 2005). 1 Categories not mutually exclusive.

Quantification and Dimensionalization of Gambling Behavior

159

a different form of gambling every day might be erroneously classified as a once-aweek gambler). Despite these limitations,Table 7.1 provides, to our knowledge, the only national perspective on self-reported gambling behaviors.

QUANTIFICATION OF OTHER ADDICTIVE BEHAVIORS Progress in the measurement of gambling has been influenced by the methodologies developed to assess other addictive behaviors, most notably alcohol.There are numerous similarities between alcohol and gambling as social phenomena. Both are legal, regulated, and considered socially acceptable for both adults and adolescents (Korn and Shaffer 1999). Key concepts in the epidemiology of alcohol consumption patterns are transferable to gambling. Researchers have quantified the consumption of alcohol across dimensions such as the proportion of drinkers in the population, the proportion who drink regularly, per capita alcohol consumption (liters per person or standard drinks per person), beverage choice, the proportion of the population who drink heavily or binge-drink, frequency of consumption, source of alcohol consumed (off- or on-premise purchase), and the drinking context (home, bars, before driving, etc.). Meaningful gambling equivalents can be created for most of these indicators. Policy development in the area of alcohol has been guided by a central principle:A direct correlation exists between alcohol consumption and risk of harm in the population (Babor 2002); numerous studies show that the risk of health and psychological problems increase with greater daily consumption (Babor 2002; Babor et al. 2003). Quantitative information on alcohol use has been used to determine how the availability of alcohol, its price, restrictions on access (e.g., age limits), and drinking circumstances (measures to deter impaired driving) are related to consumption patterns.The empirical study of this relationship has influenced harm reduction strategies, including limits on blood alcohol level while drinking, restrictions on the sale of alcohol, and the promotion of responsible drinking guidelines (Bondy et al. 1999) which place specific limits on quantity, frequency, duration, and context. No such limits exist for gambling, although a set of proposed limits derived from the CCHS-1.2 data were recently published by Currie, Hodgins, Wang, el-Guebaly, and Wynne (2006). Aiding the study of alcohol consumption patterns is the means to convert different alcoholic beverages to a standard drink. The World Health Organization and most federal health agencies agree on a common set of standard drink equivalents (Bondy et al. 1999). Although a variety of beverage choices exist, the amount of alcohol can be estimated in each drink type and aggregated per person. Sources of error notwithstanding (e.g., variations in the beverage size), the ability to equate

160

Research and Measurement Issues in Gambling Studies

drink types has been instrumental to the quantification of alcohol consumption. Without standard drink equivalents, population data on per capita consumption would be meaningless.The creation of a comparable, standard unit of gambling has not occurred (Korn and Shaffer 1999). In the area of tobacco research, the most common quantitative indicator is the number of smokers in the population (Roemer 2004). The total number of regular smokers divided by the total number of persons in the target group provides the estimated prevalence of smoking, a highly relevant statistic for monitoring the health of a population (Statistics Canada 2002b).The gambling equivalent for this measure would be percent of gamblers in the population. In terms of assessing the quantity of smoking among those who smoke, cigarettes per day is the most common indicator.This is a valid measure of present level of exposure to smoking toxins (Djordjevic 2004).The total cigarette packs per year provides an estimate of total cumulative exposure (Boyle et al. 2004). However, cigarettes per day is not without limitations, as it does not accurately capture the variable pattern of smoking seen in occasional smokers. Furthermore, different tobacco products (cigarettes, pipes, cigars, and chewing tobacco) vary in terms of the concentration of nicotine, tar, and other harmful chemicals (Djordjevic 2004). Assessment measures such as the Timeline Follow-Back (TLFB) can be used to more precisely assess smoking days in a month and cigarettes smoked per smoking day (Brown et al. 1998). It should also be noted that biochemical measures of both smoking (saliva and serum cotinine levels, carbon monoxide level in expired air) and alcohol (blood alcohol level, gamma-glutamyl transpeptidase [GGT], urine screens) are also available to validate self-reported use, quantify exposure, and serve as surrogate measures of amount consumed. Biochemical measures correlate reasonably well with the actual quantity of cigarettes or alcohol consumed (Miller, Sovereign, and Krege 1988; Stevens and Munoz 2004). The means to quantify consumption via self-report and biochemistry are unique to substance abuse. In summary, most of the quantitative measures of consumption used in tobacco and alcohol monitoring can be adapted to gambling in a meaningful way. The primary intent of quantitative indicators in these areas has been to monitor the health behaviors of individuals, measure level of risk, and assess the impact of strategies aimed at reducing high-risk health behaviors.These aims are relevant to gambling research and public policy and have driven the development of comparable quantitative indicators of gambling behaviors.

IMPORTANCE OF THE QUANTIFICATION OF GAMBLING Over the last 15 years, research on the impact of gambling has shifted focus from a purely economic perspective to an individual-behavioral perspective.

Quantification and Dimensionalization of Gambling Behavior

161

The driving force has been research on the prevalence of problem gambling in the general population (Shaffer et al. 2004). Public health officials are in need of accurate, standardized information on how much, how often, and for how long consumers gamble (Korn 2000; Korn and Shaffer 1999; Shaffer and Korn 2002). Accordingly, researchers have called for an expanded research agenda with greater emphasis on the determinants of disordered gambling, such as factors of personality, behavior, and context (Shaffer et al. 2004). “Responsible gambling” initiatives place the onus of control on the individual consumer of gambling products (Blaszczynski, Ladouceur, and Shaffer 2004). There is also now greater awareness that disordered gambling is not a simple on/off phenomenon. Rather, the risk of problem gambling exists on a quantifiable continuum. A greater understanding of individual gambling habits and their relationship to harm or risk of addiction will undoubtedly aid in responsible gambling policy development (Currie, Hodgins, Wang, el-Guebaly, and Wynne 2006). Systematic monitoring of gambling habits is highly relevant when taking a population-based perspective. However, at the level of the individual, the quantification of gambling behavior also has utility. For gamblers seeking help, the assessment of gambling quantity, frequency, and type can aid in determining the severity of addiction (see the Hodgins chapter in this volume). For example, the ratio of intended to actual expenditure on gambling (i.e., how much money the individual planned to spend versus how much he actually spent) can be a useful index of degree of control (Weinstock,Whelan, and Meyers 2004).

VARIATIONS IN SOURCES OF DATA Data on the incidence of gambling have been collected using a variety of sources. Early on, data on gambling were collected primarily by government regulatory bodies or the gambling industry. More recently, however, data on gambling have been collected by researchers using a broader range of sources and techniques, including telephone interviews, face-to-face self-reports, government reports on gambling revenue, household spending surveys, and collateral reports (Azmier 2005; Gambino 1997; Hodgins, Currie, and el-Guebaly 2001). There are strengths and weaknesses associated with each of these sources of data. A full review of these issues is beyond the scope of the present chapter.The main strengths and weaknesses of different techniques will be discussed here. Self-reports of individual gambling behavior in the form of telephone, faceto-face, or computer-based surveys provide researchers with a wealth of firsthand accounts of the extent of gambling involvement. Self-reports tend to be reasonably inexpensive and easy to use.There are a variety of instruments (e.g., the CPGI, the South Oaks Gambling Screen [SOGS], the National Opinion Research Center Diagnostic Screen [NODS]) that have been used to measure the quantity of

162

Research and Measurement Issues in Gambling Studies

gambling, with Shaffer et al. (2004) identifying 27 instruments used to identify disordered gambling alone. However, the very fact that there is such a wealth of self-reports of gambling makes it very difficult to compare the results from studies in which different instruments were used. Even when the same instruments are used, some questions have been altered for some studies. Another limitation of selfreports is the tendency for some individuals to underreport their participation in gambling activities. Underreporting can be influenced by the wording of expenditure questions or individuals’ inability to be precise (e.g., errors in recall) regarding the amount of money they spend on gambling activities. Another potential reason for underreporting is misrepresentation on the part of respondents, who may desire to provide researchers with a socially desirable answer or may even be in denial about the extent of their behavior (Stevens and Munoz 2004;Weinstock et al. 2004). Gambling revenue reports developed for governments are another source of data related to gambling.These reports tend to focus on net revenue for gambling establishments, the amount wagered, and employment benefits incurred as a result of the gambling establishments (National Council of Welfare 1996; Statistics Canada 2003). The data on household expenditures on gambling activities (Statistics Canada 2003) are particularly helpful, since they allow us to compare the household expenditures with other data collected, such as self-reports on the amount spent on gambling.There can be a large discrepancy between self-reported spending on gambling and actual revenues from gambling reported by the government. The direction of the discrepancy (overreporting vs. underreporting) varies depending on the source of the data (Wood and Williams, in press). For example, the Canadian government reported the average annual revenue from legalized gambling as approximately $1080 per household in 2004, whereas the average self-reported spending on gambling was only $272 per household (Azmier 2005). The revenue generated from tourist gambling accounts for a small part of this discrepancy. In contrast, Williams and Wood (2004) found that across the Canadian provinces, average self-reported expenditures were 2.1 times higher than actual provincial gaming revenues for the same time period when the self-reported amounts were derived from gambling behavior surveys. These data suggest that households can grossly over- or under-estimate how much family income actually goes toward gambling. Variations in the statistics related to gambling are the result of a number of factors, including the fact that different instruments are used to measure gambling in different studies. Comparable health research from the area of smoking has shown that different approaches to quantify level of risk (e.g., self-reports of cigarette consumption vs carbon monoxide levels) can impact the accuracy, sensitivity, and validity of the data gathered (Stevens and Munoz 2004).This makes it difficult to compare the statistics from different studies. For instance, prevalence rates for problem gambling vary across Canada, from a low of 2.7% in Saskatchewan to a high of 8.6% in Ontario (Williams et al. 2004). The variable prevalence rates are

Quantification and Dimensionalization of Gambling Behavior

163

thought to be due to a variety of factors: different survey instruments, the specific wording of some questions, variable response rates, socioeconomic factors, and variations in the availability of gambling across provinces. In addition, studies use different time frames when measuring the level of gambling—for instance, within the last month, within the last year, currently, and/or in a lifetime (National Council of Welfare 1996).This makes it very difficult to compare the results across various studies. Finally, individuals may interpret the wording of the question in very different ways. Researchers have determined that a relatively simple question such as “How much do you spend gambling?” can lead to a wide variety of answers that may underrepresent or overrepresent the actual amount of money spent on gambling activities (Blaszczynski, Dumlao, and Lange 1997;Wood and Williams, in press). Individual answers appear to depend on whether the respondent believes that he should include net expenditures or turnover. It is critical that clear and concise instructions be given for each question, particularly those dealing with expenditure (Williams and Wood 2004). Finally, in other fields, the use of selfreports are usually validated with other measures. Since there is no universally agreed upon metric for gambling, the information that is gathered in separate studies is not always easily compared (Weinstock et al. 2004). As mentioned earlier, in the case of alcohol, tobacco, and drugs, it is easier to make comparisons, since there are universally agreed upon biomedical and behavioral metrics.The Gambling (G)TLFB described by Weinstock et al. (2004) may be an important step in helping to identify behavioral metrics that would be agreed upon by researchers. As stated by Gambino (1997), “In the absence of definitive tests, we cannot know the true prevalence of any condition, we can only estimate it” (p. 293).

RELEVANT QUANTITATIVE DIMENSIONS OF GAMBLING BEHAVIORS Compared with alcohol and smoking, the behavioral dimensions to gambling are far more complex.Weinstock et al. (2004) note that gambling is “a heterogeneous collection of activities, and measuring it requires the assessment of various dimensions” (p. 73).To conceptualize gambling as measurable activity, a simple input–output model is proposed.As seen in Figure 7.1, several measurable inputs to gambling can be defined.These are broadly categorized as being systemic (at the community, industry, or government level) or individual (at the person level). The systemic inputs (e.g., infrastructure costs, number of employees necessary to run venues) will not be discussed in this chapter but are shown in Figure 7.1 for illustrative purposes. At the level of individual gambling, many behavioral dimensions are possible.The most relevant for describing gambling activity are participation status, frequency, expenditure, duration, and type of game played.These are described in more detail below. Individual attitudes, knowledge, and perceptions of gambling

164

Research and Measurement Issues in Gambling Studies

can also be viewed as relevant dimensions, since they may influence both the rate and the intensity of participation. Weinstock et al. (2004) further articulates contextual dimensions to gambling, such as the setting (casino, home, bar) and concurrent alcohol consumption. A brief synopsis of the relevant inputs and outputs on an individual behavioral level is provided in Table 7.2.The core dimensions are discussed in the following sections.

INPUTS Participation Status Gambling prevalence surveys broadly dichotomize the population as being gamblers or nongamblers based on having wagered money on an event with an uncertain outcome over the past year. One of the challenges inherent in determining gambling status from self-report is that perceptions differ within the general public as to what constitutes gambling. For example, many people do not consider playing the lottery a form of gambling. Furthermore, in some surveys, playing the

• Government revenue • Growth rate • Charitable benefits • # jobs created and average salaries • Gambling-related crime rates • Spin-off economic benefits (e.g., tourism) • Cost of treating PGs • Cost of regulation

System Level

• # gaming outlets (casinos, EGMs, lottery outlets, etc.) • # of visitors/year • # jobs required • Cost of maintaining venues

Gambling

Input

Output

• Gambler status • Type of game • Context • Frequency of play • Expenditure ($ wagered, percent income) • Time invested • Knowledge and attitudes

Individual Level

• Monetary win/loss $ • Psychological benefits (pleasure, socialization, satisfaction, etc.) • Harms (financial, emotional, relationship, health, job, suicide) • Pathological gambling incidence and prevalence

Figure 7.1. Conceptual model of the quantification of gambling behavior. EGM, electronic gaming machine; PG, pathological gambler.

Table 7.2. Dimensionalization of Gambling Behavior. Definition

Quantification Schemes

Data Sources

Utility

Participation in gambling

Proportion of population who gamble

Lifetime and 12-month rates of gambling participation

Population surveys

Assess norms of gambling participation in the general population; examine socioeconomic correlates of gambling

Game type

Consumer preference for gambling type among available options

% of consumers who played specific game types in the last 12 months

Population surveys

Assess the popularity of specific game types; assess market trends in gambling and shifts in preference for high- and low-risk types; examine socioeconomic correlates of gaming preference

% of consumers who play >1 type of game

G-TLFB

Most popular game types

Industry records

% of gamblers who primarily bet in casinos, bars, racetracks, home, etc.

Population surveys

Number of different gambling venues per person

G-TLFB

Per capita expenditures, losses, and net expenditures (wins – losses) among gamblers in past 12 months

Population surveys

Average, maximum, and minimum dollar output per gambling session

G-TLFB

Setting

Expenditure and losses

Venue in which gambling occurs

Amount of money spent on games of chance

Assess consumer preferences and risk level for different venues; examine contextual factors (e.g., alcohol use) influencing gambling

Determine the spending patterns of individual gamblers; examine degree of control while gambling; assess longitudinal trends in gambling expenditures; examine socioeconomic correlates of gambling expenditure

165

(Continues)

Quantification and Dimensionalization of Gambling Behavior

Parameter

Parameter

Percent of annual income diverted to games of chance

Quantification Schemes

Data Sources

Largest win and loss in lifetime, past 12 months

Government revenue

Expenditure on gambling in relation to other household expenditures

Industry records

Percent of annual personal, household, and disposable income spent on gambling

Population surveys G-TLFB Government revenue

Frequency of gambling

Time spent gambling

Past year frequency of playing games of chance

Amount of time spent on games of chance

Frequency of playing specific games in the past 12 months

Population surveys

Frequency of any gambling in the past 12 months

G-TLFB

Total time spent on all gambling activity

Population surveys

Average time spent per session on specific games

G-TLFB

Average time spent per session on any gambling GamblingNegative consequences related harms of gambling (e.g., stress, financial strain, interpersonal

Number of gambling-related harms experienced per gambler

Population surveys Crime, bankruptcy, divorce rates

Utility

Determine the economic impact of gambling in relation to the individual’s economic means; assess longitudinal trends in proportion of disposable income diverted to gambling; assess the relationship between percent income and risk of harm Assess consumer habits in gambling; determine the relationship between frequency and gambling-related harms

Determine sociological trends in how the populace spends its time; assess time among family, work, and other activities in relation to gambling; assess the health implications of spending excess time in gambling venues (e.g., inactivity, smoky environments) Assess financial and nonfinancial consequences of gambling

Research and Measurement Issues in Gambling Studies

Percent income

Definition

166

Table 7.2 (Continued)

conflict, suicide, legal problems)

Proportion of gamblers experiencing >1, >2, and >3 gambling-related harms

Emergency room and hospital admission statistics

Positive consequences Proportion of population of gambling (e.g., identifying specific excitement, socializing, psychological benefits as relief of stress, the main reason for satisfaction of gambling supporting charities, etc.) Average number of psychological benefits experienced per gambler

Population survey questions on main reasons for gambling

Assess nonmonetary incentives to gamble to understand individual gambling behavior

Problem gambling

Individuals who exceed the diagnostic threshold for problem gambling

Population surveys

Determine treatment needs in a community; assess the impact of increasing gambling opportunities on problem gambling prevalence; assess the overall health of a population

Lifetime and 12-month prevalence of problem gambling in the general population

Diagnostic interviews

Problem and pathological gambling rates

Treatment utilization

G-TLFB, Gambling Timeline Follow-Back interview.

Quantification and Dimensionalization of Gambling Behavior

Psychological benefits

167

168

Research and Measurement Issues in Gambling Studies

stock market is considered a form of gambling (Ferris and Wynne 2001; Smith and Wynne, 2000).To accurately interpret gambling participation figures derived from self-report surveys, it is important to understand the definition of gambling employed and whether some activities are excluded from the definition. When using a broad definition of gambling that includes lottery play, past-year and lifetime rates of gambling participation range between 60 and 95%, respectively, in North America (Petry 2005; Shaffer et al. 2004). Types of Gambling Gambling encompasses a broad range of activities.Types of gambling include lotteries, bingo, card games, slot machines, electronic gaming machines, sports betting, casino games, horse betting, and many others. Most of these categories can be further subdivided. Casino games include blackjack, craps, and roulette; sports betting includes baseball, hockey, and football. There are literally hundreds of different gambling activities, each with its own set of rules, odds of winning, and payout schedule. Gambling types can be categorized across several dimensions: singleplayer (e.g., blackjack) versus multiplayer (e.g., bingo); passive (e.g., sports betting) versus active (e.g., poker); continuous (e.g., slot machines) versus noncontinuous (e.g., lottery). Some forms of gambling, notably electronic gaming machines, are considered to carry a high risk of addiction, while others are considered low risk (Griffiths 1993). However, the relationship between game type and risk level is confounded by the fact that most gamblers play more than one game.The majority of Canadians (67%) play more than one type of game (Currie, Hodgins,Wang, el-Guebaly, and Wynne 2006). Furthermore, problem gamblers are more likely than nonproblem gamblers to sample a variety of game types (Petry 2005). Studying the interaction of game type with standard quantitative dimensions of amount, duration, and frequency is made more complicated with the plethora of legal and illegal gambling forms available to the consumer. Frequency How often one gambles is a concrete and useful indicator of level of participation in gambling. The CPGI uses seven categories to capture frequency of gambling, ranging from one to five times per year to daily. Other surveys have used fewer categories (daily, weekly, monthly, less than monthly) to measure frequency. A finer-grain perspective of gambling frequency can be obtained with an instrument such as the G-TLFB, which provides an estimate of the exact number of days gambled in a 30-day period (Weinstock et al. 2004).The choice of a categorical or continuous scale to quantify frequency scheme would depend on the intended purpose of the data being collected. Categorical data may be sufficient for examining population level trends in gambling but inadequate for

Quantification and Dimensionalization of Gambling Behavior

169

assessing change following treatment. Number of days gambled in the past 30 days has been shown to be reliable over time and sensitive to change for measuring the impact of treatment (Hodgins and Makarchuk 2003). Although frequency of gambling correlates with the risk of problem gambling (Currie, Hodgins, Wang, el-Guebaly, and Wynne 2006;Weinstock et al. 2004), it should not be used as a surrogate indicator of risk level. Some individuals may gamble frequently but wager small amounts (e.g., buying lottery tickets twice weekly), whereas others may wager large amounts less frequently (e.g., visiting a casino once per month). Expenditure A highly relevant dimension of gambling behavior is amount of money invested. Money spent can be subdimensionalized in terms of absolute dollars wagered, net amount gained or lost (wins minus losses), and the percent of personal or household income invested in gambling. Weinstock et al. (2004) describe an additional dimension of intent, defined as how much the gambler intends to spend in a given gambling session.The ratio of intended expenditure to actual expenditure provides an indication of the gambler’s level of control. Assessing expenditure and net money won or lost in self-report surveys can become problematic when an operational definition of this dimension is not made clear to the respondent. Most surveys ask respondents how much money they spend on gambling in a session or over the course of a specified time period.As noted, different interpretations of this question are possible, including the total money invested (not including wins or losses), net amount invested (wins minus losses), or losses only. Hence, the phrasing of expenditure questions will impact the characteristics and reliability of the data collected (Wood and Williams, in press). The percent of income spent on gambling is arguably a better indicator of gambling expenditure because it frames the amount wagered in the context of the gambler’s financial means (Weinstock et al. 2004). Spending $1000 per year on gambling has different financial implications for an individual with an annual income of $20,000 compared with $100,000. Ideally, this parameter should be expressed as a percent of disposable (after tax) income; however, it is common for household spending and gambling-specific surveys to inquire about gross income only. Because most families share expenses and income, the proportion spent on gambling should be expressed as a percentage of total household rather than personal income. Using this methodology, MacDonald et al. (2004) estimated that in 2002 Canadians spent an average of 0.9% of their total household income on gambling. Although percent income is the most useful expression of gambling expenditure, it is also the most prone to measurement error.As noted, persons responding to household expenditure surveys tend to underestimate the amount they spend on gambling compared with estimates of spending that derive from government

170

Research and Measurement Issues in Gambling Studies

revenues (Azmier 2005). In addition, some individuals are reluctant to disclose their income in surveys (Bradburn, Sudman, and Wansink 2004; Duncan and Peterson 2001), while others may not be able to provide an accurate estimate of their spouses’ income to report a total household income. Hence, both the numerator and the denominator of the percentage of income are prone to measurement error or bias (Walker et al. 2006). Duration Like frequency, time investment in gambling is game specific and often constrained by the temporal characteristics of the game. Certain forms of gambling are relatively brief in duration (e.g., buying a lottery ticket), whereas others require a longer investment of time (e.g., playing a game of bingo). Difficulty can arise when operationalizing duration for “passive” forms of gambling. For example, when an individual bets on a sporting event, should one consider duration of gambling as the time it takes to place the bet, or the full length of the sporting event until the outcome is known? Similar to expenditure, duration can be calculated in many ways: average time spent per gambling session, total time spent gambling in a month, maximum time spent gambling in a single session, and others. Surveys conducted across the United States and Canada vary in their collection of duration data. Attitudes and Cognitions An individual’s opinion of the benefits and costs associated with gambling will undoubtedly influence his or her decision to gamble. For example, an individual who believes that gambling is an opportunity to make money quickly is more likely to gamble than an individual who believes that all games of chance favor the house. Knowledge of the odds of winning should also theoretically influence gambling behavior, although not always in the expected direction (Williams and Connolly 2006). For example, there are high rates of both nonproblem and problem gambling among casino employees despite their having an excellent knowledge of gambling odds and the house advantage (Petry 2005). Other cognitions—for example, illusionary control of gambling outcome—have emerged as more important predictors of the tendency to gamble or keep gambling than has knowledge of odds (Ladouceur and Walker 1996). Self-report instruments have been developed to measure attitudes and cognitions in gamblers (Ladouceur et al. 2002).These scales have been used to research the relationship between attitudes and actual gambling behavior, as well as to measure the impact of interventions to change gambling (Williams et al. 2004). The Canadian public’s attitude toward gambling has also been assessed as part of a series of studies by the Canada West Foundation (Azmier 2000).

Quantification and Dimensionalization of Gambling Behavior

171

OUTPUTS At the level of the individual, relevant gambling outputs can be broadly categorized as being financial, legal, social, or psychological in nature. Financial The majority of gamblers, both recreational and problem, gamble with the intent to win money. Hence, the extent of wins or losses can be viewed as the main output at the level of the individual gambler. The financial outputs of gambling can be calculated in numerous ways, including average wins and losses over a specified time period, largest win or loss ever, and total wins and losses over a specified period (Walker at al. 2006).There is no standard method for collecting this information, and surveys have differed in the types of winsand-losses questions used to gather data on financial outputs (Wood and Williams, in press). As noted, a detailed perspective on wins and losses can be obtained using the G-TLFB method (Hodgins and Makarchuk 2003; Weinstock et al. 2004). However, the G-TLFB is impractical for large population surveys. The win–loss ratio could be viewed as a surrogate measure of financial harm. Over time most gamblers lose more than they win. The impact of these losses is moderated by the gambler’s financial situation. The assessment of financial harm should also include indicators such as debt load, interest charges, inability to pay other bills or provide for basic necessities, personal bankruptcy, the need for seeking additional employment, and the cost of treatment. It is worth noting that financial harm impacts not only the individual and his or her immediate family, but also the community (e.g., in terms of treatment costs, higher interest rates for everyone, social assistance costs). Legal Few gamblers experience legal ramifications of their gambling. When legal problems do emerge, however, they are often substantial. Legal consequences arising from gambling can include divorce proceedings, bankruptcy, and criminal charges if the individual has engaged in any illegal activity (e.g., theft, robbery) to obtain money to gamble or pay off debts. Population level statistics on indicators such as divorce rates and crimes attributable to gambling are collected by the government and used to assess the impact of new gambling venues on a community. At the level of the individual, survey instruments such as the CPGI and the SOGS include several items assessing both financial and legal consequences arising from gambling (Ferris and Wynne 2001; Lesieur and Blume 1987).

172

Research and Measurement Issues in Gambling Studies

Social and Psychological Harms Gambling-related harms are not limited to financial consequences. The social and psychological outputs can be either positive or negative. Instruments such as the SOGS (Lesieur and Blume 1987) and the CPGI (Ferris and Wynne 2001) collect information on a variety of psychosocial harms from excess gambling. These are discussed in more detail in Chapter 8. The quantification of psychosocial harms is complicated, yet extremely important because, unlike alcohol or smoking, the consequences of excessive gambling cannot be measured in biomedical terms.The diagnosis of problem or pathological gambling is made on the basis of exceeding a defined threshold of harms, which include social, psychological, legal, and financial consequences. Problem gambling assessment instruments to date have adopted a rather simplistic approach to quantifying levels of psychosocial harms. Both the SOGS and the National Opinion Research Center Screen for Gambling Problems based on criteria from the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) (Gerstein et al. 1999), use a dichotomous system for scoring harms. Gamblers indicate in a yes–no fashion whether they experience problems such as sleep difficulties or criticism about their gambling. The CPGI uses a four-point scale ranging from “never” to “almost always” to assess the extent to which gamblers have experienced any psychosocial harms over the past 12 months (the time frame can be modified depending on the purpose of the collection). The latter approach is an improvement over the dichotomous scoring system; however, all these instruments make the fundamental assumption that different harms are qualitatively equivalent and can be scored using the same metric. Such an assumption is untenable given the range of harms covered by gambling instruments. Can one assume that the psychological distress caused by being criticized about gambling is equivalent to entertaining thoughts of suicide? Is borrowing money from a friend to gamble equivalent to declaring bankruptcy? Endorsement of either of these harms could receive the same score. Only a minority of gamblers experience problems related to gambling. The majority of gamblers presumably find the experience pleasant and rewarding. The social and psychological benefits of gambling have been acknowledged (Korn and Shaffer 1999), but little work has been done to quantify this domain. Gamblers may experience psychological benefits from low-risk gambling such as stress reduction, socialization, and satisfaction with supporting charitable causes at “Las Vegas nights” (Desai et al. 2004). It is possible that gambling has a dose–response relationship similar to alcohol: High exposure is harmful or toxic to the individual, whereas low-dose levels may be psychologically beneficial or even protective (Kaiser 2003). To date, there has been no attempt to quantify the psychological benefits of gambling. Although population surveys have included broad questions on the reasons for gambling, no instrument has been specifically developed to assess

Quantification and Dimensionalization of Gambling Behavior

173

the benefits of gambling. Recent research suggests that some older adults do experience health benefits of recreational gambling (Desai et al. 2004; Loroz 2004). Thus far, the psychological benefits of low-risk gambling have been studied only in the elderly (Shaffer and Korn 2002).

CLINICAL USE OF QUANTITATIVE GAMBLING DATA Quantitative information on gambling is used in clinical settings for a variety of purposes. First, data on the amount and frequency of gambling are used to help identify gamblers who could benefit from treatment (Hodgins et al. 2001). Second, the assessment instruments can be used to facilitate changes in behavior.Third, the assessment instruments can be used to assess effectiveness of specific clinical interventions focused on helping individuals who have difficulties with gambling. Fourth, the assessment instruments can provide meaningful feedback to clients. Finally, the assessment instruments can help to individualize treatment for specific clients (Haynes 2006). As is evident, measuring quantity of gambling is useful in a wide variety of clinical settings.

CONCLUSIONS This chapter has provided an overview of the basic principles and challenges inherent when attempting to dimensionalize a highly complex activity such as gambling. Meaningful quantitative indicators of gambling behavior can be constructed using comparable indicators from the epidemiology of other addictive behaviors as a general guide. Quantitative measures of gambling have been used mostly in prevalence surveys aimed at tracking the extent of gambling in the general population. Many also have utility in clinical research as outcome measures following treatment, and in direct clinical applications to aid in the assessment of individual gamblers. Nevertheless, there is a lack of standardization of measures in the field of gambling. Surveys vary in the types of quantitative information collected on gambling and the wording of questions to obtain this information. Thus, estimates of key indicators such as expenditure and duration of gambling episodes can vary greatly across surveys. As a final note, none of the three primary quantitative dimensions— frequency, expenditure, and duration—are used to identify disordered gambling (Hasin 2003). In other words, gambling intensity in terms of how often, how much, and for how long is not presently considered relevant to an individual’s level of risk for harm from gambling.An individual spending relatively small amounts of money or gambling infrequently could be considered a problem gambler if

174

Research and Measurement Issues in Gambling Studies

sufficient consequences or behavioral indicators of gambling were evident. Conversely, an individual spending very large amounts of money or gambling on a daily basis may not be considered disordered if he or she reports few problems related to the gambling behavior.The identification of problem gambling is made solely on the basis of behavioral problems and consequences of gambling. Similarly, the determination of problem drinking is made on the basis of drinking consequences, not on how much someone drinks. Very few population surveys, until only recently, have bothered to collect data on the intensity of gambling involvement or other possible determinants of problem gambling. There appears to be a robust relationship between gambling frequency, expenditure, and duration, on the one hand, and risk for gambling-related problems on the other (Currie, Hodgins, Wang, el-Guebaly, Wynne, and Chen 2006). Additional research on the dose–response relationship in gambling could help to answer such questions as: How much is too much gambling? The answer to this question would have extraordinary utility in responsible gambling policy and public health initiatives. Nevertheless, the complexity of gambling will make the development and dissemination of any guidelines aimed at preventing harm by limiting gambling involvement extremely challenging. For example, the main quantitative dimensions—frequency, expenditure, and duration—are correlated and highly dependent on the type of gambling.The development of a standard unit of gambling would greatly assist future research in this area but ultimately may be untenable given the heterogenic nature of gambling.

ACKNOWLEDGMENTS Work on this chapter was funded in part by the Alberta Gaming Research Institute.The investigators wish to thank Statistics Canada for data access. However, the opinions and views expressed do not represent those of Statistics Canada.

GLOSSARY Expenditure amount of money spent on games of chance after accounting for wins and losses; the net win or loss. Frequency of gambling how often an individual gambles within a specified time period (typically one month). Gambling-related harms negative consequences of gambling (e.g., stress, financial strain, interpersonal conflict, suicide, legal problems). Game type gambling type among available options (e.g., lottery, sports betting, casino game). Participation status proportion of population who have wagered money on an event with an uncertain outcome over the past year.

Quantification and Dimensionalization of Gambling Behavior

175

Psychological benefits positive consequences of gambling (e.g., excitement, socializing, relief of stress, satisfaction of supporting charities). Setting venue in which gambling occurs (e.g., bar, casino, track, home). Time spent gambling amount of time spent gambling per session, expressed in hours or minutes.

REFERENCES Azmier, J. (2000). Canadian Gambling Behaviour and Attitudes. Calgary, AB: Canada West Foundation. ––––– . (2005). Gambling in Canada 2005: Statistics and Context. Calgary, AB: Canada West Foundation. Babor, T. F. (2002). Linking science to policy: The role of international collaborative research. Alcohol Research and Health, 26, 66–73. Babor,T. F., Caetano, R., Casswell, S., Edwards, G., Giesbrecht, N., and Graham, K. (2003). Alcohol: No Ordinary Commodity: Research and Public Policy. New York: Oxford University Press. Blaszczynski, A., Dumlao,V., and Lange, M. (1997).“How much do you spend gambling?” Ambiguities in survey questionnaire items. Journal of Gambling Studies, 13, 237–252. Blaszczynski, A., Ladouceur, R., and Shaffer, H. J. (2004). A science-based framework for responsible gambling:The Reno model. Journal of Gambling Studies, 20, 301–317. Bondy, S. J., Rehm, J., Ashley, M. J., Walsh, G., Single, E., and Room, R. (1999). Low-risk drinking guidelines: The scientific evidence. Canadian Journal of Public Health / Revue Canadienne de Sante Publique, 90, 264–270. Boyle, P., Gray, N., Henningfield, J., Seffrin, J., and Zatonksi, W. (2004). Tobacco: Science, Policy, and Public Health.Toronto: Oxford University Press. Bradburn, N., Sudman, S., and Wansink, B. (2004). Asking Questions:The Definitive Guide to Questionnaire Design: For Market Research, Political Polls, and Social and Health Questionnaires, rev. ed. San Francisco: Jossey-Bass. Brown, R.A., Burgess, E. S., Sales, S. D.,Whiteley, J.A., Evans, D. M., and Miller, I.W. (1998). Reliability and validity of a smoking timeline follow-back interview. Psychology of Addictive Behaviors, 12, 101–112. Campbell, C., and Smith, G. (1998). Canadian gambling trends and public policy issues. Annals of the American Academy of Political and Social Science, 556, 22–35. ––––– . (2003). Gambling in Canada—from vice to disease to responsibility: A negotiated history. Canadian Bulletin of Medical History, 20, 121–149. Cox, B. J., Yu, N., Afifi, T. O., and Ladouceur, R. (2005). A national survey of gambling problems in Canada. Canadian Journal of Psychiatry / Revue Canadienne de Psychiatrie, 50, 213–217. Currie, S. R., Hodgins, D. C.,Wang, J., el-Guebaly, N., and Wynne, H. (2006). Is the field of addictions ready for low risk gambling limits? Submitted. Currie, S. R., Hodgins, D. C.,Wang, J., el-Guebaly, N.,Wynne, H., and Chen, S. (2006). Risk of harm from gambling in the general population as a function of level of participation in gambling activities. Addiction, 101, 570–580. Desai, R. A., Maciejewski, P. K., Dausey, D. J., Caldarone, B. J., and Potenza, M. N. (2004). Health correlates of recreational gambling in older adults. American Journal of Psychiatry, 161, 1672–1679. Djordjevic, M. V. (2004). Nicotine dosing characteristics across tobacco products. In Boyle et al., pp. 181–206. Duncan, G. J., and Peterson, E. (2001). The long and short of asking questions about income, wealth, and labor supply. Social Science Research, 30, 248–263.

176

Research and Measurement Issues in Gambling Studies

Ferris, J., and Wynne, H. J. (2001). The Canadian Problem Gambling Index Final Report. Ottawa, ON: Canadian Centre on Substance Abuse. Gambino, B. (1997). Method, method: Who’s got the method? What can we know about the number of compulsive gamblers? Journal of Gambling Studies, 13, 291–296. Gerstein, D. R., Volberg, R. A., Toce, M. T., Harwood, H., Johnson, R. A., Buie, T., et al. (1999). Gambling Impact and Behavior Study: Report to the National Gambling Impact Study Commission. Chicago: NORC/University of Chicago. Griffiths, M. (1993). Fruit machine gambling: The importance of structural characteristics. Journal of Gambling Studies, 9, 101–120. Hasin, D. (2003). Classification of alcohol use disorders. Alcohol Research and Health, 27, 5–17. Haynes, S. N. (2006). Behavioral assessment of adults. In Handbook of Psychological Assessment (G. Goldstein and M. Hersen, eds.), pp. 423–463. Elmsford, NY: Pergamon Press. Hodgins, D. C., Currie, S. R., and el-Guebaly, N. (2001). Motivational enhancement and self-help treatments for problem gambling. Journal of Consulting and Clinical Psychology, 69, 50–57. Hodgins, D. C., and Makarchuk, K. (2003).Trusting problem gamblers: Reliability and validity of selfreported gambling behaviour. Psychology of Addictive Behaviors, 17, 244–248. Kaiser, J. (2003). Hormesis: Sipping from a poisoned chalice. Science, 302, 175–197. Korn, D. (2000). Expansion of gambling in Canada: Implications for health and social policy. Canadian Medical Association Journal, 163, 61–64. Korn, D., and Shaffer, H. J. (1999). Gambling and the health of the public: Adopting a public health perspective. Journal of Gambling Studies, 15, 289–365. Ladouceur, R., Sylvain, S., Boutin, C., and Doucet, C. (2002). Understanding and Treating Pathological Gamblers. New York:Wiley. Ladouceur, R., and Walker, M. (1996). A cognitive perspective on gambling. In Trends in Cognitive and Behavioral Therapies (P. M. Salkovski, ed.). New York:Wiley. Lesieur, H., and Blume, S. (1987). The South Oaks Gambling Screen (SOGS): A new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, 1184–1188. Loroz, P. S. (2004). Golden-age gambling: Psychological benefits and self-concept dynamics in aging consumers’ consumption experiences. Psychology and Marketing, 21, 323–349. MacDonald, M., McMullan, J. L., and Perrier, D. C. (2004). Gambling households in Canada. Journal of Gambling Studies, 20, 187–236. Marshall, K., and Wynne, H. (2003). Fighting the odds. Perspectives [Catalogue no. 75-001-XIE], 5-12. Ottawa: Statistics Canada. Miller,W. R., Sovereign, R. G., and Krege, B. (1988). Motivational interviewing with problem drinkers: II.The Drinker’s Check-up as a preventive intervention. Behavioral Psychotherapy, 16, 251–268. National Council of Welfare. (1996). Gambling in Canada. Ottawa: Minister of Supply and Services. Petry, N. M. (2005). Pathological Gambling: Etiology, Comorbidity, and Treatment.Washington, DC:American Psychological Association. Roemer, R. (2004). A brief history of legislation to control the tobacco epidemic. In Boyle et al., pp. 677–695. Shaffer, H. J., Hall, M. N., and Vander Bilt, J. (1997). Estimating the prevalence of disordered gambling behavior in the United States and Canada:A research synthesis. American Journal of Public Health, 89, 1369–1376. Shaffer, H. J., and Korn, D. (2002). Gambling and related mental disorders: A public health analysis. Annual Review of Public Health, 23, 171–212. Shaffer, H. J., LaBrie, R. A., LaPlante, D. A., Nelson, S. E., and Stanton, M.V. (2004).The road less travelled: Moving from distribution to determinants in the study of gambling epidemiology. Canadian Journal of Psychiatry / Revue Canadienne de Psychiatrie, 49, 504–516. Smith, G., and Wynne, H. (2000). A Review of Gambling Literature in the Economic and Policy Domains. Edmonton, AB.

Quantification and Dimensionalization of Gambling Behavior

177

Statistics Canada. (2002a). Canadian Community Health Survey (Cycle 1.2)—Mental Health and Well-being User’s Guide. Ottawa: Author. ––––– . (2002b). Canadian Tobacco Use Monitoring Survey Annual Results Ottawa: Author. ––––– . (2003). Survey of Household Spending. Ottawa: Author. Stevens, K. R., and Munoz, L. R. (2004). Focus on research methods: Cigarette smoking: Evidence to guide measurement. Research in Nursing and Health, 27, 281–292. Walker, M., Toneatto, T., Potenza, M., Petry, N., Ladouceur, R., Hodgins, D.C., el-Guebaly, N., Echeburua, E., and Blaszczynski, A. (2006). A framework for reporting outcomes in problem gambling treatment research:The Banff, Alberta consensus. Addiction, 101, 504–511. Weinstock, J.,Whelan, J. P., and Meyers,A.W. (2004). Behavioral assessment of gambling:An application of the Timeline Followback Method. Psychological Assessment, 16, 72–80. Williams, R. J., and Connolly, D. (2006). Does learning about the mathematics of gambling change gambling behavior? Psychology of Addictive Behaviors, 20, 62–68. Williams, R. J., Connolly, D.,Wood, R., Currie, S. R., and Davis, R. M. (2004). Program findings that inform curriculum development for the prevention of problem gambling. Gambling Research, 16, 47–69. Williams, R. J., and Wood, R.T. (2004).The proportion of gaming revenue derived from problem gamblers: Examining the issues in a Canadian context. Analyses of Social Issues and Public Policy, 4, 33–45. Wood, R., and Williams, R. J. (2007).“How much money do you spend on gambling?” The comparative validity of question wordings used to assess gambling expenditure. International Journal of Social Research Methodology:Theory and Practice.

This page intentionally left blank

CHAPTER 8

A Review of Screening and Assessment Instruments for Problem and Pathological Gambling Randy Stinchfield

Richard Govoni

Department of Psychiatry University of Minnesota Medical School Minneapolis, Minnesota

University of Windsor Windsor, Ontario, Canada

G. Ron Frisch University of Windsor Windsor, Ontario, Canada

Introduction Instruments Gamblers Anonymous 20 Questions (GA-20) South Oaks Gambling Screen (SOGS) Massachusetts Gambling Screen (MAGS) DSM-IV-MR (MR = Multiple Response) Diagnostic Interview for Gambling Severity (DIGS) Gambling Treatment Outcome Monitoring System (GAMTOMS) National Opinion Research Center DSM-IV Screen for Gambling Problems (NODS) Lie/Bet Questionnaire Gambling Assessment Module (GAM) Canadian Problem Gambling Index (CPGI) Gambling Behavior Interview (GBI) Clinical Global Impression Scale (CGI) Pathological Gambling Adaptation of the Yale Brown Obsessive-Compulsive Scale (PG-YBOCS) Gambling Symptom Assessment Scale (G-SAS) Structured Clinical Interview for Pathological Gambling (SCI-PG) Conclusions and Future Research Directions 179

180

Research and Measurement Issues in Gambling Studies

INTRODUCTION Problem gambling has vexed humans for centuries, but it has only been within the last few decades that assessment instruments have been developed to identify problem and pathological gamblers. Legalized gambling as an industry has experienced unprecedented growth and expansion over the past three decades, and along with this rapid growth have come concerns about problem gambling.There is a need to identify problem gamblers in the general population in order to determine the extent of the problem in society and to aid public policy planning, such as the provision of treatment and prevention programs for problem gambling. Furthermore, mental health care agencies need to be able to accurately screen for and diagnose pathological gambling (PG) in order to provide appropriate treatment services. In 1990, a published critical review of existing instruments included only two instruments (Volberg and Banks 1990). There now exist over a dozen problem-gambling instruments that have been developed for a variety of purposes, including screening, assessment, diagnosis, epidemiological surveys, research, treatment planning, and treatment outcome monitoring. These instruments range in length from as few as two items to more than one hundred items. Since new instruments continue to be developed, this is an opportune time to examine what instruments are available, compare the strengths and limitations of existing instruments, and make recommendations for future refinement of existing instruments as well as for future instrument development. It should also be noted that there have been fairly large differences in reported prevalence rates in epidemiological surveys of problem gambling—from as low as less than 1% to as high as 10%—and at least part of this disparity may be attributed to a lack of precision in current measurement efforts. The current diagnostic criteria for PG have been established by the American Psychiatric Association (APA) (1994) in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). In general, the APA has taken an objective, behavioral approach to diagnostic criteria. Although it has placed PG in the impulse control disorder section of the DSM, the diagnostic criteria are very similar to substance use disorders diagnostic criteria and share a number of the signs and symptoms found in substance use disorders, such as tolerance and withdrawal. Instruments based on DSM diagnostic criteria can be expected to inquire about consequences of gambling, attempts at controlling one’s gambling, and changes in gambling behavior that may indicate tolerance and withdrawal syndromes. A number of assessment instruments included in this review are based on DSM diagnostic criteria or include some of the diagnostic criteria. Most problem-gambling instruments are relatively new and have not undergone rigorous reliability, validity, and classification accuracy evaluation (National Research Council 1999).There is also a paucity of research on the measurement of problem gambling among special populations, such as youth. At this

Screening and Assessment Instruments for Problem Gambling

181

point, the assessment of special populations has been conducted either by making revisions to instruments developed for adults, as in the case of youth gambling, or by assuming that existing adult instruments are appropriate, as in the case of seniors.The clinical experience of those who work with special populations suggests that the signs and symptoms of problem gambling may be somewhat different in these segments of the population. Both researchers and clinicians are confronted with the challenge of selecting from among existing instruments, many of which have little, if any, reliability or validity information for the task at hand. The primary aim of this review is to describe the instruments currently available and to provide information about each instrument, including development, author(s), year of development, content, number of items, administration method and time, intended purpose of the instrument, psychometric properties (reliability, validity, and classification accuracy), norms, scoring instructions, interpretation of scores, and strengths and limitations. See Table 8.1 for a description of each instrument. Due to space limitations, this review is confined to instruments designed for adults and that are in current use and have at least minimal evidence of reliability and validity. For reviews of other instruments, including youth instruments, the reader is referred to Stinchfield, Govoni, and Frisch (2001, 2004).

INSTRUMENTS GAMBLERS ANONYMOUS 20 QUESTIONS (GA-20) Gamblers Anonymous (GA) uses a set of 20 questions for the purpose of identifying compulsive gamblers. A score of 7 or higher indicates that the respondent is a compulsive gambler. The items address behaviors related to compulsive gambling, such as remorse over gambling, gambling to forget problems, borrowing money to gamble, and difficulty sleeping, to name a few. Although the GA-20 is commonly used, there is little psychometric and classification accuracy information available. Ursua and Uribelarrea (1998) note that there are no published reports describing the development of the GA-20 and only two studies that report any psychometric information. The earliest known validity evidence for the GA-20 was reported by Kuley and Jacobs (1988), who found that the GA-20 yielded high correlations with frequency of gambling and with dissociative experiences. Ursua and Uribelarrea (1998) conducted a study of the psychometric properties of the GA-20 in a sample of 127 problem gamblers who came for treatment at two self-help agencies in Madrid, Spain, and they also administered the GA-20 to a comparison sample of 142 nonproblem, social gamblers matched on age and gender with the problem gamblers. The internal consistency of the GA-20, using Cronbach’s (1951)

182

Table 8.1 Descriptions of Instruments. Content Areas

Number of Items and Response Options

Administration Time and Method

Scoring Instructions, Score Range, Cut Scores, and Interpretation of Scores

Gamblers Anonymous 20 questions (GA-20)

Signs and symptoms of compulsive gambling; negative consequences

20 items; true/false response option

10-minute paperand-pencil or interview

One point for each item; score of 7 or more indicates compulsive gambler.

South Oaks Gambling Screen (SOGS) (1987)

Games played; signs and symptoms of problem gambling; negative consequences; sources of money to gamble

20 scored items; response options vary

10–20-minute paperand-pencil questionnaire

One point for each item; score range 0–20; score of 5 or more indicates PPG.

Massachusetts Gambling Screen (MAGS) (1994)

Signs and symptoms of pathological gambling; psychological and social problems associated with gambling; this study also includes a 12-item measure of DSM-IV diagnostic criteria

14 items (7 items are scored)

5–10-minute paper-and- 7 MAGS items are scored by multiplying pencil questionnaire each item times a discriminant function coefficient; then sum and add a constant; score range 0–2 = transitional or potential pathological gambler; score >2 = pathological gambling

DSM-IV-MR (2000)

DSM-IV diagnostic criteria

10 items, one for each criterion; four-point response options for most items

5-minute questionnaire

One point for each item; score range 0–10; score of 3–4 (including at least on point from criteria 8, 9, or 10) is a problem gambler; score of 5 or more is severe problem gambler.

Diagnostic Interview for Gambling Schedule (DIGS)

Demographics, gambling 20 diagnostic symptom involvement, treatment items to measure the history, onset of gambling, 10 DSM-IV diagnostic

30-minute interview

If respondent endorses either of the two items per criterion, the criterion is considered endorsed. One point

Research and Measurement Issues in Gambling Studies

Name of Instrument (year)

criteria.Two items for each criterion

for each of the 10 criteria. Score range 0–10; cut score of >5 indicates PG.

Gambling Treatment The Gambling Treatment Outcome Monitoring Admission Questionnaire System (GAMTOMS) includes a 10-item measure of DSM-IV diagnostic criteria for PG, as well as other measures of gambling problem severity, including the SOGS, gambling frequency, gamblingrelated financial problems, and legal problems.

142-item Gambling Treatment Admission Questionnaire has a 10-item measure of DSM-IV diagnostic criteria

30–45-minute paperand-pencil questionnaire

The DSM-IV diagnostic criteria items are one point each and are summed. Score range is 0–10; cut score of >5 indicates PG.

National Opinion Research Center DSM-IV Screen for Gambling Problems (NODS) (1999)

17 items

5–10-minute interview for NODS

NODS is scored one point for each DSM criterion. Score range 0–10; score of 0 = low-risk gambler; 1 or 2 = at-risk gambler; 3 or 4 = problem gambler; and >5 = PG

DSM-IV diagnostic criteria for diagnosing PG including lifetime and past-year time frames. A filtering question of losing $100 or more was used prior to administration of NODS.

Screening and Assessment Instruments for Problem Gambling

gambling frequency, amounts of money bet and lost, sources of borrowed money, financial problems, legal problems, mental health screen, other impulse disorders, medical status, family and social functioning, and diagnostic symptoms (lifetime and past year)

(Continues)

183

184

Table 8.1 (Continued) Content Areas

Number of Items and Response Options

Administration Time and Method

Scoring Instructions, Score Range, Cut Scores, and Interpretation of Scores

Lie/Bet (1997)

Lie to people about your gambling; bet more and more money

2 items; yes/no response option

1-minute interview

Answering yes to one or both items indicates PG.

Gambling Assessment Module (GAM) and Computerized GAM

Structured gambling Demographics section = diagnostic interview that 27 items; gambling has three modules: section = 40 items; Demographics, Gambling, interviewer and Interviewer observations = observations.The gambling 7 items module includes items assessing gambling frequency and diagnostic criteria

30–60-minutes; interview (paperand-pencil or computerized)

A score of >5 out of 10 DSM-IV criteria indicates PG; 11 algorithms for the activity-specific diagnoses

Canadian Problem cates Gambling Index (CPGI) (2001)

Gambling involvement,

15-minute interview

Score range is 0–27. Score of 0 indi-

Gambling Behavior Interview (GBI) (2001)

Clinical interview to measure 76 items, including 20 signs and symptoms of PG, SOGS, 10 DSM-IV including gambling diagnostic criteria, and frequency, amount of time 32 research items

problem gambling, adverse consequences, family history of gambling, comorbid disorders, and distorted cognitions

31 total; 9-item problem gambling scale; four response options: never = 0; sometimes = 1; most of the time = 2; and almost always = 3

nonproblem gambling; 1–2 indicates low risk gambling; 3–7 indicates moderate risk gambling; and 8 or more indicates problem gambling. 30–60-minute interview

DSM score of 5 or more indicates PG; 20-item research scale uses item weights; 5-item screen score of >2 indicates probable PG.

Research and Measurement Issues in Gambling Studies

Name of Instrument (year)

and money spent gambling, the SOGS, DSM-IV, and 32 research items with a past-year time frame Three items: (a) severity of illness, (b) rating of improvement, and (c) efficacy index. It is used primarily in pharmacological studies

Three items; improvement Clinician-administered item is rated on a 7-point interview that takes Likert scale from very about 5 minutes much improved to very much worse

NA

Pathological Gambling Severity of pathological Adaptation of the gambling symptoms over a Yale-Brown recent time period (usually Obsessive-Compulsive within the past 1 or 2 Scale (PG-YBOCS) weeks). Gambling thoughts/ urges and behavior

Ten items, rated on a Clinician-administered 5-point Likert scale interview that takes ranging from least severe about 10 minutes (0) to most severe (4).

Each set of questions is totaled separately as well as one total score. No interpretation of scores was provided since it is used as a measure of change.

Gambling Symptom Assessment Scale (G-SAS) (2001)

Gambling urges, thoughts, feelings, and behavior

10 items; four-point Likert response options

Clinician-administered interview that takes about 10 minutes

Scores range from 0–40.

Structured Clinical Interview for Pathological Gambling (SCI-PG) (2004)

DSM-IV diagnostic criteria for pathological gambling

11 screening items; 33 PG diagnostic items

10–20-minute interview

Five or more of the ten DSM diagnostic criteria and evidence that the gambling is not better accounted for by a manic episode indicate PG (Continues)

Screening and Assessment Instruments for Problem Gambling

Clinical Global Impression (CGI) (1976)

185

Name of Instrument Reliability α = .94 (Ursua and Uribelarrea, 1998)

Classification Accuracy Indices Sample Characteristics, Criterion, Base Rate, Sensitivity, Specificity, and Hit Rate

Kuley and Jacobs (1988) Ursua and Uribelarrea (1998). Criterion is group membership, 127 problem gamblers report that the GA-20 142 nonproblem social gamblers; base rate = .47; sensitivity = .98; specificity = .99; yielded high correlations hit rate = .99. It should be noted that these classification accuracy indices are based with frequency of gambling upon a sample with a base rate of about 50%, which inflates classification accuracy and with dissociative indices. experiences; GA-20 was highly correlated with the SOGS (r = .94). Ursua and Uribelarrea (1998)

SOGS

α = .97; onemonth testretest reliability r = .71

Correlations with counselor assessments (r = .86), family member assessment (r = .60), and DSM-III-R pathological gambling diagnosis (r = .94)

GA members (n=213), university students (n=384), and hospital employees (n=152). Criterion was DSM-III-R diagnosis of PG hit rates among GA members (.98), university students (.95), and hospital employees (.99);

MAGS

MAGS 7-item scale α = .84; DSM-IV 12item scale α = .89

MAGS total discriminant score was correlated with total DSM-IV score, r = .83

NA

DSMIV-MR

α = .79

Discriminated between NA regular and nonregular gamblers and between problem and social gamblers

Research and Measurement Issues in Gambling Studies

GA-20

Validity

186

Table 8.1 (Continued) Psychometrics

α = .92

The total diagnostic score (0–10) exhibited significant correlations with the following measures of gambling problem severity: gambling frequency, r = .39; highest amount gambled in one day, r = .42; current gambling debt, r = .47; number of financial problems, r = .40; number of borrowing sources, r = .31; and legal problems, r = .50.

NA

GAMTOMS

Internal consistency reliability: DSMIV diagnostic criteria (α = 89), SOGS (α = .85), and financial problems (α = .78); 1week test–retest yielded correlations of r = .74 for DSM-IV; r = .91 for SOGS;

Validity of the DSM-IV diagnostic criteria was measured by correlations with the following measures of gambling problem severity: SOGS (r = .83); gambling frequency (r = .43); and number of financial problems (r = .40).

DSM-IV diagnosis of PG was used to classify clinical versus nonclinical cases: base rate = .20; hit rate = .96; sensitivity = .96; specificity = .95; false positive rate = .01; and false negative rate = .14. DSM-IV diagnosis of PG was used to classify SOGS PPG versus non-PPG cases: base rate = .79; hit rate = .98; sensitivity = .97; specificity = 1.00; false positive rate = .00; and false negative rate = .10.

NODS

2-4 week test–retest coefficients of r = .99 and r = .98 for lifetime and past year, respectively

NODS was administered to 40 individuals in outpatient problem-gambling treatment programs. Of these 40, 38 scored >5 on the lifetime NODS and two obtained scores of 4. For past year NODS, 30 scored >5, 5 scored 3 or 4, and 5 scored