Alcohol, Young Drivers, and Traffic Accidents

Alcohol, Young Drivers, and Traffic Accidents Effects of Minimum-Age Laws

Alexander C. Wagenaar The University of Michigan

Lexington Books D.C. Heath and Company Lexington, Massachusetts Toronto

Fig

Tal Acl

Library of Congress Cataloging in Publication Data

Chapter 1

Int]

Chapter 2

Y 01

Chapter 3

Ale Dri Res

Wagenaar, Alexander C. Alcohol, young drivers, and traffic accidents.

Res

Bibliography: p. 135 Includes index. I. Drinking and traffic accidents-United States. 2. Youth-United States-Alcohol use. 1. Title. HE5620.D7W33 1983 363.1'251 83-47515 ISBN 0-669-06696-6

Mo Spe

Chapter 4

Res of,

Res Dat Dat

Copyright © 1983 by D.C. Heath and Company

Chapter 5

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 or retrieval system, without permission in writing from the publisher.

Dis Eff

Leg Enl

Third printing, June 1985 Published simultaneously in Canada

Chapter 6

Sell

Printed in the United States of America on acid-free paper

Appendix A

Res Stal

International Standard Book Number: 0-669-06696-6

Int(

Library of Congress Catalog Card Number: 83-47515

COl

Ext

J.

Contents

m Data

Figures

vii

Tables

ix

Acknowledgments

xi

Chapter 1

Introduction

Chapter 2

Youth, Alcohol, and Highway Safety

Chapter 3

Alcohol Availability and the Minimum LegalDrinking Age Research on the Legal-Drinking Age and Highway Crashes Research on the Legal-Drinking Age and Alcohol Consumption Model of the Effects of Changes in the LegalDrinking Age Specific Research Questions

30 35

Research Design, Data Collection, and Methods of Analysis Research Design Data Collection Data-Analysis Methods

37 37 41 50

Chapter 4

Chapter 5

eproduced or transincluding system, without

9 17 18 27

53

Discussion of Results Effects of the Raised Drinking Age on Crash Involvement Legal-Drinking Age and Beverage-Alcohol Sales Enforcement of the Minimum Legal-Drinking Age

53 80 95

Chapter 6

Selecting the Best Minimum Drinking Age

101

Appendix A

Research-Design Validity Statistical-Conclusion Validity Internal Validity Construct Validity External Validity

113 113 115 119 121

v

vi Appendix B

Alcohol, Young Drivers, and Traffic Accidents Data-Analysis Methods

125

Bibliography

135

Index

147

About the Author

153

___________h_"

3-1

Model of Drinking

4-1

Nonequiv

4-2

First LeVl

4-3

Second L

4-4

Third Le'

5-1

Police-Re Involved

5-2

SVNM H Damage (

5-3

Police-Re Involved

5-4

SVNMH Crashes iJ

5-5

Police-Re Involved.

5-6

SVNM 24 Crashes il

5-7

Police-Re Vehicle C

5-8

Police-Re Involved.

5-9

SVNM IB Damage (

5-10

Police-Re Involved]

5-11

SVNM 16 Damage (

5-12

SVNM 21 Damage (

5-13

Male 18-2 Daytime,

_

Traffic Accidents

Figures

125 135

147 153

Model of the Impact of Changes in the LegalDrinking Age on Motor-Vehicle-Crash Involvement

32

4-1

Nonequivalent Multiple-Time-Series Design

38

4-2

First Level of Research-Design Comparisons

38

4-3

Second Level of Research-Design Comparisons

39

4-4

Third Level of Research-Design Comparisons

40

5-1

Police-Reported 18-20-Year-Old Drinking Drivers Involved in Property-Damage Crashes in Michigan

56

SVNM 18-20-Year-Old Drivers Involved in PropertyDamage Crashes in Michigan

57

Police-Reported 18-20-Year-Old Drinking Drivers Involved in Injury Crashes in Michigan

59

SVNM 18-20-Year-Old Drivers Involved in Injury Crashes in Michigan

60

Police-Reported 24-45-Year-Old Drinking Drivers Involved in Injury Crashes in Michigan

61

SVNM 24-45-Year-Old Drivers Involved in Injury Crashes in Michigan

62

Police-Reported Drinking Drivers Involved in MotorVehicle Crashes in Michigan

63

Police-Reported 18-19-Year-Old Drinking Drivers Involved in Property-Damage Crashes in Maine

65

SVNM 18-19-Year-Old Drivers Involved in PropertyDamage Crashes in Maine

66

Police-Reported 16-17-Year-Old Drinking Drivers Involved in Property-Damage Crashes in Maine

67

SVNM 16-17-Year-Old Drivers Involved in PropertyDamage Crashes in Maine

67

SVNM 21-23-Year-Old Drivers Involved in PropertyDamage Crashes in New York

72

Male 18-20-Year-Old Drivers Involved in Single-Vehicle, Daytime, Property-Damage Crashes in New York

73

3-1

5-2 5-3 5-4 5-5 5-6 5-7 5-8 5-9 5-10 5-11 5-12 5-13

vii

viii

Alcohol, Young Drivers, and Traffic Accidents

5-14

Package-Beer Distribution in Michigan

83

5-15

Draft-Beer Distribution in Michigan

84

5-16

Beer Distribution in Maine

88

5-17

Wine Distribution in Maine

89

5-18

Package-Beer Distribution in New Hampshire

91

5-19

Draft-Beer Distribution in New Hampshire

92

5-20

Draft-Beer Distribution in the United States

94

5-21

Number of Citations for Selling or Allowing Minors to Consume Alcoholic Beverages, Michigan

96

5-22 B-1 B-2 B-3 B-4 B-5 B-6

1-1

States L( 1970-197

1-2

States R; 1976-198

1-3

Current

4-1

Missing-

4-2

Missing-:

4-3

Missing-:

Number of Citations for Selling or Allowing Minors to Consume Alcoholic Beverages, Maine

97

4-4

Negative Impact Pattern Estimated by the rsb (1,0,0) Transfer-Function Model with a Step-Function Input

Missing1979

129

5-1

Positive Impact Pattern Estimated by the rsb (1,0,0) Transfer-Function Model with a Step-Function Input

129

Estimate Involven Drinking

5-2

Estimate Associat MichigaJ

Negative Impact Pattern Estimated by the rsb (0,0,0) Transfer-Function Model with a Step-Function Input

130

Positive Impact Pattern Estimated by the rsb (0,0,0) Transfer-Function Model with a Step-Function Input

130

5-3

Negative Impact Pattern Estimated by the rsb (1,0,0) Transfer-Function Model with a Pulse-Function Input

Number Influenc

131

5-4

Positive Impact Pattern Estimated by the rsb (1,0,0) Transfer-Function Model with a Pulse-Function Input

Estimate Involven

132

5-5

Estimate Associat Maine

5-6

Estimat( Involver Maine R

5-7

Estimat( New Yo Legal-D

5-8

Estimate Involver Michiga

1-

Traffic Accidents

Tables

83 84 88

1-1

States Lowering the Minimum Legal-Drinking Age, 1970-1975

3

States Raising the Minimum Legal-Drinking Age, 1976-1983

4

1-3

Current Minimum Legal-Drinking Ages, January 1983

6

4-1

Missing-Data Rates for the State of Michigan, 1978

45

4-2

Missing-Data Rates for the State of Maine, 1979

46

4-3

Missing-Data Rates for the State of New York, 1979

47

4-4

Missing-Data Rates for the State of Pennsylvania, 1979

49

Estimated Change in Property-Damage-Crash Involvement Associated with Raising the LegalDrinking Age in Michigan

55

Estimated Change in Injury-Crash Involvement Associated with Raising the Legal-Drinking Age in Michigan

58

Number of Drivers Arrested for Driving under the Influence of Liquor in Michigan

62

Estimated Change in Property-Damage-Crash Involvement Associated with Raising the LegalDrinking Age in Maine

64

Estimated Change in Injury-Crash Involvement Associated with Raising the Legal-Drinking Age in Maine

68

Estimated Change in Property-Damage-Crash Involvement in New York after October 1977, when Maine Raised the Legal-Drinking Age

69

Estimated Change in Injury-Crash Involvement in New York after October 1977, when Maine Raised the Legal-Drinking Age

70

Estimated Change in Property-Damage-Crash Involvement in New York after December 1978, when Michigan Raised the Legal-Drinking Age

71

89 pshire

91

lire

92

:ates

94

wing Minors igan

96

,wing Minors

e

97

he rsb (1,0,0) unction Input

129

Ie rsb (1,0,0) unction Input

129

1-2

5-1

5-2

he rsb (0,0,0) unction Input

130

Ie rsb (0,0,0) unction Input

130

5-3

:he rsb (1,0,0) Function Input

131

5-4

Ie rsb (1,0,0) Function Input

132

5-5

5-6

5-7

5-8

ix

x 5-9

5-10

5-11

5-12

5-]3

5-14 5-15 5-16 5-17

Alcohol, Young Drivers, and Traffic Accidents Estimated Change in Injury-Crash Involvement in New York after December 1978, when Michigan Raised the Legal-Drinking Age

74

Estimated Change in Property-Damage-Crash Involvement in Pennsylvania after October 1977, when Maine Raised the Legal-Drinking Age

75

Estimated Change in Injury-Crash Involvement in Pennsylvania after October 1977, when Maine Raised the Legal-Drinking Age

76

Estimated Change in Property-Damage-Crash Involvement in Pennsylvania after December 1978, when Michigan Raised the Legal-Drinking Age

77

Estimated Change in Injury-Crash Involvement in Pennsylvania after December 1978, when Michigan Raised the Legal-Drinking Age

78

Estimated Change in Average Monthly BeverageAlcohol Sales in Michigan

82

Estimated Change in Average Monthly BeverageAlcohol Sales in Maine

87

Estimated Change in Average Monthly BeverageAlcohol Sales in New Hampshire

90

Estimated Change in Average Monthly BeverageAlcohol Sales in the United States

93

The help of several i this book. The rese States National Insti vehicle crashes wen ment of Transporta York Division of P Department of Trar vided by the Alcoholic dation. Data on cit. to minors were prov Michigan Liquor 0 Several people I book, including Ly Laurence Ross, lohl go to Robert L. He University of Michil during the final pre The contributiol acknowledged. The main solely my owr Permission was in the following jOt: "Effects of an I oj Public Health PI "Aggregate Legal Minimum Dr posit Law in Mid

Studies on Alcohol Alcohol, Inc., New "Preventing Hi. Drinking: An Emp:

Journal ojSajety R Ltd. "Public Policy Hampshire: 1970-1;

L

1..-

iffic Accidents

Acknowledgments

ment in ligan Raised 74 ash Involvewhen Maine 75 ment in line Raised 76 ash Involvei, when

77

ment in \1ichigan 78 verage82 verage87 verage90

:verage-

The help of several individuals and organizations aided in the completion of this book. The research reported here was partially funded by the United States National Institute on Alcohol Abuse and Alcoholism. Data on motorvehicle crashes were obtained with the cooperation of the Maine Department of Transportation, the Michigan Department of State Police, the New York Division of Alcoholism and Alcohol Abuse, and the Pennsylvania Department of Transportation. Data on alcoholic-beverage sales were provided by the United States Brewers Association, the Maine Bureau of Alcoholic Beverages, and the Michigan Beer and Wine Wholesalers Association. Data on citations for violations of state liquor laws concerning sales to minors were provided by the Maine Department of Public Safety and the Michigan Liquor Control Commission. Several people provided helpful comments on an earlier draft of this book, including Lyle D. Filkins, James Haney, William T. Pollock, H. Laurence Ross, John Wagenaar, and Theodore C. Wagenaar. Special thanks go to Robert L. Hess, William E. McCormick, and James O'Day, of the University of Michigan Transportation Research Institute, for their support during the final preparation of the manuscript. The contribution of all these individuals and organizations is gratefully acknowledged. The findings and conclusions reported here, however, remain solely my own. Permission was kindly given to reproduce material that I have published in the following journal articles:

93 "Effects of an Increase in the Legal Minimum Drinking Age," Journal of Public Health Policy 2(3) (1981):2J6-225. "Aggregate Beer and Wine Consumption: Effects of Changes in the Legal Minimum Drinking Age and a Mandatory Beverage Container Deposit Law in Michigan." Reprinted with permission from Journal of Studies on Alcohol 43 (1982):469-487. Copyright by Journal of Studies on Alcohol, Inc., New Brunswick, N.J. "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Reprinted with permission from Journal ofSafety Research 13 (1982):57-71. Copyright by Pergamon Press, Ltd. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980," Contemporary Drug Problems 11 (1982):3-202. xi

xii


"Raising the Legal Drinking Age in Maine: Effects on Traffic Accidents Among Young Drivers," International Journal of the Addictions 18 (1983):365-377. Reprinted with permission of Marcel Dekker, Inc., N.Y.

• The massive socia come under incre; estimates vary, th cost U.S. society t damage, and lost and Brand 1980).. fering are associat Efforts to pre' centrated on youn: lems and reduce proaches that have among young peo] is designed to redu ways that might r phasizes restrictin populations. This in legislation that The core ques legal-drinking age those under the minimum-age laVi related automobil focus. First, mot( and death among the past decade cc ing age has center of accidents amOl Minimum-age nineteenth centur: population; drink high (Aaron and spirits were consi alcohol distributil to alcoholic beve their drinking wal

L t h o , e in the growi

j Traffic Accidents

Effects on Traffic Acmal of the Addictions 18 :cel Dekker, Inc., N.Y.

1

Introduction

The massive social costs associated with the use of beverage alcohol have come under increased critical scrutiny in recent years. Although specific estimates vary, there is universal agreement that alcohol-related problems cost V.S. society billions of dollars in health care, social services, property damage, and lost production (Berry and Boland 1977; Schifrin, Hartsog, and Brand 1980). In addition to these economic costs, untold pain and suffering are associated with the use of alcoholic beverages. Efforts to prevent alcohol-related problems have frequently been concentrated on young people, to combat development of alcohol-related problems and reduce their effects. Education and law are two general approaches that have been used in efforts to prevent alcohol-related problems among young people. The educational approach emphasizes persuasion and is designed to reduce alcohol consumption or to encourage use of alcohol in ways that might reduce alcohol-related problems. The legal approach emphasizes restricting the availability of beverage alcohol to specific age populations. This book focuses on the legal approach, commonly expressed in legislation that specifies the legal-drinking age. l The core question is: Does state legislation that establishes a minimum legal-drinking age reduce the incidence of alcohol-related problems among those under the age limit? This book focuses on the effectiveness of minimum-age laws in reducing young people's involvement in alcoholrelated automobile crashes. Two factors highlight the importance of this focus. First, motor-vehicle crashes are the leading cause of serious injury and death among young people (National Safety Council 1980). Second, in the past decade controversy over the establishment of an appropriate drinking age has centered on whether it would exacerbate or reduce the problem of accidents among youth. Minimum-age laws are fairly new. In the late eighteenth and early nineteenth centuries, beverage alcohol was readily available to the V.S. population; drinking was frequent; and per-capita-consumption rates were high (Aaron and Musto 1981; Clark 1976; Mosher 1980).2 Beer, cider, and spirits were considered wholesome. There was little formal regulation of alcohol distribution: in particular, there were no legal restrictions on access to alcoholic beverages by young people. Most young people drank, and their drinking was not seen as a problem. In the late nineteenth century even those in the growing temperance movement did not concentrate their efforts

1

2

"

I


on preventing access to alcohol by youth. Instead, temperance advocates focused on restricting the availability of alcohol to the general population through regulation of hours of sale, location of alcohol outlets, and high license fees. Concern about the availability of alcoholic beverages to youth emerged at the beginning of the twentieth century. Legal restrictions were initiated as part of a general trend toward increasing state controls on adolescent behavior. Although early laws prohibited sales of alcohol to young people, they did not prohibit drinking by youth nor did they prohibit parents or other adults from providing alcoholic beverages to minors. Strict minimumage laws were implemented only after repeal of Prohibition. All fifty states then passed laws that specified a minimum drinking age; most states set the age at twenty-one. Furthermore, most states prohibited virtually all drinking by underage individuals, making it illegal for any person, including parents, to provide youth with alcoholic beverages. 3 Minimum-age laws attracted little attention from 1940 to 1970. The controversy surrounding the drinking age began in 1970, when Congress passed the Twenty-sixth Amendment to the U.S. Constitution, which extended the right to vote in federal elections to citizens between the ages of eighteen and twenty-one. This amendment was subsequently ratified by the states. A movement to extend the rights and privileges of adulthood to youth aged eighteen and over began. Within three years, all fifty states extended the right to vote in state elections to eighteen-year-olds. Consistent with the trend toward lower legal ages for voting, participation in binding contracts, and other matters, twenty-nine states reduced their minimum legal-drinking ages between 1970 and 1975. 4 Twenty-three states lowered the drinking age for all alcoholic beverages, usually from age twenty-one to age eighteen or nineteen. Of these states, Iowa and Montana first lowered the drinking age from twenty-one to nineteen, and later to eighteen; Illinois and Maryland reduced the legal age for beer and wine only; South Dakota lowered the age for low-alcohol-content beer only; and Virginia reduced the drinking age for regular beer only. West Virginia reduced the legal age for distilled spirits to age eighteen; the age for purchase of beer and wine had been eighteen since 1935. Wisconsin lowered the drinking age to eighteen for all beverages except beer, which eighteenyear-oIds had been legally permitted to use since 1933. In Oklahoma the drinking age for low-alcohol-content beer was lowered to age eighteen for male drinkers. The drinking age for men had been twenty-one, While the age for women was eighteen. The U.S. Supreme Court ruled that this differential legal-drinking age was unconstitutional. The effect of the court ruling was a reduction in the drinking age for men to age eighteen. This change is not included in the twenty-nine-state total because it was not initiated by the state. Table 1-1 lists states that lowered the drinking age, the effective dates

Introduction Table 1-1 States Lowering the

]\I

State Alabama Alaska Arizona Connecticut Delaware Florida Georgia Hawaii Idaho Illinois Iowa Iowa Maine Maryland Massachusetts Michigan Minnesota Montana Montana Nebraska New Hampshire New Jersey Rhode Island South Dakota Tennessee Texas Vermont Virginia West Virginia Wisconsin Wyoming

of the new laws, and (beer, wine, spirits). drinking age in the ea Following these r wisdom of lowering tl political, and industri: United States 1973; B, in the political debate vehicle crashes among ing age (Michigan COl these arguments were 1

Traffic Accidents , temperance advocates ) the general population lcohol outlets, and high erages to youth emerged rictions were initiated as controls on adolescent tlcohol to young people, they prohibit parents or minors. Strict minimum)hibition. All fifty states g age; most states set the bited virtually all drinkr any person, including 3

'rom 1940 to 1970. The in 1970, when Congress Constitution, which exzens between the ages of )sequently ratified by the lvileges of adulthood to years, all fifty states ex:en-year-olds. Consistent participation in binding reduced their minimum

Ir all alcoholic beverages, lineteen. Of these states, from twenty-one to ninereduced the legal age for : for low-alcohol-content ,r regular beer only. West age eighteen; the age for 1935. Wisconsin lowered beer, which eighteen1933. In Oklahoma the vered to age eighteen for twenty-one, while the age t ruled that this differenof the court ruling : eighteen. This change is it was not initiated by the 19 age, the effective dates

Introduction

3

Table 1-1 States Lowering the Minimum Legal-Drinking Age, 1970-1975 State

Effective Date

Alabama Alaska Arizona Connecticut Delaware Florida Georgia Hawaii Idaho

7/75 9/70 8/72 10/72 7/72 7/73 7/72 3/72 7/72

lllinois Iowa Iowa Maine Maryland Massachusetts Michigan Minnesota Montana Montana Nebraska New Hampshire New Jersey Rhode Island South Dakota Tennessee Texas Vermont Virginia West Virginia

9/73 4/72 7/73 6/72 7/74 3/73 1/72 6/73 7/71 7/73 6/72 6/73 1/73 3/72 7/72 5/71 8/73 7/71 7/74 6/72

Wisconsin

3/72

Wyoming

5/73

Description of Change 21 to 19-all beverages 21 to 19-all beverages 21 to 19-all beverages 21 to 18-all beverages 21 to 20-all beverages 21 to 18-all beverages 21 to 18-all beverages 20 to 18-all beverages 21 to 19-wine and distilled spirits; 20 to 19-beer 21 to 19-beer and wine only 21 to 19-all beverages 19 to 18-all beverages 20 to 18-all beverages 21 to 18-beer and light wine only 21 to 18-all beverages 21 to 18-all beverages 21 to 18-all beverages 21 to 19-all beverages 19 to 18-all beverages 20 to 19-all beverages 21 to 18-all beverages 21 to 18-all beverages 21 to 18-all beverages 19 to 18-3.2 beer only 21 to 18-all beverages 21 to 18-all beverages 21 to 18-all beverages 21 to 18-beer only 21 to 18-distilled spirits (beer and wine were 18 since 1935) 21 to 18-all beverages except beer, which has been 18 since 1933 21 to 19-all beverages

of the new laws, and beverage categories for which the age was lowered (beer, wine, spirits). All ten Canadian provinces also reduced the legaldrinking age in the early 1970s. Following these reductions, considerable controversy concerning the wisdom of lowering the drinking age arose in academic, law-enforcement, political, and industrial circles (Works 1973; Distilled Spirits Council of the United States 1973; Bowen and Kagay 1973; Zylman 1973). Some partisans in the political debate argued that huge increases in alcohol-related motorvehicle crashes among youth occurred immediately after lowering the drinking age (Michigan Council on Alcohol Problems 1973). However, many of these arguments were based on uncontrolled analyses of data collected during

i I

4


the first few months following implementation of lower drinking ages. Others argued .that observed increases in crashes resulted from changes in the reporting practices of police, growth in the population of young drivers, and long-term trends in both alcohol consumption and the incidence of traffic accidents (Zylman 1974). By the mid-1970s, controlled studies of the effects of lowered drinking ages began appearing in the United States and Canada. Although the magnitude of the impact estimates varied among the states and provinces analyzed, most investigations concluded that reductions in the legal-drinking age led to significant increases in alcohol-related crashes among young drivers. Several studies also reported increased consumption of alcoholic beverages following implementation of a lower drinking age. The trend toward reducing the drinking age reversed as evidence documenting the adverse effects of lower drinking ages accumulated. Since 1975 no state has lowered its drinking age. In fact, between September 1976 and January 1983 sixteen states increased the minimum age for purchase of all alcoholic beverages (table 1-2). Rhode Island and New Jersey increased the drinking age in two steps. Rhode Island increased the age from eighteen to nineteen in 1980 and from nineteen to twenty in 1981; New Jersey increased the age from eighteen to nineteen in 1980 and from nineteen to twenty-one

Table 1-2 States Raising the Minimum Legal-Drinking Age, 1976-1983 State Connecticut Florida Georgia Illinois Iowa Maine Maryland Massachusetts Michigan Minnesota Montana Nebraska New Hampshire New Jersey New Jersey New York Rhode Island Rhode Island Tennessee Texas Virginia

Effective Date 7/82 10/80 9/80 1/80 7/78 10/77 7/82 4/79 12/78 9/76 1/79 5/80 5/79 1/80 1/83 12182 7/80 7/81 6/79 9/81 7/81

Description of Change 18 to 18 to 18 to 19 to 18 to 18 to 18 to 18 to 18 to 18 to 18 to 19 to 18 to 18 to 19 to 18 to 18 to 19 to 18 to 18 to 18 to

19-all beverages 19-all beverages 19-all beverages 21-beer and wine only 19-all beverages 20-all beverages 21-all beverages 20-all beverages 21-all beverages 19-all beverages 19-all beverages 20-all beverages 20-all beverages 19-all beverages 21-all beverages 19-all beverages 19-all beverages 20-all beverages 19-all beverages 19-all beverages 19-off-premise beer only

Introduction

in 1983. Three other sl categories. IIlinois and twenty-one (the age for one), and Virginia increi purchase of beer for off states had increased the provinces have also rece Saskatchewan). Legislat sideration in several ott As of January alcoholic beverages was nineteen in fourteen sta District of Columbia ha' such as beer and wine, t for distilled spirits. Th alcoholic beverages in t states have established t hibition, the variability strates that the nation 1 minimum age for legal Several studies of tt made. Most report signi volved in motor-vehicle many studies contain me reported in this book We ing age, as the laws an alcohol-related problen youth. Chapter 2 examines drinking among youth. particular concern to p< reviews past research 0 alcohol availability, Th< may affect the practices is also discussed. Chapt methods used to evalu Maine and Michigan lowered their drinking a higher legal ages in the: drinking age from legal age to twenty, effe its drinking age from tw returned the legal age to I

j

Traffic Accidents

of lower drinking ages. :esulted from changes in ulation of young drivers, and the incidence of trafltrolled studies of the efn the United States and :imates varied among the lS concluded that reduc:reases in alcohol-related reported increased conlementation of a lower

reversed as evidence ages accumulated. Since between September 1976 num age for purchase of nd New Jersey increased 1the age from eighteen to 81; New Jersey increased 11 nineteen to twenty-one

1976-1983 Description of Change

) 19-all beverages ) 19-all beverages ) 19-all beverages ) 21-beer and wine only ) 19-all beverages ) 20-all beverages ) 21-all beverages ) 20-all beverages , 21-all beverages , 19-all beverages , 19-all beverages , 20-all beverages , 20-all beverages , 19-all beverages I 21-all beverages I 19-all beverages I 19-all beverages I 20-all beverages I 19-all beverages I 19-all beverages I 19-off-premise beer only

Introduction

5

in 1983. Three other states raised the drinking age for specific beverage categories. Illinois and Maryland raised the age for beer and wine to age twenty-one (the age for distilled spirits in both states was already twentyone), and Virginia increased the minimum age from eighteen to nineteen for purchase of beer for off-premise consumption. As of January 1983 nineteen states had increased the minimum drinking age since 1975. Two Canadian provinces have also recently increased their legal-drinking age (Ontario and Saskatchewan). Legislation to raise the drinking age is currently under consideration in several other states. As of January 1983 the minimum legal-drinking age for all types of alcoholic beverages was twenty-one in sixteen states, twenty in six states, nineteen in fourteen states, and eighteen in five states. Nine states and the District of Columbia had a mixed legal age, with a lower age for beverages, such as beer and wine, that have a lower alcohol content, and a higher age for distilled spirits. The legal ages for each of the major categories of alcoholic beverages in the fifty states are listed in table 1-3. Although all states have established the age between eighteen and twenty-one since Prohibition, the variability of minimum-age laws within this range demonstrates that the nation has yet to achieve a consensus on the appropriate minimum age for legal drinking. Several studies of the recent increases in legal-drinking ages have been made. Most report significant reductions in the number of young people involved in motor-vehicle crashes since implementation of the new laws, but many studies contain methodological flaws. The specific goal of the research reported in this book was to determine whether increases in the legal-drinking age, as the laws are currently implemented, are effective in reducing alcohol-related problems-in particular, motor-vehicle crashes-among youth. Chapter 2 examines literature on patterns of drinking and driving after drinking among youth. The aim is to clarify why these issues have been of particular concern to policymakers and citizens in recent years. Chapter 3 reviews past research on minimum drinking ages and the related issue of alcohol availability. The mechanism through which a simple change i'R law may affect the practices of drinking and driving after drinking among youth is also discussed. Chapter 4 presents the research design and data-analysis methods used to evaluate the impact of raising the legal-drinking age. Maine and Michigan were selected as the primary-study states. Both lowered their drinking ages in the early 1970s and returned to their original, higher legal ages in the late 1970s. As noted in table 1-1, Maine lowered its drinking age from twenty to eighteen, effective June 9, 1972, and returned the legal age to twenty, effective October 23, 1977. Similarly, Michigan lowered its drinking age from twenty-one to eighteen, effective January 1, 1972, and returned the legal age to twenty-one, effective December 23, 1978. When the

6


Table 1-3 Current Minimum Legal-Drinking Ages, January 1983

State

Beer Over 3.2% 3.2% or Alcohol Less Alcohol

Light

Wine Fortified

Distilled Spirits

19 19 19 19 19 Alabama 19 19 19 19 19 Alaska 19 19 19 19 19 Arizona 21 21 21 21 21 Arkansas 21 21 21 21 21 California 21 21 21 21 18 Colorado 19 19 19 19 19 Connecticut 20 20 20 20 20 Delaware 21 18 21 18 18 District of Columbia 19 19 19 19 19 Florida 19 19 19 19 19 Georgia 18 18 18 18 18 Hawaii 19 19 19 19 19 Idaho 21 21 21 21 21 Illinois 21 21 21 21 21 Indiana 19 19 19 19 19 Iowa 21 21 21 21 18 Kansas 21 21 21 21 21 Kentucky 18 18 18 18 18 Louisiana 20 20 20 20 20 Maine 21 21 21 21 21 Maryland 20 20 20 20 20 Massachusetts 21 21 21 21 21 Michigan 19 19 19 19 19 Minnesota 21 21 18a 21 18a Mississippi 21 21 21 21 21 Missouri 19 19 19 19 19 Montana 20 20 20 20 20 Nebraska 21 21 21 21 21 Nevada 20 20 20 20 20 New Hampshire 21 21 21 21 21 New Jersey 21 21 21 21 21 New Mexico 19 19 19 19 19 New York 21 21 18 18 18 North Carolina 21 21 21 21 21 North Dakota 21 21 11 18 21 Ohio 21 21 21 18 b 21 Oklahoma 21 21 21 21 21 Oregon 21 21 21 21 21 Pennsylvania 20 20 20 20 20 Rhode Island 21 18 18 18 18 South Carolina 21 21 21 21 18 South Dakota 19 19 19 19 19 Tennessee 19 19 19 19 19 Texas 21 21 21 21 21 Utah 18 18 18 18 18 Vermont 21 21 21 18c 18 c Virginia 21 21 21 21 21 Washington 18 18 18 18 d 18 West Virginia 18 18 18 18 18 Wisconsin 19 19 19 19 19 Wyoming aDrinking age is 18 for beer or wine that is 4 percent or less alcohol by weight. bprior to December 1976, the age was 18 for females and 21 for males. This sex discrimination was ruled unconstitutional by the U.S. Supreme Court, having the effect of lowering the age to 18 for males. Craig v. Boren, U.S. Ok!. 97 S. Ct. 451 (1976). CEighteen for on-premise consumption, 19 for off-premise consumption. dEffective June 1980; beer over 3.2 percent was illegal in West Virginia before that date.

Introduction

drinking age was whereby young pf drinking age WOUII ing ages went into purchase alcoholic These two states w the lowered drinki the higher drinkinl drinking age in b< and (3) adequate rl were available. Chapter 5 repo both on the invol' alcoholic-beverage reported findings fl ing and includes Sl designed to reduce

Notes

1. The exact tel to state. In anum beverage alcohol b accurately describel prohibit more than youth is illegal in IT purchase the alcol parental supervisic beverages by youth legal nomenclatun drinking age, and refer to the minim I youth for regular c 2. This discussi history of drinking 3. Most states statutes, permittin! physician or as par 4. Published lit< formation on the n the nature of thos provided here is ba ducted by Wagenac

Traffic Accidents

1983

11

Wine Forlified

Dislilled Spirits

19 19 19 19 19 19 21 21 21 21 21 21 19 19 20 20 21 21 19 19 19 19 18 18 19 19 21 21 21 21 19 19 21 21 21 21 18 18 20 20 21 21 20 20 21 21 19 19 21 21 21 21 19 19 20 20 21 21 20 20 21 21 21 21 19 19 21 21 21 21 21 21 21 21 21 21 21 21 20 20 18 21 21 21 19 19 19 19 21 21 18 18 21 21 21 21 18 18 18 18 19 19 lhol by weight. nales. This sex discrimination e effect of lowering the age to

Imption. lirginia before that date.

Introduction

7

drinking age was raised, neither state included a grandfather clause whereby young people who could legally drink prior to the increase in drinking age would continue to possess that right. After the raised drinking ages went into effect, young people who previously had the right to purchase alcoholic beverages were no longer legally allowed to do so. These two states were selected for intensive study because (1) both reversed the lowered drinking age at one time, rather than gradually returning to the higher drinking age in several steps, (2) the experience with the lower drinking age in both states had previously been scientifically evaluated, and (3) adequate records on both fatal and nonfatal motor-vehicle crashes were available. Chapter 5 reports findings on the effects of the higher drinking-age laws both on the involvement of youth in motor-vehicle crashes and on the alcoholic-beverage market. Finally, chapter 6 discusses the implications of reported findings for determining the optimal minimum legal age for drinking and includes suggestions for the development of other public policies designed to reduce alcohol-related problems among youth.

Notes

1. The exact terminology of drinking-age laws varies widely from state to state. In a number of states the law specifically prohibits purchase of beverage alcohol by those under a given age and might therefore be more accurately described as minimum-purchase-age laws. Most states, however, prohibit more than purchase of alcohol by youth. Provision of alcohol to youth is illegal in most states, regardless of whether young people or others purchase the alcoholic beverages. Consumption of alcohol, even under parental supervision, is illegal in many states. Possession of alcoholic beverages by youth is also frequently restricted. Because of the variability in legal nomenclature, the terms minimum legal-drinking age, minimum drinking age, and drinking age are used interchangeably in this book to refer to the minimum age at which beverage alcohol is legally available to youth for regular consumption as an adult. 2. This discussion is based primarily on the excellent review of the early history of drinking-age laws provided by Mosher (1980). 3. Most states allow for some exceptions to minimum-drinking-age statutes, permitting consumption of alcohol by youth if prescribed by a physician or as part of a religious ceremony. 4. Published literature on the legal-drinking age provides conflicting information on the number of states that have changed the drinking age and the nature of those changes. Information on legal-drinking-age changes provided here is based on a comprehensive survey of the fifty states conducted by Wagenaar (1981-1982).

i'

,

I

I

,:

r

,II'

The most prominent i the impact of modific in alcohol-related mo widespread drinking overrepresentation 0: Therefore, recent trel alcohol in traffic cras first. A review of pr chapter 3. Most young peop erages. Blane and He thirteen to eighteen) cluded that the preval drink?") was increasi junior- and senior-h drinkers over the per revealed for lifetime p drunk?"), which incre ing the period from l' year period. Prevalen do you become drunk about 19 percent duri number of surveys the the conclusions that Cl and Hewitt also coulG adolescents over the I frequency. Their be! drinkers aged thirteen Although these es surveys, only 14 of 1 defined populations. adolescents in the Uni many studies over an of adolescents drink I quently become intox:

_ _ _ _1_-_

2


The most prominent issue in the debate over the legal-drinking age has been the impact of modifications in the drinking age on the involvement of youth in alcohol-related motor-vehicle crashes. This concern is not only a result of widespread drinking among youth, but is also a result of the substantial overrepresentation of young people among drivers involved in crashes. Therefore, recent trends in drinking patterns among youth and the role of alcohol in traffic crashes, especially among young drivers, will be examined first. A review of previous research on the legal-drinking age follows in chapter 3. Most young people in the United States regularly drink alcoholic beverages. Blane and Hewitt (1977) reviewed 120 surveys of adolescent (ages thirteen to eighteen) drinking practices conducted since 1941. They concluded that the prevalence of young drinkers (that is, "have you ever had a drink?") was increasing before the mid-1960s and that about 70 percent of junior- and senior-high-school students were consistently identified as drinkers over the period from 1966 through 1975. A similar pattern was revealed for lifetime prevalence of intoxication (that is, "have you ever been drunk?"), which increased from 19 percent prior to 1966 to 45 percent during the period from 1966 to 1975 and remained stable during the latter tenyear period. Prevalence of self-reported monthly intoxication ("how often do you become drunk?") similarly increased from 10 percent before 1966 to about 19 percent during the period from 1966 to 1975, although the small number of surveys that assessed prevalence of monthly intoxication limited the conclusions that could be made concerning trends in recent years. Blane and Hewitt also could not identify trends in frequency of drinking among adolescents over the past two decades because of inconsistent measures of frequency. Their best estimate of average drinking frequency among drinkers aged thirteen to eighteen was three drinking occasions per month. Although these estimates were based on a comprehensive review of 120 surveys, only 14 of those studies used probability samples from clearly defined populations. As a result, estimates of the drinking practices of adolescents in the United States should be viewed cautiously. Nevertheless, many studies over an extended period have indicated that the great majority of adolescents drink regularly and that a substantial proportion also frequently become intoxicated. 9

10

Youth, Alcoho


than 5 percent, and road defects) caused Research in Public S speed, tailgating, dr examined. The resea in causing accidents or environmental c< human error most 0 Human conditic human errors that c, experienced, emotio disciplinary investig; alcohol impairment causal factor in cras nite or probable cam motor-vehicle crashe in these studies, the r minor property-dam amination of nightti much higher rates of The epidemiolog crashes the crash. Only abOl crashes have blood-< about 15 percent of have BACs of 0.05 p of drivers involved iJ greater, and the alcohol impairment, least 0.10 percent. TI vestigations that incll matched in time and studies have found 1 rapidly at BACs over Jones and Joscelyn 1 Numerous indivi attitudinal, personali the human condition predisposing charactl the best predictors a (aged fifteen to twent crashes. Significantly than would be expeCl

The above figures have been restricted to the drinking practices of junior- and senior-high-school youth. The literature on college students, also reviewed by Blane and Hewitt (1977), is even more limited. Existing surveys of college students indicate that the prevalence of drinkers ("have you ever had a drink?") has been continually increasing since World War II. It is estimated that about 89 percent of all college students are drinkers. There are indications that the frequency of intoxication among college students has increased in the past quarter century. Furthermore, people consume more beverage alcohol between the ages of eighteen and twenty-five than at any other period in the life cycle, and they drink larger quantities of alcohol per occasion than do older drinkers (Blane and Hewitt 1977; National Institute on Alcohol Abuse and Alcoholism 1978). The ongoing nationwide probability surveys conducted by Johnston, Bachman, and O'Malley (1981) provide the most recent information concerning drinking practices among young people. They reported that 93 percent of high-school seniors surveyed in 1981 had at least tried beverage alcohol, 71 percent reported use within the past month, and 41 percent reported binge drinking, that is, consuming five or more drinks on at least one occasion in the previous two weeks. Furthermore, similar surveys conducted each year since 1975 revealed that, while the prevalence of drinkers ("ever used alcohol") has remained stable in recent years, the prevalence of highschool seniors who frequently become intoxicated increased from 37 percent in 1975 to 41 percent in 1979. The rate remained at 41 percent in 1980 and 1981, the most recent years for which data are available. I These recent data confirm and extend the conclusion drawn by Blane and Hewitt on the basis of their review of surveys conducted prior to 1975: the prevalence of drinkers among older adolescents and young adults has apparently reached a plateau, with about 90 percent of this group identifying themselves as drinkers. Until recently, however, the prevalence of young people who frequently become intoxicated has been increasing, with current data indicating that between one-third and one-half of the young people in the United States become intoxicated at least once every fourteen days (Johnston, Bachman, and O'Malley 1981, p. 46). Frequent intoxication by a sizable proportion of American adolescents creates the potential for serious mortality and injury consequences if young drinkers operate motor vehicles while in an alcohol-impaired state. As noted in chapter 1, motor-vehicle crashes are the leading cause of death among youth aged fifteen to twenty-four and claimed 18,800 lives in the United States in 1980 (National Safety Council 1981). A large number of interacting factors, including alcohol consumption, have been identified as contributing to the occurrence of automobile crashes. Intensive investigations of random samples of crashes have revealed that vehicular factors (such as brake failures and tire blowouts) caused less

, I

:I

I

,I

1. _

L

Traffic Accidents

Ie drinking practices of lre on college students, 1 more limited. Existing lence of drinkers ("have easing since World War ge students are drinkers. )xication among college people coneighteen and twenty-five :lrink larger quantities of le and Hewitt 1977; Nal 1978). conducted by Johnston, recent information conhey reported that 93 perit least tried beverage alonth, and 41 percent re.ore drinks on at least one similar surveys conducted alence of drinkers ("ever S, the prevalence of high:l increased from 37 perned at 41 percent in 1980 e available. l nclusion drawn by Blane conducted prior to 1975: nts and young adults has :nt of this group identifyr, the prevalence of young n increasing, with current l1f of the young people in mce every fourteen days Frequent intoxication by creates the potential for 19 drinkers operate motor

; are the leading cause of nd claimed 18,800 lives in 11981). A large number of n, have been identified as ;hes. of crashes have revealed tire blowouts) caused less


11

than 5 percent, and environmental factors (slick roads, reduced visibility, road defects) caused less than 20 percent, of crashes examined (Institute for Research in Public Safety 1975). Direct human behaviors (such as excessive speed, tailgating, driver inattention) caused over 75 percent of the crashes examined. The researchers emphasized the dominant role of human factors in causing accidents and pointed out that even in cases in which a vehicular or environmental cause was clearly evident, such factors combined with human error most often brought about the collision. 2 Human conditions at the time of the crash largely account for the human errors that cause most collisions. For example, the driver may be inexperienced, emotionally upset, fatigued, or impaired by drugs. Multidisciplinary investigations of the causes of traffic collisions indicate that alcohol impairment is the human condition most frequently identified as a causal factor in crashes. Alcohol impairment has been identified as a definite or probable cause in about 7 percent of collisions (Treat 1977). Because motor-vehicle crashes at all times of the day and week were used as samples in these studies, the majority of the investigated collisions were the relatively minor property-damage crashes that occur during daytime rush hours. Examination of nighttime and weekend crashes would, undoubtedly, reveal much higher rates of alcohol impairment. The epidemiological literature on the role of beverage alcohol in traffic crashes demonstrates that the role of alcohol increases with the severity of the crash. Only about 10 percent of the drivers in minor property-damage crashes have blood-alcohol concentrations (BACs) over 0.05 percent, and about 15 percent of drivers involved in extensive property-damage crashes have BACs of 0.05 percent or greater. However, approximately 25 percent of drivers involved in serious-injury crashes have BACs of 0.10 percent or greater, and the most-serious crashes, fatalities, have the highest rates of alcohol impairment, with about one-half the drivers having a BAC of at least 0.10 percent. The findings of these studies are supported by other investigations that included control groups of drivers not involved in crashes, matched in time and place to samples of drivers involved in crashes. Such studies have found that the risk of being involved in a crash accelerates rapidly at BACs over 0.08 percent (Borkenstein et al. 1964; Cameron 1977; Jones and Joscelyn 1978). Numerous individual characteristics (for example, sociodemographic, attitudinal, personality, and socioenvironmental) result in predisposition to the human conditions that often lead to driver error and collisions. Of all predisposing characteristics, age and sex of driver are consistently among the best predictors of crash involvement (Cameron 1977). Young people (aged fifteen to twenty-four), especially males, are overrepresented in traffic crashes. Significantly more young people are involved in automobile crashes than would be expected based on the total number of young people in the

12

! I


population, the number of young people licensed to drive, or the amount of automobile travel among youth. For example, U.S. youth aged fifteen to twenty-four have a motor-vehicle-crash death rate of about 43 per 100,000, more than twice the rate for those aged twenty-five and over (National Safety Council 1981). Similarly, the total (fatal and nonfatal) crash rate per 100,000 licensed drivers is about twice as high for eighteen-to-twenty-fouryear-olds than for drivers aged twenty-five and over (Michigan State Police 1982). The overrepresentation of youth in automobile crashes is even more pronounced when the amount of automobile travel is taken into account. The fatality rate per 100 million miles traveled (as drivers or passengers) among women aged eighteen to twenty-four is three times the rate for women aged thirty to sixty; the corresponding rate for young men is six times the rate for older men (Carsten 1981). Similar patterns of youth overrepresentation in motor-vehicle crashes have been found in numerous developed countries; crash rates for youth are typically from two to ten times the rates for drivers of other age groups (Organization for Economic Cooperation and Development 1975). Many exposure variables have been suggested as explanations for the overrepresentation of youth among crash-involved drivers, especially involvement in the more serious injury-producing collisions. Exposure factors include (1) driving at more hazardous times and locations, for example, driving at night and on weekends; (2) more frequent driving with passengers present, which increases the probability of distraction; (3) driving vehicles that are in poor condition, which increases the probability of vehicle malfunction; and (4) more frequent use of two-wheeled vehicles. Although much more work needs to be done on the effects of differential exposure, studies to date indicate that, after allowing for higher exposure to crashes, young drivers are clearly overrepresented in the crash-involved population (Organization for Economic Cooperation and Development 1975; Preusser, Oates, and Orban 1975). Comparisons of the proportion of crashes among young drivers that involve alcohol with the corresponding proportion for older drivers provide further insight into the role of alcohol in youth crashes. Several studies conducted in the 1960s and early 1970s found that among young drivers a lower proportion of crashes involved alcohol than among older drivers (Borkenstein et al. 1964; Filkins and Flora 1970; Waller et al. 1970; Perrine, Waller, and Harris 1971; Farris, Malone, and Lilliefore 1976). For example, Waller examined a sample of fatal crashes in California and found that 49 percent of the fifteen-to-nineteen-year-old drivers had been drinking, compared with 61 percent of the drivers aged twenty and over. None of these earlier studies examined youth residing in a state with a legal-drinking age below twenty-one. Flora and others (1978) examined fatal-crash involvement in a state with a drinking age of eighteen. Although the ratio of alcohol-related

Youth, Alcohol

to creased when the dri higher for drivers un one and over. In spite of the 0 tively conclude that ( been drinking than al difference is a result careful driving after the former. Roadside of all young drivers 0 than, the proportion BACs (Perrine, Wallt Farris, Malone, and eighteen-to-twenty-ye alcohol. The differen narrowed for youth (Douglass et al. 1974: It is clearly not consumption of alcol crash involvement at ' aged drivers. A in a crash than is an 0 creases more sharply other ages (Perrine, " and Lilliefore 1976). The particularly compared with older of extensive experienc an explanation is SUp] among drinkers of all of crash involvement although youth have 1979), their recent inl them sufficient expel drinking to enable th risk of an alcohol-relt A second explana BAC on the risk of cr exacerbates the preexi behavior that is chara Pelz and Schuman 19 ably valid, with the

ci Traffic Accidents

o drive, or the amount of .S. youth aged fifteen to of about 43 per 100,000, and over (National Safety nonfatal) crash rate per eighteen-to-twenty-fourer (Michigan State Police >bile crashes is even more tel is taken into account. as drivers or passengers) three times the rate for ate for young men is six ar patterns of youth overeen found in numerous ypically from two to ten rganization for Economic

d as explanations for the 'ed drivers, especially in>llisions. Exposure factors 1 locations, for example, nt driving with passengers ction; (3) driving vehicles robability of vehicle malleeled vehicles. Although ; of differential exposure, igher exposure to crashes, :rash-involved population velopment 1975; Preusser,

ong young drivers that infor older drivers provide ashes. Several studies conlong young drivers a lower mg older drivers (Borken,al. 1970; Perrine, Waller, 976). For example, Waller and found that 49 percent been drinking, compared ,ver. None of these earlier l legal-drinking age below tal-crash involvement in a he ratio of alcohol-related


13

to non-alcohol-related crashes among youth under twenty-one years old increased when the drinking age was lowered, the ratio was not consistently higher for drivers under the age of twenty-one than for drivers aged twentyone and over. In spite of the obvious limitations of available data, one might tentatively conclude that a young driver involved in a crash is less likely to have been drinking than an older driver. The question remains as to whether this difference is a result of less driving after drinking among youth or of more careful driving after drinking. Recent research indicates that the answer is the former. Roadside-breath-test surveys have revealed that the proportion of all young drivers on the road with elevated BACs is the same as, or lower than, the proportion of drivers in their thirties or forties who have elevated BACs (Perrine, Waller, and Harris 1971; Preusser, Oates, and Orban 1975; Farris, Malone, and Lilliefore 1976; Wolfe 1975). Again, most of the eighteen-to-twenty-year-old drivers studied did not have legal access to alcohol. The differences in the amount of impaired driving appear to have narrowed for youth who gained legal access to alcohol in the 1970s (Douglass et al. 1974). It is clearly not the case that youth drive more carefully than adults after consumption of alcoholic beverages. Just the opposite is true. The risk of crash involvement at various BACs is higher for youth than it is for middleaged drivers. A young driver with a given BAC is more likely to be involved in a crash than is an older driver at the same level, and the risk of a crash increases more sharply with increasing BACs for youth than for drivers of other ages (Perrine, Waller, and Harris 1971; Zylman 1972; Farris, Malone, and Lilliefore 1976). The particularly high susceptibility to traffic crashes among youth, compared with older drivers at identical BACs, may be the result of a lack of extensive experience with drinking and with driving after drinking. Such an explanation is supported by the work of Hurst (1973), who found that, among drinkers of all ages, those who drink infrequently have a higher risk of crash involvement at a given BAC than do frequent drinkers. Thus, although youth have been characterized as frequent heavy drinkers (Blane 1979), their recent initiation into regular drinking may not have afforded them sufficient experience with drinking effects and with driving after drinking to enable them to develop compensatory actions that reduce the risk of an alcohol-related collision. A second explanation for the particularly serious effect of an elevated BAC on the risk of crash involvement among young drivers is that alcohol exacerbates the preexisting impulsiveness and propensity toward risk-taking behavior that is characteristic of adolescents and young adults (Klein 1971; Pelz and Schuman 1971; Makela 1978). Both these explanations are probably valid, with the high risk of crash involvement of young drivers at

14


relatively moderate BACs a result of their inexperience with alcohol and their higher threshold of acceptable risk. In summary, literature on motor-vehicle crashes has revealed that, of the multiple environmental, vehicular, and human causes of collisions, human error is generally thought to be a central factor in most traffic crashes. These human errors frequently result from the alcohol-impaired condition of the driver. Drinking patterns among youth, characterized by frequent consumption of large quantities of alcoholic beverages per occasion, and increased sensitivity to impairment at a given BAC of young drivers as compared with older drivers combine to make youth particularly susceptible to involvement in alcohol-related crashes. Although the proportion of all crashes that involve alcohol may be slightly lower for young drivers (particularly those under the legal drinking age), the fact that overall crash rates for youth are several times higher than those for older drivers means that death, injury, and property damage resulting from alcohol-related crashes are also unusually high for young drivers. Therefore, a combination of (1) high rates of motor-vehicle collisions regardless of alcohol involvement, reflecting inexperience with driving, and (2) high probability of a crash when driving after drinking, reflecting inexperience with drinking, indicates that young drinking drivers are an appropriate high-risk target group for the prevention of death and injury resulting from alcohol-related traffic crashes. Because of the extent of death, injury, and other social costs caused by alcohol-impaired driving among youth, a higher minimum legal-drinking age has been proposed as one way to reduce this major social problem.

Notes 1. Some writers have suggested that the minimum drinking age is of little significance given the high prevalence of young drinkers under the legal age. Those who supported reductions in legal age in the early 1970s used this argument, and it is still occasionally heard today (Plaut 1967; Keller 1980; Clay, quoted in Guy 1978). According to this line of reasoning, because so many youth drink anyway, lowering the legal-drinking age would have little effect on drinking practices and alcohol-related problems among young people. Subsequent research proved this assumption to be false. After the legal age was lowered, drinking problems among youth increased. In spite of the high prevalence of young drinkers who have tried alcohol at least once, the legal-drinking age has potential for reducing the frequency of drinking occasions, the quantity of alcohol consumed, the frequency of intoxication, and the prevalence of alcohol-related problems from the levels that would prevail without a minimum legal-drinking age.

•",J

Youth, Alcohol, a

2. Attribution of ca by an investigator's pn pie, instead of attributi that the roadway was n tion. Attributing the CI victim" (Ryan 1976). I adequately deal with til< crash to vehicle or envi existed, given our pres(

{·1 I,

!I

i

Id Traffic Accidents

)erience with alcohol and es has revealed that, of the LUses of collisions, human most traffic crashes. These impaired condition of the ized by frequent consump:r occasion, and increased drivers as compared with susceptible to involvement 1 of all crashes that involve ,articularly those under the for youth are several times eath, injury, and property :e also unusually high for tigh rates of motor-vehicle ting inexperience with drivIg after drinking, reflecting :lrinking drivers are an aption of death and injury lUse of the extent of death, II-impaired driving among proposed as one way to

mum drinking age is of litIg drinkers under the legal in the early 1970s used this r (Plaut 1967; Keller 1980; e of reasoning, because so nking age would have little :l problems among young Ition to be false. After the youth increased. In spite have tried alcohol at least :educing the frequency of lImed, the frequency of in1 problems from the levels nking age.


!

15

2. Attribution of cause for a particular motor-vehicle crash is influenced by an investigator's preexisting theories of the cause of crashes. For example, instead of attributing a collision to driver inattention, one might argue that the roadway was not designed well enough to retain the driver's attention. Attributing the crash to driver error can be viewed as "blaming the victim" (Ryan 1976). In this discussion human error refers to a failure to adequately deal with the usual vehicle and roadway situation. Attributing a crash to vehicle or environmental factors means that an unusual condition existed, given our present highway and vehicle systems.

3

,I

This chapter begins wit alcohol availability and t beverages to prevent alco presents a comprehensive ducted in the past decade lowering the drinking agt fects of returning to high; on effects of different ] youth. Based on the litera effects of changes in leg research questions addre: Laws and regulatior alcoholic beverages, of \ have attracted increased prevention of alcohol-re considerable detail how, problems have been vie' disease called alcoholisr establishment and group: those addicted to alcoh health and social probl Although the specific cat was generally thought t possessed a particular co that made them suscepti disease, by implication, I ety at large, and for mm Dominance of the co the role of laws and regl social and health probleJ resulting from the diseas because they have partil public policy affecting a attempting to reduce the lems. If acute public-hea motor-vehicle crashes, a

II

II I, I I

'III

II

,I

,

I

,..I.e

3

Alcohol Availability and the Minimum Legal-Drinking Age

This chapter begins with an introduction to the broader literature on alcohol availability and the potential of a policy on the control of alcoholic beverages to prevent alcohol-related problems. The main part of the chapter presents a comprehensive review of the research on minimum-age laws conducted in the past decade. The literature includes (1) studies of the effects of lowering the drinking age on youth crash involvement, (2) studies of the effects of returning to higher drinking ages, and (3) the more-limited research on effects of different legal-drinking ages on drinking patterns among youth. Based on the literature reviewed, a conceptual model of hypothesized effects of changes in legal-drinking age is presented. Finally, the specific research questions addressed in the present investigation are discussed. Laws and regulations that affect the availability or accessibility of alcoholic beverages, of which the minimum drinking age is one example, have attracted increased attention in recent years as a strategy for the prevention of alcohol-related problems. Beauchamp (1980) describes in considerable detail how, since the repeal of Prohibition, alcohol-related problems have been viewed as symptoms or consequences of a specific disease called alcoholism. With the emergence of a medical-treatment establishment and groups, such as Alcoholics Anonymous, which focus on those addicted to alcohol, individuals who experienced alcohol-related health and social problems were viewed as alcoholic or prealcoholic. Although the specific causes of alcoholism remained poorly understood, it was generally thought that those experiencing alcohol-related problems possessed a particular constellation of physiological or psychological traits that made them susceptible to the disease. The view of alcoholism as a disease, by implication, proclaimed alcohol to be nonproblematic for society at large, and for most beverage-alcohol consumers. Dominance of the concept of alcoholism as a disease led to neglect of the role of laws and regulations in controlling alcohol use and associated social and health problems. If all alcohol-related problems are viewed as resulting from the disease of alcoholism, which strikes certain individuals because they have particular traits that most people do not have, then public policy affecting alcohol availability will be seen as irrelevant when attempting to reduce the prevalence or incidence of alcohol-related problems. If acute public-health problems, such as death and injury caused by motor-vehicle crashes, are seen as symptomatic of alcoholism, solutions

17

18 1:1:


: i

are focused on defining, identifying, and treating alcoholics, not investigating the effects of controls on alcohol availability. As a result of these perspectives, little empirical research on the effects of alcohol-control laws was conducted in the United States between 1930 and 1970. One important exception to this paucity of research is a series of papers produced by the Moreland Commission of New York State in 1963 (New York State 1963, 1964; Bacon 1971). The commission concluded that, in general, beverage-control laws did not have beneficial effects in reducing alcohol problems, and it recommended relaxing restrictions on the marketing of alcoholic beverages. Although the commission's conclusions were of limited scientific merit and were based on admittedly poor data, they were accepted for several years, resulting in little detailed examination of beverage-control laws during the subsequent decade. During the 1970s and early 1980s, the issue of alcohol availability received increasing attention, particularly in Europe and Canada, but also in the United States (Bruun et al. 1975; Medicine in the Public Interest 1979; Harford, Parker, and Light 1980; Moser 1980; Frankel and Whitehead 1981). Although U.S. studies in the 1970s focused on the effects of specific changes in availability, such as the drinking age, by late in the decade the role of general availability of alcohol was receiving more attention. Reports by the federal government discussed the potential of alcohol-beveragecontrol laws as one strategy for the prevention of alcohol-related problems (National Institute on Alcohol Abuse and Alcoholism 1978; Alcohol, Drug Abuse, and Mental Health Administration 1981). The National Academy of Sciences recently issued a report on the prevention of problems related to alcohol use; the role of controls on the availability of alcohol in such prevention efforts was discussed extensively (Moore and Gerstein 1981). These recent reports indicate that the issues of alcohol availability and alcohol-beverage-controllaws are receiving increasing attention among researchers and policymakers. Although research on alcohol availability is increasing, a focus on alcoholism and chronic heavy drinking continues. Most recent research on relationships between public policy, alcohol availability, alcohol consumption, and alcohol-related public-health problems has used cirrhosis mortality as the dependent variable (Bruun et al. 1975). The effects of changing alcohol availability by lowering or raising the minimum legal-drinking age is one area where the focus has been on acute alcohol problems, particularly alcohol-related traffic crashes.

I

Research on the Legal-Drinking Age and Highway Crashes

iii

q

Ii 1

i I

1f: .

1

,I 1

Iii jil

!II Iii

i

i! I[ 'pi

I

II'

As noted in chapter 1, after many states and Canadian provinces lowered the legal-drinking age in the early 1970s, numerous evaluations were con-

Alcohol Availabi

ducted of how legal motor-vehicle collisiOI were based on compar before and after a ree other studies included: laws had been change( by the legal change (su parison jurisdictions tl Williams and othc crashes among fifteen old drivers in Michigal age had been lowered three years prior to an< the frequency of fatal and Minnesota. DrinkiJ during the time period Significant increase both fifteen-to-seventee the jurisdictions that ha single-vehicle and nigh known to be alcohol rc analyses of all fatal cra. was substantially great( smaller effect in WiSCOl in the legal availability eighteen-to-twenty-yeal simply extended that p Naor and Nashold Wisconsin on highway fatalities increased con, fatally injured drivers significantly. Naor anc reduced drinking age However, since beer, legally available prior t tion alone should not lowered legal-drinking Cucchiaro, Ferreira traffic crashes to eval Massachusetts. They ( fifteen-to-seventeen, ei twenty-four and over. twenty-year-old driven crashes, alcohol-related

j

Traffic Accidents

ting alcoholics, not inailability. As a result of of alcohol-control 1930 and 1970. earch is a series of papers York State in 1963 (New ission concluded that, in :ficial effects in reducing jng restrictions on the :ommission's conclusions m admittedly poor data, ittle detailed examination .ecade. e of alcohol availability >pe and Canada, but also 11 the Public Interest 1979; Frankel and Whitehead I on the effects of specific by late in the decade the g more attention. Reports tial of alcohol-beverage. alcohol-related problems llism 1978; Alcohol, Drug The National Academy of : prevention of problems availability of alcohol in ely (Moore and Gerstein les of alcohol availability Ilcreasing attention among

is increasing, a focus on ;. Most recent research on lability, alcohol consumps has used cirrhosis morl). The effects of changing imum legal-drinking age is hoI problems, particularly

madian provinces lowered ous evaluations were con-

Alcohol Availability

19

ducted of how legal changes affected the frequency of involvement in motor-vehicle collisions among young drivers. Most of the investigations were based on comparisons between indices of crash involvement by youth before and after a reduction in legal-drinking age took effect. Numerous other studies included an assessment of crash involvement prior to and after laws had been changed for (1) comparison age groups not directly affected by the legal change (such as drivers over the age of twenty-one) or (2) comparison jurisdictions that had not changed the legal-drinking age. Williams and others (1974) examined the frequency of fatal traffic crashes among fifteen-to-seventeen-year-old and eighteen-to-twenty-yearold drivers in Michigan, Wisconsin, and Ontario, where the legal-drinking age had been lowered. The frequency of fatal crashes in those areas for three years prior to and one year after the legal changes was compared with the frequency of fatal crashes in the contiguous states of Indiana, Illinois, and Minnesota. Drinking-age laws in the comparison states were not changed during the time period studied. Significant increases in the number of fatal crashes were found among both fifteen-to-seventeen-year-old and eighteen-to-twenty-year-old drivers in the jurisdictions that had lowered the legal-drinking age. Separate analyses of single-vehicle and nighttime fatal crashes, a large proportion of which are known to be alcohol related, revealed larger increases than were found by analyses of all fatal crashes. Increased involvement in fatal crashes by youth was substantially greater for Michigan and Ontario than for Wisconsin. The smaller effect in Wisconsin was most likely a result of the less drastic change the legal availability of alcohol. In Wisconsin, prior to the legal change, elghteen-to-twenty-year-olds could legally purchase beer; the new law simply extended that privilege to other types of alcoholic beverages. Naor and Nashold (1975) also studied the impact of the legal change in Wisconsin on highway fatalities. Although the frequency of alcohol-related fatalities increased concomitant with the legal change, the proportion of all fatally injured drivers with elevated blood-alcohol levels did not change significantly. Naor and Nashold used the latter finding to argue that the reduced drinking age had no effect on traffic crashes among youth. However, since beer, the beverage of choice among young people, was legally available prior to the drinking-age change evaluated, this investigation alone should not be considered an adequate test of the effects of a lowered legal-drinking age. Cucchiaro, Ferreira, and Sicherman (1974) used monthly time series of traffic crashes to evaluate the impact of reducing the drinking age in Massachusetts. They examined crash involvement among drivers aged fifteen-to-seventeen, eighteen-to-twenty, twenty-one-to-twenty-three, and twenty-four and over. After the drinking age was lowered, eighteen-totwenty-year-old drivers experienced significant increases in total fatal crashes, alcohol-related fatal crashes, and alcohol-related property-damage

I

i

20


crashes. None of the crash measures changed significantly for drivers aged twenty-one-to-twenty-three and twenty-four and over. Douglass and others (1974) also used monthly time series of traffic crashes to assess the effects of reduced drinking ages in Maine, Michigan, and Vermont. Collision involvement of young drivers in these states was compared with collision involvement of older drivers within the same state and with young drivers in Louisiana, Pennsylvania, and Texas, which held the drinking age constant over the study period. Analyses revealed significant increases in involvement in alcohol-related crashes among young drivers in both Michigan and Maine. No significant increases in alcoholrelated crashes among youth were observed in any of the comparison states, nor were there any significant shifts for older drivers within Maine or Michigan. The researchers suggested that the lack of significant changes in the frequency of traffic crashes in Vermont, which also lowered its drinking age, may have been a result of the relative ease with which eighteen-totwenty-year-olds in Vermont could obtain alcoholic beverages prior to the reduced drinking age by driving to New York, which had a drinking age of eighteen at the time. Douglass and Freedman (1977) replicated some of the earlier analyses, using four years of crash data after the legal age was lowered in Michigan. Results demonstrated that increased alcohol-related crashes among Michigan youth, identified in the 1974 research, persisted over the first four years with the lower drinking age (1972 through 1975). Evaluation of the Michigan experience continued with analyses of fatal crashes in Michigan by Flora, Filkins, and Compton (1978). Although these authors did not use the same analytical techniques used by Douglass, the impact of the 1972 reduction in the legal-drinking age on alcohol-related traffic crashes among youth was again observed. An increase in alcohol-related collisions following Ontario's reduction in drinking age was also reported by Schmidt and Kornaczewski (1975), who examined crash data from 1967 through 1971. The lack of monthly data and the failure to analyze separately only eighteen-to-twenty-year-old drivers made this study a conservative test of the effects of a reduced drinking age. Nonetheless, the researchers found a significant increase in crash involvement among drivers aged sixteen to nineteen after the law was changed. Whitehead and others (1975) examined crash involvement among drivers aged sixteen to twenty and twenty-four years in London, Ontario, for the period from 1968 through 1973. After Ontario's drinking age was lowered, alcohol-related crashes among drivers aged eighteen to twenty increased 150 to 300 percent. In contrast, for the first year after the legal change, twenty-four-year-old drivers experienced only a 20 percent increase in alcohol-related crashes, and their collision frequency returned to the

Alcohol Availabilit

prechange level the secm In a follow-up study, W of collision data. A tota: reduction in drinking ag alcohol-related collision: minimum drinking age. Warren and others (1 on traffic fatalities in A ewan. Only those fatali; was administered were i related fatalities for fift reduction in drinking ag creases in fatalities amor in the study jurisdictio However, since the bloc driving in Canada was re was lowered, the author were confounded with ef insufficient numbers of quately control for the r; from year to year. As a [ occurred after the drinkj quivocally attributed to One of the province! studied by Shattuck anI lowered from twenty-one old drivers exhibited incr After the drinking age w; this same group of drive crashes. Thus, two reduc alcohol-related crashes al underage population. Bako, MacKenzie, al with blood-alcohol conc fatally injured in the pr< observed in the incidence aged fifteen to nineteen concluded that their finl ages lead to increased al The reduction in the (from twenty-one to ninl Transportation (1977). ( Illinois and five compari

and Traffic Accidents

ignificantly for drivers aged ld over. ,nthly time series of traffic 19 ages in Maine, Michigan, : drivers in these states was lrivers within the same state ania, and Texas, which held :I. Analyses revealed signifilted crashes among young ificant increases in alcoholny of the comparison states, er drivers within Maine or lck of significant changes in ich also lowered its drinking ase with which eighteen-toholic beverages prior to the which had a drinking age of

orne of the earlier analyses, :e was lowered in Michigan. col-related crashes among persisted over the first four 1975). Evaluation of the If fatal crashes in Michigan gh these authors did not use ass, the impact of the 1972 traffic crashes among

[lowing Ontario's reduction and Kornaczewski (1975), 1971. The lack of monthly eighteen-to-twenty-year-old e effects of a reduced drink;ignificant increase in crash lineteen after the law was

crash involvement among years in London, Ontario, Ontario's drinking age was aged eighteen to twenty inle first year after the legal d only a 20 percent increase frequency returned to the


21

prechange level the second year after the reduced drinking age took effect. In a follow-up study, Whitehead (1977) examined an additional two years of collision data. A total of four years of crash-involvement data after the reduction in drinking age demonstrated the permanence of higher rates of alcohol-related collisions among young drivers after a reduction in the minimum drinking age. Warren and others (1977) evaluated the impact of reduced drinking ages on traffic fatalities in Alberta, Manitoba, New Brunswick, and Saskatchewan. Only those fatalities for which a blood-alcohol-concentration test was administered were included in the analyses. Frequencies of alcoholrelated fatalities for fifteen-to-twenty-year-old drivers before and after a reduction in drinking age were compared within each province. Some increases in fatalities among fifteen-to-twenty-year-old drivers were observed in the study jurisdictions at the time the drinking age was lowered. However, since the blood-alcohol concentration legally defined as drunk driving in Canada was reduced at about the same time that the drinking age was lowered, the authors pointed out that effects of the BAC legislation were confounded with effects of the lower legal-drinking age. Furthermore, insufficient numbers of prechange observations were available to adequately control for the random variation in the number of traffic fatalities from year to year. As a result, although increases in fatalities among youth occurred after the drinking age was lowered, the increases cannot be unequivocally attributed to the lower drinking age. One of the provinces investigated by Warren, Saskatchewan, was also studied by Shattuck and Whitehead (1976). After the drinking age was lowered from twenty-one to nineteen in April 1970, sixteen-to-twenty-yearold drivers exhibited increases of 20 to 50 percent in alcohol-related crashes. After the drinking age was lowered from nineteen to eighteen in June 1972, this same group of drivers experienced further increases in alcohol-related crashes. Thus, two reductions in the legal age were associated with increased alcohol-related crashes among both the newly enfranchised drinkers and the underage population. Bako, MacKenzie, and Smith (1976) examined the frequency of drivers with blood-alcohol concentrations of 0.08 percent or greater among those fatally injured in the province of Alberta. An increase of 118 percent was observed in the incidence of fatal collisions related to alcohol among drivers aged fifteen to nineteen after the drinking age was lowered. The researchers concluded that their findings support the argument that lowered drinking ages lead to increased alcohol-related collisions among youth. The reduction in the legal-drinking age for beer and wine in Illinois (from twenty-one to nineteen) was evaluated by the Illinois Department of Transportation (1977). Comparisons between numbers of crash fatalities in Illinois and five comparison states were used as the basis for the conclusion

22


that the lowered drinking age in Illinois resulted in a small increase in fatalities among drivers aged nineteen and twenty. The Alcohol Epidemiologic Data System (1980) examined annualtraffic-fatality counts in the fifty states from 1970 through 1978. Although increased motor-vehicle deaths were seen in some states following a reduction in drinking age, it was concluded that "differences in highway fatalities with the change in drinking laws do not appear significantly large." The authors readily admit, however, that their analyses did not adequately control for the effects of numerous confounding factors that occurred in the 1970s (fuel shortages, speed-limit reductions, and so on). After Alabama lowered its drinking age from twenty-one to nineteen in 1975, alcohol-related crashes increased significantly among drivers aged eighteen to twenty according to Brown and Maghsoodloo (1981). Koch (1981) pointed out the methodological limitations of Brown and Maghsoodloo's study and argued that it did not establish a causal connection between the drinking age and crashes. The study did not include adequate comparisons between alcohol-related and non-alcohol-related crashes and between states that changed the drinking age and those that did not. However, the association between lowered drinking age and increased alcohol-related crashes in Alabama is consistent with the results from studies of other states and provided additional support for the conclusion that lower drinking ages increase alcohol-related crashes. Lynn (1981) analyzed annual counts of alcohol-related crashes in Virginia from 1969 through 1979. Drivers aged sixteen to twenty experienced an identifiable increase in involvement in alcohol-related crashes beginning in 1974, when the drinking age for beer was lowered to eighteen. In contrast, drivers aged twenty-five and over experienced a decrease in alcoholrelated crashes during the same period. The author concluded that the lowered drinking age was responsible for increased alcohol-related crashes among young drivers in Virginia in the late 1970s and recommended a gradual return to the legal-drinking age of twenty-one. Cook and Tauchen (1982) estimated the effect of providing youth legal access to alcohol through lowering the drinking age, using nationwide U. S. data on the number of young people killed in motor-vehicle crashes. Analyses of fatality data pooled across states and years revealed that a 7 percent increase in the number of youth killed in automobile crashes was associated with reductions in the legal age. The 7 percent estimate applies to youth killed in all motor-vehicle crashes, whether drivers or passengers, drinking or not drinking, at the time of the crash. Crashes in which a young driver contributed to the death of an older person but in which a young person was not killed were not taken into account by Cook and Tauchen. Some evidence of a spillover effect of lower drinking ages was also found. An estimated 6 percent increase in deaths among sixteen-to-seventeen-year-olds was associated with reductions in the drinking age.

I,

I"

IlII

I!

II II

\

Alcohol Availabil

L

The literature revie impact of lower legal-( sions have found sil previously underage dr laws (usually eighteen-1 demonstrated substant drivers (usually sixteen minimum drinking age have methodological li groups, (2) lack of extel limited coverage with crashes or only crashes The pattern of results using a variety of anal drinking ages typically safety among young dI The view that low youth in crashes is not of a causal interpreta1 1974, 1976, 1977) has has argued that observf among youth after the changes in drinking agf in traffic-crash time sel consumption and alco] prior to the legal chang fic offenses by law-enf, however, indicates tha1 sophisticated investigat evident. For example, ; trends and random fluc ing age. Although Zyln on analyses of police-rl drinking ages (althougt reports) have been c involvement not influe analyses of single-vehic In addition to evall have been made on the f (1979) compared the nu fluence of liquor (DUll legal age was raised VI. drinking-age change. I to conclude that the ra

md Traffic Accidents lted in a small increase in lty. (1980) examined annual70 through 1978. Although le states following a reducin highway fatalities lr significantly large." The ses did not adequately con'actors that occurred in the ld so on). n twenty-one to nineteen in ;antly among drivers aged Ilaghsoodloo (1981). Koch ions of Brown and Maghlish a causal connection ber did not include adequate 3.lcohol-related crashes and and those that did not. rinking age and increased ent with the results from support for the conclusion d crashes. alcohol-related crashes in ixteen to twenty experienced )l-related crashes beginning Iwered to eighteen. In conneed a decrease in alcoholauthor concluded that the lsed alcohol-related crashes 1970s and recommended a 1ty-one. of providing youth legal age, using nationwide U.s. in motor-vehicle crashes. md years revealed that a 7 in automobile crashes was , percent estimate applies to ther drivers or passengers, 1. Crashes in which a young n but in which a young pery Cook and Tauchen. Some g ages was also found. An teen-to-seventeen-year-olds age.


23

The literature reviewed here shows that most of the investigations of the impact of lower legal-drinking ages on involvement in motor-vehicle collisions have found significant increases in crash involvement among previously underage drivers who acquired the right to drink under the new laws (usually eighteen-to-twenty-year-old drivers). Several studies have also demonstrated substantial increases in crash involvement among underage drivers (usually sixteen and seventeen years old) following reductions in the minimum drinking age. Many of the evaluations of lowered drinking ages have methodological limitations, such as (1) lack of adequate comparison groups, (2) lack of extended longitudinal data, and (3) outcome measures of limited coverage with small numbers of cases (for example, only fatal crashes or only crashes for which a chemical BAC test was administered). The pattern of results across studies of numerous jurisdictions, however, using a variety of analytic methods, leads to the conclusion that lowered drinking ages typically result in an increase in problems related to highway safety among young drivers. The view that lower legal-drinking ages increase the involvement of youth in crashes is not universally held. Zylman is a well-known opponent of a causal interpretation of the observed relationships. Zylman (1973, 1974, 1976, 1977) has criticized several of the studies reviewed above. He has argued that observed increases in involvement in alcohol-related crashes among youth after the drinking age was lowered were not the result of the changes in drinking age, but rather were a result of (1) random fluctuations in traffic-crash time series, (2) continuation of trends of increasing alcohol consumption and alcohol-related crashes among youth that were evident prior to the legal changes, or (3) increased attention to alcohol-related traffic offenses by law-enforcement officers. A careful review of the research, however, indicates that these factors were taken into account in the more sophisticated investigations and that the effect of the drinking age was still evident. For example, studies that explicitly controlled for both long-term trends and random fluctuations found significant effects of the lower drinking age. Although Zylman correctly points out limitations of relying solely on analyses of police-reported alcohol involvement, the effects of lowered drinking ages (although of smaller magnitude than analyses based on police reports) have been observed using alternative measures of alcoholinvolvement not influenced by the reporting practices of police (such as analyses of single-vehicle, nighttime, and weekend crashes). In addition to evaluations of the lowered drinking age, several reports have been made on the effects of raising the drinking age. Roy and Greenblatt (1979) compared the number of teenagers charged with driving under the influence of liquor (DUlL) who appeared in Massachusetts courts before the legal age was raised with similar data for a one-month period after the drinking-age change.! Small increases in youthful DUlL arrests were used to conclude that the raised drinking age led to increased drinking-driving

24

I

I

II I I

I Ii

I II

I

,1,.11


problems among youth. This study, however, is flawed in the following respects: (1) DUlL arrests alone are an inadequate measure of alcoholrelated highway-safety hazards, because young drivers are more likely than older drivers to be involved in an alcohol-related crash but are less likely to be arrested for DUlL (Organization for Economic Cooperation and Development 1975);2 (2) the design is a single-group pretest-posttest, inherently characterized by low internal validity because of its lack of comparison groups and an extended series of observations (Cook and Campbell 1979); (3) related to the basic inadequacy of the research design is the lack of any statistical control on time-ordered trends, seasonal cycles, or random fluctuations in the frequency of DUlL arrests. As a result, this study provides little useful information concerning the effects of a raised drinking age. Hingson and others (1983) also evaluated the effects of the increase from eighteen to twenty in the drinking age in Massachusetts (effective April 1979). Traffic-fatality data from April 1976 through April 1981 were analyzed for Massachusetts and compared with similar data for New York, which had not raised the drinking age. After the drinking age was raised, fatal crashes among eighteen-to-nineteen-year-old Massachusetts drivers decreased 15 percent, versus a 16 percent increase among eighteen-to-nineteen-year-olds in New York. A similar effect was not found for sixteen-to-seventeen-year-old drivers. Telephone surveys of Massachusetts and New York youth before and after the new law revealed decreases in self-reported nonfatal-crash involvement in both states. Analyses of sixteen-to-nineteen-year-olds as a single group indicated that the decrease in Massachusetts was not significantly greater than the decrease in New York. Responses to questions concerning the frequency of driving after drinking revealed no change in driving after heavy drinking that was attributable to the higher drinking age. However, the proportion of sixteen-to-nineteen-year-old Massachusetts youth who reported driving after any drinking (as opposed to driving after heavy drinking) decreased from 51 percent before the raised drinking age to 42 percent after; no decrease was found in New York. Separate results for the focal group of interest, eighteen-to-nineteen-year-olds, were not reported. Voas and Moulden (1980) briefly reported analyses of the effect of raising the drinking age in Maine from eighteen to twenty. Their initial analyses of eighteen-year-old drivers indicated larger reductions in single-vehicle nighttime crashes by men (many of which involve alcohol) than in singlevehicle daytime crashes (few of which involve alcohol). They estimated an average reduction of 2.4 crashes per month among eighteen-year-old drivers over the first year with the new law. No effect was found for nineteen-yearolds. The authors pointed out that these results were based on early analyses for a short postchange period in only one small state and that additional results from other states are needed before any general conclusions on the effects of higher drinking ages are made.

Alcohol Availabil

Klein (1981) also e' age. Involvement in n nighttime crashes arne crashes. Time-series ar significant reductions i after Maine's higher Ie Maxwell (1981) cor from 1977 through 198 single-vehicle nighttim twenty-year-old drivers to twenty-one (effectiv in these measures were the age of twenty. Filkins and Flora ( youth in Michigan. A twenty-one significant I of involvement in alco old drivers. Separate significant changes ass concluded that the 0' drinking age influence young people. Wagenaar (1981) aJ motor-vehicle crashes i 1979. Using a multiplecrashes among young alcohol-related crashes, pared with that of olde cent reduction in invc drivers was associated Separate analyses wen results did not reveal an small subset of all cras] Williams and others to 1980 in nine states t neighboring states with Two of the states raised eluded grandfather c1m continue to drink; cras twelve months after the decreases in youth invol age was raised; crash re estimated reduction apr: percent reduction in you

1

nd Traffic Accidents

is flawed in the following luate measure of alcohol[rivers are more likely than . crash but are less likely to jc Cooperation and Develpretest-posttest, inherently of its lack of comparison :ook and Campbell 1979); :h design is the lack of any lal cycles, or random flucresult, this study provides f a raised drinking age. effects of the increase from lsetts (effective April 1979). )ril 1981 were analyzed for r New York, which had not : was raised, fatal crashes ts drivers decreased 15 per:en-to-nineteen-year-olds in xteen-to-seventeen-year-old 'lew York youth before and ted nonfatal-crash involveeteen-year-olds as a single setts was not significantly to questions concerning the lange in driving after heavy jng age. However, the prousetts youth who reported dng after heavy drinking) jng age to 42 percent after; sults for the focal group of reported. lalyses of the effect of rais{enty. Their initial analyses in single-vehicle ,ve alcohol) than in single[cohol). They estimated an 19 eighteen-year-old drivers 1S found for nineteen-year'ere based on early analyses I state and that additional general conclusions on the


25

Klein (1981) also evaluated the effect of Maine's increase in the drinking age. Involvement in nighttime crashes among males and in single-vehicle nighttime crashes among males were used as proxies for alcohol-related crashes. Time-series analyses using data from 1974 through 1979 revealed significant reductions in crash involvement from 14 percent to 20 percent after Maine's higher legal age was implemented. Maxwell (1981) conducted a similar time-series analysis on Illinois data from 1977 through 1980. A significant 9 percent decrease in involvement in single-vehicle nighttime crashes by males was found among nineteen-totwenty-year-old drivers after Illinois raised the drinking age from nineteen to twenty-one (effective January 1980). In contrast, no significant changes in these measures were found for drivers under the age of nineteen or over the age of twenty. Filkins and Flora (1981) examined random samples of crash-involved youth in Michigan. After the drinking age was raised from eighteen to twenty-one significant reductions of 11 to 24 percent were found in the rate of involvement in alcohol-related crashes among eighteen-to-twenty-yearold drivers. Separate analyses of fatal crashes, however, identified no significant changes associated with the higher drinking age. The authors concluded that the overall results clearly demonstrated that the legaldrinking age influences the amount of alcohol-impaired driving among young people. Wagenaar (1981) analyzed a 20 percent random sample of all reported motor-vehicle crashes in Michigan between January 1972 and December 1979. Using a multiple-time-series design, the frequency of alcohol-related crashes among young drivers was compared with the frequency of nonalcohol-related crashes, and crash involvement of young drivers was compared with that of older drivers. Results showed that an estimated 18 percent reduction in involvement in alcohol-related crashes among young drivers was associated with Michigan's increase in the drinking age. Separate analyses were also conducted of fatal-crash involvement. The results did not reveal any significant effect of the raised drinking age on this small subset of all crashes. Williams and others (1981) analyzed fatal-crash involvement from 1975 to 1980 in nine states that raised the drinking age, comparing them with neighboring states with unchanged drinking ages during the period studied. Two of the states raised the drinking age from eighteen to nineteen and included grandfather clauses that permitted those already aged eighteen to continue to drink; crash data in these states were excluded for the first twelve months after the legal changes. Eight of the nine states experienced decreases in youth involvement in nighttime fatal crashes after the drinking age was raised; crash reductions ranged from 6 to 75 percent. The smallest estimated reduction applied to Massachusetts, for which a nonsignificant 6 percent reduction in youth involvement in nighttime fatal crashes was found.

26

I I

I II I,

II I:


Averaging across the nine states examined, the authors concluded that raising the drinking age in any given state should result in a 28 percent reduction in involvement in nighttime fatal crashes among the age group affected by the legal change. Vingilis and Smart (1981) assessed the initial effects of Ontario's increase in the drinking age from eighteen to nineteen. The law included a grandfather clause, permitting those aged eighteen before implementation of the higher legal age to continue to consume alcoholic beverages. Aggregate analyses of drinking-driving convictions and fatal-crash involvement among sixteen-to-twenty-one-year-olds for a multiyear period revealed no significant changes associated with the increase in drinking age. These results should be interpreted cautiously for several reasons. First, only a small effect of the new law was expected, because the legal age was increased by only one year, with a grandfather clause that excluded eighteen-year-olds who already had the right to drink from the provisions of the new law. With a grandfather clause in the law, and analyses of only the first year under the new law, only a small proportion of eighteen-year-old Ontario drivers were denied easy access to alcohol during the study period. Finally, convictions and fatalities, being relatively rare events, are measures of limited sensitivity and respond only to major shifts in drinking-driving behavior of a sizable proportion of the population. Because of these factors it is important to consider these results preliminary findings, as Vingilis and Smart pointed out. The National Highway Traffic Safety Administration (1982) compared serious-injury rates (including both motor-vehicle and non-motor-vehicle injuries) across states with low drinking ages (eighteen or nineteen) and states with high drinking ages (twenty or twenty-one), using hospitalemergency-room data. States with higher drinking ages had lower nighttime serious-injury rates among youth under twenty-one years old than did states with lower drinking ages. Most of the studies reported to date have found significant effects of the minimum legal-drinking age on youth involvement in alcohol-related crashes. Significant effects of raising the legal age appear less likely when the drinking age is changed by only one year (for example, from eighteen to nineteen), when a grandfather clause is included in the new law, and when limited short-term postchange data are available. Analyses based on small numbers of drivers (for example, only those involved in fatal crashes or only those convicted of drinking and driving) also are less likely to identify a significant effect of a change in law. Given these considerations, the comprehensive evaluation of the increase in legal-drinking age in Maine and Michigan reported later in this book examined all reported crashinvolved drivers in Maine and Michigan from January 1972 through December 1979.

Alcohol Availab

Research on the and Alcohol Consum

Implicit in research c problems among you amount or the patten the impact of the dri less methodologically of drinking-age chan: ference is that age-sp( tion are usually not a alcohol consumption consumption data: (1 sumption patterns an gregate sales volumes Surveys of the effl tion among youth ha' (1973) surveyed Michi age was lowered, and substantial increase i quantity of alcohol CI was lowered. The au consistent with the I therefore could not drinking age. Smart and Schmit of Toronto junior- an drinking age, 40 perc( terns, 20 percent repc and 9 percent indicate had been reduced (the also surveyed college: frequency or quantit report increased patn legal change. McFadden and V 1965, 1970, and tween 1965 and 1970, as between 1970 and l twenty-one to eightee students in 1977 and drinking age consume with a high drinking

I

Traffic Accidents

lOrs concluded that raislt in a 28 percent reduc19 the age group affected effects of Ontario's ineen. The law included a 1 before implementation l1coholic beverages. Agand fatal-crash involvenultiyear period revealed e in drinking age. These al reasons. First, only a le legal age was increased :luded eighteen-year-olds ons of the new law. With ly the first year under the -old Ontario drivers were iod. Finally, convictions ures of limited sensitivity ing behavior of a sizable actors it is important to Igilis and Smart pointed

;tration (1982) compared e and non-motor-vehicle ghteen or nineteen) and tty-one), using hospitalages had lower nighttime years old than did states

md significant effects of ement in alcohol-related e appear less likely when example, from eighteen led in the new law, and [able. Analyses based on involved in fatal crashes Iso are less likely to identhese considerations, the .I-drinking age in Maine ined all reported crashJanuary 1972 through


27

Research on the Legal-Drinking Age and Alcohol Consumption

Implicit in research on the effects of the drinking age on alcohol-related problems among youth is the assumption that the legal age influences the amount or the pattern of alcohol consumption by young people. Studies of the impact of the drinking age on alcohol consumption among youth are less methodologically sophisticated than is research that assesses the impact of drinking-age changes on traffic crashes. The main reason for this difference is that age-specific data on longitudinal trends in alcohol consumption are usually not available. Studies of the effects of the drinking age on alcohol consumption among youth have used three main types of alcoholconsumption data: (1) self-reported consumption, (2) perceptions of consumption patterns among youth reported by school officials, and (3) aggregate sales volumes. Surveys of the effects of lowering the drinking age on alcohol consumption among youth have produced conflicting results. Wolfe and Chapman (1973) surveyed Michigan high-school students in 1971, before the drinking age was lowered, and again in 1973, after the reduction in drinking age. A substantial increase in the frequency of drinking and an increase in the quantity of alcohol consumed per occasion were found after the legal age was lowered. The authors pointed out, however, that the increases were consistent with the preexisting trends in alcohol use among youth and therefore could not be unambiguously attributed to the lowered legaldrinking age. Smart and Schmidt (1975) conducted a similar before-and-after survey of Toronto junior- and senior-high-school students. After a reduction in the drinking age, 40 percent of the students reported no change in drinking patterns, 20 percent reported drinking more, 4 percent reported drinking less, and 9 percent indicated that they had started drinking since the drinking age had been reduced (the other 27 percent were abstainers). Smart and Schmidt also surveyed college students, the majority of whom reported no change in frequency or quantity of alcohol consumption, although 55 percent did report increased patronization of public drinking establishments since the legal change. McFadden and Wechsler (1979) surveyed Massachusetts teenagers in 1965, 1970, and 1974. Alcohol consumption for this group increased between 1965 and 1970, when there was no change in the drinking age, as well as between 1970 and 1974, when there was a reduction in the legal age from twenty-one to eighteen. Those authors also surveyed New England college students in 1977 and found that students from states with a low legaldrinking age consumed alcohol more frequently than students from states with a high drinking age.

'I

28

- I

i

I

I -I

1,1

'il

I",Iii

Alcohol Availabili1


Rooney and Swartz (1977) surveyed high-school students in three selected states with minimum legal-drinking ages at eighteen and two selected states with higher drinking ages, one at twenty and the other at twenty-one. The survey samples were not necessarily representative of the high-school-aged population in the states examined. A slightly higher proportion of students in states with high drinking ages than in states with lower drinking ages reported at least weekly consumption of beer (47 percent versus 42 percent). Furthermore, a lower prevalence of abstainers (19 percent versus 24 percent) and a higher incidence of alcohol-related problems was found in states with high drinking ages. The authors concluded that a high drinking age has no beneficial effect in controlling alcohol consumption among young people and that it may even have adverse effects. Opposite results were obtained by Maisto and Rachal (1980) in their analyses of a probability sample representative of high-school students in the United States. They found that students in states with a higher legaldrinking age were more likely to be abstainers, less likely to be heavy drinkers, and experienced intoxication less frequently than students in states with lower drinking ages. The authors concluded that legal availability of beverage alcohol, as reflected in the drinking age, is associated with drinking practices of young people. Two survey studies of the effect of increasing the drinking age on drinking patterns among youth have been reported. Vingilis and Smart (1981) surveyed junior- and senior-high-school students in Ontario in 1977, when the drinking age was eighteen, and again in 1979, just two months after the legal age had been increased to nineteen. The number of drinkers aged eighteen and nineteen and the proportion who reported getting "tight" in the previous month decreased significantly after the higher age limit was implemented. In contrast, the number of drinkers aged seventeen and under and the proportion in this group who reported getting "tight" increased significantly from 1977 to 1979. Three considerations should be kept in mind when interpreting the results of this study. First, students were questioned just two months after the drinking-age change; results therefore only apply to the short-term impact of the new law. Second, eighteen-year-olds, the only age category directly affected by the drinking-age increase from eighteen to nineteen, were not analyzed separately, but were combined with nineteen-year-olds in the analyses. Third, Ontario's new drinking-age law included a grandfather clause; as a result, few eighteen-year-olds actually were barred from drinking in the first two months after the law took effect. Hingson and others (1983) surveyed Massachusetts youth before and after an increase in the legal-drinking age. No significant changes in reported alcohol consumption among sixteen-to-nineteen-year-olds (analyzed as a single group) occurred after the drinking age was raised from eighteen

i

to twenty. Separate reSl nineteen-year-olds, were Perceptions of scho( changes in alcohol cons, the drinking age. Hamm over three hundred Mich dent drinking following the Toronto area found' students at school functi Schmidt 1975). The surv after the drinking age w Smart 1981). Although tl survey reported no appre did perceive less drinkin months after Michigan n 830 high-school administ effects of the new law 1979). Of the 50 percent t cent reported fewer alco: drinking among students The third major type ing ages on alcohol-consl and Schmidt (1975) com and after a reduction in tl five months after the the prechange figures. Inc on-premise sales, strengtl was at least partially resp Aggregate distilled-s1 Barsby and Marshall ( drinking ages on spirit sal sions, however, by notin! in distilled-spirits consur would have to be large t figures (which include spi tle is known about chang change in location or qu lowered drinking age, fo consequences, independe analyses applied only to , popular beverages among the study was short, inch legal changes.

Id Traffic Accidents school students in three at eighteen and two twenty and the other at uily representative of the ed. A slightly higher proages than in states with mmption of beer (47 perof abstainers (19 : of alcohol-related prob>. The authors concluded 11 controlling alcohol conven have adverse effects. Id Rachal (1980) in their If high-school students in tates with a higher legal, less likely to be heavy Itly than students in states I that legal availability of , is associated with drink:ing the drinking age on rted. Vingilis and Smart students in Ontario in again in 1979, just two nineteen. The number of )ortion who reported getIlificantly after the higher .umber of drinkers aged oup who reported getting 79. Three considerations suits of this study. First, the drinking-age change; pact of the new law. Secdirectly affected by the een, were not analyzed ear-olds in the analyses. 1 grandfather clause; as a from drinking in the ilusetts youth before and o significant changes in Ileteen-year-olds (analyzed was raised from eighteen


29

to twenty. Separate results for the focal group of interest, eighteen- and nineteen-year-olds, were not presented. Perceptions of school officials have also been used as an indicator of changes in alcohol consumption among youth concomitant with changing the drinking age. Hammond (1973) found that the majority of a sample of over three hundred Michigan high-school principals reported increased student drinking following the reduction in drinking age. A similar survey in the Toronto area found that vice-principals reported more drinking among students at school functions after the drinking age was lowered (Smart and Schmidt 1975). The survey of Ontario vice-principals was repeated in 1980 after the drinking age was raised from eighteen to nineteen (Vingilis and Smart 1981). Although the majority of vice-principals who responded to the survey reported no appreciable change in student drinking, some 40 percent did perceive less drinking and drunkenness among students. Finally, six months after Michigan raised its drinking age from eighteen to twenty-one, 830 high-school administrators were questioned on their perceptions of the effects of the new law (American Businessmen's Research Foundation 1979). Of the 50 percent that responded to the mailed questionnaire, 74 percent reported fewer alcohol-related problems and 69 percent reported less drinking among students after the higher drinking age was passed. The third major type of data used to assess the impact of reduced drinking ages on alcohol-consumption patterns is aggregate sales volumes. Smart and Schmidt (1975) compared Ontario beverage-alcohol shipments before and after a reduction in the legal age. Apparent consumption during the first five months after the legal change was higher than expected on the basis of the prechange figures. Increased alcohol sales were particularly noticeable for on-premise sales, strengthening the argument that the lowered drinking age was at least partially responsible for the observed changes. Aggregate distilled-spirits sales in twenty-five states were examined by Barsby and Marshall (1977). No significant effects of lowered legaldrinking ages on spirit sales were found. The authors tempered their conclusions, however, by noting four limitations of their study. First, any change in distilled-spirits consumption by youth following drinking-age changes would have to be large before the impact would be seen in the aggregate figures (which include spirits consumed by drinkers of all ages). Second, little is known about changes in consumption patterns after legal changes; a change in location or quantity consumed per occasion as a result of the lowered drinking age, for example, could have significant adverse health consequences, independent of the total quantity consumed. Third, the analyses applied only to distilled spirits, not beer or wine, which are more popular beverages among young drinkers. Fourth, the time span covered by the study was short, including only one year before and one year after the legal changes.

30


In their study, Douglass and Freedman (1977) avoided the last two design limitations by examining the monthly aggregate sales of draft beer, packaged beer, wine, and distilled spirits in Michigan over an eight-year period. A statistically significant increase in draft-beer sales was associated with lowering the drinking age. 3 The authors attributed the shift in draft-beer sales to the lowered drinking age because no other confounding factors were identified that could have plausibly accounted for the observed relationship. No significant shifts were identified for any of the other beverage categories. Per-capita beer consumption in Canada was examined by Smart and Finley (1977). Eight provinces that lowered their drinking ages were compared with two that had not changed during the study period. Although increased beer sales were evident in the comparisons for three provinces that experienced a reduction in the drinking age, the increases were similar in magnitude to the experience of the two comparison provinces. Moreover, beer sales decreased in the other five provinces that had reduced the drinking age. The study's findings do not allow any general conclusion as to the effect of a lowered drinking age on total beer sales. Finally, Smart (1977) compared sales of beer, wine, and distilled spirits in twenty-five states that reduced the drinking age with twenty-five states with unchanged drinking ages. 4 Although no significant differences between the states were identified for wine or distilled spirits, increases in beer sales were about 6 percent greater in the states with lowered drinking ages than in the states with an unchanged legal age. In summary, literature on the effects of modified legal-drinking ages clearly indicates that there is an inverse relationship between the minimum age for purchase and consumption of alcoholic beverages and involvement in alcohol-related crashes among young drivers. States that lower the minimum age frequently experience increased rates of youth involvement in traffic crashes, whereas states that raise the minimum age generally have decreased crash rates among youth. The magnitude of the estimated impact" varies across jurisdictions, however, and depends to some extent on the sophistication of the research methods used. In contrast to the relatively Consistent pattern of findings across most crash studies, literature concerning the effects of changing the legal-drinking age on beverage-alcohol consumption among youth provides inconsistent results and has major methodological limitations. Additional research is required before any general conclusion can be made about the effects of the drinking age on drinking practices among young people.

I

Ii

I

I",

Model of the Effects of Changes in the Legal-Drinking Age The impact of changes in the legal-drinking age on involvement in traffic crashes is not direct, but rather is mediated by a variety of intervening

....

Alcohol Availab

variables. One obviot crash-involvement ou1 broader model of th drinking age might sented in figure 3-1. 11 fluence drinking behal in social norms concer marketing activities oj availability of alcohol Drinking norms r law (Bonnie 1980; Me age may be perceived encouraged for youn established. Young : drinkers before the pressure to drink as drinking and as they I alcohol is an integral the model, might reSl reduction in the legalpected to have opposi ing among youth and young drinkers. Disa age might cause the 1 (bar and tavern drink drink because alcohol support for some of t (1980), who analyzed and found that youth peer approval of drinl reported by students Marketing activit to depend on the lega to result in advertisi outlets oriented tow locating a substantia college campuses). A practices specifically therefore alter the s drinking. Changing the leg beverage alcohol to alcohol availability h;

d Traffic Accidents

voided the last two design es of draft beer, packaged an eight-year period. A vas associated with lowerhift in draft-beer sales to mding factors were iden)bserved relationship. No :r beverage categories. examined by Smart and drinking ages were comudy period. Although in; for three provinces that increases were similar in on provinces. Moreover, It had reduced the drinkleral conclusion as to the wine, and distilled spirits .e with twenty-five states differences bel spirits, increases in beer th lowered drinking ages lified legal-drinking ages ip between the minimum :rages and involvement in i that lower the minimum h involvement in traffic generally have decJeased estimated impact varies extent on the sophisticae relatively consistent patconcerning the effects of onsumption among youth >logicallimitations. Addi.on can be made about the mg young people.

involvement in traffic a variety of intervening

III


31

variables. One obvious variable between a drinking-age modification and crash-involvement outcomes already discussed is drinking among youth. A broader model of the mechanism through which changes in the legaldrinking age might cause changes in involvement in traffic crashes is presented in figure 3-1. It is proposed that changes in the legal-drinking age influence drinking behavior and alcohol-related crashes by causing (1) changes in social norms concerning drinking among young people, (2) changes in the marketing activities of the beverage-alcohol industry, and (3) changes in the availability of alcohol to the target age group. Drinking norms may change because of the symbolic function of the law (Bonnie 1980; Mosher 1980); that is, a reduction in the legal-drinking age may be perceived as an indicator that alcohol use is acceptable or even encouraged for young people. As a result, new patterns of drinking are established. Young people who were nondrinkers or only occasional drinkers before the lowered drinking age experience increased social pressure to drink as more of their friends and associates increase their drinking and as they participate in more social situations in which beverage alcohol is an integral part. Such changes in drinking norms, according to the model, might result in increased drinking among young people after a reduction in the legal-drinking age. Conversely, a higher drinking age is expected to have opposite effects, symbolizing society's disapproval of drinking among youth and concern over the extent of alcohol problems among young drinkers. Disapproval of drinking symbolized by a raised drinking age might cause the reduction or elimination of certain drinking patterns (bar and tavern drinking, for example) and a reduction in social pressure to drink because alcohol would be present in fewer social situations. Empirical support for some of these propositions was provided by Maisto and Rachal (1980), who analyzed data on a nationwide sample of high-school students and found that youth in states with higher legal-drinking ages reported less peer approval of drinking and less perceived drinking among peers than was reported by students in states with lower drinking ages. Marketing activities of the beverage-alcohol industry are also expected to depend on the legal-drinking age. A low drinking age would be expected to result in advertising campaigns and location and design of drinking outlets oriented toward the young drinking population (for example, locating a substantial number of establishments with entertainment near college campuses). A higher drinking age might reduce such marketing practices specifically designed to encourage drinking among youth and therefore alter the social and ecological structures that induce teenage drinking. Changing the legal-drinking age also results in altered availability of beverage alcohol to the affected population. The concept of beveragealcohol availability has numerous dimensions and has been defined in many

32


C

Q)

.cE

E

enQ) 0>

(1)

g 00

C

00

0-

N

..s:::

o

I

"U

0 0

UI

:§

(1)

U

C

'>

00.'-

>....

«._ 0

I

o

o

'C'

1 .:Jt:.o C.c .- Q)

:.;;2

Olen C C .0 .:Jt:..C-

.- 0 :J

0::

(/)

>.

Olen

cO

.c..c

Co

Q)Q)en :J

o

oo::::_C

«

gf :£2 .S ....

gjl

OeD

":::0-

(1)

-

.S E

d

o OJ)

01 0

01-

;>

"0 ......

'-'

OQ)OI

5

0:=

I:

Q

o

E E '2

....:l

i=:

"0

.....

..c: rn

ro I-< UI designated as drinking by 78). The SVNM indicator is reliability over time and the Longitudinal consistency in icularly important for time)ort measure is potentially

'.

Research Design and Analysis

43

unreliable because changes in the legal-drinking age might affect the way in which police officers report the presence of alcohol among young, crashinvolved drivers rather than the actual phenomenon of involvement in alcohol-related crashes among youth. Because it is unlikely that police bias would affect the reporting of the driver's sex, number of vehicles involved, or time of crash, the SVNM indicator provides a more reliable alternative to reported alcohol involvement based on the subjective assessments of investigating police officers. Although the SVNM measure is perhaps less valid than the alcohol-involvement measure reported by the police, its use ensures that observed effects of the drinking age are not simply reflecting changes in the reporting of alcohol involvement. Because of the uncertain reliability of police-reported alcohol involvement many individuals dismiss any results based on police reports of drinking. It is important to recognize wide differences across states in the quality of police-reported information on drinking among crash-involved drivers and to use the data as one indicator of alcohol involvement in those states where standardized systems of recording are adequately developed. Because of these considerations, both drinking and SVNM crashes were used to measure involvement in alcohol-related crashes. The consistency of the findings across both measures provided evidence of the validity of each. Data-collection efforts were aimed at the acquisition of complete records on all crash-involved drivers reported to police authorities in each of the four states between January 1972 and December 1979. A census of all reported crashes was successfully obtained for Maine, Michigan, and Pennsylvania. A more limited set of data for New York was acquired. Two crash files were obtained, containing subsets of all reported crashes in New York. The first file included all crashes that occurred between January 1975 and December 1979 in which the investigating police officer reported alcohol as a contributing causal factor. The second New York file contained all singlevehicle crashes that involved male drivers, also for the period from 1975 to 1979. Thus, both indicators of alcohol-related crashes were available, but no indicators of the frequency of non-alcohol-related crashes were analyzed for New York. Because measures of alcohol-related crashes among youth were the most important variables for comparison states that did not change the drinking age, the limitations of the New York file were of minor consequence for interpretation of the study findings. Variables used from the original crash-data files acquired from the four states included (1) the police officer's judgment concerning whether the driver had been drinking; (2) the type of vehicle (automobile, motorcycle, pickup truck, heavy truck, and so on); (3) the age of the driver; (4) the sex of the driver; (5) the time of day at which the crash occurred; (6) the month in which the crash occurred; (7) the severity of the crash (seriousness of in-

44


juries and property damage resulting from the crash); and (8) the number of vehicles involved in the crash. The operational definitions of most of these variables-age and sex of driver and time of crash, for example-are straightforward and comparable across states. Some differences in the definition of vehicle type, number of vehicles involved in the crash, and reported alcohol involvement occurred across states. The vehicle-type variable was used to exclude a variety of miscellaneous traffic units included in comprehensive crash files. The goal was to include drivers of automobiles, light trucks, and motorcycles, but to exclude pedestrians and drivers of other vehicles, such as farm equipment, snowmobiles, bicycles, buses, and heavy trucks. One of the major reasons for the examination of single-vehicle crashes was that fault for the crash can be more readily attributed to the driver. In multivehicle crashes, determination of fault is frequently impossible. For the purposes of this study, a single-vehicle crash was defined as a crash involving one vehicle in transport. For example, one moving vehicle striking an automobile stopped in traffic is not a single-vehicle crash; one vehicle striking a parked vehicle is a single-vehicle crash. Vehicles striking pedestrians or bicyclists were also not considered single-vehicle crashes because pedestrians and bicyclists are moving traffic units and frequently cause the crash. The desired operational definition of police-reported drinking was the officer's simple judgment as to whether the driver had been drinking at the time of the crash, as opposed to the more complex judgment by the officer as to the extent to which alcohol consumption was a primary or contributory cause of the crash. Furthermore, results of chemical tests for the presence of alcohol in a driver's body were not used as an indicator for the incidence of alcohol-related crashes because only a small fraction of all drinking drivers involved in crashes are chemically tested for alcohol. Crash files acquired from the several states had a variety of data structures and file formats and therefore required different processing protocols. Details for each state follow.

Michigan Crash Data The Michigan Department of State Police supplied comprehensive crashdata files on all reported motor-vehicle crashes in the state. The data were reformatted for research studies, using the Automated Data Access and Analysis System, a series of computer programs specifically designed for the management of data on motor-vehicle crashes. Because over half a million individuals each year are involved in motor-vehicle crashes reported in Michigan, 20 percent random samples of all crashes were selected for

Research Design and

each year from 1972 througl reduction in the legal-drink available files were used in tl time period. For 1978 anc Michigan were constructed ;; Because the files for 1972 to resulting crash frequencies . them comparable to the 19/ All the crash files were i light trucks (such as pickup buses, farm and constructi, were excluded from the am The crash data for Mi, data rates on such type of vehicle. Missing-da1 table 4-1. Driver records with research design were exclud item on the crash-report for to make a decision about d vehicle. As a result, the states. The exact phrasing ( form is: "Driver had takeI alcohol or drugs." "Unkno vestigating officer leaves thi other drugs as well as alcohc as "Had taken alcohol or ,

Maine Crash Data Maine's Department of Tra involved drivers in the stat Table 4-1 Missing-Data Rates for Variable

Police reported had been dri Driver sex Driver age Vehicle type Time of crash Month of crash

Id Traffic Accidents ;h); and (8) the number of :finitions of most of these for example-are straight:rences in the definition of 'ash, and reported alcohol a variety of miscellaneous i. The goal was to include ;ycles, but to exclude pelrm equipment, snowmoof single-vehicle crashes ttributed to the driver. In 'equently impossible. For was defined as a crash inIe moving vehicle striking vehicle crash; one vehicle . Vehicles striking pedesIe-vehicle crashes because : and frequently cause the 1

'eported drinking was the . had been drinking at the x judgment by the officer as a primary or contribu)f chemical tests for the ed as an indicator for the y a small fraction of all ly tested for alcohol. ld a variety of data strucrent processing protocols.

ed comprehensive crashthe state. The data were )mated Data Access and specifically designed for les. Because over half a r-vehicle crashes reported :rashes were selected for


45

each year from 1972 through 1977 and used in earlier research on Michigan's reduction in the legal-drinking age. To reduce data-processing costs, these available files were used in the present investigation for the Michigan baseline time period. For 1978 and 1979, census files of all reported crashes in Michigan were constructed and used as the basis for the dependent variables. Because the files for 1972 to 1977 contained only 20 percent of all crashes, the resulting crash frequencies for these years were multiplied by five to make them comparable to the 1978 and 1979 census data. All the crash files were filtered to include only drivers of passenger cars, light trucks (such as pickups, jeeps, and vans), and motorcycles. Drivers of buses, farm and construction machinery, and other miscellaneous vehicles were excluded from the analyses. The crash data for Michigan were characterized by minimal missingdata rates on such variables as police-reported drinking, age of driver, and type of vehicle. Missing-data rates for the 1978 Michigan file are shown in table 4-1. Driver records with missing data on any of the variables required for the research design were excluded from analyses. As noted earlier, the drinking item on the crash-report form in Michigan requires the investigating officer to make a decision about drinking for every driver of every crash-involved vehicle. As a result, the missing-data rate is low compared with that of other states. The exact phrasing of the drinking item on Michigan's crash-report form is: "Driver had taken alcohol or drugs" or "Driver had not taken alcohol or drugs." "Unknown" is coded in the data files only when the investigating officer leaves this item blank. Although the item includes use of other drugs as well as alcohol, the overwhelming majority of drivers recorded as "Had taken alcohol or drugs" were using alcohol alone. Maine Crash Data

Maine's Department of Transportation provided data on all reported crashinvolved drivers in the state from 1972 through 1979. Unfortunately, the Table 4-1 Missing-Data Rates for the State of Michigan, 1978 Variable Police reported had been drinking Driver sex Driver age Vehicle type Time of crash Month of crash

Percent Missing 8,7

0.0 2.9 4.7 0.0 0.0

Ii

, I I

I

46


original Maine files did not contain the necessary information to link specific records of drivers with the corresponding records of the vehicles driven at the time of the crash. As a result, miscellaneous vehicles could not be filtered out, and the Maine data used in subsequent analyses included all types of motor vehicles recorded in the original files. The condition-of-driver variable used to measure alcohol-related crashes was coded using the following six categories: (1) apparently normal, (2) had been drinking, (3) under the influence of liquor, (4) under the influence of drugs, (5) asleep, and (6) fatigued. Any driver identified as "had been drinking" or "under the influence of liquor" was considered a drinking driver for this study. No unusual missing-data problems were identified in the Maine crash files (table 4-2). Those cases missing information or with code values undefined for any of the required variables were excluded from subsequent analyses.

New York Crash Data Data on involvement in motor-vehicle crashes in New York were obtained from the New York Division of Alcohol Abuse and Alcoholism. Two sets of files were obtained for the period from 1975 through 1979. The first contained all crash-involved drivers for which the investigating police officer indicated that alcohol was a contributing causal factor in the crash. The second set of files included all single-vehicle crashes that involved male drivers. Time series of alcohol-.related crashes reported by the police and SVNM crashes were analyzed, but frequencies of all daytime crashes or crashes with no police-reported drinking were not available. The frequency of daytime single-vehicle crashes among males was used as an indicator of non-alcohol-related crashes for New York, whereas all daytime crashes were used for the other three states.

Table 4-2 Missing-Data Rates for the State of Maine, 1979 Variable

Condition of driver Driver sex Driver age Time of crash Month of crash

Research Design and The frequency of crash was relatively low in the Ne volved in crashes in New Y parent factor in the crash; 0' Michigan in 1979 were code of the nature of the item use question put to the officer is all. In New York the investi plex judgment of whether a the crash. In New York alcc tive factors from which offil most significant. Immediate are frequently coded rathe driver, which often is the Ul Information on the type included in the data files pre types of vehicles were inclu of all types of vehicles in the because over 92 percent of types included in this inves cycles). An important change iJ fected interpretation of the amount of property damag creased from $250 to $400. damage crashes decreased. trolled in subsequent analys Missing-data rates for t table 4-3. The lack of because of the nature of the apparent causative factors) apparent causal factor was: volved in crashes in New Yc Table 4-3 Missing-Data Rates for

Percent Missing

0.0 0.6 1.2

0.0 0.0

Variable

Alcohol as causative factor Driver sex Driver age Time of crash Month of crash

d Traffic Accidents ;ary information to link 19 records of the vehicles aneous vehicles could not uent analyses included all 'iles. Ire alcohol-related crashes pparently normal, (2) had (4) under the influence of identified as "had been 'as considered a drinking ltified in the Maine crash ion or with code values from subsequent

New York were obtained d Alcoholism. Two sets of ough 1979. The first conlVestigating police officer factor in the crash. The 'ashes that involved male by the police and of all daytime crashes or t available. The frequency as used as an indicator of ereas all daytime crashes

Research Design and Analysis "

'I

47

The frequency of crashes reported by the police to be alcohol related was relatively low in the New York data. About 3 percent of all drivers involved in crashes in New York in 1979 were coded with alcohol as an apparent factor in the crash; over 9 percent of all drivers involved in crashes in Michigan in 1979 were coded as drinking. Part of the difference is a result of the nature of the item used to code alcohol involvement. In Michigan the question put to the officer is simply whether the driver had been drinking at all. In New York the investigating officer must make the much more complex judgment of whether alcohol involvement was a contributing cause of the crash. In New York alcohol involvement is one of forty possible causative factors from which officers must select the two that they believe are the most significant. Immediate causes, such as excess speed or failure to yield, are frequently coded rather than the alcohol-impaired condition of the driver, which often is the underlying cause of the crash. Information on the type of vehicle involved in reported crashes was not included in the data files provided by New York. As in Maine, therefore, all types of vehicles were included in the final time-series variables. Inclusion of all types of vehicles in the analyses was not expected to affect the results because over 92 percent of all crashed vehicles in New York were vehicle types included in this investigation (automobiles, light trucks, or motorcycles). An important change in crash-reporting procedures in New York affected interpretation of the data. In September 1978 the minimum dollar amount of property damage for mandatory reporting of a crash was increased from $250 to $400. As a result, the number of recorded propertydamage crashes decreased. This change in the reporting system was controlled in subsequent analyses. Missing-data rates for the crash variables in New York are shown in table 4-3. The lack of any missing data on the drinking variable was because of the nature of the variable. The valid choice' 'None" (that is, no apparent causative factors) precludes the need for a missing-data code. No apparent causal factor was recorded for about 65 percent of all drivers involved in crashes in New York. Table 4-3 Missing-Data Rates for the State of New York, 1979

1979 Percent Missing

0.0 0.6

1.2 0.0 0.0

Variable

Alcohol as causative factor Driver sex Driver age Time of crash Month of crash

Percent Missing

0.0 4.5 4.5 0.0 0.0

,,

iI , I,

ii, Ii

I

48

I

Research Design and


Finally, since New York City is an unusual crash-reporting jurisdiction, the final time-series variables were constructed in two sets. 2 The first set of variables consisted of crashes that occurred in the entire state, as for Maine, Michigan, and Pennsylvania. The second set of variables excluded all crash reports from New York City. Analyses of both sets of variables revealed that exclusion of New York City had little effect on the findings of this study.

Table 4-4 Missing-Data Rates for Variable

Alcohol as causative factor Driver sex Driver age Vehicle type Time of crash Month of crash

Pennsylvania Crash Data

The drinking variable much like that of New Yc causative factor, selected fr proportion of all crashes re: 1979). In addition, the alco the officer judged alcohol item could not be linked 1 crash. Time series were thl drivers involved in alcoholthe frequency of drinking portant limitation, alcoho police were not analyzed ir

A census of all reported motor-vehicle crashes in the state of Pennsylvania from January 1972 through December 1979 was obtained from the Pennsylvania Department of Transportation. Changes in the criteria for reporting a crash were an important characteristic of the Pennsylvania data. Prior to July 1977, local police departments were not required to use a statewide standardized crash-report form, nor were they required to submit the reports to the State Department of Transportation. Beginning in July 1977, a standard reporting form was adopted, and local police departments were required to submit these reports to the Pennsylvania Department of Transportation. At the same time, criteria for reporting drivers involved in crashes also changed. Prior to July 1977, drivers involved in any crash resulting in damage amounting to $200 or more had to report the crash to the local police. After this date only crashes that resulted in a vehicle being towed from the scene of the crash had to be reported. Crashes that involved personal injury were always required to be reported. These changes in data collection affected the frequency of reported crashes in two important ways. First, the number of reported injury crashes increased, because all injury crashes were now reported to the state office on standard forms. Second, the frequency of reported property-damage crashes decreased substantially, because property-damage crashes were now reported only if a vehicle was towed from the scene of the crash. These effects of changes in the reporting system were controlled in subsequent analyses. Because data received from Pennsylvania contained all reported crashes, the data were filtered to include only drivers of passenger cars, light trucks, and motorcycles. Because of incompatibilities between computer systems, approximately five crashes per year were deleted prior to analysis. This small loss of data was a random occurrence and thus was not expected to affect the results. As in the other states, rates of missing data for Pennsylvania were quite low; rates for calendar 1979 are shown in table 4-4.

Data on Alcohol Sales ana In addition to the primary were conducted of aggreg2 of the higher drinking age. distribution in Maine fron tained from the Maine B monthly wholesale distrib Michigan from January the Michigan Beer and W monthly distribution of dJ the United States as a wh Maine and Michigan,wen addition to analyses of a citations for selling or al were examined briefly. . before the administrative 1979 were provided by thl

I

;Ii

cI Traffic Accidents


sh-reporting jurisdiction, two sets. 2 The first set of :ntire state, as for Maine, riables excluded all crash .ets of variables revealed t on the findings of this

Table 4-4 Missing-Data Rates for the State of Pennsylvania, 1979

he state of Pennsylvania >btained from the Pennthe criteria for reporting msylvania data. Prior to uired to use a statewide red to submit the reports ling in July 1977, a stan::e departments were rea Department of Transing drivers involved in nvolved in any crash reo report the crash to the I in a vehicle being towed ashes that involved per-

frequency of reported f reported injury crashes ted to the state office on orted property-damage image crashes were now : of the crash. These efmtrolled in subsequent

:ontained all reported :of passenger cars, light [ties between computer eleted prior to analysis. :l thus was not expected

)ennsylvania were quite

Variable Alcohol as causative factor Driver sex Driver age Vehicle type Time of crash Month of crash

49

Percent Missing 0.6 1.3 4.6 3.6 0.7 0.0

The drinking variable reported by police in Pennsylvania was coded much like that of New York. Investigating officers identify alcohol as a causative factor, selected from a list of ninety-one possible causes. Thus the proportion of all crashes reported as alcohol related was low (2.6 percent in 1979). In addition, the alcohol item in the Pennsylvania data indicated that the officer judged alcohol to be a causal factor in the crash; this alcohol item could not be linked to a particular driver who was involved in the crash. Time series were thus constructed of the frequency of Pennsylvania drivers involved in alcohol-related crashes, not, as in the other three states, the frequency of drinking drivers involved in crashes. Because of this important limitation, alcohol-related crashes in Pennsylvania reported by police were not analyzed in great detail.

Data on Alcohol Sales and Enforcement In addition to the primary emphasis on traffic crashes, additional analyses were conducted of aggregate alcoholic-beverage sales and the enforcement of the higher drinking age. Data on monthly beer, wine, and distilled-spirits distribution in Maine from January 1972 through February 1981 were obtained from the Maine Bureau of Alcoholic Beverages. The figures for monthly wholesale distribution of wine, draft beer, and package beer in Michigan from January 1969 through September 1980 were obtained from the Michigan Beer and Wine Wholesalers Association. Data on wholesale monthly distribution of draft and package beer in New Hampshire and in the United States as a whole, used for comparison with the results from Maine and Michigan, were obtained from the U.S. Brewers Association. In addition to analyses of aggregate beverage sales, annual frequencies of citations for selling or allowing minors to consume alcoholic beverages were examined briefly. Annual frequencies of such citations brought before the administrative court in the state of Maine from 1971 through 1979 were provided by the Maine Department of Public Safety, Bureau of

I,

ii' I

. I

!

I

'I

:1 I:

i

i

i:

50


Liquor Enforcement. Similar data for Michigan for the period from 1970 through 1980 were provided by the Michigan Liquor Control Commission. Data-Analysis Methods The number of drivers involved in crashes per month was examined for an extended period of time for each of the crash categories included in the research design. Long series of observations were required to assess the degree to which crash-involvement frequencies in 1978 and 1979 (after the drinking age was raised) were different from the level expected, given regular patterns in crash involvement throughout the decade of the 1970s. Examination of both the raw plots of crash involvement and the series smoothed with a simple twelve-month moving average provided preliminary evidence concerning effects of the legal changes. This examination also revealed major long-term trends, periodic seasonal cycles, and variability of each series. 3 The same analytic strategy was applied to the data on aggregate beverage sales. crash and The objective of the analyses was to estimate shifts in alcohol-sales time series associated with the legal interventions in 1977 (Maine) and 1978 (Michigan). To estimate such shifts beginning the first month after a legal change was implemented, long-term trends and seasonal cycles must first be controlled. For example, the increase in crash involvement in 1977 over 1976 (see figure 5-1) is not a significant shift in the time series because the increase was simply a part of the continuous upward trend throughout the mid-1970s. Another example is comparison of crash involvement for any given April-September period with the subsequent October-March period, where a large decrease is usually seen. Again, the change is not a significant shift in the crash time series, but merely a reflection of a long-standing seasonal cycle in motor-vehicle crashes (high rates in summer and low rates in winter). The Box-Jenkins and Box-Tiao (Box and Tiao 1975; Box and Jenkins 1976) intervention-analysis methods were used. The methods combine iterative univariate (baseline) Auto-Regressive Integrated Moving Average (ARIMA) model identification, estimation, and evaluation techniques with simple transfer functions (intervention-impact models). The Box-Jenkins approach is a versatile time-series-modeling strategy that can model a wide variety of trend, seasonal, and other recurring patterns. On a conceptual level, the analytic strategy involves explaining as much of the variance in crash involvement as possible on the basis of the past history of crash involvement, before attributing any of the variance to another variable, such as a change in the minimum legal-drinking age. Comparative studies have found that, in most cases, the Box-Jenkins methods more accurately account

Research Design anc

for regularities in time serie than do alternative stratef Newbold and Granger 1974 tion analysis was particular objective was to identify si/ age distribution associated regularities in the history 0 In addition to controllin time series, effects of specifi of reported crashes were ah Such events included the f1 speed limit of early 1974 a motor-vehicle crashes. To facilitate interpretal series models were used to I variable after a legal changl represent the difference in drinking-age modification fl not been changed. 4 In short, results discus: models for each dependent involvement or beverage dis controlling for the effects 0: factors known to influence or the amount of alcoholic· discussion of the statistical specific models used in the I and Jenkins (1976), for a co: and to McCleary and Hay ( Notes 1. Previous studies of tl alcohol-concentration (BAC: BAC tests are not an optima drinking drivers involved in Michigan, more than 75 pe drinking at the time of the c: Compton 1978). 2. Two major features tl state in terms of crash reporl less-serious crashes and lowe in crashes.

md Traffic Accidents

n for the period from 1970 iquor Control Commission.

nonth was examined for an ategories included in the rerequired to assess the degree and 1979 (after the drinking :xpected, given regular patide of the 1970s. Examinant and the series smoothed wided preliminary evidence imination also revealed mas, and variability of each d to the data on aggregate

He shifts in crash and legal interventions in 1977 :h shifts beginning the first ng-term trends and seasonal Ie increase in crash involvesignificant shift in the time of the continuous upward lple is comparison of crash leriod with the subsequent is usually seen. Again, the e series, but merely a reflecvehicle crashes (high rates in

'iao 1975; Box and Jenkins The methods combine [ntegrated Moving Average l evaluation techniques with models). The Box-Jenkins itegy that can model a wide patterns. On a conceptual as much of the variance in he past history of crash in:to another variable, such as Comparative studies have Jds more accurately account


51

for regularities in time series (as reflected in lower residual-error variances) than do alternative strategies of analysis (Reid, cited in Kendall 1976' Newbold and Granger 1974; Vigderhous 1977). This approach of tion analysis was particularly appropriate for the present study because the objective was to identify significant shifts in crash involvement and beverage distribution associated with legal changes, independent of observed regularities in the history of each variable. In addition to controlling for long-term trends and cycles in the baseline time series, effects of specific events known to have influenced the frequency of reported crashes were also explicitly controlled in the statistical models. Such events included the fuel shortage and decreased national maximumspeed limit of early 1974 and major changes in procedures for reporting motor-vehicle crashes. To facilitate interpretation of the findings, estimates from the timeseries models were used to calculate the percent change in each dependent variable after a legal change was implemented. The percent-change figures represent the difference in crash involvement or alcohol sales after a drinking-age modification from what would have been expected had the law not been changed. 4 In short, results discussed in chapter 5 are based on comprehensive models for each dependent variable. The models estimated shifts in crash or beverage distribution associated with drinking-age changes, controlhng for the effects of long-term trends, seasonal cycles, and specific factors known to influence the numbers of motor-vehicle crashes reported or the amount of alcoholic beverages sold. Appendix B contains a detailed discussion of the statistical methods and additional information on the specific models used in the present study. The reader is also referred to Box and Jenkins (1976), for a comprehensive treatment of the analytic methods, and to McCleary and Hay (1980), for a more readable introduction. Notes

1. Previous studies of the drinking age have used the results of bloodalcohol-concentration (BAC) tests as an indicator of alcohol-related crashes. BAC tests are not an optimal measure of alcohol-related crashes because few drinking drivers involved in crashes are tested for BAC. For example in Michigan, more than 75 percent of all crash-involved drivers reported as drinking at the time of the crash are not tested for BAC (Flora, Filkins, and Compton 1978). 2. Two major features that set New York City apart from the rest of the state in terms of crash reporting are the significantly lower rates of recording less-serious crashes and lower rates of recording the involvement of alcohol in crashes.

I I:

I' I

52


3. The moving-average-trend lines shown in chapter 5 are simple, frontended twelve-month moving averages calculated as follows:

x/

=

Y/ + Yt

- 1

+ Yt -

2

+ Y/ - 3 + ... Y t -

11

12

4. Percent-change figures were based on the first twelve months after an increase in drinking age and were calculated as follows: percent change

12w 12 E

f

(lOO)

-12w

i = 1

where w is the shift in frequency of crash involvement or alcohol distribution estimated by the time-series models and f is the actual monthly frequency of crash involvement or alcohol consumption.

5

The results revealed signifi to alcohol among young ( Decreases were most prOf under the influence of alco ing age was raised. Analyse changes associated with ml specific consumption dat straightforward interpreta findings are presented in tl the effect of the drinking a: ings concerning the relatic the drinking age are prese the minimum drinking ag( Effects of the Raised Drin on Crash Involvement

Findings from analyses 0 York, and Pennsylvania 1 variable, the estimated ave traffic crashes per month series were mathematicall) tion over time in the fn parameter estimates canne of crash-involved drivers· estimates, the summary t estimates calculated as per involved drivers after the number expected had ther The percent figures fc tified in the tables with statistically significant 1-1 beginning the first month that are so large that they E thermore, changes in ill'.

and Traffic Accidents chapter 5 are simple, fronted as follows:

11

+ ... Y, -

11

e first twelve months after an s follows: (100) 12w

vement or alcohol distribuis the actual monthly frelption.

5

Discussion of Results

The results revealed significant reductions in motor-vehicle crashes related to alcohol among young drivers after increases in the legal-drinking age. Decreases were most pronounced in Michigan, where arrests for driving under the influence of alcohol among youth also decreased when the drinking age was raised. Analyses of alcoholic-beverage sales revealed significant changes associated with modifying the drinking age. However, lack of agespecific consumption data and several confounding factors prevented straightforward interpretation of the consumption results. The detailed findings are presented in three sections. First, the results of the main study, the effect of the drinking age on traffic crashes, are discussed. Second, findings concerning the relationship between sales of alcoholic beverages and the drinking age are presented. Finally, available data on enforcement of the minimum drinking age are briefly examined.

Effects of the Raised Drinking Age on Crash Involvement

Findings from analyses of crash involvement in Maine, Michigan, New York, and Pennsylvania are summarized in tables that follow. For each variable, the estimated average change in the number of drivers involved in traffic crashes per month after a change in drinking age is shown. Some series were mathematically transformed because of the nature of the variation over time in the frequency of crash involvement. In such cases, parameter estimates cannot be directly interpreted as the change in number of crash-involved drivers per month. In addition to the time-series-model estimates, the summary tables include t-ratios for the estimates and the estimates calculated as percent change between the actual number of crashinvolved drivers after the implementation of a raised drinking age and the number expected had there been no change in drinking age. The percent figures for those estimates with significant t-ratios, identified in the tables with asterisks, are most important. Estimates with statistically significant t-ratios represent changes in crash involvement beginning the first month after implementation of a higher drinking age that are so large that they are not likely to be the result of chance alone. Furthermore, changes in involvement in alcohol-related crashes should be

53

I

54


compared with changes in involvement in non-alcohol-related crashes, and changes in involvement for young drivers should be compared with the figures for older drivers. These results' were extracted from comprehensive statistical models for each variable. The estimates, therefore, represent net changes in crash involvement associated with raising the drinking age, controlling for long-term trends, seasonal cycles, and other major factors that affect the frequency of reported crash involvement, including fuel shortages, speed-limit changes, and modifications in crash-reporting systems. Taking all these factors into account produced models that fit the data well and in most cases explained over 60 percent of the variation in the original crash series. Findings for each state are now examined in detail. Michigan Crash Findings

Net changes in involvement in property-damage crashes associated with the December 1978 increase in the drinking age in Michigan are summarized in table 5-1. Several findings emerge. First, involvement in property-damage crashes related to alcohol decreased significantly for drivers aged eighteen to twenty, the target of the increase in the drinking age. The pronounced decrease in involvement in property-damage crashes related to alcohol was clearly evident for drivers reported by the police as drinking (figure 5-1) and for single-vehicle nighttime male (SVNM) drivers (figure 5-2). The unusually large number of drivers involved in crashes in December 1978 may reflect one last binge on the part of young drivers before alcohol was no longer legally available, beginning December 23. The apparent upward trend during 1979 does not indicate that the effect of the legal change lasted only a few months, but rather reflects the recurring seasonal pattern of low involvement in alcohol-related crashes in the winter months early in the year and high involvement in the summer and fall months. This seasonal cycle was filtered out before estimating the crash reduction attributable to the raised drinking age. Involvement in property-damage crashes not related to alcohol also decreased significantly for those aged eighteen to twenty, but the magnitude of the decrease was only half as large as the decrease in alcohol-related crashes (table 5-1). Subtracting the percent decrease in involvement in crashes not related to alcohol from the percent decrease in involvement in crashes related to alcohol produces a net reduction of 11 to 22 percent in involvement in alcohol-related property-damage crashes attributable to implementation of the higher drinking age. The 11 percent estimate is based on a comparison between SVNM and daytime crashes, and the 22 percent estimate is based on a comparison between crashes for which police reported drinking and those for which no drinking was reported. The 11 to 22 percent reduction represents 725 to 1,617 fewer young Michigan drivers

Discussion of Resul

Table 5·1 Estimated Change in Prop. Raising the Legal-Drinkinl Type of Crash Alcohol-related indicators Police-reported drinking Estimate (-ratio Percent Single-vehicle nighttime male Estimate (-ratio Percent Non-alcohol-related indicators No police-reported drinking Estimate (-ratio Percent Daytime Estimate (-ratio Percent Source: Wagenaar, A.C. "Preve for Drinking: An Empirical C Reprinted with permission. Note: • Significant at p < .01, t\ tEstimate based on logaritl

involved in property-dama months with the higher Ie: drinking age not been rais( responsible for the crash n the older age groups. Driv four to forty-five exhibitec contrast to those aged eig change. A second major findinl fect on the frequency of p teen and seventeen (table 5· alcohol related decreased drinking decreased by abo alcohol-related crashes can

·s, and Traffic Accidents n-alcohol-related crashes, and hould be compared with the from comprehensive nates, therefore, represent net raising the drinking age, con. and other major factors that nent, including fuel shortages, ish-reporting systems. Taking Is that fit the data well and in variation in the original crash ined in detail.

.ge crashes associated with the I Michigan are summarized in olvement in property-damage ntly for drivers aged eighteen rinking age. The pronounced related to alcohol was ce as drinking (figure 5-1) and ers (figure 5-2). The unusually n December 1978 may reflect before alcohol was no longer e apparent upward trend durthe legal change lasted only a seasonal pattern of low ininter months early in the year l months. This seasonal cycle reduction attributable to the

not related to alcohol also to twenty, but the magnitude e decrease in alcohol-related decrease in involvement in at decrease in involvement in :tion of 11 to 22 percent in ine crashes attributable to im1 percent estimate is based on crashes, and the 22 percent n crashes for which police king was reported. The 11 to ewer young Michigan drivers

55


Table 5-1 Estimated Change in Property-Damage-Crash Involvement Associated with Raising the Legal-Drinking Age in Michigan Type of Crash

16-17

Age of Driver 18-20 21-23

24-45

Alcohol-related indicators Police-reported drinking Estimate

t-ratio Percent Single-vehicle nighttime male Estimate

t-ratio Percent

-21.4 3.0-14.7

-208.2 7.2-34.0

-7.6 0.5 -1.7

44.9 1.2 3.8

-34.9 2.6 -15.5

-129.4 5.1-24.0

-31.2 2.3 -8.8

-20.7 0.5 -2.5

-400.2 3.7 -13.3

-O.l t 3.0-12.2

-O.l t 2.4 -10.4

-324.2 3.4-13.3

-539.9 2.9-12.8

Non-alcohol-related indicators No police-reported drinking Estimate

t-ratio Percent Daytime Estimate

t-ratio Percent

-315.5 2.0 -9.6

-O.l t 2.3 -9.5

-1186.7 2.4 -10.0

Source: Wagenaar, A.C. "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Journal of Sqfety Research 13 (1982):65. Reprinted with permission. Note: -Significant at p < .01, two-tailed test. tEstimate based on logarithmically transformed series.

involved in property-damage crashes related to alcohol over the first twelve months with the higher legal age than would have been expected had the drinking age not been raised. The conclusion that the raised drinking age is responsible for the crash reductions observed is strengthened by results for the older age groups. Drivers aged twenty-one to twenty-three and twentyfour to forty-five exhibited no significant change in crash involvement, in contrast to those aged eighteen to twenty, the target of the drinking-age change. A second major finding was that the raised drinking age had no clear effect on the frequency of property-damage crashes among drivers aged sixteen and seventeen (table 5-1). Although crashes reported by the police to be alcohol related decreased significantly, crashes with no police-reported drinking decreased by about the same amount. As a result, the decrease in alcohol-related crashes cannot be attributed to the increase in drinking age.

!:

'.

:i.

iW

56

Discussion of Resul

Alcohol, Young Drivers, and Traffic Accidents 800

800

___ ACTUAL FREQI ___ MOVING AVERI

____ ACTUAL FREQUENCY _ _ MOVING AVERAGE

700

700

600

600

300

Drinking Age Raised

200

1976

1977

1978

1979

Source: Wagenaar, A.C. "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Journal of Safety Research 13 (1982):65. Reprinted with permission.

Figure 5-1. Police-Reported 18-20- Year-Old Drinking Drivers Involved in Property-Damage Crashes in Michigan

A third trend evident in table 5-1 is that the parameter estimates for non-alcohol-related-crash indicators for all age groups are negative and have relatively large I-ratios even if not significant at the .01 level. Furthermore, all the estimated crash reductions are in the range of 10 to 13 percent. Reduced economic activity in Michigan in 1979 is probably the main cause of the uniform reduction in the numbers of non-alcohol-related crashes across all age groups. The effect of the raised drinking age was also evident from analyses of injury-crash involvement in Michigan (table 5-2). Police-reported alcoholrelated crashes among eighteen-to-twenty-year-olds decreased by 28 percent when the drinking age was raised (figure 5-3); SVNM crashes decreased by 22 percent (figure 5-4). Daytime crashes involving injury also decreased significantly for this age group, but the magnitude of the decrease was only half as large as the decrease in nighttime crashes. Subtracting the 11 percent decrease in daytime crashes from the 22 percent decrease in SVNM crashes

1972

197

Source: Wagenaar, A.C. "Prevel for Drinking: An Empirical 0 . Reprinted with permission.

Figure 5-2. SVNM 18-20-' Crashes in Mil

leaves an 11 percent reduct to the raised drinking age. with no police-reported dri reported by the police to 1 estimate of the effect of thl ing age in Michigan resultc ment in injury crashes rela 1,726 fewer young drivers months after the drinkingthe law not been changed result of the drinking-age c reductions in alcohol-relatl aged twenty-one to twentyfor drivers aged twenty-fOl

S,



57

800 _ _ ACTUAL FREQUENCY ___ MOVING AVERAGE

700

600

500

400

300

Drinking Age Raised

Drinking Age Raised

200 1978

1979

by Raising the Legal Minimum Age of Safety Research 13 (1982):65.

)rinking Drivers Involved in higan

1972

1978

1979

Source: Wagenaar, A.C. "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Journal of Safety Research 13 (1982):65. . Reprinted with permission.

Figure 5-2. SVNM 18-20-Year-Old Drivers Involved in Property-Damage Crashes in Michigan

the parameter estimates for ge groups are negative and :ant at the .01 level. Further:he range of 10 to 13 percent. is probably the main cause non-alcohol-related crashes Iso evident from analyses of ·2). Police-reported alcohololds decreased by 28 percent SVNM crashes decreased by lIving injury also decreased Jde of the decrease was only ;. Subtracting the 11 percent t decrease in SVNM crashes

leaves an 11 percent reduction in alcohol-related crashes that is attributable to the raised drinking age. Since there was no significant change in crashes with no police-reported drinking, the 28 percent reduction in injury crashes reported by the police to be related to alcohol also can be considered an estimate of the effect of the raised drinking age. In short, raising the drinking age in Michigan resulted in a reduction of 11 to 28 percent in involvement in injury crashes related to alcohol. This reduction represents 373 to 1,726 fewer young drivers involved in injury crashes over the first twelve months after the drinking-age increase than would have been expected had the law not been changed.) The conclusion that these reductions were a result of the drinking-age change was strengthened by finding no significant reductions in alcohol-related or non-alcohol-related crashes among drivers aged twenty-one to twenty-three or twenty-four to forty-five. In fact, results for drivers aged twenty-four to forty-five showed increases in involvement

58

Discussion of Resl

Alcohol, Young Drivers, and Traffic Accidents 60(1

Table 5-2 Estimated Change in Injury-Crash Involvement Associated with Raising the Legal-Drinking Age in Michigan Type of Crash

16-17


500

_ ACTUAL FREI ___ MOVING AVE

24-45

Alcohol-related indicators Police-reported drinking Estimate t-ratio Percent

-20.2 2.5 -16.8

-143.8 4.6· -27.8

9.6 0.6 2.5

88.2 2.3 8.7

Single-vehicle nighttime male Estimate t-ratio Percent

-12.7 2.3 -12.6

-63.0 3.6· -21.5

9.3 1.3 5.1

38.6 2.4 9.8

400

300

Non-alcohol-related indicators No police-reported drinking Estimate t-ratio Percent Daytime Estimate t-ratio Percent

200 -100.8 2.5 -8.5

-160.4 1.8 -7.8

-60.3 0.8 -3.8

-263.0 1.0 -4.8

-73.2 2.1 -7.7

-187.7 3.0· -10.9

-53.0 0.8 -4.0

-219.9 1.0 -4.4

100

o'

1!

Figure 5-3. Police-Repol Injury Crast

The substantial effel analyses of multiple time amination of the frequer , quor (DUlL) in Michiga drivers decreased 8.5 P( twenty-one and over eXT quency of arrest (table 5 drivers aged twenty-one l tion to the drinking-driv however, was caused by police officers to arrest ir (Public Laws 384 and 39 could not arrest a person i

I

1973

Source: Wagenaar, A.C. "Pre' for Drinking: An Empirical Reprinted with permission.

in alcohol-related crashes, in contrast to the decreases evident for eighteento-twenty-year-old drivers. Although the increases among drivers aged twenty-four to forty-five were not significant at the .01 level, examination of the time'series plots reveals higher involvement in injury crashes related to alcohol in 1979 than previous trends would have indicated (figures 5-5 and 5-6). No unequivocal effects -of the raised drinking age on injury-crash involvement among Michigan drivers aged sixteen and seventeen were found. Some evidence, however, indicates that the raised drinking age resulted in fewer alcohol-related crashes among underage drivers. Although not statistically significant at the .01 level, sixteen-to-seventeen-year-olds experienced reductions of 13 to 17 percent in alcohol-related crashes and a reduction of only 8 percent in non-alcohol-related crashes. The larger decrease in alcohol-related crashes may be viewed as an effect of the higher drinking age.

I

i i

1972

Source: Wagenaar, A.C. "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Journal of Safety Research 13 (1982):66. Reprinted with permission. Note: ·Significant at p < .01, two-tailed test.

,nd Traffic Accidents

59

Discussion of Results 600

lssociated with Raising the ge of Driver )

500

21-23

___ ACTUAL FREQUENCY ___ MOVING AVERAGE

24-45 .."

8 6* 8

9.6 0.6 2.5

88.2 2.3 8.7

0 6* 5

9.3 1.3 5.1

38.6 2.4 9.8

"

I

400

300

200 4 8 8 7 0* 9

-60.3 0.8 -3.8 -53.0 0.8 -4.0

-263.0 1.0 -4.8 -219.9 1.0 -4.4

the Legal Minimum Age Safety Research 13 (1982):66.

evident for eighteenlses among drivers aged ,he .01 level, examination t in injury crashes related ave indicated (figures 5-5

g age on injury-crash innd seventeen were found. I drinking age resulted in drivers. Although not o-seventeen-year-olds exlol-related crashes and a lted crashes. The larger as an effect of the higher

100

Drinking Age Raised

0 1972

1973

1974

1975

1976

1978

1979


Figure 5-3. Police-Reported 18-20-Year-Old Drinking Drivers Involved in Injury Crashes in Michigan The substantial effect of the raised drinking age identified through analyses of multiple time series was corroborated by a more rudimentary examination of the frequency of arrests for driving under the influence of liquor (DUlL) in Michigan. 2 DUlL arrests for eighteen-to-twenty-year-old drivers decreased 8.5 percent from 1978 to 1979, whereas drivers aged twenty-one and over experienced a substantial 20 percent increase in frequency of arrest (table 5-3). A small part of the higher arrest rates among drivers aged twenty-one and over might have resulted from increased attention to the drinking-driving problem in recent years. Most of the increase, however, was caused by a change in Michigan law in 1978, which allowed police officers to arrest intoxicated drivers at crash scenes without a warrant (Public Laws 384 and 391 of 1978). Prior to this change in policy, officers could not arrest a person for DUlL unless they personally witnessed a driver

Discussion of Resu


60

1,600

500 _ _ ACTUAL FREQUENCY _ _ MOVING AVERAGE

ACTUAL FRE MOVING AVE

Drinking Age Raised

1,400

400 1,200

300

--

1,000

800

200

600 100 1972 1972

1973

1979

1973

19

Figure 5-5. Police-Report Injury Crashe


Nevertheless, the unexpect provided further evidenci youth was significantly ree One question not adel of a raised drinking age. S or two years of postchang monthly crash data for th< totals of crash-involved substantial decrease in i drivers aged eighteen to t 1 drivers aged twenty-one t< from 1978 to 1979. It is also clear that I twenty-four age groups ex crashes in 1980 and 1981. . for drivers of all ages, weI sion in Michigan as well

Figure 5-4. SVNM 18-20-Year-Old Drivers Involved in Injury Crashes in Michigan

operating a motor vehicle while under the influence. Thus, prior to August 1978, numerous crash-involved drivers who had obviously been driving under the influence prior to a crash were not arrested because the officer did not personally witness the impaired driving. It might be safely assumed that drivers aged eighteen to twenty would have experienced the same 20 percent increase in DUlL arrests experienced by twenty-one-to-twenty-three-year-olds as a result of the warrantless-arrest law if there had been no reduction in drinking and driving attributable to implementation of the higher drinking age. Given this assumption, the actual number of DUlL arrests among eighteen-to-twenty-year-old drivers in 1979 was an estimated 29 percent lower than would have been expected had the drinking age remained the same. The arrest findings are less sure than the crash findings because of lack of analyses of extended time series. I, I



61

I ,I

1,600

____ ACTUAL FREQUENCY _ _ MOVING AVERAGE

Drinking Age Raised

I

. I

!

1,400

!

:I

I

1,200

1,000

800

Drinking Age Raised

600 1972 1977

1978

1979

Raising the Legal Minimum Age

1973

1974

1975

1976

1977

1978

1979

Figure 5-5. Police-Reported 24-45-Year-Old Drinking Drivers Involved in Injury Crashes in Michigan

. Safety Research 13 (1982):66.

:>lved in Injury Crashes in

Ice. Thus, prior to August ,d obviously been driving Ited because the officer did

eighteen to twenty would DUlL arrests experienced lt of the warrantless-arrest Ind driving attributable to n this assumption, the actwenty-year-old drivers in ld have been expected had findings are less sure than of extended time series.

Nevertheless, the unexpected decrease in DUlL arrests among young drivers provided further evidence that the amount of impaired driving among youth was significantly reduced by the increase in drinking age. One question not adequately addressed concerns the long-term effects of a raised drinking age. Studies to date have necessarily been limited to one or two years of postchange data. Although detailed time-series analyses of monthly crash data for the early 1980s have not yet been completed, annual totals of crash-involved drinking drivers are shown in figure 5-7. The substantial decrease in involvement in alcohol-related crashes among drivers aged eighteen to twenty is again clearly evident and contrasts with drivers aged twenty-one to twenty-four, who continued their upward trend from 1978 to 1979. It is also clear that both the eighteen-to-twenty and twenty-one-totwenty-four age groups experienced reduced involvement in alcohol-related crashes in 1980 and 1981. The reductions for 1980 and 1981, which occurred for drivers of all ages, were probably a result of the severe economic recession in Michigan as well as other highway-safety programs during those


62

Discussion of Resu

7001

13,000

_ _ ACTUAL FREQUENCY _ _ MOVING AVERAGE

12,500

600

12,000 500

en

11,500

0:

w

> 0:

11,000

0

u.

a

300

0:

w

co

1972

1973

1974

1975

1976

1978

10,000

::>

z

Drinking Age Raised

200

10.500

1979

Figure 5-6. SVNM 24-45-Year-Old Drivers Involved in Injury Crashes in Michigan

/

.

,/ ,/ ,/ ,/ ,/

9,500 9,000 8,500 8,000

years. In any event, the crash reductions among eighteen-to-twenty-year-old drivers resulting from the drinking-age change do not appear to be a temporary phenomenon. These crash reductions were maintained in 1980 and 1981, and young drivers experienced further crash reductions similar to those experienced by drivers of all ages in Michigan. These results are con-

7,500

1975

Table 5-3 Number of Drivers Arrested for Driving under the Influence of Liquor in Michigan Year Age

1978

1979

Percent Change

13-17 18-20 21-23 24-45

706 4,412 4,283 15,993

789 4,035 5,156 19,456

11.8 -8.5 20.4 21.6


L

Source: Wagenaar, A.C. "Preve for Drinking: An Empirical C Reprinted with permission

Figure 5-7. Police-Reportl Cr'ashes in Mi, sistent with the findings 01 the drinking age in the ear ment in alcohol-related I drinking ages were found.



63

13,000

,,/\

"""

12,500 12,000 / (j)

a:

11,500 /

UJ

>

a:

11,000

Cl lJ..

0

/

10,500

a:

UJ a:l

:::i:

1978

10,000

:J

z

'rinklng Age Raised

1977

/

1979

rolved in Injury Crashes in

eighteen-to-twenty-year-old do not appear to be a temere maintained in 1980 and ;rash reductions similar to igan. These results are con-

9,500

/

"

'79

Percent Change

789 [}35 156 456

11.8 -8.5 20.4 21.6

Raising the Legal Minimum Age . Safety Research 13 (1982):67.

/

/

/

\

\

"

\

\

\

\

\

\

"

"

9,000

Legal Drinking Age Raised

8,500 8,000 7,500

-

DRIVERS AGE 18-20 DRIVERS AGE 21-24

1975

the Influence of Liquor in

/

/

/

/

/

/

/

',,,, ,\

1976

1977

1978

1979

1980

1981

YEAR Source: Wagenaar, A.C. "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Journal of Safety Research 13 (1982):67. Reprinted with permission

Figure 5-7. Police-Reported Drinking Drivers Involved in Motor-Vehicle Crashes in Michigan sistent with the findings of multiyear follow-ups of the effects of lowering the drinking age in the early 1970s, where permanent increases in involvement in alcohol-related crashes after implementation of lower legaldrinking ages were found.

64


Maine Crash Findings

Discussion of Resul 50

Estimated effects of the raised drinking age on minor, property-damage crashes in Maine are shown in table 5-4. Drivers aged eighteen to nineteen experienced a significant 17 percent reduction in involvement in propertydamage crashes reported by the police to be alcohol related and a 22 percent reduction in SVNM crashes. Figures 5-8 and 5-9 show crash frequencies significantly lower than expected; the moving-average-trend line levels off after the drinking age was raised, in contrast to the previous long-term upward trend. Drivers aged eighteen to nineteen showed no significant change in involvement in property-damage crashes not related to alcohol, whether measured by police reports or daytime crashes (table 5-4). In addition, no significant changes in any of the four crash series were observed for drivers aged twenty to twenty-one or twenty-two to forty-five in Maine. A signifi-

_ _ ACTUAL FREQUE _ _ MOVING AVERA(

40

30

20

10

Table 5-4 Estimated Change in Property-Damage-Crash Involvement Associated with Raising the Legal-Drinking Age in Maine

, 1972

16-17


Police-reported drinking Estimate t-ratio Percent

-0.3 t 1.4 -24.4

-5.1 3.7· -16.8

-3.0 1.5 -10.5

-7.0 1.8 -6.9

Single-vehicle nighttime male Estimate t-ratio Percent

-2.8 2.2 -15.9

-7.5 3.5· -21.5

-0.9 0.5 -3.6

0.2 14.9

Type of Crash

0'

22-45

Alcohol-related indicators

1.1

Non-alcohol-related indicators No police-reported drinking Estimate t-ratio Percent

5.8 0.5 3.0

-19.8 1.2 -6.4

-12.5 0.6 -4.8

-28.8 0.3 -1.9

Daytime Estimate t-ratio Percent

3.5 0.3 2.2

-10.8 0.7 -4.3

-8.0 0.5 -3.6

-31.4 0.3 -2.3

Source: Adapted from Wagenaar, A.C. "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Marcel Dekker, Inc., New York. Reprinted with permission. Note: ·Significant at p < .01, two-tailed test. tEstimate based on logarithmically transformed series.

, 1974

Source: Wagenaar, A.C. "Raisin, cidents among Young Drivers." Dekker, Inc., New York. Reprint

Figure 5-8. Police-Reporte Property-Dam cant reduction in alcoho drivers in Maine at the til parable reductions in eitl drivers or alcohol-related CI the drinking age apparent damage crashes related tc represents sixty-one to nine the first year with the high basis of previous trends. Although the time-se property-damage crashes re seventeen, the estimates we chosen for this study. HOWl to-seventeen-year-olds in M in the time series may mask drinkers. Examination of tt

md Traffic Accidents 50

minor, property-damage ; aged eighteen to nineteen 1 involvement in property101 related and a 22 percent -9 show crash frequencies verage-trend line levels off the previous long-term up)wed no significant change 'elated to alcohol, whether table 5-4). In addition, no ; were observed for drivers y-five in Maine. A signifi-

65


1


40

30

20

10

Drinking Age Raised

volvement Associated with Ol----..,....---....,....----,----r----r---""'"r-----,r---1972

ge of Driver

•

1979

20-21

22-45

-3.0 1.5 -10.5

-7.0 1.8 -6.9

Figure 5-8. Police-Reported 18-19-Year-Old Drinking Drivers Involved in Property-Damage Crashes in Maine

-0.9 0.5 -3.6

1.1 0.2 14.9

-12.5 0.6 -4.8

-28.8 0.3 -1.9

-8.0 0.5 -3.6

-31.4 0.3 -2.3

cant reduction in alcohol-related property-damage crashes for young drivers in Maine at the time the drinking age was raised, with no comparable reductions in either non-alcohol-related crashes among young drivers or alcohol-related crashes among older drivers, indicates that raising the drinking age apparently caused the observed reduction in propertydamage crashes related to alcohol. The reduction of 17 to 22 percent represents sixty-one to ninety fewer young drivers involved in such crashes the first year with the higher drinking age than would be expected on the basis of previous trends. Although the time-series-model estimates indicated reductions in property-damage crashes related to alcohol among drivers aged sixteen and seventeen, the estimates were not significant using the .01 probability level chosen for this study. However, the small monthly crash counts for sixteento-seventeen-year-olds in Maine and the resulting large random component in the time series may mask an effect of the raised drinking age on underage drinkers. Examination of the raw frequency plots (figures 5-10 and 5-11) in-

inking Age in Maine: Impact on nal of the Addictions 18 (1983).

Source: Wagenaar, A.C. "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Marcel Dekker, Inc., New York. Reprinted with permission.


66

Discussion of Resul 20

60

ACTUAL FREQUENCY MOVING AVERAGE

50

_ _ ACTUAL FREQ _ _ MOVING AVER

15

40

10

30

20 0' 10

Drinking Age Raised

1972

,

i

197'

Figure 5-10. Police-Repor in Property-I

oI

,

1972

,

,

i

,

I,

I

1979

Source: Wagenaar, A.C. "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Marcel Dekker, Inc., New York. Reprinted with permission.

30

_ _ ACTUAL FREQI _ _ MOVING AVER,

25

Figure 5-9. SVNM 18-19-Year-Old Drivers Involved in Property-Damage Crashes in Maine 20

dicated a decline in crash involvement among underage drinkers for the first year after the drinWJ.ng age was raised; however, the crash series resumed its long-term upward trend in 1979. Analyses of serious crashes, that is, those involving at least one injury or fatality, revealed no significant shifts in involvement in alcohol-related crashes among Maine drivers aged eighteen tonineteen (table 5-5). Similarly, there were no measurable changes in the frequency of serious crashes for drivers aged twenty-two to forty-five and no changes in involvement in serious crashes that were not alcohol related for drivers aged twenty to twenty-one. However, drivers aged eighteen to nineteen were involved in 24 percent more serious daytime crashes after October 1977 than previous trends indicate would be expected. 3 Maine drivers aged twenty to twentyone experienced a significant increase in alcohol-related crashes, averaging 4.9 drivers per month (21 percent) for the police-reported-drinking' measure, and 2.4 drivers per month (14 percent) for the SVNM measure.

15

10

5

0'

,

1972

1973

,

197'

Figure 5-11. SVNM 16-17 Crashes in

I

I nd Traffic Accidents

67

Discussion of Results 20

____ ACTUAL FREQUENCY _ _ MOVING AVERAGE Drinking Age Raised

15

10

5

0,L-_-r----r---r----r---r----r----,.---

1979

1972

aised

Figure 5-10. Police-Reported 16-17-Year-Old Drinking Drivers Involved in Property-Damage Crashes in Maine 1977

1979

30

Maine: Impact on Traffic Ace Addictions 18 (1983). Marcel

I


25

Ived in Property-Damage

drinkers for the first le crash series resumed its

olving at least one injury vement in alcohol-related teen (table 5-5). Similarly, lCy of serious crashes for langes in involvement in r drivers aged twenty to leteen were involved in 24 ober 1977 than previous s aged twenty to twentyelated crashes, averaging police-reported-drinking or the SVNM measure.

20

15

10

5 Drinking Age Raised

oL----r--..,---.,....---r----'T--J.,---...,...--1972

1973

1974

1979

Figure 5-11. SVNM 16-17-Year-Old Drivers Involved in Property-Damage Crashes in Maine

I

I

68


A significant increase in involvement in serious daytime crashes among eighteen-to-nineteen-year-old drivers in Maine, with no significant change in involvement in serious SVNM crashes, might indicate that the raised drinking age prevented an increase in involvement in SVNM crashes that would have occurred had there been no legal change. Such a conclusion is strengthened by the significant increase in involvement in alcohol-related crashes for drivers aged twenty to twenty-one, the proximal peers of the focal group. In addition, although involvement in serious SVNM crashes among drivers aged eighteen to nineteen showed no statistically significant change, the results of the time-series-model estimation indicate an 18.4 percent decrease when the drinking age was raised. Because the magnitude and direction of the estimated change in SVNM crashes among drivers aged eighteen to nineteen was consistent with hypothesized drinking-age effects, the results might be interpreted as evidence that Maine's raised drinking age affected involvement in serious crashes. Given the tenuousness of the argument concerning SVNM and serious daytime crashes, and no changes in serious crashes with or without police-reported drinking, the most conserTable 5-5 Estimated Change in Injury-Crash Involvement Associated with Raising the Legal-Drinking Age in Maine Type of Crash

16-17


22-45

Alcohol-related indicators Police-reported drinking Estimate t-ratio Percent

2.7 1.6 30.5

1.9 0.8 7.9

4.9 2.8* 21.4

3.2 0.8 3.5

0.4 0.3 2.8

-3.2 1.4 -18.4

2.4 2.8* 14.2

2.5 0.9 4.4

Percent

0.8 0.2 l.l

-1.8 0.2 -1.5

Daytime Estimate t-ratio Percent

2.1 0.5 3.2

Single-vehicle nighttime male Estimate t-ratio Percent Non-alcohol-related indicators No police-reported drinking Estimate

t-ratio

-1.4 0.2 -1.4

-18.3 0.7 -3.5

5.7 l.l 7.2

-26.4 1.0 -5.4

Discussion of Resu vative conclusion is that th raised drinking age on ill among eighteen-to-ninetee Analyses of Maine d significant changes in in: drinking age.

New York Crash Findings

The strategy of plied to New York to del crashes among young pee Maine or Michigan raised which did not change its c ment similar to those in M perimental states could be the raised drinking age. Estimated changes in i York, beginning in Noven are summarized in table changes in any of the eras: Table 5-6 Estimated Change in Pro after October 1977, when Type of Crash Alcohol-related indicators Police-reported drinking Estimate t-ratio Percent Single-vehicle nighttime male Estimate t-ratio Percent Non-alcohol-related indicator

19.0 4.4* 24.0

Source: Adapted from Wagenaar, A.C. "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Marcel Dekker, Inc., New York. Reprinted with permission. Note: *Significant atp < .01, two-tailed test.

Single-vehicle daytime male Estimate

t-ratio Percent

Note: *Significant at p < .01, t tEstimate based on logaril

nd Traffic Accidents daytime crashes among vith no significant change t indicate that the raised nt in SVNM crashes that .nge. Such a conclusion is vement in alcohol-related he proximal peers of the in serious SVNM crashes no statistically significant Ition indicate an 18.4 perecause the magnitude and shes among drivers aged ,ized drinking-age effects, aine's raised drinking age : tenuousness of the arguIshes, and no changes in rinking, the most conserLIS

I\ssociated with Raising


69

vative conclusion is that there was no clearly demonstrable effect of Maine's raised drinking age on involvement in injury crashes related to alcohol among eighteen-to-nineteen-year-old drivers. Analyses of Maine drivers aged sixteen and seventeen revealed no significant changes in injury-crash involvement after Maine raised the drinking age. New York Crash Findings The strategy of time-series analysis used for Maine and Michigan was applied to New York to determine whether involvement in alcohol-related crashes among young people in that state decreased significantly when Maine or Michigan raised the minimum legal-drinking age. If New York, which did not change its drinking age, showed decreases in crash involvement similar to those in Maine or Michigan, observed reductions in the experimental states could be considered the result of some factor unrelated to the raised drinking age. Estimated changes in involvement in property-damage crashes in New York, beginning in November 1977 (when Maine raised the drinking age), are summarized in table 5-6. For New York, there were no significant changes in any of the crash-involvement categories for eighteen-to-twenty-

'e of Driver 20-21

22-45

4.9 2.8* 21.4

3.2 0.8 3.5

2.4 2.8* 14.2

2.5 0.9 4.4

-1.4 0.2 -1.4

-18.3 0.7 -3.5

5.7 1.1 7.2

-26.4 1.0 -5.4

Table 5-6 Estimated Change in Property-Damage-Crash Involvement in New York after October 1977, when Maine Raised the' Legal-Drinking Age Age of Driver Type of Crash

16-17

18-20

0.2 0.2 3.1

21-23

24-45

3.5 2.1 12.7

1.6 0.9 6.9

12.0 2.5 15.3

-0.3 t 1.5 -23.7

-OAt 6.7* 29.5

-0.7t 4.1* -47.8

Alcohol-related indicators Police-reported drinking Estimate t-ratio Percent Single-vehicle nighttime male Estimate t-ratio Percent

-0.7 t 3.7* -52.3

nking Age in Maine: Impact on of the Addictions 18 (1983).

101

I

I

'1\il

Non-alcohol-related indiCator Single-vehicle daytime male Estimate t-ratio Percent

I

.

-0.3 t 2.5 -28.8

-0.3 t 1.9 -22.9

Note: *Significant at p < .01, two-tailed test. tEstimate based on logarithmically transformed series.

-OAt 3.7* -32.3

-0.6t 4.7* -45.1

'I

i

70

Discussion of Resul


year-old drivers, in contrast to Maine, where eighteen-to-nineteen-year-old drivers experienced significant reductions in property-damage crashes related to aicohol beginning in November 1977. Parameter estimates for SVNM and single-vehicle daytime male crashes were negative for all age groups and were significant for drivers aged twenty-one to twenty-three and twenty-four to forty-five. Furthermore, the reported-crash reductions were substantial, ranging from 23 to 52 percent. The fact that reported involvement in property-damage crashes was lower than expected across all age groups for both daytime and nighttime single-vehicle male crashes was a result of an important change in crash-reporting criteria mentioned in chapter 4. After September 1978, any crash that resulted in property damage of $400 or more had to be reported to the police; prior to this date, any crash that resulted in damage of $250 or more had to be reported. This change, which occurred within the first year after the October 1977 drinking-age change in Maine, apparently caused significant reductions in property-damage crashes reported in New York. Estimated changes in involvement in injury crashes in New York concomitant with Maine's increase in the drinking age are shown in table 5-7. As with property-damage crashes, no significant changes in any crash category for eighteen-to-twenty-year-old drivers were found. The other three age groups experienced significant increases in crashes reported by the Table 5-7 Estimated Change in Injury-Crash Involvement in New York after October 1977, when Maine Raised the Legal-Drinking Age Type of Crash

16-17


5.3 4.0· 18.1

Single-vehicle nighttime male Estimate (-ratio Percent

5.6 1.7 10.5

5.4 0.4 3.1 -3.4 0.4 -1.3

21.5 5.0· 18.0 -5.3 0.6 -3.0

27.5 2.8· 7.3 7.9 0.5 2.3

Non-alcohol-related indicator Single-vehicle daytime male Estimate (-ratio Percent

-3.4 0.7 -3.9

Note: ·Significant at p < .01, two-tailed test.

-0.1 0.0 -0.5

- 1.0 0.2 -0.9

Table 5-8 Estimated Change in Pro after December 1978, whe Type of Crash

24-45

Alcohol-related indicators Police-reported drinking Estimate (-ratio Percent

police to be alcohol York is not based on a sim is based on the police offic in the crash. Using this i crashes is probably quite se of alcohol-related crashes more attention to alcohol observed increases in invo to be alcohol related may the involvement of alcoho Time-series models fo crash involvement beginni drinking age. Results for 1= 5-8. No significant reduct damage crashes occurred a York in 1979, in contrast t provided further support Michigan were a direct res general factors that influe this period.

-32.7 7.1· -8.9

Alcohol-related indicators Police-reported drinking Estimate (-ratio Percent Single-vehicle nighttime male Estimate (-ratio Percent Non-alcohol-related indicator Single-vehicle daytime male Estimate (-ratio Percent

Note: ·Significant atp < .01, t tEstimate based on logaril


d Traffic Accidents lteen-to-nineteen-year-old . property-damage crashes Parameter estimates for were negative for all age y-one to twenty-three and ted-crash reductions were 'act that reported involven expected across all age male crashes was a ng criteria mentioned in .hat resulted in property : police; prior to this date, e had to be reported. This after the October 1977 I significant reductions in :rashes in New York conge are shown in table 5-7. .nt changes in any crash ; were found. The other in crashes reported by the

n New York after October ge 21-23

21.5 5.0· 18.0 -5.3 0.6 -3.0

-1.0 0.2 -0.9

police to be alcohol related. Reporting of alcohol-related crashes in New York is not based on a simple drinking versus nondrinking item, but rather is based on the police officer's selection of alcohol as a major causal factor in the crash. Using this indicator, frequency of reported alcohol-related crashes is probably quite sensitive to increased public interest in the problem of alcohol-related crashes. The media and policymakers have been giving more attention to alcohol and highway-safety issues in recent years, and observed increases in involvement in serious crashes reported by the police to be alcohol related may very well be the result of improved reporting of the involvement of alcohol in New York crashes. Time-series models for New York also included estimates of shifts in crash involvement beginning in December 1978, when Michigan raised its drinking age. Results for property-damage crashes are summarized in table 5-8. No significant reductions in involvement in alcohol-related propertydamage crashes occurred among eighteen-to-twenty-year-old drivers in New York in 1979, in contrast to significant decreases in Michigan. This finding provided further support for the hypothesis that the crash reductions in Michigan were a direct result of the higher drinking age, not of other more general factors that influenced crash involvement in various states during this period. Table 5-8 Estimated Change in Property-Damage-Crash Involvement in New York after December 1978, when Michigan Raised the Legal-Drinking Age Type of Crash

ge of Driver 24-45

27.5 2.8· 7.3 7.9 0.5 2.3

-32.7 7.1· -8.9

71

16-17


24-45

Alcohol-related indicators Police-reported drinking Estimate (-ratio Percent Single-vehicle nighttime male Estimate (-ratio Percent

-1.5 0.8 -4.7

4.2 1.9 16.4

2.1 0.5 2.6

O.ot 0.1 3.0

0.3 t 1.7 35.0

0.3 t 7.2· 35.0

O.lt 0.0 6.2

O.ot 0.2 2.0

O.4t 3.1· 46.2

O.ot 0.0 4.1

O.lt 0.6 7.2

2.0 1.6 34.6

I

I

I


Note: ·Significant at p < .01, two-tailed test. tEstimate based on logarithmically transformed series.

II !I

I1j

!i, "

:1

72


The only significant time-series-model estimates in table 5-8 were for SVNM crashes among drivers aged twenty-one to twenty-three and singlevehicle daytime male crashes among drivers aged eighteen to twenty. In both of these cases, the significant positive time-series-model estimates were a result of the leveling off of a preexisting downward trend in crash frequency (figures 5-12 and 5-13). Table 5-9 shows estimated changes from preexisting trends in involvement in serious crashes in New York in 1979. Again, in contrast to the Michigan results, no significant decreases in the frequency of injury and fatal crashes were found among eighteen-to-twenty-year-old drivers. The lack of reduction in involvement in injury-producing crashes among youth in New York when Michigan raised the drinking age and experienced significant decreases in youth involvement provided further evidence that Michigan's crash reductions among youth resulted from the increase in drinking age. Significant increases in alcohol-related crashes reported by the police for three of the four age groups might be the result of improved police

Discussion of Results reporting of alcohol involv gested explanation of the in in New York reported by tl 1977, when Maine raised table 5-7, however, table 5-' indicator, SVNM crash invo the groups aged eighteen tc estimated increase for tv significant at the less conSI positive estimates, with five that a real increase in invol' curred in 1979 among New

Pennsylvania Crash

Crash involvement in Pen analyzed for comparison wi! reported by the police to b

140 _ ACTUAL FREQUENCY _ _ MOVING AVERAGE

120

160

_ _ ACTUAL FREQl _ _ MOVING

140 120

40 20

o

20

----"""T-----r-----...,...------,----1975 I 1976 I 1977 i i 1978 1979

L..'

Figure 5-12. SVNM 21-23-Year-Old Drivers Involved in Property-Damage Crashes in New York

0'

1975

,

1976

Figure 5-13. Male 18-20-' Daytime, Pn

Traffic Accidents s in table 5-8 were for wenty-three and singleeighteen to twenty. In $-model estimates were trend in crash frequency sting trends in involve;ain, in contrast to the 'equency of injury and y-year-old drivers. The 19 crashes among youth g age and experienced d further evidence that d from the increase in

73


reporting of alcohol involvement. Improved reporting was also the suggested explanation of the increased frequency of property-damage crashes in New York reported by the police to be alcohol related after November 1977, when Maine raised its drinking age. Unlike the pattern of results in table 5-7, however, table 5-9 reveals that the more reliable alcohol-related indicator, SVNM crash involvement, also exhibited significant increases for the groups aged eighteen to twenty and twenty-four to forty-five and an estimated increase for twenty-one-to-twenty-three-year-olds that was significant at the less conservative .05 level. The pattern of consistently positive estimates, with five of the eight significant at the .01 level, suggests that a real increase in involvement in injury crashes related to alcohol occurred in 1979 among New York drivers.

Pennsylvania Crash Findings reported by the police ult of improved police

Crash involvement in Pennsylvania from 1972 through 1979 was also analyzed for comparison with Michigan and Maine. The measure of crashes reported by the police to be alcohol related available in the Pennsylvania II

160

_____ ACTUAL FREQUENCY __ MOVING AVERAGE

I,

140 120 100 80

60 40 20

978

1979

ed in Property-Damage

1975

1976

1977

1978

1979

Figure 5-13. Male 18-20-Year-Old Drivers Involved in Single-Vehicle, Daytime, Property-Damage Crashes in New York

74

Discussion of Resul


Table 5·10 Estimated Change in Prope after October 1977, when 1

Table 5-9 Estimated Change in Injury-Crash Involvement in New York after December 1978, when Michigan Raised the Legal-Drinking Age Type of Crash

16-17

Age of Driver 21-23

18-20

24-45

Alcohol-related indicator

Alcohol-related indicators Police-reported drinking Estimate (-ratio Percent Single-vehicle nighttime male Estimate (-ratio Percent

11.2 7.0* 33.1

23.7 2.3 11.9

31.1 6.6*

i1.2

56.8 5.0* 13.7

3.2 0.9 5.2

32.7 3.6* 12.2

24.3 2.6 14.0

59.6 4.6* 17.5

1.1 0.2 1.2

7.0 1.0 4.7

4.0 1.0 3.7

7.6 1.5 2.2

Single-vehicle nighttime male Estimate (-ratio Percent Non-alcohol-related indicator


Type of Crash

Daytime Estimate (-ratio Percent Total

Note: *Significant at p < .01, two-tailed test.

files was not well suited for this research. Few crashes were identified as alcohol related, and for those that were, the alcohol item applied to crashes, not to drivers. The item indicated whether there was at least one driftking driver involved in the crash but did not identify the particular driver that had been drinking. To avoid use of the police-reported alcohol indicator, crashes with police-reported drinking and those with no reported drinking were not analyzed separately. Instead, the total frequency of crash involvement was analyzed. The results, shown in table 5-10, indicate that Pennsylvania drivers experienced no change in involvement in property-damage crashes in 1977, when Maine raised its drinking age and experienced a significant decrease in involvement of youth in property-damage crashes related to alcohol. Again, a state without a drinking-age change experienced no change in crash involvement among youth while a state that raised its drinking age experienced significant reductions in crash involvement. Analyses of Pennsylvania injury crashes revealed no significant changes beginning in 1977 for any of the crash-category-age-group combinations (table 5-11). When Michigan raised its drinking age and experienced significant decreases in involvement of young drivers in property-damage crashes related to alcohol, Pennsylvania showed no change in youth involvement in

Estimate (-ratio Percent

Source: Adapted from Wagenaar, Traffic Accidents among Young I Marcel Dekker, Inc., New York. Note: tEstimate based on logaritt

such crashes (table 5-12) sylvania among drivers 0 however. Similar reductio groups were found in Mic general factor affecting CI economic recession in 197 non-alcohol-related crashe found in the Maine results not related to alcohol also analyzed, only New York in property-damage crashl Analyses of involveme significant changes in 197 binations (table 5-13). The related to alcohol among] significant decreases in yc causal interpretation of th and fewer motor-vehicle c

75


,d Traffic Accidents

Table 5-10 Estimated Change in Property-Damage-Crash Involvement in Pennsylvania after October 1977. when Maine Raised the Legal-Drinking Age

n New York after I-Drinking Age

Age of Driver

e of Driver 21-23

24-45

Type of Crash

16-17

18-19

20-21

22-45

0.2 0.0 0.1

-17.2 0.8 -8.9

-12.7 0.8 -7.3

-77.1 1.6 -12.6

Alcohol-related indicator 31.1 6.6·

:J1.2

56.8 5.0· 13.7

24.3 2.6 14.0

59.6 4.6· 17.5

Single-vehicle nighttime male Estimate (-ratio Percent Non-alcohol-related indicator Daytime Estimate (-ratio Percent

-O.lt 1.9 -12.2

-O.ot 0.5 -3.4

-O.lt 0.7 -4.8

-O.ot 0.3 -2.1

::

Total 4.0 1.0 3.7

7.6 1.5 2.2

:rashes were identified as )1 item applied to crashes, was at least one drinking the particular driver that ported alcohol indicator, vith no reported drinking equency of crash involve-

Pennsylvania drivers exdamage crashes in 1977, d a significant decrease in related to alcohol. Again, d no change in crash in. drinking age experienced

ed no significant changes ·age-group combinations

: experienced significant lroperty-damage crashes in youth involvement in

Estimate (-ratio Percent

-O.lt 1.7 -10.6

-O.lt 0.8 -5.0

-O.ot 0.5 -2.9

-O.ot 0.3 -2.0

Source: Adapted from Wagenaar, A.C. "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Marcel Dekker, Inc., New York. Reprinted with permission. Note: tEstimate based on logarithmically transformed series.

such crashes (table 5-12). Daytime property-damage crashes in Pennsylvania among drivers of all ages were down 9 to 17 percent in 1979, however. Similar reductions in non-alcohol-related crashes across all age groups were found in Michigan. These findings indicate that some moregeneral factor affecting crash involvement in multiple states, such as the economic recession in 1979, may be the cause of reduced involvement in non-alcohol-related crashes. Further support for such an argument can be found in the Maine results, where involvement in property-damage crashes not related to alcohol also decreased significantly in 1979. Ofthe four states analyzed, only New York did not exhibit obvious decreases in involvement in property-damage crashes not related to alcohol in 1979. Analyses of involvement in serious crashes in Pennsylvania revealed no significant changes in 1979 for any of the crash-category-age-group combinations (table 5-13). The lack of change in involvement in serious crashes related to alcohol among Pennsylvania youth in 1979, when Michigan had significant decreases in youth involvement in such crashes, strengthens a causal interpretation of the association between a higher legal-drinking age and fewer motor-vehicle crashes.

I

i . I

I

I

/

·1

76

i

i:

,

!

I

I:

I

Discussion of Resul1


Table 5-11 Estimated Change in Injury-Crash Involvement in Pennsylvania after October 1977, when Maine Raised the Legal-Drinking Age

Table 5-12 Estimated Change in Propel after December 1978, whel1

Age of Driver Type of Crash

16-17

18-19

20-21

22-45

Alcohol-related indicator Single-vehicle nighttime male Estimate t-ratio Percent

Alcohol-related indicator - O.l t

0.7 -7.5

-O.ot 0.1 -0.6

-0.2t 2.1 -14.8

-O.ot 0.5 -3.4

Non-alcohol-related indicator Daytime Estimate t-ratio Percent

Single-vehicle nighttime male Estimate t-ratio Percent Non-alcohol-related indicator

20.8 0.5 3.1

31.7 0.5 3.7

-6.0 0.1 -0.7

133.7 0.5 2.8

-34.5 0.5 -3.5

25.1 0.3 1.9

-76.5

170.1 0.5 2.6

Total Estimate t-ratio Percent

Type of Crash

Daytime Estimate t-ratio Percent Total

1.2

-5.7

Source: Adapted from Wagenaar, A.C. "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Marcel Dekker, Inc., New York. Reprinted with permission. Note: tEstimate based on logarithmically transformed series.

Discussion of Crash Findings Analyses of extended time series produced estimates of the net shift in the frequency of crash involvement associated with increases in the minimum legal-drinking age in Maine and Michigan. The estimates are independent of the effects of long-term trends, seasonal cycles, and other factors, such as implementation of the national speed limit of fifty-five miles per hour and reporting-system changes. In both Maine and Michigan, involvement among young drivers in property-damage crashes related to alcohol after an increase in the drinking age was significantly lower than previous trends would have indicated. Similar reductions in involvement in propertydamage crashes related to alcohol were not found for older drivers within either experimental state or for young drivers in the two comparison states that did not change the drinking age. Furthermore, identified reductions in property-damage crashes not related to alcohol during this period were less than half the magnitude of decreases in alcohol-related crashes. The same analyses applied to injury-producing crashes related to alcohol demonstrated significant reductions among young drivers associated with the higher drinking age in Michigan, with no clear effect observed in Maine.

Estimate t-ratio Percent

Note: ·Significant at p < .01, t\1 tEstimate based on logaritl

Analyses of older drivers iJ states revealed no signifi alcohol. The most plausibl raising the legal-drinking in property-damage crashe Michigan and Maine and injury-producing crashes il Taking into account groups, and indicators of a the magnitude of the effect drivers aged eighteen to 20 percent in the frequen related to alcohol as a resu percent means that about I in the first twelve months been expected had the drivers were involved in to alcohol after the drinkir 1,500 crash-involved drive

ld Traffic Accidents Pennsylvania after Iking Age II


77

Table 5-12 Estimated Change in Property-Damage-Crash Involvement in Pennsylvania after December 1978, when Michigan Raised the Legal-Drinking Age

of Driver 20-21

22-45

Type of Crash

16-17


-11.8 0.7 -11.0

-20.7 1.0 -10.8

12.7 0.8 8.4

42.4 0.9 8.4

-0.2t 3.8· -16.6

-0.2t 3.1· -14.4

-0.2 t 3.3· -16.5

-O.lt 2.1· -9.1

-0.2t 3.3· -13.9

-O.lt 2.6 -10.7

-O.lt 2.5 -10.2

-O.lt 1.3 -5.5

22-45

Alcohol-related indicator -0.2 t 2.1 -14.8

-O.ot 0.5 -3.4

Single-vehicle nighttime male Estimate t-ratio Percent Non-alcohol-related indicator

-6.0 0.1 -0.7

133.7 0.5 2.8

-76.5 1.2 -5.7

170.1 0.5 2.6

Daytime Estimate I-ratio Percent Total

lking Age in Maine: Impact on at of the Addictions 18 (1983).

:es of the net shift in the lcreases in the minimum nates are independent of Id other factors, such as '-five miles per hour and Michigan, involvement to alcohol after an 'er than previous trends volvement in propertyfor older drivers within e two comparison states identified reductions in ing this period were less lated crashes. The same lted to alcohol demoners associated with the fect observed in Maine.

Estimate I-ratio Percent

Note: ·Significant at p < .01, two-tailed test. tEstimate based on logarithmically transformed series.

Analyses of older drivers in Michigan and young drivers in the comparison states revealed no significant reductions in serious crashes related to alcohol. The most plausible explanation for this pattern of findings is that raising the legal-drinking age caused significant reductions in involvement in property-damage crashes related to alcohol among young drivers in both Michigan and Maine and caused significant reductions in involvement in injury-producing crashes in Michigan. Taking into account results from analyses of multiple states, age groups, and indicators of alcohol involvement, the best overall estimates of the magnitude of the effects of a raised drinking age follow. First, Michigan drivers aged eighteen to twenty experienced a net reduction of approximately 20 percent in the frequency of involvement in injury-producing crashes related to alcohol as a result of the higher drinking age. A reduction of 20 percent means that about 1,100 fewer young Michigan drivers were injured in the first twelve months with the higher drinking age than would have been expected had the legal age not been raised. Second, young Michigan drivers were involved in 17 percent fewer property-damage crashes related to alcohol after the drinking-age change, representing a reduction of about 1,500 crash-involved drivers per year. Third, eighteen-to-nineteen-year-old

I

I 1

II I

i I

78


Table 5-13 Estimated Change in Injury-Crash Involvement in Pennsylvania after December 1978, when Michigan Raised the Legal-Drinking Age Age of Driver Type of Crash

16-17

18-19

20-21

22-45

Alcohol-related indicator Single-vehicle nighttime male Estimate (-ratio Percent

O.ot 0.5 3.9

O.l t 0.7 5.1

O.l t 1.1 4.6

O.l t 1.9 9.0

Non-alcohol-related indicator Daytime Estimate (-ratio Percent

-52.6 1.2 -8.5

-20.3 0.3 -2.6

-8.3 0.2 -1.2

64.1 0.2

-104.0 1.6 -11.3

-27.7 0.4 -2.2

-91.3 1.4 -7.4

108.0 0.3

1.5

Total Estimate (-ratio Percent

1.7

Note: tEstimate based on logarithmically transformed series.

Maine drivers were involved in approximately 20 percent fewer propertydamage crashes related to alcohol after the drinking age was raised; that is, 75 fewer young drivers were involved in property-damage crashes than would have been expected had the law not been changed. Despite the large estimates of public-health benefits and reduced social costs of injury and property damage, the benefits of higher drinking ages identified in this research are understated. The present investigation examined crash-involved drivers only; reductions in injuries to passengers of young drivers have not been taken into account. Maine's raised drinking age had a smaller effect on serious crashes than on property-damage crashes. A possible explanation for this finding follows. Previous research has indicated a positive relationship between crash severity and blood-alcohol concentration at the time of the crash; in other words, more-severe crashes are likely to involve drivers with high BACs (Jones and Joscelyn 1978; Warren et al. 1981). Therefore, a plausible assumption is that young drinking drivers involved in injury and fatal crashes are likely to be heavier drinkers (that is, high quantity per occasion) than drinking drivers involved in property-damage crashes. Findings reported for Maine could then be thought to indicate that the drinking-driving behavior of heavy drinkers is less affected by a change in drinking-age law than is the

Discussion of

drinking-driving behavior oj fects of the raised drinking a of severity were seen only iJ reduced the frequency of research is required to furtt relationship between the Ie. volvement. No unequivocal efifects ( and seventeen) drivers were fects of higher drinking ages. ing age in Ontario revealed 1 year of the new law. Durir alcohol-related crashes incre effects occurred after raisin! research with extended mult The time-series design us most competing hypotheses between the legal-drinking a: observed relationship related property-damage cra study is the reporting hypoth related crashes associated wi of (changes in the reportin changes in the incidence of ! reporting of the presence of ; the findings because the SV alternative indicators of inv< alcohol-related crashs; these tices in the reporting of alc< A second potential repol reported at all. Some prope nighttime crashes involving police. This situation is not i findings of this investigati, changes with modifications volved in crashes might be 11 the drinking age has made change when they could leg; ably not a plausible alternat drinking age and the incideI reason. Incentives are substa crash, because of the pena11 The additional penalty for

ld Traffic Accidents

n Pennsylvania after I-Drinking Age

e of Driver 20-21

O.lt 1.1 4.6

22-45

O.lt 1.9 9.0

-8.3 0.2 -1.2

64.1 0.2 1.5

-91.3 1.4 -7.4

108.0 0.3 1.7

o percent fewer

propertying age was raised; that is, crashes than changed. enefits and reduced social ts of higher drinking ages present investigation ext injuries to passengers of .on serious crashes than on I for this finding follows. hip between crash severity he crash; in other words, ith high BACs (Jones and a plausible assumption is l fatal crashes are likely to occasion) than drinking dings reported for Maine king-driving behavior of nking-age law than is the


79

drinking-driving behavior of moderate drinkers. However, differential effects of the raised drinking age on the incidence of crashes at various levels of severity were seen only in Maine. In Michigan the raised drinking age reduced the frequency of both serious and minor crashes. Additional research is required to further illuminate the role of crash severity in the relationship between the legal-drinking age and motor-vehicle-crash involvement. No unequivocal eflfects of the raised drinking age on underage (sixteen and seventeen) drivers were found in this evaluation of the short-term effects of higher drinking ages. Earlier research on the reduction of the drinking age in Ontario revealed little effect on underage drivers during the first year of the new law. During the second year of the new law, however, alcohol-related crashes increased (Whitehead 1977). Whether such delayed effects occurred after raising the drinking age remains an issue for further research with extended multiyear-foIlow-up data. The time-series design using three levels of comparison groups rules out most competing hypotheses to a causal interpretation of the relationship between the legal-drinking age and crash outcomes. One explanation of the observed relationship between the drinking age and involvement in alcoholrelated property-damage crashes that was not completely controlled in this study is the reporting hypothesis. That is, decreased involvement in alcoholrelated crashes associated with raising the legal-drinking age may be a result of (changes in the reporting of crash-involved drinking drivers, not of changes in the incidence of such crashes. However, modifications in police reporting of the presence of alcohol in crash-involved drivers cannot explain the findings because the SVNM and daytime-crash series were analyzed as alternative indicators of involvement in alcohol-relatep crashes and in nonalcohol-related crashs; these measures cannot be influenced by police practices in the reporting of alcohol involvement. A second potential reporting problem is the extent to which crashes are reported at all. Some property-damage crashes, particularly single-vehicle nighttime crashes involving a drinking driver, may not be reported to police. This situation is not in itself a threat to a causal interpretation of the findings of this investigation unless the extent of such underreporting changes with modifications in the drinking age. For example, drivers involved in crashes might be less willing to report a crash after an increase in the drinking age has made them underage, than -before the drinking-age change when they could legally drink. Such a reporting hypothesis is probably not a plausible alternative explanation of the relationship between the drinking age and the incidence of alcohol-related crashes for the following reason. Incentives are substantial for a drinking driver not to report a minor crash, because of the penalties for driving under the influence of alcohol. The additional penalty for underage drinking is modest in comparison to

80


the penalties for driving under the influence of alcohol. 4 An increase in the legal-drinking age is therefore unlikely to significantly affect the reporting of motor-vehicle crashes; if the raised drinking age does reduce the reporting of crashes, this effect probably accounts for only a small proportion of the 11 to 28 percent reductions in crashes associated with increases in drinking age. Observed reductions in involvement in alcohol-related crashes among young drivers support the hypothesis that there is a causal relationship between increases in the legal-drinking age and reduced involvement in motor-vehicle crashes. The conclusion that the minimum legal-drinking age affects crash involvement among youth is strengthened by comparing results of this research on the raised drinking age with the results of earlier research on the effects of a lowered drinking age. Douglass and Freedman (1977) used a time-series design to analyze a subset of Michigan jurisdictions with complete crash reporting from 1%8 through 1975. The results revealed a 17 percent (p < .06) increase in total (property-damage and injury-producing) SVNM crash involvement among drivers aged eighteen to twenty associated with the lowered drinking age in 1972. Crash involvement of drinking drivers reported by the police increased 35 percent (p < .01) after the drinking age was reduced. Similar analyses for Maine revealed a 29 percent (P < .02) increase in reported alcohol-related crashes and a 16 percent (P < .10) increase in SVNM crash involvement associated with Maine's reduction in the drinking age from twenty to eighteen in 1972 (Douglass, Filkins, and Clark 1974). Comparisons between these earlier findings and results of the present investigation reveal that raising the drinking age reverses the effect of prior reductions in the drinking age. Estimates of the increase in involvement in alcohol-related crashes among young drivers associated with Michigan and Maine's lowered drinking ages ranged from 16 to 35 percent, remarkably similar to the estimates of 11 to 34 percent obtained in the present study of decreased involvement in alcohol-related crashes associated with raising the drinking age in these two states.

Legal-Drinking Age and Beverage-Alcohol Sales A major intervening variable between changes in the legal-drinking age and traffic crashes is the drinking practices of young people. In an attempt to identify the effects of changing the drinking age on alcohol consumption, aggregate sales of alcoholic beverages were examined. Sales data for the entire decade of the 1970s were readily available, permitting simultaneous analyses of lowering and raising the drinking age in Michigan and Maine. Because information on distribution of alcoholic beverages to specific age groups was not available, comparisons across age groups could not be

Discussion of Result

made. In addition, other container-deposit laws, occ imately the same time that 1 results, therefore, are mon clusive evidence of the effec alcohol by young people. III beverage sales revealed maj It is hoped that these resul1 public policies, such as th alcoholic beverages.

Michigan Sales Findings

A major public-policy char sumption occurred in Mich age was raised. That change containers (Public Acts of 1977). A deposit of at leas! purchase of bottled or cam returned. The decrease in a penditure required for acq convenience of returning err package-beer sales. Package-beer sales migl crease (that is, controllinl resulting from the ban c Legislature 1979).5 controversial, the weight 0 between price of beer and Popham, Schmidt, and del retail prices of package bee law were expected to resul have in fact been observi mandatory-container-depm The mandatory-deposit package to draft beer, parI cant increase in the price c draft beer lowers the relath chases from package to dra Because the ban on I Michigan at the same time t policy changes were confOl

rraffic Accidents 10l. 4

An increase in the ly affect the reporting loes reduce the reportr a small proportion of increases in drinkllcohol-related crashes is a causal relation"educed involvement in

g age affects crash inparing results of this f earlier research on the reedman (1977) used a jurisdictions with comrevealed a 17 perand injury-producing) en to twenty associated volvement of drinking nt (p < .01) after the Ie revealed a 29 percent shes and a 16 percent ssociated with Maine's een in 1972 (Douglass, ese earlier findings and .ising the drinking age 19 age. Estimates of the ; among young drivers Iking ages ranged from , of 11 to 34 percent obnent in alcohol-related these two states.

: legal-drinking age and :ople. In an attempt to I alcohol consumption, 1. Sales data for the enermitting simultaneous l Michigan and Maine. :verages to specific age : groups could not be


81

made. In addition, other events, such as implementation of beveragecontainer-deposit laws, occurred in both Maine and Michigan at approximately the same time that the higher drinking ages were implemented. The results, therefore, are more difficult to interpret and do not provide conclusive evidence of the effect of raising the drinking age on consumption of alcohol by young people. In spite of these limitations, analyses of aggregate beverage sales revealed major changes in beverage markets in recent years. It is hoped that these results will stimulate further research on the role of public policies, such as the minimum drinking age, on consumption of alcoholic beverages.

i I

Michigan Sales Findings

A major public-policy change that might influence aggregate alcohol consumption occurred in Michigan during the same month that the drinking age was raised. That change was a prohibition of beer sales in nonreturnable containers (Public Acts of 1976, Initiated Measure A; Public Act 270 of 1977). A deposit of at least five cents per container is now required upon purchase of bottled or canned beer and is refunded when the container is returned. The decrease in alcohol availability because of the increased expenditure required for acquisition of package beer, coupled with the inconvenience of returning empty containers, might have resulted in decreased package-beer sales. Package-beer sales might also decline as a result of a 10 percent real increase (that is, controlling for the effects of inflation) in retail price resulting from the ban on nonreturnable containers (Michigan State Legislature 1979).5 Although the degree of price elasticity of beer remains controversial, the weight of the evidence indicates an inverse relationship between price of beer and quantity purchased or consumed (Lau 1975; Popham, Schmidt, and deLint 1976; Ornstein 1980). Therefore, increased retail prices of package beer after implementation of a mandatory-deposit law were expected to result in reduced sales. Reduced package-beer sales have in fact been observed in other states that recently implemented mandatory-container-deposit laws (Branch 1980). The mandatory-deposit law also might be expected to cause a shift from package to draft beer, particularly among on-premise drinkers. A significant increase in the price of package beer with no change in the price of draft beer lowers the relative price of draft, leading consumers to shift purchases from package to draft beer. Because the ban on nonreturnable containers was implemented in Michigan at the same time the drinking age was raised, the effects of the two policy changes were confounded and observed shifts in beverage distribu-

:1 ,I i

'l

1

I

82

Discussion of Resultl


tion in 1979 to 1980 cannot be attributed simply to the raised drinking age or to the ban on nonreturnable containers. The two policy changes are conceptually similar, however, because since both caused a reduction in the availability of alcohol, that is, both increased the effort required for acquisition of alcoholic beverages. The raised drinking age reduced the legal availability of alcohol to an important segment of alcohol consumers-youth. The mandatory-deposit law reduced availability to all consumers by causing increased inconvenience of returnable containers and increased retail prices. Results of the time-series modeling of alcoholic-beverage distribution in Michigan are summarized in table 5-14. The estimated average change in the amount of beverage (in kiloliters) distributed per month in Michigan is shown for each variable. The estimates represent the net shift in beverage distribution associated with a specific legal change, independent of the effects of long-term trends and seasonal cycles. In addition to the estimated change in monthly beverage sales, table 5-14 includes the t-ratio for each estimate and the estimated change in beverage distribution calculated as the

percent change between act distribution expected had thl Wine distribution did no ing age was lowered, or in mandatory-container-deposi also did not change significal kiloliters per month OCCllI mandatory-deposit law anc Package-beer distribution in 1972, but decreased 7,659 ki A large, one-month dw] October 1978 is evident in fi likely caused by wholesale ar an attempt to eliminate their the December deadline, whel The 1979-1980 decrease larger than the decrease in to of the decline in package-bf

Table 5-14 Estimated Change in Average Monthly Beverage-Alcohol Sales in Michigan

80,000

_ _ ACTUAL DIS _ _ MOVINGAV

Intervention

Type of Beverage

January 1972: Michigan Lowered Drinking Age

January 1979: Michigan Raised Drinking Age

70,000

en

Draft beer Estimate t-ratio Percent

1321.4 6.3· 18.9

Package beer Estimate t-ratio Percent

1049.2 0.6 1.9

-7658.7 4.3· -11.5

50,000 '

Total beer Estimate t-ratio Percent

1462.0 0.8 2.4

- 5383.5 3.0· -7.3

40,000

Wine Estimate t-ratio Percent

165.9 1.2 0.2

1600.2 11.9· 19.8

86.8 0.7 0.1

Source: Wagenaar, A.C. "Aggregate Beer and Wine Consumption: Effects of Changes in the Minimum Legal Drinking Age and a Mandatory Beverage Container Deposit Law in Michigan." Journal of Studies on Alcohol 43 (1982):477. Reprinted with permission. Note: ·Significant at p < .01, two-tailed test.

0:: W

60,000

9 S2

Dri

1969 . 1970 ' 1971 . 19

Source: Wagenaar, A.C. Minimum Legal Drinking Age Michigan." Journal of Studies on

Figure 5-14. Pac

o the raised drinking age o policy changes are conlUsed a reduction in the e effort required for acing age reduced the legal ment of alcohol coned availability to all conmabIe containers and in-

:-beverage distribution in ted average change in the r month in Michigan is the net shift in beverage independent of the eflddition to the estimated lldes the t-ratio for each calculated as the

83


d Traffic Accidents

percent change between actual distribution after a legal change and the distribution expected had there been no change in law. Wine distribution did not change significantly in 1972, after the drinking age was lowered, or in 1979, after the drinking age was raised and a mandatory-container-deposit law was implemented. Total beer distribution also did not change significantly in 1972, but a significant decrease of 5,384 kiloliters per month occurred beginning in January 1979, when the mandatory-deposit law and the raised drinking age went into effect. Package-beer distribution in Michigan also showed no significant change in 1972, but decreased 7,659 kiloliters per month beginning in January 1979. A large, one-month dwp in wholesale distribution of package beer in October 1978 is evident in figure 5-14. This temporary reduction was most likely caused by wholesale and retail dealers who reduced their purchases in an attempt to eliminate their inventory of nonreturnable containers before the December deadline, when the mandatory-deposit law went into effect. The 1979-1980 decrease in package-beer distribution is substantially larger than the decrease in total beer sales. This situation indicates that part of the decline in package-beer sales is offset by an increase in draft-beer

80,000

Sales in Michigan

,

: I

_ _ ACTUAL DISTRIBUTION _ _ MOVING AVERAGE

ervention January 1979: Michigan Raised Drinking Age

70,000

(/)

1600.2 11.9· 19.8

a:

w

g

60,000

52

-7658.7 4.3· -1l.5 -5383.5 3.0· -7.3

,I

Drinking Age Lowered Drinking Age Raised

t I

,tion: Effects of Changes in the e Container Deposit Law in rinted with permission.

f

I

86.8 0.7 0.1

1979

1970

Source: Wagenaar, A.C. "Aggregate Beer and Wine Consumption: Effects of Changes in the M!niI?um Legal Drinking Age and a Mandatory Beverage Container Deposit Law in Michigan." Journal of Studies on Alcohol 43 (1982):475. Reprinted with permission.

Figure 5-14. Package-Beer Distribution in Michigan

"

;

i

: i

,1

,

II I

I il

I'

i !'

! ',.

.

84

sales, which was confirmed by separate analyses of the draft-beer series. Results for 1979 and 1980 indicate dramatic increases in draft-beer sales in Michigan. Distribution in 1979 was 1,600 kiloliters per month higher than expected, given baseline patterns; the first nine months of 1980 exhibited an additional increase of 1,194 kiloliters per month (figure 5-15). In spite of the large increases in draft-beer distribution, total beer distribution is down significantly for 1979-1980 because package beer, the distribution of which decreased in 1979-1980, accounts for over three-fourths of all beer distributed in Michigan. Results also revealed a significant temporary increase in draft-beer sales beginning in January 1972, when the drinking age was lowered. The effect dissipated by the end of 1972, however, and no significant permanent shift in draft-beer sales associated with the lowered drinking age was observed (note the pattern of the moving-average-trend line in figure 5-15). The significant increase in draft-beer distribution in 1972, immediately after the drinking age was reduced from twenty-one to eighteen, can be interpreted as a result of the sudden expansion in the population of legal

I'

---l


15,000 _ _ ACTUAL DISTRIBUTION _ _ MOVING AVERAGE

13,000

11,000 UJ f-

:::;

9

52

Source: Wagenaar, A.C. "Aggregate Beer and Wine Consumption: Effects of Changes in the Minimum Legal Drinking Age and a Mandatory Beverage Container Deposit Law in Michigan." Journal of Studies on Alcohol 43 (1982):476. Reprinted with permission.

Figure 5-15. Draft-Beer Distribution in Michigan

Discussion of Result

drinkers. Why the positive clear. One might speculate t sudden experimentation wit wearing off after the first y practiced before the drinkir plausible, however, given t health problems related to from reductions in drinking A plausible hypothesis lowered drinking age, cause, youth aged eighteen to twen drinking population, a sub group could be easily mask( other drinkers. Therefore, t sales in 1972, concomitant \ offset by a small reduction caused by other, unidentifiel A third possible explan. distribution in 1972 is that tl ing age. Given the temporar) no significant changes in pac be concluded that lowering 1 the aggregate quantity of all total quantity of alcoholic mechanism by which the 10' volvement, perhaps a chang change in quantity consumec that followed the lowered drl leads to increased drinking j Since personal automobiles a to and from such drinking I the frequency of driving aft( old group, even with no cha The substantial decreas when the drinking age was ra also has multiple explanatio count for a portion of the d( t1'at other factors, which al consumers, were operating. deposit law. As noted earIiel cent increase in the real pric convenience of returnable cc addition to reduced consumI

1 I

d Traffic Accidents

of the draft-beer series. ases in draft-beer sales in per month higher than nths of 1980 exhibited an gure 5-15). In spite of the leer distribution is down the distribution of which lree-fourths of all beer

llcrease in draft-beer sales was lowered. The effect gnificant permanent shift age was observed e in figure 5-15). tion in 1972, immediately ,ne to eighteen, can be in1 the population of legal

inking Age Raised

Effects of Changes in the age Container Deposit Law in eprinted with permission.

n in Michigan


.

85

drinkers. Why the positive shift in draft-beer sales rapidly decayed is not clear. One might speculate that the effect ofthe reduced drinking age was a sudden experimentation with consumption of draft beer, with the novelty wearing off after the first year, followed by a return to drinking patterns practiced before the drinking age was lowered. This argument seems less plausible, however, given the permanent significant increases in publichealth problems related to alcohol, such as traffic collisions, that resulted from reductions in drinking age. A plausible hypothesis is that other exogenous factors, unrelated to the lowered drinking age, caused draft-beer sales to decrease in 1973. Because youth aged eighteen to twenty constitute only a fraction of the total beerdrinking population, a substantial increase in consumption by this age group could be easily masked by a small reduction in consumption by all other drinkers. Therefore, the significant increase in aggregate draft-beer sales in 1972, concomitant with the lowered drinking age, may have been offset by a small reduction in consumption among all drinkers in 1973, caused by other, unidentified factors. A third possible explanation of the temporary increase in draft-beer distribution in 1972 is that the increase is not a result of the lowered drinking age. Given the temporary nature of the observed draft-beer increase and no significant changes in package-beer or wine distribution in 1972, it might be concluded that lowering the drinking age in Michigan had no effect on the aggregate quantity of alcoholic beverages consumed. If an increase in total quantity of alcoholic beverages consumed is not the intervening mechanism by which the lower drinking age causes increases in crash involvement, perhaps a change in drinking patterns among youth, with no change in quantity consumed, accounts for the increased crash involvement that followed the lowered drinking age. For example, lowering the legal age leads to increased drinking in bars and taverns by the age group affected. Since personal automobiles are likely to be the usual mode of transportation to and from such drinking places, lowering the drinking age may increase the frequency of driving after drinking among the eighteen-to-twenty-yearold group, even with no change in total quantity of alcohol consumed. The substantial decrease in package-beer distribution in 1979-1980, when the drinking age was raised and nonreturnable containers were banned, also has multiple explanations. Although the raised drinking age may account for a portion of the decrease, the magnitude of the decrease suggests tbat other factors, which affected the entire population of package-beer consumers, were operating. One such factor is the mandatory-containerdeposit law. As noted earlier, one result of the container law was a 10 percent increase in the real price of package beer. The price jump and the inconvenience of returnable containers may explain the decrease in sales. In addition to reduced consumption of package beer as a result of its increased

I I

I

I I

.!

I

II I

II

86

r

I i

i

I


cost (both direct monetary costs and in terms of inconvenience), some consumers who reside near Michigan's borders may have shifted their packagebeer purchases to retailers in bordering states. Since most of Michigan's population does not reside near the state's borders, however, they do not have easy access to adjoining states with nonreturnable containers and lower package-beer prices. It is therefore unlikely that cross-border purchases account for much of the decrease in package-beer sales in Michigan. Cross-border purchase of beer does account for some of the observed decrease in package-beer sales in Michigan, however; in 1981 the Michigan Commerce Department announced a crackdown on persons crossing the state's borders to purchase package beer, arguing that a smuggling epidemic was harming Michigan business (Detroit Free Press 1981). Given substantial increases in draft-beer distribution during the same period, it appears that a number of beer drinkers shifted some of their consumption from package to draft beer. The shift from package to draft beer is particularly likely to have occurred among on-premise consumers. Draft beer has traditionally been the least expensive on-premise alcoholic beverage;6 the increase in package-beer prices following the mandatorycontainer law further reduced the relative price of draft beer and caused some consumers to shift their on-premise purchases from package beer to draft beer. Michigan's economic recession, with associated unemployment and reductions in disposable income, may also have contributed to the shift to less-expensive draft beer. 7 Furthermore, bar and tavern managers may have encouraged sales of draft beer over package beer to reduce both the cost and inconvenience associated with storing and returning large numbers of empty bottles and cans. Finally, there may have been a shift from package to draft beer among those who purchase large quantities of beer for private parties. As with bar and tavern owners, the inconvenience of returning containers and the increasing cost of package beer may have changed their buying patterns. In addition, Michigan's raised drinking age may have had a larger effect in reducing package-beer consumption among youth than in reducing draftbeer consumption because the law did not prohibit young people aged eighteen to twenty from patronizing public drinking establishments; only purchase and consumption of alcoholic beverages were prohibited. In a crowded bar or tavern a legal drinker might easily purchase alcoholic beverages for friends under the age of twenty-one. As a result, draft-beer sales may be higher than expected if those under the age of twenty-one were not allowed to frequent public drinking establishments. Finally, part of the increase in draft-beer sales in 1979-to-1980 may be a result of an increased number of keg parties among eighteen-to-twenty-year-olds, where a legal drinker purchases a keg of draft beer for consumption off-premise at a party attended by underage drinkers.

Discussion of Resl Maine Sales Findings

The impact of lowering beverage sales was also changes in alcoholic-bevc lowered (table 5-15). Afte decreased an average I, I than-expected sales after] Separate analyses of dn because those data were n Wine sales did not ch 5-15 shows an increase in \ raised in late 1977; the p < .01, the significance marginal significance of t (figure 5-17) indicated tJ dramatically different thaI Results for spirits sal associated with either lowe On the basis of these drinking age in Maine resu

Table 5-15 Estimated Change in Aver:

Type of Beverage Total beer Estimate I-ratio Percent Wine Estimate I-ratio Percent Spirits Estimate I-ratio Percent

Source: Wagenaar, A.C. "Public F Hampshire: 1970-1980." Conlempo Note: ·Significant at p < .01, two.

'attic Accidents

lvenience), some conshifted their packagemost of Michigan's 10wever, they do not tlable containers and lat cross-border pureer sales in Michigan. Jme of the observed in 1981 the Michigan persons crossing the a smuggling epidemic

1981).

ltion during the same ted some of their conpackage to draft beer nise consumers. Draft on-premise alcoholic Iwing the mandatoryiraft beer and caused from package beer to )ciated unemployment :ontributed to the shift tavern managers may eer to reduce both the :turning large numbers een a shift from packquantities of beer for convenience of returneer may have changed

Maine Sales Findings

The impact of lowering and raising the legal-drinking age on aggregate beverage sales was also examined in the state of Maine. No significant changes in alcoholic-beverage sales occurred when the drinking age was lowered (table 5-15). After the drinking age was raised, however, beer sales decreased an average 1,114 kiloliters per month (13 percent). The lowerthan-expected sales after November 1977 are clearly evident in figure 5-16. Separate analyses of draft- and package-beer sales were not possible because those data were not available for the state of Maine. Wine sales did not change when the drinking age was lowered. Table 5-15 shows an increase in wine sales of 10 percent after the drinking age was raised in late 1977; the increase is significant at p < .05, but not at p < .01, the significance level selected for this study. Consistent with the marginal significance of the estimate, examination of the wine-sales plot (figure 5-17) indicated that wine sales after October 1977 were not dramatically different than expected given the long-term upward trend. Results for spirits sales in Maine revealed no measurable changes associated with either lowering or raising the legal-drinking age. On the basis of these findings it might be concluded that the raised drinking age in Maine resulted in significantly lower alcohol consumption Table 5-15 Estimated Change in Average Monthly Beverage-Alcohol Sales in Maine Intervention June 1972:· Maine Lowered Drinking Age

November 1977: Maine Raised Drinking Age

Total beer Estimate I-ratio Percent

179.4 1.1 2.6

-1,114.2 6.8· -12.9

Wine Estimate t-ratio Percent

-14.2 0.7 -4.1

46.9 2.5 10.2

Spirits Estimate t-ratio Percent

-23.3 1.2 -3.7

25.7 1.3 3.6

Type of Beverage

lave had a larger effect than in reducing draft)ung people aged eighotblishments; only pur·ohibited. In a crowded tlcoholic beverages for raft-beer sales may be I-one were not allowed part of the increase in tl increased number of :re a legal drinker purise at a party attended

87


Source: Wagenaar, A.C. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems 11 (1982). Reprinted with permission. Note: • Significant at p < .01, two-tailed test.

!

88

ii

I,

i

Discussion of RE


12,000

900

I

,"

_


,.1

ACTUA

_ _ MOVIN

1

1:1

,I

10,000

700

1,1

(f)

a:: w

w

--.' /"

8,000

g

9

;Z

500

I 'I 6,000

300

Drinking Age Lowered

4,000

Drinking J

Drinking Age Raised

100

\

1971' 1972 . 197: Source: Wagenaar, A.C. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems 11 (1982). Reprinted with permission.

Figure 5-16. Beer Distribution in Maine

Source: Wagenaar, A.C. "PI Hampshire: 1970-1980."

among youth because sales of beer, the beverage of choice among young drinkers, decreased significantly after the raised drinking age was implemented. The interrupted-time-series design employed here permits causal interpretation of the results with a moderately high level of confidence. As with the Michigan results, however, a serious threat to a causal interpretation of the findings is contemporaneous history, that is, other events occurring at about the same time as the raised drinking age that may account for the observed decrease in beer sales. The major confounding factor was a mandatory-beverage-container-deposit law, implemented in Maine in January 1978, just two months after the higher drinking age went into effect (Maine Revised Statutes Annotated, Chapter 27, section 1751, through Chapter 28, section 1871). The inconvenience of returnable containers along with substantial increases in the price of canned and bottled beer may have caused the reduction in beer sales. Reduced sales, in turn, may have resulted from reduced consumption or a shift from purchasing beer in Maine to purchases in neighboring states, particularly New Hampshire, where beer prices are substantially lower. Beer prices have traditionally been lower in New Hampshire than in Maine because of lower tax rates. Maine's

taxes on beer include a general sales tax of 5 pel teen cents per gallon, aT United States 1980). Ac ferential has widened n deposit law and Maine r Hampshire beer. 8 In an attempt to sep beverage-container-dep( Hampshire were analyz Maine, and it has been s some of their beer purct implemented and beer '}: count for the decrease iT crease in the drinking a Hampshire beer sales 1 Time-series modeling oj

Figure

"po,

89


Traffic Accidents

900 _ _ ACTUAL DISTRIBUTION _ _ MOVING AVERAGE 700

w

9

500

52

300


Isumption in Maine and New ). Reprinted with permission.

Maine

)f choice among young drinking age was imIyed here permits causal level of confidence. As t to a causal interpretat is, other events occurthat may account for tfounding factor was a ented in Maine in JanuIg age went into effect section 1751, through . returnable containers and bottled beer may ales, in turn, may have )m purchasing beer in .larly New Hampshire, have traditionally been ower tax rates. Maine's

100

Drinking Age Raised

1971

1972

1973

1974

1975

1976

1977

1976

1979

1960

Source: Wagenaar, A.C. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems 11 (1982). Reprinted with permission.

Figure 5-17. Wine Distribution in Maine taxes on beer include an excise tax of twenty-five cents per gallon and a general sales tax of 5 percent; in contrast, New Hampshire's excise tax is fifteen cents per gallon, and it has no sales tax (Distilled Spirits Council of the United States 1980). According to Maine state officials, the beer-price differential has widened noticeably since implementation of the mandatorydeposit law and Maine residents increasingly purchase the lower-priced New Hampshire beer. 8 In an attempt to separate the effects of the drinking-age change and the beverage-container-deposit law in Maine, beverage sales in the state of New Hampshire were analyzed. New Hampshire is the only state that borders Maine, and it has been suggested that Maine beer drinkers may have shifted some of their beer purchases to New Hampshire after the container law was implemented and beer prices increased. 9 If such cross-border purchases account for the decrease in Maine beer sales that was concomitant with the increase in the drinking age and implementation of the container law, New Hampshire beer sales should have increased by a comparable amount. Time-series modeling of New Hampshire beer sales, however, revealed no

90

i I

! i

I

II II II I


Discussion of Rl

significant change when the Maine container law was implemented (table 5-16). Thus observed reductions in Maine beer sales are most likely the result of the increase in the drinking age and not of cross-border purchases after implementation of the mandatory-container-deposit law. New Hampshire lowered the minimum legal-drinking age from twentyone to eighteen for all alcoholic beverages in June 1973. Parameters for this legal change were included in the time-series models for beer sales in New Hampshire (table 5-16). Results showed a significant 8 percent increase in package-beer sales after the drinking age was lowered (figure 5-18). Draftbeer sales were up an estimated 11 percent (figure 5-19). Based on these results, it appears that lowering the drinking age in New Hampshire contributed to an increase in sales of package and draft beer .10 To ensure that shifts in beer sales associated with policy changes in Michigan, Maine, and New Hampshire were not simply reflecting nationwide trends, beer sales in the entire United States from 1970 through 1980 were analyzed (table 5-17).11 Results showed no significant changes in nationwide draft-beer sales after January 1972, when Michigan lowered the drinking age and experienced a temporary increase in draft-beer sales. There were also no changes in nationwide sales of draft or package beer after October 1977, when Maine raised the drinking age and experienced a 13 percent reduction in beer sales. Finally, nationwide sales of package beer did not change significantly after December 1978, when Michigan raised the drinking age and experienced a significant 12 percent reduction in packagebeer sales. In each of these cases, nationwide beer sales showed no

11,500

_ _ ACTU _ _ MOVI

9,500

CJ)

a: w

I-

:J

7,500

9;;::

Source: Wagenaar, A.C. "PI Hampshire: 1970-1980." Con

Figure 5-18. Pa

./

i

i

measurable change at a age increased or decrea. Draft-beer sales in t in January 1979, when experienced a significar tern of U.S. draft-beer Michigan; that is, draf pected sales, given previ 1980 over the already h wide sales of draft beel (table 5-17), while sales (table 5-14). It is clear Michigan in recent yean creases in nationwide the successful introduct substantially higher inc] wide increases are appal

Table 5-16 Estimated Change in Average Monthly Beverage-Alcohol Sales in New Hampshire Intervention

Type of Beverage

Draft beer Estimate t-ratio Percent Package beer Estimate t-ratio Percent

June 1973: New Hampshire Lowered Drinking Age

January 1978: Maine MandatoryDeposit Law

67.8 2.5 10.7

11.3 0.8 1.5

539.7 2.8* 7.7

258.6 1.5 3.1

Source: Wagenaar, A.C. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems 11 (1982). Reprinted with permission. Note: *Significant at p < .01, two-tailed test.

,

1

as implemented (table es are most likely the :ross-border purchases posit law. king age from twenty73. Parameters for this , for beer sales in New It 8 percent increase in :d (figure 5-18). Draft5-19). Based on these New Hampshire conbeer. 10 ¥ith policy changes in nply reflecting nationom 1970 through 1980 lificant changes in naMichigan lowered the se in draft-beer sales. draft or package beer age and experienced a le sales of package beer ten Michigan raised the t reduction in packagebeer sales showed no

leohoI Sales

'ntion January 1978: Maine MandatoryDeposit Law 11.3 0.8 1.5

258.6 1.5 3.1 Isumption in Maine and New ,). Reprinted with permission.

91


rraffic Accidents

11,500 _ _ ACTUAL DISTRIBUTION _ _ MOVING AVERAGE

9,500

,.

en

/

a:

w I:::;

7,500

o...J >2


1979

Source: Wagenaar, A.C. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems II (1982). Reprinted with permission.

Figure 5-18. Package-Beer Distribution in New Hampshire

measurable change at a time when sales in a state that modified the drinking age increased or decreased significantly. Draft-beer sales in the United States did increase significantly beginning in January 1979, when Michigan implemented the higher drinking age and experienced a significant increase in draft-beer sales (figure 5-20). The pattern of U.S. draft-beer sales in recent years was similar to that found in Michigan; that is, draft-beer sales increased significantly in 1979 over expected sales, given previous trends, with an additional significant increase in 1980 over the already higher-than-expected 1979 figures. However, nationwide sales of draft beer in 1979 were only 6 percent higher than expected (table 5-17), while sales of draft beer in Michigan increased by 19 percent (table 5-14). It is clear that part of the increase in sales of draft beer in Michigan in recent years is a reflection of a nationwide trend. Significant increases in nationwide sales of draft beer may be a result of such factors as the successful introduction and marketing of low-calorie (light) beers. The substantially higher increases in Michigan draft-beer sales over the nationwide increases are apparently the result of factors unique to Michigan, such

92

Discussion of RE


Table 5-17 Estimated Change in ) in the United States

1,100 _ ACTUAL DISTRIBUTION _ _ MOVING AVERAGE

900 Mi

Type of Beverage Draft beer Estimate t-ratio Percent

(/)

a:

::i

700

9 52

Package beer Estimate t-ratio Percent

500

Note: *Significant at p < .0 Drinking Age Lowered

300

late 1970s were apparel of less-expensive New I plemented a mandator tionwide sales of beer, that observed increases decreases when the age' This result strengthens the drinking age and The results for be' however. The lowered ( associated with signific Maine's age was not. A in the late 1970s were n( by Maine residents, may have caused a economic conditions h Michigan may also be i are not yet fully unden Finally, youth are drinkers. They would 11 tion to account for the number of people aged to be 41,507 (U.S. De (1981) survey results, as drinking at least once a group reduced their be drinking age was raised

Drinking Age Raised

1976

1977

1978' 1979

Source: Wagenaar, A.C. "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems 11 (1982). Reprinted with permission.

Figure 5-19. Draft-Beer Distribution in New Hampshire

as the increase in the drinking age, the mandatory-container-deposit law, and the unusually severe economic recession. 12

Discussion of Sales Findings Controlling for the effects of long-term trends and regular seasonal patterns, the present study found significant changes in aggregate sales of alcoholic beverages concomitant with modifications in the minimum legaldrinking age. In all states examined, beer, the beverage of choice among young drinkers, was the beverage category most affected by changes in the drinking age. In Michigan and New Hampshire, reductions in the drinking age from twenty-one to eighteen were associated with significant increases in beer sales. In Maine and Michigan, increases in the drinking age from eighteen to twenty and from eighteen to twenty-one, respectively, were associated with significant decreases in total beer sales. Analyses of beer sales in New Hampshire revealed that decreased beer sales in Maine in the

I

I

1

Traffic Accidents

93

Discussion of Results Table 5-17 Estimated Change in Average Monthly Beverage-Alcohol Sales in the United States

Type of Beverage Draft beer Estimate t-ratio Percent Package beer Estimate t-ratio Percent

January 1972: Michigan Lowered Drinking Age

Intervention November 1977: Maine Raised Drinking Age

1,045.8 0.5 0.6

-2,765.2 1.4 -1.5

30,431.0 2.0 2.8

17,099.6 1.0 1.2

January 1979: Michigan Raised Drinking Age 11,515.2 5.8· 5.9 -13,407.8 0.9 -0.9

Note: ·Significant at p < .01, two-tailed test.

Drinking Age Raised

176

1977

1978

1979

Isumption in Maine and New ). Reprinted with permission.

Hampshire

'-container-deposit law,

d regular seasonal pats in aggregate sales of s in the minimum legal'erage of choice among :ected by changes in the :luctions in the drinking ith significant increases the drinking age from one, respectively, were sales. Analyses of beer er sales in Maine in the

late 1970s were apparently not caused by additional cross-border purchase of less-expensive New Hampshire beer by Maine residents after Maine implemented a mandatory-container-deposit law. Results of analyses of nationwide sales of beer, compared with the state-specific analyses, revealed that observed increases in beer sales after the drinking age was lowered and decreases when the age was raised were not a reflection of nationwide trends. This result strengthens the argument that observed relationships between the drinking age and beer sales reflect a causal effect of the drinking age. The results for beverage sales must still be interpreted with caution, however. The lowered drinking ages in Michigan and New Hampshire were associated with significant increases in beer sales, whereas the reduction in Maine's age was not. Although it appears that reduced beer sales in Maine in the late 1970s were not a result of increased cross-border purchase of beer by Maine residents, the container-deposit law and resulting price increases may have caused a significant reduction in beer consumption. Poor economic conditions in the late 1970s and early 1980s in states such as Michigan may also be influencing the beverage-alcohol market in ways that are not yet fully understood. Finally, youth are only a small proportion of the total population of drinkers. They would therefore have to dramatically reduce their consumption to account for the observed decreases in beer sales. For example, the number of people aged eighteen and nineteen in Maine in 1978 is estimated to be 41,507 (U.S. Department of Commerce 1973).13 Using Johnston's (1981) survey results, assume that 72 percent are regular drinkers, defined as drinking at least once a month. If the estimated 29,885 drinkers in this age group reduced their beer drinking by 1,114 kiloliters per month after the drinking age was raised (from table 5-15), the quantity of beer consumed by

94


275,000

250,000


225,000 Cf)

a: w

I-

::::;

g

200,000

175,000

150,000

125,000

1970 . 1971 . 1972 . 1973 . 1974

1975

1976' 1977 ' 1978' 1979 - 1980

Figure 5-20. Draft-Beer Distribution in the United States each 18- or 19-year-old drinker in Maine decreased an average of 37 liters per month, or, about 105 twelve-ounce servings of beer per person per month. Although these estimates are admittedly crude and part of the reduction in beer distribution may have resulted from reduced consumption among drinkers under the age of eighteen, these calculations illustrate the implausibility of attributing the entire decrease in sales of beer in Maine beginning in late 1977 and early 1978 to the increase in the minimum legaldrinking age. These complications illustrate the limitations of analyses of aggregate sales data. Without age-specific consumption data, the differential effects of drinking-age changes and other policy changes, such as the mandatorydeposit law, cannot be unambiguously determined. Furthermore, detailed information on the drinking practices of various subpopulations, identified by stratification variables such as income level and employment status, is needed to assess the influence of economic conditions and beverage-specific price changes on individual drinking patterns. In light of the above considerations, the most reasonable conclusion is that reduced beer sales in Maine and Michigan in the late 1970s were caused by a combination of

Discussion of Re

reductions in beer const reduced by raising the. tion among all drinkers tation of beverage-cont Although plausibl, presented, they remain studies are completed, noted in designing fur gregate beverage sales j experimental design th. age groups (as has been traffic accidents) would tributable to the drink Such age-specific consu pIe surveys, and inform (for example, frequene social context of drink volving on-premise vers ing practices of various as income level and em consumed. Although cc and after a change in foresight and resources which the minimum Ie alcohol-related crashes. A second approach sources rather than on s discussion of possible ex effects of economic fac ment, and personal inco beverages, monthly dau Incorporation of these' explains aggregate bevel on alcohol sales of gem changes.

Enforcement of the Mil

Another implicit intervi crash involvement is the no detailed analyses of d of enforcement citation

--------".

Traffic Accidents

1977 I 1978 I 1979


1980

United States

an average of 37 liters f beer per person per rude and part of the l reduced consumption culations illustrate the :ales of beer in Maine in the minimum legalanalyses of aggregate the differential effects Jch as the mandatoryFurthermore, detailed )opulations, identified employment status, is : and beverage-specific ght of the above conreduced beer sales in by a combination of

95

reductions in beer consumption among young drinkers after availability was reduced by raising the legal-drinking age and reductions in beer consumption among all drinkers as a result of higher prices that followed implementation of beverage-container-deposit laws. Although plausible explanations of the sales findings have been presented, they remain at the level of informed speculation until further studies are completed. The weaknesses of the present design should be noted in designing further research. The present investigation used aggregate beverage sales for all age groups as dependent variables. A quasiexperimental design that compares trends in alcohol consumption across age groups (as has been done in studies of the effect of the drinking age on traffic accidents) would significantly increase the ability to isolate effects attributable to the drinking age from numerous alternative explanations. Such age-specific consumption data would most likely be collected by sample surveys, and information could also be collected on (1) drinking patterns (for example, frequency and degree of intoxication), (2) the immediate social context of drinking (including proportion of drinking episodes involving on-premise versus off-premise consumption of alcohol), (3) drinking practices of various subgroups identified by stratification variables such as income level and employment status, and (4) total quantity of alcohol consumed. Although conducting such a survey at repeated intervals before and after a change in the legal-drinking age would require considerable foresight and resources, the potential for illuminating the mechanism by which the minimum legal-drinking age influences youth involvement in alcohol-related crashes is substantial. A second approach for further analyses would rely on existing data sources rather than on a costly original-data-collection effort. Much of the discussion of possible explanations for the results reported here included the effects of economic factors, such as beverage prices, inflation, unemployment, and personal income. With the exception of retail prices for alcoholic beverages, monthly data on these economic indicators are readily available. Incorporation of these variables into a comprehensive structural model that explains aggregate beverage sales would help clarify the differential effects on alcohol sales of general economic conditions and specific public-policy changes.

i I

!

Enforcement of the Minimum Legal-Drinking Age Another implicit intervening variable between the legal-drinking age and crash involvement is the level of enforcement of the drinking age. Although no detailed analyses of drinking-age enforcement were conducted, frequency of enforcement citations for violation of laws concerning selling or pro-

iI


96

Discussion of RE

viding alcoholic beverages to underage persons in Michigan and Maine were briefly examined. The objective was simply to provide a preliminary impression of enforcement of the drinking age, focused on the providers of alcoholic beverages. Citations for underage individuals who purchase or consume alcoholic beverages were not included. The substantial effect of Michigan's higher drinking age in reducing alcohol-related crashes among youth in the first six months after the law was implemented was apparently not a result of strict enforcement of the law. In the early months after Michigan's increase in drinking age, enforcement was minimal, and the new law was flagrantly disobeyed according to an advisory panel established by Michigan's governor (Grand Rapids Press 1979). Efforts to increase enforcement in late 1979 and 1980 were reflected in the frequency of citations issued by the Michigan Liquor Control Commission for selling, serving, or allowing minors to consume alcoholic beverages. Although the number of citations was 68 percent higher in 1979 than in 1978, citation frequency in 1980 was 156 percent higher than in 1978 (figure 5-21). Higher citation frequencies in 1979 and 1980 were in contrast to the relatively constant citation frequency of three hundred to four hundred per year in the mid-1970s. In spite of the increase in citations in 1980, underage youth were still able to acquire alcoholic beverages from licensed

1000

outlets. For example, I and September 1980 f( fluence of liquor (by team), 24 percent repOI 1981). The frequency of ( Maine for selling or shown in figure 5-22. , quency the first two ye was similar to that in 1\ tially smaller. Citation: sons increased by 17 pe to 1979. Based on the minimal enforcement tl efforts to increase enf creases in citation freq1

300

250

800 200

600

400

150

Drinking Age Lowered

200

Drinking Age Lo Drinking Age Raised

oI 1970

I

i i i

1971

1972

1973

i

I

I

1974

1975

1976

I

i i '

1977

1978

1979

i

1980

Figure 5-21. Number of Citations for Selling or Allowing Minors to Consume Alcoholic Beverages, Michigan

1971

1972

1973

Figure 5-22. Number· sume Ale

I'

l'

rraffic Accidents

97


;higan and Maine were a preliminary impresI on the providers of uals who purchase or

nking age in reducing . months after the law ict enforcement of the drinking age, enforcelisobeyed according to ,r (Grand Rapids Press nd 1980 were reflected Liquor Control Comto consume alcoholic percent higher in 1979 ent higher than in 1978 11980 were in contrast : hundred to four hunIse in citations in 1980, everages from licensed

outlets. For example, of forty-one underage youth arrested between April and September 1980 for driving while intoxicated or driving under the influence of liquor (by an Oakland County, Michigan, special enforcement team), 24 percent reported that they had been at a bar prior to arrest (Wolfe 1981). The frequency of citations brought before the administrative court in Maine for selling or allowing minors to consume alcoholic beverages is shown in figure 5-22. Although the pattern of the increase in citation frequency the first two years after implementation of the higher drinking age was similar to that in Michigan, the magnitude of the increase was substantially smaller. Citations for providing alcoholic beverages to underage persons increased by 17 percent from 1977 to 1978 and by 68 percent from 1977 to 1979. Based on these data, it appears that Maine also may have had minimal enforcement the first year after the new law was implemented, with efforts to increase enforcement during the second year. The smaller increases in citation frequency in Maine might also indicate less-rigorous en-

300

250

200


I Age Raised

77

1978

1979

1980

owing Minors to Con-

1971

1972

1973

1974

1975

1976

1977

1978

1979

Figure 5-22. Number of Citations for Selling or Allowing Minors to Consume Alcoholic Beverages, Maine

98


Discussion of RE

forcement of the law in Maine. The lower enforcement efforts may account for the lack of a clear effect of the higher drinking age on injury crashes in Maine, in contrast to the significant effect of the increased drinking age on injury crashes in Mighican. This brief examination of the enforcement of the minimum legaldrinking age was based only on enforcement focused on the providers of alcoholic beverages, primarily off-premise retail outlets, bars and taverns, and restaurants. Enforcement of the drinking age focused on the individual underage drinker, and enforcement of laws against drinking and driving also would be expected to influence the frequency of involvement in . alcohol-related crashes, the main dependent variable of the present study. Finally, the frequency of citations is only a gross indicator of the level of enforcement because information was not available on (1) the number of citations in relation to total number of citable violations, or (2) the perceived risk of experiencing sanctions for providing alcohol to minors. The role of enforcement of the drinking-age laws in reducing highway crashes and other alcohol-related problems is not well understood and deserves the attention of a separate study. It is worth noting, however, that results from this study, indicating substantial effects of the raised drinking age in spite of relatively lax enforcement of the new laws, are consistent with Burkett and Carrither's (1980) findings from a survey of high-school students. As expected, both peer disapproval of drinking and perceived certainty of legal sanctions for drinking were related to drinking behavior. In addition, the young people's moral commitment to laws also significantly affected drinking practices, independent of the effects of social pressure and legal deterrents. Thus reductions in drinking and driving that follow increases in the drinking age probably result from the combined effects of fear of punishment for illegal drinking and the desire to abide by the law because it is the right thing to do. The present research found pronounced reductions in involvement in traffic crashes following increases in the minimum drinking age and some evidence that alcohol consumption is lower when the legal age is higher. From a public-health perspective, these results clearly indicate that higher legal-drinking ages should be endorsed. Implications of the findings from this study and other research on the drinking age, and other factors to consider when establishing a minimum legal age for drinking, are discussed next.

drivers (for example, 3' range of estimates are reported drinking versu result, similar percent variables produce diffe 2. The assistance Transportation Researc tion of the DUIL-arres 3. As in Michiga drivers aged eighteer presumably because of safety. 4. For example, tl in Maine is as follows: imprisonment up to ni mandatory attendanc{ alcoholism treatment a and $2,000, mandatory mandatory six-month alcohol-education pro! mandatory imprisonm{ two-year license SUSp{ rehabilitation (Maine penalty for a minor in I fense, fine up to $100; : offense, fine of $100 C 5. The political mandatory-container dustries were justified prices did increase. 6. A telephone su and restaurants reveale than any other on-pren state, the retail price of of on-premise (Douglass et al. 1980). 7. Package-beer p 1979 (Michigan State I creased almost 13 perc( percent (Verway 1980). come in Michigan decr 8. Beer prices incr' law was implemented,

Notes

1. The range of estimated-percent crash reductions (for example, 11 to 28 percent) and the range of estimated reductions in terms of the number of

1

raffic Accidents It efforts may account :e on injury crashes in eased drinking age on

. the minimum legal:d on the providers of lets, bars and taverns, used on the individual drinking and driving of involvement in : of the present study. :ator of the level of en(1) the number of citas, or (2) the perceived to minors. The role of Iway crashes and other deserves the attention

this study, indicating e of relatively lax enIrkett and Carrither's ItS. As expected, both of legal sanctions for m, the young people's drinking practices, ineterrents. Thus reducthe drinking age probmnishment for illegal s the right thing to do. :ms in involvement in Tinking age and some le legal age is higher. .y indicate that higher ; of the findings from I other factors to con'inking, are discussed

ns (for example, 11 to of the number of


99

drivers (for example, 373 to 1726) may not always appear comparable. The range of estimates are based on multiple indicators (for example, policereported drinking versus SVNM) that have different baseline averages. As a result, similar percent changes in the level of crash involvement across variables produce different numerical estimates of crash reductions. 2. The assistance of Lyle D. Filkins of The University of Michigan Transportation Research Institute with collection, analysis, and interpretation of the DUlL-arrest data is gratefully acknowledged. 3. As in Michigan, involvement in daytime crashes among Maine drivers aged eighteen to nineteen decreased significantly in 1979, presumably because of the economic recession and efforts to increase traffic safety. 4. For example, the penalty for driving under the influence of alcohol in Maine is as follows: first offense, fine between $250 and $1,000, possible imprisonment up to ninety days, mandatory thirty-day license suspension, mandatory attendance at an alcohol-education program, and possible alcoholism treatment and rehabilitation; second offense, fine between $250 and $2,000, mandatory imprisonment from twenty-four hours to six months, mandatory six-month license suspension, mandatory attendance at an alcohol-education program; third offense, fine between $250 and $2,500, mandatory imprisonment from forty-eight hours to ten months, mandatory two-year license suspension, and mandatory alcoholism treatment and rehabilitation (Maine Revised Statutes Annotated, Title 29, 1981). The penalty for a minor in possession of alcoholic beverages in Maine is: first offense, fine up to $100; second offense, fine between $50 and $100; and third offense, fine of $100 (Maine Revised Statutes Annotated, Title 28, 1981). 5. The political debate surrounding the implementation of the mandatory-container law included discussion as to whether the beverage industries were justified in raising prices; it is simply noted here that retail prices did increase. 6. A telephone survey of a random sample of Michigan bars, taverns, and restaurants revealed that draft beer has a significantly lower retail price than any other on-premise alcoholic drink. Depending on the region of the state, the retail price of draft beer was 17 to 40 percent lower than the price of on-premise package beer, usually the second-least-expensive drink (Douglass et al. 1980). 7. Package-beer prices increased about 23 percent between 1978 and 1979 (Michigan State Legislature 1979), and the Consumer Price Index increased almost 13 percent, but per-capita personal income increased only 9 percent (Verway 1980). Controlling for the effects of inflation, personal income in Michigan decreased when package-beer prices increased. 8. Beer prices increased "considerably" after the mandatory-container law was implemented, and consumers complained "bitterly" according to a

I

I

I

100


personal conversation with S. Redfield, Maine Department of Agriculture, on June 1, 1981. 9. Although there might be beer purchases across Maine's northern border with Canada, the great majority of Maine's population resides in the southern region of the state. 10. The effects of New Hampshire's increase in the drinking age from eighteen to twenty in May 1979 could not be evaluated because data on beer sales were available only through 1979, providing only seven postchange observations. II. The time-series models included parameters for January 1972 (Michigan lowered drinking age), November 1977 (Maine raised drinking age), and January 1979 (Michigan raised drinking age). Parameters for the drinking-age reductions in Maine and New Hampshire were not included because they occurred within a relatively short time. Inclusion of parameters for all of the drinking-age reductions would have resulted in multicollinearity problems when estimating the parameters of the full timeseries model. (See Pindyck and Rubinfeld 1976, pp. 66-68 for an introduction to the multicollinearity issue.) 12. Subtracting the 6 percent reduction in U.S. total draft-beer sales from the 19 percent reduction observed in Michigan leaves a 13 percent reduction in Michigan that cannot be directly attributed to factors that caused the nationwide increase in draft-beer sales. 13. The population aged eighteen to nineteen in 1978 was estimated from the population aged ten to eleven identified by the 1970 U.S. Census.

E

This study clearly den ages in preventing alcc Approximately 20 pel drivers can be preven1 These findings are cor increases in drinking a have been repeatedly j and decreases in crash the drinking age is rai1 or three years (that is, include grandfather cl Although the effe< crashes among young warranted before a bl drinking age can cou among youth. The effl across states. In this r obvious than the effec producing crashes. As in Massachusetts four crease in the drinking Williams et al. 1981). One possible reas Michigan is that fou minimum drinking ag{ chusetts implemented beverage alcohol to l\r other states that raised legal supply of alcoho . and others (1983) provi Massachusetts youth I drinking age was raisl differential minimum I chase of alcohol by y< found an overrepresel residents in alcohol-n

Traffic Accidents rtment of Agriculture,

ross Maine's northern )pulation resides in the

the drinking age from d because data on beer )nly seven postchange

for January 1972 Maine raised drinking :e). Parameters for the lire were not included : time. Inclusion of 'ould have resulted in neters of the full time66-68 for an introduc-

TS

total draft-beer sales n leaves a 13 percent buted to factors that

n 1978 was estimated the 1970 U.S. Census.

6

Selecting the Best Minimum Drinking Age

This study clearly demonstrates the efficacy of higher minimum drinking ages in preventing alcohol-related automobile crashes among young drivers. Approximately 20 percent of all alcohol-related crashes involving young drivers can be prevented by removing legal access to alcoholic beverages. These findings are corroborated by multiple studies of both decreases and increases in drinking age; increases in crash involvement of 10 to 30 percent have been repeatedly found immediately following reductions in legal age, and decreases in crash involvement of 10 to 30 percent typically occur when the drinking age is raised. Such effects follow drinking-age changes of two or three years (that is, from eighteen to twenty or twenty-one) that do not include grandfather clauses. Although the effects of the raised drinking age in reducing automobile crashes among young people are now clearly documented, some caution is warranted before a blanket statement is made that any state raising the drinking age can count on a 20 percent decrease in crash involvement among youth. The effect of higher drinking ages is not necessarily uniform across states. In this research, the effect in Michigan was larger and more obvious than the effect in Maine, particularly for the more-serious, injuryproducing crashes. As noted in chapter 3, studies of fatal-crash involvement in Massachusetts found relatively small reductions associated with an increase in the drinking age from eighteen to twenty (Hingson et al. 1983; Williams et al. 1981). One possible reason for a smaller effect in Massachusetts than in Michigan is that four of the five states bordering Massachusetts had minimum drinking ages of eighteen for all alcoholic beverages after Massachusetts implemented a higher drinking age. l Thus the availability of beverage alcohol to Massachusetts youth was not reduced as much as in other states that raised the drinking age because Massachusetts youth had a legal supply of alcohol in contiguous states. The survey results of Hingson and others (1983) provided some support for this line of reasoning; underage Massachusetts youth reported little difficulty obtaining alcohol after the drinking age was raised. Additional evidence that contiguous states with differential minimum drinking ages create problems with cross-border purchase of alcohol by youth was provided by Lillis and others (1981). They found an overrepresentation of eighteen-to-twenty-year-old Pennsylvania residents in alcohol-related crashes that occurred in New York counties 101

!

, I

I

I

Ii !

102


contiguous with Pennsylvania. 2 It should be noted, however, that the magnitude of the cross-border problem was quite small, a finding also reported by Cook and Tauchen (1982). In light of such cross-border problems, the development of a nationwide consensus for a drinking age, with uniform enforcement of the law across states, might be considered. Potential crossborder purchase of alcohol must be considered when evaluating the effects of state-specific drinking-age laws. Another potential explanation of the small observed effect of the higher drinking age in Massachusetts is related to the data analyzed. In several studies of drinking-age changes reviewed earlier, the dependent variable, fatal-crash involvement, had low frequencies. The number of alcoholrelated-crash fatalities among a limited age group within one state is relatively small for analytic purposes. Thus the large random variation in the number of fatalities from month to month, or even year to year, makes it difficult to identify significant effects of a policy change such as the drinking age. Even in Michigan, where substantial reductions in involvement in both injury-producing and property-damage crashes resulted from the raised drinking age, no significant effect of the raised drinking age was discernible when fatal crashes alone were analyzed (Wagenaar 1980). Others have also recognized the importance of not limiting policy evaluations to analyses of fatal crashes (Whitehead 1980). The problems with low crash counts for analysis might also emerge for nonfatal crashes in less populous states, such as Maine, making it more difficult to detect the effects of policy changes. Therefore, whereas evidence to date clearly demonstrates the effects of raising the drinking age in numerous states, reductions in crashes and injuries may not always be clearly evident in less-populous jurisdictions. One implication of these findings for future evaluations of the drinkingage or other public-policy changes is that analyses should not be limited only to fatalities, but should also include the much larger numbers of injury and property-damage crashes. Although the effort and costs associated with analyzing nonfatal crashes is substantially higher, such analyses may avoid incorrect conclusions that a policy change had no effect on the outcome of interest. 3 In spite of the fact that the link between the legal-drinking age and involvement in traffic crashes is now well established, a number of questions remain for further research. First, follow-up studies should be conducted to assess whether the effects observed to date increase or decrease over time. The research reported here was limited to an examination of the first two years after higher drinking ages were implemented. One hypothesis might be that the long-term effect will be larger than the short-term effect identified here because the eighteen-to-twenty-year-old group the first few years after a legal change included individuals who had had the right to drink

The Best MinimL

prior to implementatio ing legally prior to the drinking habits than la effect of raising the dri crashes for the transiti effect. A second area for vironment of youth. A mediate the relationst comes. One main inter cluding both quantity in which predriving dri tices of youth should f after drinking. Third, analyses oj should be conducted.. age on the aggregate involved youth in Mich youth in Maine. Analy termination of an optiI effect of a higher drink alcohol-related crashes tie effect on twenty-yel vide most of the benef: restricting the freedom istic of crash-involved As discussed in chapter problem of male drivel ential drinking (and pe: and females are decn Alcoholism 1980). Uttl crash-involved driver. Fourth, other majc for public-health effec alcohol to young drink found to have a clear other changes in alcoh, effects. Government regulatory changes, or alcohol availability. FOi beverage prices, change. ordinance modification: sequences for alcohol-n

'affic Accidents

)wever, that the magfinding also reported border problems, the llg age, with uniform ered. Potential crossevaluating the effects

effect of the higher analyzed. In several : dependent variable, number of alcoholwithin one state is random variation in n year to year, makes r change such as the eductions in involvecrashes resulted from ised drinking age was 1genaar 1980). Others policy evaluations to blems with low crash lshes in less populous :t the effects of policy demonstrates the efreductions in crashes ss-populous jurisdic-

,tions ofthe drinkingId not be limited only umbers of injury and ;osts associated with h analyses may avoid :ct on the outcome of

-drinking age and innumber of questions ,ould be conducted to r decrease over time. ltion of the first two )ne hypothesis might lort-term effect idenIUp the first few years ld the right to drink

r iI

I

The Best Minimum Drinking Age

103

prior to implementation of the higher drinking age. Those who were drinking legally prior to the new laws were probably less likely to give up their drinking habits than later groups who never had the right to drink. Thus the effect of raising the drinking age on reducing involvement in alcohol-related crashes for the transitional-age group might be smaller than its long-term effect. A second area for futher research involves the predriving drinking environment of youth. As discussed in chapter 3, several intervening factors mediate the relationship between the legal-drinking age and crash outcomes. One main intervening factor is drinking behavior among youth, including both quantity and frequency of consumption and social situations in which predriving drinking occurs. Further research on the drinking practices of youth should focus on the behavioral patterns that precede driving after drinking. Third, analyses of specific subpopulations of crash-involved drivers should be conducted. This investigation assessed the effect of the drinking age on the aggregate of all reported eighteen-to-twenty-year-old crashinvolved youth in Michigan and eighteen-to-nineteen-year-old crash-involved youth in Maine. Analyses of single-year age categories would aid in the determination of an optimal legal-drinking age. For example, if the beneficial effect of a higher drinking age is largely the result of reduced involvement in alcohol-related crashes among eighteen-to-nineteen-year-old youth, with little effect on twenty-year-olds, a minimum drinking age of twenty may provide most of the benefits of the higher legal age at a lower cost in terms of restricting the freedom of young people. A second demographic characteristic of crash-involved drivers that should be taken into account is gender. As discussed in chapter 2, the problem of alcohol-related crashes is largely a problem of male drivers. There is some evidence, however, that the differential drinking (and perhaps driving-after-drinking) patterns between males and females are decreasing (National Institute on Alcohol Abuse and Alcoholism 1980). Little research to date has focused on the young, female, crash-involved driver. Fourth, other major changes in alcohol availability should be evaluated for public-health effects. Since substantial changes in the availability of alcohol to young drinkers, as reflected in the legal-drinking age, have been found to have a clear impact on a major alcohol-related health problem, other changes in alcohol availability should be examined for public-health effects. Government actions, either through administrative policy, regulatory changes, or legislation, frequently have direct implications for alcohol availability. For example, deregulation or other changes in alcoholicbeverage prices, changes in alcohol-tax formulas, and zoning and other localordinance modifications should be adequately evaluated regarding their consequences for alcohol-related injury and death. Some of these policies, which

104


The Best Miniml

may appear to have no direct connection with alcohol policy, such as mandatory-container-deposit laws, were found in this research to be associated with major changes in aggregate alcoholic-beverage sales. The effects of such new legislation and regulations should be regularly measured, and the results should be used to guide the formulation of public policies designed to prevent alcohol abuse and other alcohol-related problems. In spite of the ever-present need for further research, findings to date clearly demonstrate the efficacy in preventing alcohol-related problems of government regulation of the supply of alcoholic beverages to young people. These findings might be used as the basis for an unqualified recommendation for all jurisdictions to establish a legal-drinking age of twentyone. Based solely on public-health considerations, such a recommendation appears warranted. 4 Several other factors must also be considered when establishing the optimal legal age for drinking. One consideration not related to public health is whether restricting access to alcohol for young people that are otherwise considered adult citizens is constitutionally permitted. Is it not unconstitutional to discriminate against a certain class of adult citizens on the basis of age? This question has been treated in detail in a case that challenged the constitutionality of Michigan's increase in the drinking age from eighteen to twenty-one (Guy 1978). The court ruled that legal-drinking ages are not unconstitutional; this ruling was based on three major legal arguments. First, discriminatory classifications require intensive court review if the law restricts a "fundamental" right. The court held that consumption of alcoholic beverages is not by any means a fundamental right guaranteed by the Constitution. As a result, legal restrictions on access to alcoholic beverages do not warrant intensive review by the courts. Second, discriminatory classifications require intensive review if they are based on inherently "suspect" criteria. Legal precedent has not held age to be such a suspect criteria for discrimination, again obviating the need for intensive review of high drinking-age laws. s Given that the raised drinking age does not affect a fundamental right nor involve a suspect criteria for discrimination, the state is not required to demonstrate a "compelling need" for the higher drinking age. The state must only show that the age discrimination has a "rational basis," or that it is "reasonably related" to a legitimate state objective. The legitimate state objective was reduction of alcohol-related crashes among youth. The court ruled, on the basis of scientific evidence that linked lower drinking ages with increased crash involvement among youth, that the drinking-age discrimination was reasonably related to the state objective of reducing highway crashes. Thus the higher drinking age withstood the constitutional challenge on all three issues: (1) drinking alcohol is not a fundamental right, (2) age is not inherently a suspect criteria for discrimination, and (3) using the drinking age to prevent highway crashes has a rational basis in available scientific

evidence. Several othe tutional were also men manent disability, but havior for the young Pi to regulate the distribl first Amendment. Th( regulations, has a "strc High drinking ages they appear to be cons ing age closed, then? N democracy such as the is support among thl referenda in Michigan 1980 supported the twe conducted in 1975 rev( that the drinking age : should be lowered. Thl undecided on the issue port for higher legal-d Some questions rer and popular support fl age for all rights and J ception of legal acces: noted that legal ages fo of adulthood are not varies from twelve to t1 mobile, sale and use 0 consent for contracepti sent, marriage with arl, contracts, service on jl for public-welfare sUPI In addition, minimum resentatives, governors legal ages are not uni behavior involved. Sett with this general practi Although there are with the use of beverag of alcohol must also 1 made to bar young pel alcohol serves a variety Perhaps the most freqt which imparts convivi

J.

raffic Accidents

:ohol policy, such as this research to be :>everage sales. The efIe regularly measured, jon of public policies -related problems. arch, findings to date ll-related problems of rages to young people. 1 unqualified recom'inking age of twentych a recommendation l be considered when

Nhether restricting acnsidered adult citizens ional to discriminate 1ge? This question has e constitutionality of n to twenty-one (Guy unconstitutional; this First, discriminatory lW restricts a "fundaalcoholic beverages is the Constitution. As a ges do not warrant inclassifications require spect" criteria. Legal 'ia for discrimination, drinking-age laws. 5 :t a fundamental right :tate is not required to inking age. The state ional basis," or that it The legitimate state long youth. The court Ner drinking ages with drinking-age discrimof reducing highway mstitutional challenge nental right, (2) age is ld (3) using the drink; in available scientific

The Best Minimum Drinking Age

105

evidence. Several other reasons that a higher drinking age is not unconstitutional were also mentioned. The higher drinking age does not cause a permanent disability, but is only a temporary postponement of a specific behavior for the young person's own protection, and states have broad powers to regulate the distribution and use of beverage

Figure B-1. Negative I Function I cant correlations at tl parameter estimates sh tions and known Inadequacies in tb tected when evaluating

Interve

C

Q)Q)

"0_

'c.c Q)lU

0.';:

Q)lU

c>

Figure B-2. Positive In Function

'raffic Accidents

I

129

Appendix B

simple forms of the

In model were used in ry-impact model is :1 component, 0 is the d as a pulse function. 'ith II defined as a step : models are labeled as : auto-regressive comImponent, and b is the of the intervention b was included in the s were expected in the Icing age. 4 The impact B-1 through B-6. All $ were modeled with except the increase in Iction in the drinking , model (figure B-6). ntified on the basis of ·autocorrelation funclOnents were added, lentified model were :stimated autocorrelaIlg values for the comuse the models are inluardt method to obs. The unconditionalthm was used rather se seasonal movingns (1976) recommend Is. ins modeling strategy lated model must be ty to account for all are several considera:d parameters should for the pars 1976). Second, the ld be significantly dif,arameters should not ified. Fourth, overall residuals should be i Pierce 1970; Ljung uld hot reveal signifi-

Intervention Start

C

CD CD '0_ CoO

CDC

a:c CDc c>

Time

Figure B-1. Negative Impact Pattern Estimated by the rsb (1,0,0) TransferFunction Model with a Step-Function Input

cant correlations at the first few lags or the first seasonal lag. Sixth, parameter estimates should be interpretable in terms of theoretical expectations and known characteristics of the dependent variable. Inadequacies in the combined ARIMA-transfer-function model, detected when evaluating the estimation results, were used to guide respecifi-

Intervention Start

C

CD CD

'0_

'c.c CDCU a:c CDcu

c>

, Time

Figure B-2. Positive Impact Pattern Estimated by the rsb (1,0,0) TransferFunction Model with a Step-Function Input

!

130

rI

L i:

rI

,

Inter

Intervention Start

I'

r

Appendix B


'E

e

Q)Q) '0_

Q)Q) '0_

a:.:

Q)CU

a:.:

c>

c>

eoC

eoC

Q)CU Q)CU

Q)CU

Time

I

I

Figure B-3. Negative Impact Pattern Estimated by the rsb (0,0,0) TransferFunction Model with a Step-Function Input

Figure B-S. Negative II Function

cation of the model. After respecification, maximum-likelihood estimates were obtained, the revised model was evaluated according to the above criteria. If the model was still inadequate, the specification, estimation, and evaluation steps were repeated again; if more than one model was adequate from by these criteria, the model with the lowest sum of squared

the fitted model was Sell estimates, along with UJ to determine the existe tervention effects were c impact in terms of the n by the intervention. All specific models model, which views a . discrete linear stochastic a realization of a discn tions are made: (1) the stant means, that is, the in overall level, (2) all dependently from the sa by constant variance, a depending only on the assumption of normall: strict stationarity. The original series is not 1 propriate provided that differencing factors on after differencing, the Sl behavior" (Box and JeI An important strenJ assessment of the extent

Intervention Start

e

Q)Q) '0_

eoC

Q)CU

a:.:

•

I

Q)CU

c>

Time

Figure B-4. Positive Impact Pattern Estimated by the rsb (0,0,0) TransferFunction Model with a Step-Function Input

Appendix B

"raffic Accidents

131

Intervention Start

C

CIlCll

'0_

COO CIlIll

0.';: CIlIll

c>

Time

Ie rsb (0,0,0) Transfertlput

Figure B-S. Negative Impact Pattern Estimated by the rsb (1,0,0) Transfer-

m-likelihood estimates according to the above ;ation, estimation, and Ie model was adequate luared deviatio.Qs from

the fitted model was selected. The values ofthe transfer-function-parameter estimates, along with unbiased estimates of their standard errors, were used to determine the existence of effects of the interventions and, where intervention effects were evident, to assess the direction and magnitude of the impact in terms of the number of crashes apparently caused by or prevented by the intervention. All specific models fit to the series were variations of the underlying model, which views a particular time series as a realization of a general discrete linear stochastic process (Nelson 1973). In modeling a time series as a realization of a discrete linear stochastic process, the following assumptions are made: (1) the time series is stationary and characterized by a constant means, that is, the series does not exhibit a substantial trend or change in overall level, (2) all random errors (u t in equation A.l) are drawn independently from the same distribution over time and thus are characterized by constant variance, and (3) the autocovariances are constant over time, depending only on the extent of lag between the observations. Adding the assumption of normally distributed errors achieves what is referred to as strict stationarity. The assumption of a stationary mean or level in the original series is not strictly required, because the model remains appropriate provided that stationarity is obtained after using the appropriate differencing factors on the original series. If a stationary level is obtained after differencing, the series is said to exhibit "homogeneous nonstationary behavior" (Box and Jenkins 1976). An important strength of the Box-Jenkins modeling strategy is that an assessment of the extent to which the assumptions that underlie the analyses

,e rsb (0,0,0) TransferIput

Function Model with a Pulse-Function Input

132


Appendix B Notes

Intervention Start

c:

!

II, ;

!

i

II ,I

II, II

Aaron, P. and D. l' Historical Over' Alcohol and Pu 127-181. Washir Alcohol Drug Abuse vention Policy I Government Pri' Alcohol Epidemiolo: ence of Changes the Deaths of the Electric Co:, American Businessrr tions Becoming (1978):11-16. ___ . "Drinking Line on Alcohol Bacon, S.D. "Alcoh Snyder, eds., So bondale, Ill.: So ___ . "The Role and Abuse." In 29th Internatior. Sidney, Australi Bako, G., W.C. Mal Lowering of the Young Drivers i Health 67 (1976: Barsby, S.L. and G Lower Minimur Alcohol 38 (197' Beauchamp, D.E. B Philadelphia: Te Berk, R.A. and M. ment of Statistil aI. , eds. , Evalu Beverly Hills, C Berry, R.E., Jr. and New York: Free

Bibliography

Aaron, P. and D. Musto. "Temperance and Prohibition in America: A Historical Overview." In M.H. Moore and D.R. Gerstein, eds., Alcohol and Public Policy: Beyond the Shadow of Prohibition, pp. 127-181. Washington, D.C.: National Academy Press, 1981. Alcohol Drug Abuse and Mental Health Administration. ADAMHA Prevention Policy and Programs, 1979-1982. Washington, D.C.: U.S. Government Printing Office, 1981. Alcohol Epidemiologic Data System. "Data Computations on the Influence of Changes in State Drinking Age Laws on Highway Fatalities in the Deaths of the 70s." Wor king paper 37. Gaithersburg, Md.: General Electric 1980. American Businessmen's Research Foundation. "Drinking Age Implications Becoming Clearer." The Bottom Line on Alcohol in Society 2 (1978): 11-16. ___ . "Drinking Age Implications Becoming Clearer." The Bottom Line on Alcohol in Society 3 (1979):16-20. Bacon, S.D. "Alcohol and Complex Society." In D.J. Pittman and C.R. Snyder, eds., Society, Culture, and Drinking Patterns, pp. 78-93. Carbondale, Ill.: Southern Illinois University Press, 1962. ___ . "The Role of Law in Meeting Problems of Alcohol and Drug Use and Abuse." In A.G. Kiloh and D.S. Bell, eds., Proceedings of the 29th International Congress on Alcoholism and Drug Dependence. Sidney, Australia: Butterworths, 1971. Bako, G., W.C. MacKenzie, and E.S.O. Smith. "The Effect of Legislated Lowering of the Drinking"Age on Total Highway Accidents Among Young Drivers in Alberta, 1970-1972." Canadian Journal of Public Health 67 (1976):161-163. Barsby, S.L. and G.L. Marshall. "Short-term Consumption Effects of a Lower Minimum Alcohol Purchasing Age." Journal of Studies on Alcohol 38 (1977):1665-1679. Beauchamp, D.E. Beyond Alcoholism: Alcohol and Public Health Policy. Philadelphia: Temple University Press, 1980. Berk, R.A. and M. Brewer. "Feet of Clay in Hobnail Boots: An Assessment of Statistical Inference in Applied Research." In T.D. Cook et aI., eds., Evaluation Studies Review Annual, vol. 3, pp. 190-214. Beverly Hills, Calif.: Sage Publications, 1978. Berry, R.E., Jr. and J.P. Boland. The Economic Cost of Alcohol Abuse. New York: Free Press, 1977.

135

.ij

.Ij

·1

136

i

'I


Blane, H.T. "Middle-aged Alcoholics and Young Drinkers." In H.T. Blane and M.E. Chafetz, eds., Youth, Alcohol, and Social Policy, pp. 5-38. New York: Plenum Press, 1979. Blane, H.T. and L.E. Hewitt. Alcohol and Youth: An Analysis of the Literature, 1960-1975. Report prepared for the National Institute on Alcohol Abuse and Alcoholism. Pittsburgh: The University of Pennsylvania, 1977. Bonnie, R.J. "Discouraging Unhealthy Personal Choices Through Government Regulation: Some Thoughts About the Minimum Drinking Age." In H. Wechsler, ed., Minimum-Drinking-Age Laws, pp. 39-58. Lexington, Mass.: D.C. Heath and Co., 1980. Borg, T.M., H.L. Michael, A.A. Gadallah, and G.K. Stafford. "Evaluation of the 55 MPH Speed Limit." In D.G. Shurig, ed., Proceedings of the 62nd Annual Road School, pp. 160-184. School held at Purdue University, March 9-11, 1976. Borkenstein, R.F., R.F. Crowther, R.P. Schumate, et al. The Role of the Drinking Driver in Traffic Accidents. Bloomington, Ind.: Indiana University, Department of Police Administration, 1964. Bowen, B.D. and M.R. Kagay. Report to the White House Conference on Youth: The Impact ofLowering the Age of Majority to 18. Washington, D.C.: White House Conference on Youth, 1973. Box, G.E.P. and G.M. Jenkins. Time Series Analysis: Forecasting and Control, Revised Edition. San Francisco: Holden-Day, 1976. Box, G.E.P. and D.A. Pierce. "Distribution of Residual Autocorrelations in Autoregressive-integrated Moving Average Time Series Models." Journal of the American Statistical Association 64 (1970):1509-1526. Box, G.E.P. and G.c. Tiao. "Intervention Analysis with Applications to Economic and Environmental Problems." Journal of the American Statistical Association 70 (1975):70-79. Branch, B. "The Economic Outlook for Beer Wholesalers." Brewers Digest 55 (1980):32-33, 42. Brissett, D. "Toward an Interactionist Understanding of Heavy Drinking." Pacific Sociological Review 21 (1978):3-20. Brown, D.B. and S. Maghsoodloo. "A Study of Alcohol Involvement in Young Driver Accidents with the Lowering of the Legal Age of Drinking in Alabama." Accident Analysis and Prevention 13 (1981):319-322. Bruun, K., G. Edwards, M. Lumio and others. Alcohol Control Policies in Public Health Perspective. Helsinki: The Finnish Foundation for Alcohol Studies, 1975. Burkett, S.R. and W.T. Carrithers. "Adolescents' Drinking and Perceptions of Legal and Informal Sanctions." Journal of Studies on Alcohol 41 (1980):839-853. Burritt, B.E., A. Moghrabi, and J .S. Matthias. "An Analysis of the Relationships Between Accidents and the 55 MPH Speed Limit on the

Bibliography

Arizona Highway search Board 55th 1975. Cahalan, D. and I.H. in the U.S." In B. holism, Volume York: Plenum Pre Cameron, T. "Alcohc Casualties and Cr School of Public I Campbell, D.T. "Ref (1969) :409-429. Campbell, D.T. and. Designs for Resea, Carsten, O. "Use of Calculate Exposur Cavan, S. Liquor LiCE Chu, B. and G.E. Nul' Fatalities During t 8 (1976):145-150. Clark, N.H. Deliver {. hibition. New Yor Cook, T.D. and D: Analysis Issues fo Cook, P.J. "Alcohol' of Addiction 77 (1 Cook, P.J. and G. Tal lation on Youthft script, 1982. Cucchiaro, S., J. Fen year-old Drinking chusetts Institute ( Dart, O.K. "Effects c forcement on Traf Record 643 (1977) Detroit Free Press. " December 6, 1981 Distilled Spirits COUll! Law Experience 0 December 1973. _ _ . 1978 Public D.C., 1980. Douglass, R.L., L.D. Drinking Ages on versity of Michiga

Traffic Accidents

Irinkers." In H.T. Blane Social Policy, pp. 5-38.

'th: An Analysis of the le National Institute on 'he University of Penn-

loices Through Governlnimum Drinking Age." Laws, pp. 39-58. Lex-

[.K. Stafford. "Evalua[rig, ed., Proceedings of School held at Purdue

, et aI. The Role of the :ton, Ind.: Indiana Uni1964. e House Conference on 'Jrity to 18. Washington,

3.

alysis: Forecasting and len-Day, 1976. sidual Autocorrelations Time Series Models."

164 (1970):1509-1526.

js with Applications to urnal of the American

)lesalers." Brewers Diding of Heavy Drink-

O.

, Alcohol Involvement ng of the Legal Age 'is and Prevention 13

fcohol Control Policies 'innish Foundation for Drinking and PercepI of Studies on Alcohol

n Analysis of the Rela-I Speed Limit on the

Bi bl iography

137

Arizona Highway System." Paper presented at the Transportation Research Board 55th Annual Meeting, Washington, D.C., January 19-23, 1975. Cahalan, D. and I.H. Cisin. "Drinking Behavior and Drinking Problems in the U.S." In B. Kissin and H. Begleiter, eds., The Biology of Alcoholism, Volume 4: Social Aspects of Alcoholism, pp. 77-115. New York: Plenum Press, 1976. Cameron, T. "Alcohol and Traffic." In M. Aarens et aI., eds., Alcohol, Casualties and Crime, pp. 120-288. Berkeley: Social Research Group, School of Public Health, University of California, 1977. Campbell, D.T. "Reforms as Experiments." American Psychologist 24 (1969):409-429. Campbell, D.T. and J.C. Stanley. Experimental and Quasi-experimental Designs for Research. Chicago: Rand McNally, 1966. Carsten, O. "Use of the Nationwide Personal Transportation Study to Calculate Exposure." HSRI Research Review 11 (1981):1-8. Cavan, S. Liquor License. Chicago: Aldine, 1966. Chu, B. and G.E. Nunn. "An Analysis of the Decline in California Traffic Fatalities During the Energy Crisis." Accident Analysis and Prevention 8 (1976):145-150. Clark, N.H. Deliver Us From Evil: An Interpretation of American Prohihibition. New York: Norton, 1976. Cook, T.D. and D.T. Campbell. Quasi-experimentation: Design and Analysis Issues for Field Settings. Chicago: Rand McNally, 1979. Cook, P.J. "Alcohol Taxes as a Public Health Measure." British Journal of Addiction 77 (1982):245-250. Cook, P.J. and G. Tauchen. "The Effect of Minimum Drinking Age Legislation on Youthful Auto Fatalities, 1970-1977." Unpublished manuscript, 1982. Cucchiaro, S., J. Ferreira, Jr., and A. Sicherman. The Effect of the 18year-old Drinking Age on Auto Accidents. Cambridge, Mass.: Massachusetts Institute of Technology, Operations Research Center, 1974. Dart, O.K. "Effects of the 88.5 km/h (55 mph) Speed Limit and Its Enforcement on Traffic Speeds and Accidents." Transportation Research Record 643 (1977):23-32. Detroit Free Press. "Beer War on the Michigan-Indiana Line." P. 3A. December 6, 1981. Distilled Spirits Council of the United States. "Survey of Minimum Age Law Experience on Drinking/Driving." DISCUS Newsletter, no. 330, December 1973. ___ . 1978 Public Revenues from Alcoholic Beverages. Washington, D.C., 1980. Douglass, R.L., L.D. Filkins, and F.A. Clark. The Effect of Lower Legal Drinking Ages on Youth Crash Involvement. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1974.

138

: I

[I

,

1:,1

IL


Douglass, R.L. and J.A. Freedman. Alcohol-related Casualties and Alcohol Beverage Market Response to Beverage Alcohol Availability Policies in Michigan. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1977. Douglass, R.L., P.M. Barkey, A.C. Wagenaar, and P. DeBiasi. Retail Price Distribution of Alcoholic Beverages for On-premise Consumption in Michigan. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1980. Dunn, O.J. and V.A. Clark. Applied Statistics: Analysis of Variance and Regression. New York: John Wiley and Sons, 1974. Farris, R., T.R Malone, and H. Lilliefore. A Comparison of Alcohol Involvement in Exposed and Injured Drivers. Alexandria, Va.: Essex Corporation, 1976. Filkins, L.D., C.D. Clark, C.A. Rosenblatt, W.L. Carlson, M.W. Kerlan, and H. Manson. Alcohol Abuse and Traffic Safety: A Study of Fatalities, D. W.I. Offenders, Alcoholics, and Court-related Treatment Approaches. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1970. Filkins, L.D. and J.D. Flora. Alcohol-related Accidents and DUlL Arrests in Michigan: 1978-1979. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1981. Flora, J.D., L.D. Filkins, and C.P. Compton. Alcohol Involvement in Michigan Fatal Accidents: 1968-1976. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1978. Frankel, RG. and P.C. Whitehead. Drinking and Damage: Theoretical Advances and Implications for Prevention. New Brunswick, N.J.: Rutgers Center of Alcohol Studies, 1981. Gottman, J .M. Time-series Analysis: A Comprehensive Introduction for Social Scientists. New York: Cambridge University Press, 1981. Grand Rapids Press. "Tougher Stand on Underage Drinking Is Urged by Study Panel." P. AlO, May 2, 1979. Guy, Hon. Ralph, Jr. Opinion. Ref. Civil No. 8-73015 and Civil No. 8-73159. U.S. District Court, Eastern District of Michigan, Southern Division, Detroit, Mich., December 22, 1978. Hammond, R.L. "Legal Drinking at 18 or 21-Does it Make Any Difference?" Journal of Alcohol and Drug Education 18 (1973):9-13. Harford, T.C., D.A. Parker, and L. Light, eds. Normative Approaches to the Prevention ofAlcohol Abuse and Alcoholism. (Proceedings of a symposium, April 26-28, 1977, San Diego.) Washington, D.C.: U.S. Government Printing Office, 1980. Haugh, L.D. and G.E.P. Box. "Identification of Dynamic Regression (Distributed Lag) Models Connecting Two Time Series." Journal of the American Statistical Association 72 (1977): 121-130.

Bibliography

Heath, D.B. "Drinkin and C.R. Snyder, 22-36. Carbondah Hibbs, D.A., Jr. "On Jenkins and Box-' Heise, ed., SocioJ cisco: Jossey-Bass Hingson, R., N. Scotcl the Legal Drinkin Journal of Public Hurst, P.M. "Epidem Citations." Journ Illinois Department of ment of the Effec Age in Illinois. N< Institute for Research Traffic Accidents, University, 1975. Jessor, R. and S.L. J Drinking." Journl Johnston, L.D., J.G. : dent Drug Use in ernment Printing I Jones, R.K. and K.B. the State of Kno Highway Safety R Kahane, C.J. Lower S ington, D.C.: Nati of Statistics and A Keller, M. "Alcohol a Adolescence and. Publishing Co., Kendall, M. Time-seri, Klein, D. "Adolescent The Young Drivel North Carolina Sy Klein, T. The Effect oj Accidents in the, Highway Traffic S Koch, G.G. "Commer of a Social Policy Accident Analysis Lau, H-H. "Cost of A

Traffic Accidents

'ted Casualties and AIre Alcohol Availability ;ity of Michigan, Highand P. DeBiasi. Retail On-premise Consumpof Michigan, Highway

'1alysis of Variance and 1974.

1parison of Alcohol In!\lexandria, Va.: Essex Carlson, M.W. Kerlan, {ety: A Study ofFatali-related Treatment Aphigan, Highway Safety

rents and DUlL Arrests niversity of Michigan, Ilcohol Involvement in bor: The University of 1978. t Damage: Theoretical \lew Brunswick, N.J.: msive Introduction for :sity Press, 1981. : Drinking Is Urged by l-73015 and Civil No. of Michigan, Southern

es it Make Any Differ18 (1973):9-13. Vormative Approaches lism. (Proceedings of a ashington, D.C.: U.S. 1

f Dynamic Regression le Series." Journal of 121-130.

Bibliography

139

Heath, D.B. "Drinking Patterns of the Bolivian Camba." In D.J. Pittman and C.R. Snyder, eds., Society, Culture, and Drinking Patterns, pp. 22-36. Carbondale, Ill.: Southern Illinois University Press, 1962. Hibbs, D.A., Jr. "On Analyzing the Effects of Policy Interventions: BoxJenkins and Box-Tiao Versus Structural Equation Models." In D.R. Heise, ed., Sociological Methodology 1977, pp. 137-179. San Francisco: Jossey-Bass, 1977. Hingson, R., N. Scotch, T. Mangione, et al. "Impact of Legislation Raising the Legal Drinking Age in Massachusetts from 18 to 20." American Journal of Public Health 73 (1983):163-170. Hurst, P.M. "Epidemiological Aspects of Alcohol in Driver Crashes and Citations." Journal of Safety Research 5 (1973):130-148. Illinois Department of Transportation, Division of Traffic Safety. Assessment of the Effect on Traffic Safety of Lowering the Legal Drinking Age in Illinois. November 1977. Institute for Research in Public Safety. Tri-Ievel Study of the Causes of Traffic Accidents, Volume I, Research Findings. Bloomington: Indiana University, 1975. Jessor, R. and S.L. Jessor. "Adolescent Development and the Onset of Drinking." Journal of Studies on Alcohol 36 (1975):27-51. Johnston, L.D., J.G. Bachman, and P.M. O'Malley. Highlightsfrom Student Drug Use in America: 1975-1981. Washington, D.C.: U.S. Government Printing Office, 1981. Jones, R.K. and K.B. Joscelyn. Alcohol and Highway Safety: A Review of the State of Knowledge. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1978. Kahane, C.J. Lower Speed Limits, Reduced Speeds, Fewer Deaths. Washington, D.C.: National Highway Traffic Safety Administration, Office of Statistics and Analysis, 1975. Keller, M. "Alcohol and Youth." In J.E. Mayer and W.J. Filstead, eds., Adolescence and Alcohol, pp. 245-256. Cambridge, Mass.: Ballinger Publishing Co., 1980. Kendall, M. Time-series, Second Edition. New York: Hafner Press, 1976. Klein, D. "Adolescent Driving as Deviant Behavior." In P.F. Waller, ed., The Young Driver: Reckless or Unprepared?, pp. 2-19. Chapel Hill: North Carolina Symposium on Highway Safety, Fall 1971. Klein, T. The Effect ofRaising the Minimum Legal Drinking Age on Traffic Accidents in the State of Maine. Washington, D.C.: U.S. National Highway Traffic Safety Administration, 1981. Koch, G.G. "Commentary on Statistical Issues Underlying the Evaluation of a Social Policy Change on Motor Vehicle Accident Involvement." Accident Analysis and Prevention 13 (1981):323-329. Lau, H-H. "Cost of Alcoholic Beverages as a Determinant of Alcohol Con-

I!

i'

&:

f ;i

I'

, i

Ii •

140


sumption." In R. Gibbons, ed., Research Advances in Alcohol and Drug Problems, vol. 2, pp. 211-246. New York: John Wiley and Sons, 1975. Lillis, R., T. Williams, and W. Williford. Reported Alcohol Crashes Involving 18-21 Year Old Pennsylvania Drivers in Ten New York Border Counties. Albany: New York State Division of Alcoholism and Alcohol Abuse, Bureau of Alcohol and Highway Safety (Research Report Series Number 10), 1981. Liu, L-M. User's Manual for BMDQ2T (TSPACK) Time Series Analysis (Box-Jenkins). Technical report 57. Los Angeles: University of California, Los Angeles, Department of Biomathematics, 1979. Ljung, G. and G.E.P. Box. Studies in the Modeling ofDiscrete Time Series 3: A Modification of the Overall Chi-square Test for Lack of Fit in Time Series Model. Technical report 477. Madison, Wise.: University of Wisconsin, Department of Statistics, 1976. Lynn, C. The Effects of Lowering the Legal Drinking Age in Virginia. Charlottesville, Va.: Virginia Highway and Transportation Research Council, 1981. MacAndrew, C. and R. Edgerton. Drunken Comportment. Chicago: Aldine, 1969. Maddox, G.L. and B.C. McCall. Drinking Among Teenagers. New Brunswick, N.J.: Rutgers Center of Alcohol Studies, 1964. Maisto, S.A. and J. V. Rachal. "Indications of the Relationship Among Adolescent Drinking Practices, Related Behaviors, and Drinking Age Laws." In H. Wechsler, ed., Minimum-Drinking-Age Laws, pp. 155-176. Lexington, Mass.: D.C. Heath and Co., 1980. Makela, K. "Level of Consumption and Social Consequences of Drinking." In Y. Israel et al., eds., Research Advances in Alcohol and Drug Problems, vol. 4, pp. 303-348. New York: Plenum Press, 1978. Maxwell, D.M. Impact Analysis of the Raised Legal Drinking Age in Illinois. Washington, D.C.: U.S. National Highway Traffic Safety Administration, 1981. McFadden, M. and H. Wechsler. "Minimum Drinking Age Laws and Teenage Drinking." Psychiatric Opinion 16 (1979):22-23,26-28. McCleary, R. and R.A Hay, Jr. Applied Time Series Analysis for the Social Sciences. Beverly Hills, Calif.: Sage Publications, 1980. Medicine in the Public Interest. The Effects of Alcoholic Beverage Control Laws. Washington, D.C., 1979. Merton, R.K. "A Paradigm for Functional Analysis in Sociology." In L.A. Coser and B. Rosenberg, eds., Sociological Theory: A Book of Readings, 4th ed., pp. 527-534. New York: Macmillan, 1976. Merton, R.K. and AS. Kitt. "Reference Groups." In L.A Coser and B. Rosenberg, eds., Sociological Theory: A Book of Readings, 4th ed., pp. 243-250. New York: Macmillan, 1976.

Bibliography

Michigan Council on P 1973. Michigan State Legislat

Report of the Spec erage Container DE Lansing, Mich.: St Michigan State Police. Moore, M.H. and D.R the Shadow of P, Press, 1981. Mosher, J.F. "Dram S Problems." Jourm ___ . "The History rent Policy. In H. 11-38. Lexington, I Moser, Joy. Prevention view of Preventive. diction Research F< Naor, E.M. and R.D.. duction in the Leg (1975):74-79. National Highway Tra Washington, D.C., ___ . Evaluation oj Electronic Injury S National Institute on A to the U. S. Congre Government Printi ___ . Fourth Speci£ Health. Washingto National Safety Counc National Transportatio H-82-18. Washingt Nelson, C.R. Applied San Francisco: Hol Newbold, P. and C.W.. Time Series and thl Statistical Society I New York State. MorE trol Law Study Pap through 3. New Y( O'Day, J., D.J. and 55 MPH Speel Michigan,

Traffic Accidents

(vances in Alcohol and : John Wiley and Sons,

Alcohol Crashes InTen New York Border \.lcoholism and Alcohol (Research Report Series !

o Time Series Analysis

les: University of Calinatics, 1979. ofDiscrete Time Series Test for Lack of Fit in ison, Wise.: University

nking Age in Virginia. ransportation Research Jmportment. Chicago:

Teenagers. New Bruns1964. e Relationship Among ors, and Drinking Age ;nking-Age Laws, pp. J., 1980. equences of Drinking. " !cohol and Drug Prob:>ress, 1978. 'Jrinking Age in Illinois. raffic Safety Adminis-

rlg Age Laws and Teen:2-23, 26-28. :es Analysis for the Soations, 1980. coholic Beverage Con-

vsis in Sociology." In 'al Theory: A Book of cmillan, 1976. " In L.A. Coser and k of Readings, 4th ed.,

Bibliography

141

Michigan Council on Alcohol Problems. Micap Recap, no. 35, August 14, 1973. Michigan State Legislature. Michigan's Deposit Law: First Year, An Interim Report of the Special Joint Committee to Study the Impact of the Beverage Container Deposit Law, Senator Stephen Monsma, Chairperson. Lansing, Mich.: State of Michigan, 1979. Michigan State Police. Unpublished data, 1982. Moore, M.H. and D.R. Gerstein, eds., Alcohol and Public Policy: Beyond the Shadow of Prohibition. Washington, D.C.: National Academy Press, 1981. Mosher, J.F. "Dram Shop Liability and the Prevention of Alcohol-related Problems." Journal of Studies on Alcohol 40 (1979):773-798. ___ . "The History of Youthful-Drinking Laws: Implications for Current Policy. In H. Wechsler, ed., Minimum-Drinking-Age Laws, pp. 11-38. Lexington, Mass.: D.C. Heath and Co., 1980. Moser, Joy. Prevention ofAlcohol-related Problems-An International Review ofPreventive Measures, Policies, and Programmes. Toronto: Addiction Research Foundation, 1980. Naor, E.M. and R.D. Nashold. "Teenage Driver Fatalities Following Reduction in the Legal Drinking Age." Journal of Safety Research, 7 (1975):74-79. National Highway Traffic Safety Administration. 55 MPH Fact Book. Washington, D.C., 1978. ___ . Evaluation of Minimum Drinking Age Laws Using the National Electronic Injury Surveillance System. Washington, D.C., 1982. National Institute on Alcohol Abuse and Alcoholism. Third Special Report to the U.S. Congress on Alcohol and Health. Washington, D.C.: U.S. Government Printing Office, 1978. ___ . Fourth Special Report to the U.S. Congress on Alcohol and Health. Washington, D.C.: U.S. Government Printing Office, 1980. National Safety Council. Accident Facts, 1980 Edition. Chicago, 1980. National Transportation Safety Board. "Safety Recommendations," no. H-82-18. Washington, D.C., 1982. Nelson, C.R. Applied Time Series Analysis for Managerial Forecasting. San Francisco: Holden-Day, 1973. Newbold, P. and C.W.J. Granger. "Experience with Forecasting Univariate Time Series and the Combination of Forecasts." Journal of the Royal Statistical Society 137 (1974):131-165. New York State. Moreland Commission on the Alcoholic Beverage Control Law Study Papers 1 through 5 and Report and Recommendations 1 through 3. New York, 1963 and 1964. O'Day, J., D.J. Minahan, and D. Golomb. The Effects of the Energy Crisis and 55 MPH Speed Limit in Michigan. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1975.

142

I

i l

I


Bibliography

Organization for Economic Cooperation and Development. Young Driver Accidents. Paris, 1975. Ornstein, S.1. "Control of Alcohol Consumption Through Price Increases." Journal of Studies on Alcohol 41 (1980):807-817. Pelz, D.C. and S.H. Schuman. "Motivational Factors in Crashes and Violations of Young Drivers." Presented at the American Public Health Association Annual Meeting, Minneapolis, October 13, 1971. Perrine, M.W., J.A. Waller, and L.S. Harris. Alcohol and Highway Safety: Behavioral and Medical Aspects. Burlington, Vt.: University of Vermont, 1971. Pindyck, R.S. and D.L. Rubinfeld. Econometric Models and Economic Forecasts. New York: McGraw-Hill, 1976. Plaut, T. Alcohol Problems: A Report to the Nation by the Cooperative Commission on the Study of Alcoholism. New York: Oxford University Press, 1967. Popham, R.E., W. Schmidt, and J. deLint. "The Effects of Legal Restraint on Drinking." In B. Kissin and H. Begleiter, eds., The Biology of Alcoholism, Volume 4: Social Aspects of Alcoholism, pp. 579-626. New York: Plenum Press, 1976. Preusser, D.F., J.F. Oates, Jr., and M.S. Orban. Identification of Countermeasures for the Youth Crash Problem Related to Alcohol: Review of Existing Literature and Preliminary Hypotheses. Darien, Conn.: Dunlap and Associates, 1975. Rooney, J.F. and S.M. Schwartz. "The Effects of Minimum Drinking Age Laws Upon Adolescent Alcohol Use and Problems." Contemporary Drug Problems 6 (1977):569-583. Roy, M.B. and E. Greenblatt. Driving Under the Influence of Liquor: Followup Study of Age, Sex, and Simultaneous Offenses. Boston: Commonwealth of Massachusetts, Office of Commissioner of Probation, November 27, 1979. Ryan, W. Blaming the Victim. New York: Random House, 1976. Schifrin, L.G., C.E. Hartsog, and D.H. Brand. "Costs of Alcoholism and Alcohol Abuse and Their Relation to Alcohol Research." In Institute of Medicine, ed., Alcoholism, Alcohol Abuse, and Related Problems: Opportunities for Research, pp. 165-186. Washington, D.C.: National Academy Press, 1980. Schmidt, W. and A Kornaczewski. "The Effect of Lowering the Legal Drinking Age in Ontario on Alcohol-related Motor Vehicle Accidents." In Alcohol, Drugs, and Traffic Safety, Proceedings of the Sixth International Conference on Alcohol, Drugs, and Traffic Safety, pp. 763-770, Toronto, September 8-13,1974. Toronto: Addiction Research Foundation, 1975. Seila, AF., M.A. Entsminger, and C.Z. Silva. An Examination of the Effects of the Lowered Maximum Speed Limit and Fuel Shortages in

North Carolina. C Safety Research 0 Shattuck, D. and P .C. ewan: The Effect ( Department of He: Smart, R.G. "Change: the Legal Drinkinl Abuse 4 (1977):101 Smart, R.G. and J. F Beer Consumption (1976):393-402. Smart, R.G. and W. After a Reduction: Addiction 70 (1975 Smart, R.G. and M.S. ( Purchasing Age on pirical Studies." J( Treat, J .R. "Tri-Ievel of Final Results." Association for Au, Columbia, Septem1 United States Departm of Population, Vol Maine. 1973. Verway, D.I., ed. Mid Michigan State Un tion, 1980. Vigderhous, G. Jenkins Methodolo ciological Methodo 1977. Vingilis, E. and R.G. 5 in Ontario." Britisj Voas, R.B. and Moul by Young America Laws, pp. 59-72. L Wagenaar, A.C. The M Evaluation. Ph.D. ___ . "Effects ,of al Journal of Public 1 ___ . "Legal Minirr 1970-1981." A lcohl ___ . "Aggregate Be the Minimum Lega

1

I Traffic Accidents

:lopment. Young Driver

hrough Price Increases." 17. tors in Crashes and Vio\.merican Public Health tober 13, 1971.

701 and Highway Safety: Vt.: University of Ver-

Models and Economic

lion by the Cooperative

{ York: Oxford Univer-

Le Effects of Legal Reeds., The Biology lcoholism, pp. 579-626.

Identification of Counlted to Alcohol: Review theses. Darien, Conn.:

\1inimum Drinking Age blems." Contemporary

e Influence of Liquor: Offenses. Boston: ommissioner of Proba-

House, 1976. 'osts of Alcoholism and Research." In Institute

[1

and Related Problems: ington, D.C.: National of Lowering the Legal tor Vehicle Accidents."

?dings of the Sixth In'1d Traffic Safety, pp. to: Addiction Research

n Examination of the and Fuel Shortages in

Bibliography

143

North Carolina. Chapel Hill: University of North Carolina, Highway Safety Research Center, 1977. Shattuck, D. and P .C. Whitehead. Lowering the Drinking Age in Saskatchewan: The Effect on Collisions Among Young Drivers. Saskatchewan Department of Health, May 1976. Smart, R.G. "Changes in Alcoholic Beverage Sales After Reductions in the Legal Drinking Age." American Journal of Drug and Alcohol Abuse 4 (1977):101-108. Smart, R.G. and J. Finley. "Changes in Drinking Age and Per Capita Beer Consumption in Ten fanadian Provinces." Addictive Diseases 2 (1976):393-402. . . Smart, R.G. and W. Schmidt. "Drinking and Problems From Dnnkmg After a Reduction in the Minimum Drinking Age." British Journal of Addiction 70 (1975):347-358. Smart, R.G. and M.S. Goodstadt. "Effects of Reducing the Legal AlcoholPurchasing Age on Drinking and Drinking Problems: A Review of Empirical Studies." Journal of Studies on Alcohol 38 (1977):1313-1323. Treat, J .R. "Tri-level Study of Causes of Traffic Accidents: An Overview of Final Results." In D.F. Huelke, ed., Proceedings of the American Association for Automotive Medicine, pp. 391-403. Vancouver, British Columbia, September 15-17, 1977. United States Department of Commerce, Bureau of Census. 1970 Census of Population, Volume 1, Characteristics of the Population, Part 21, Maine. 1973. Verway, D.L, ed. Michigan Statistical Abstract, 15th ed. Lansing, Mich.: Michigan State University, Graduate School of Business Administration, 1980. Vigderhous, G. "Forecasting Sociological Phenomena: Application of BoxJenkins Methodology to Suicide Rates." In K.F. Schuessler, ed., Sociological Methodology 1978, pp. 20-51. San Francisco: Jossey-Bass, 1977. Vingilis, E. and R.G. Smart. "Effects of Raising the Legal Drinking Age in Ontario." British Journal of Addiction 76 (1981):415-424. Voas, R.B. and l Moulden. "Historical Trends in Alcohol Use and Driving by Young Americans." In H. Wechsler, ed., Minimum-Drinking-Age and Co., 1980. Laws, pp. 59-72. Lexington, Mass.: D.C. Wagenaar, A.C. The Minimum Legal Drinking Age: A Time Series Impact Evaluation. Ph.D. dissertation, The University of Michigan, 1980. "Effects of an Increase in the Legal Minimum Drinking Age." of Public Health Policy 2 (1981):206-225. ___ . "Legal Minimum Drinking Age Changes in the United States: 1970-1981." Alcohol Health and Research World 6 (1981-1982):21-26. ___ . "Aggregate Beer and Wine Consumption: Effects of Changes in the Minimum Legal Drinking Age and a Mandatory Beverage Con-

144


tainer Deposit Law in Michigan." Journal of Studies on Alcohol 43 (1982a) :469-487. ___ . "Preventing Highway Crashes by Raising the Legal Minimum Age for Drinking: An Empirical Confirmation." Journal of Safety Research 13 (1982b):57-71. ___ . "Public Policy Effects on Alcohol Consumption in Maine and New Hampshire: 1970-1980." Contemporary Drug Problems 11 (1982c). ___ . "Raising the Legal Drinking Age in Maine: Impact on Traffic Accidents among Young Drivers." International Journal of the Addictions 18 (1983). Waller, J.A., E.M. King, G. Nielson, and H.W. Turkel. "Alcohol and Other Factors in California Highway Fatalities." In Proceedings of the lith Annual Meeting of the American Association for Automotive Medicine. Springfield, Ill.: Charles C. Thomas, 1970. Warren, R.A., H.M. Simpson, L. Page-Valin, and D. Collard. Point Zero Eight and the Change in Drinking Age: One Step Forward and Two Steps Backward? Ottawa: Traffic Injury Research Foundation of Canada, 1977. Warren, R.A., H.M. Simpson, M.A. Buhlman, L.A. Bourgeois, and L.D. Chattaway. "Relationship of Driver Blood Alcohol to Injury Severity." In Proceedings of the 25th Conference of the American Association for Automotive Medicine, pp. 133-144. Morton Grove, Ill.: American Association for Automotive Medicine, 1981. Whitehead, P.C. Alcohol and Young Drivers: Impact and Implications of Lowering the Drinking Age. Ottawa: Department of National Health and Welfare, Health Protection Branch, Non-medical Use of Drugs Directorate, Research Bureau, 1977. ___ . "Research Strategies to Evaluate the Impact of Changes in the Legal Drinking Age." In H. Wechsler, ed., Minimum-Drinking-Age Laws, pp. 73-92. Lexington, Mass.: D.C. Heath and Co., 1980. Whitehead, P.C., J. Craig, N. Langford, C. MacArthur, B. Stanton, and R.G. Ferrence. "Collision Behavior of Young Drivers: Impact of the Change in the Age of Majority." Journal of Studies on Alcohol 36 (1975): 1208-1223.

Williams, A.F., R.F. Rich, P.L. Zador, and L.S. Robertson. The Legal Minimum Drinking Age and Fatal Motor Vehicle Crashes. Washington, D.C.: Insurance Institute for Highway Safety, 1974. Williams, A.F., P.L. Zador, S.S. Harris, and R.S. Karpf. "The Effects of Raising the LegalMinimum Drinking Age on Fatal Crash Involvement." Washington, D.C.: Insurance Institute for Highway Safety, 1981. Wiorkowski, J.J. and R.F. Heckard. "The Use of Time Series Analysis and Intervention Analysis to Assess the Effects of External Factors on Traffic Indices: A Case Study of the Effects of the Speed Limit Reduc-

Bibliography

tion and Energy ( Prevention 9 (197' Wolfe, A.C. "Charac the Society of Aut. Detroit, Mich., 0, ___ . First Evaluat, mentlEducation 1 Highway Safety R Wolfe, A.C. and M.T\tenaw County Michigan, Highwa Works, D.A. "Statem and Drug Educati, Zylman, R. "Age is M ment of Young an 7-8, 34.

___ . "When it is Journal of Traffic ___ . "Drinking an lem Real?" Journ, ___ . "Collison Bel Whitehead et al." ___ . "The Conseq related Crash Invc Douglass." Jourm

? .,

I

r

Traffic Accidents

Studies on Alcohol 43

19 the Legal Minimum

In.'' Journal of Safety mmption in Maine and rug Problems 11 (1982c). Line: Impact on Traffic al Journal of the AddicTurkel. "Alcohol and ," In Proceedings of the r:iation for Automotive t, 1970. D. Collard. Point Zero Step Forward and Two Foundation of 1\.. Bourgeois, and L.D. lhol to Injury Severity."

rnerican Association for Grove, Ill.: American act and Implications of lent of National Health -medical Use of Drugs pact of Changes in the 4inimum-Drinking-Age :h and Co., 1980. lrthur, B. Stanton, and Drivers: Impact of the Studies on Alcohol 36 Robertson. The Legal e Crashes. Washington, 1974. Karpf. "The Effects of al Crash Involvement." way Safety, 1981. f Time Series Analysis of External Factors on he Speed Limit Reduc-

Bibliography

145

tion and Energy Crisis in the State of Texas." Accident Analysis and Prevention 9 (1977):229-247. Wolfe, A.C. "Characteristics of Alcohol-impaired Drivers." Presented at the Society of Automotive Engineers, Automobile Engineering Meeting, Detroit, Mich., October 13-17, 1975. ___ . First Evaluation Report on the Oakland County Alcohol Enforcement/Education Project. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1981. Wolfe, A.C. and M.M. Chapman. 1971 and 1973 ASAP Surveys: Washtenaw County High School Students. Ann Arbor: The University of Michigan, Highway Safety Research Institute, 1973. Works, D.A. "Statement on 18 Year Old Drinking." Journal of Alcohol and Drug Education 18 (1973):14. Zylman, R. "Age is More Important than Alcohol in the Collision Involvement of Young and Old Drivers." Journal of Traffic Safety 20 (1972): 7-8, 34. ___ . "When it is Legal to Drink at 18: What Should We Expect?" Journal of Traffic Safety Education 20 (1973):9-10. ___ . "Drinking and Driving After its Legal to Drink at 18: Is the Problem Real?" Journal of Alcohol and Drug Education 20 (1974):48-52. ___ . "Collison Behavior of Young Drivers: Comment on the Study by Whitehead et al." Journal of Studies on Alcohol 37 (1976):393-401. ___ . "The Consequences of Lower Legal Drinking Ages on Alcoholrelated Crash Involvement of Young People. Rebuttal to Dr. Richard Douglass." Journal of Traffic Safety Education 24 (1977):13-15.

Aaron, P., and D. Mu Abrupt-permanent-imp model, 128 Alabama, 22 Alberta, 21 Alcohol, benefits of, II Alcohol, Drug Abuse, tal Health Administl Alcohol availability, 31 Alcohol consumption a drinking age, 27-30, Alcohol Epidemiologic System, 22 Alcohol marketing pral Alcohol sales, 29-30; do 49-50; research See also Draft-beer! Package-beer sales Alcoholism, 17-18 American Businessmen Foundation, 29, 106 Auto-Regressive Integn Average (ARIMA) n 125-133

I

I I

II

Bacon, S.D., 18, 106 Bako, G., w.e. MacKI E.S.O. Smith, 21 Barsby, S.L., and G.L. 29 Beauchamp, D.E., 17 Beer prices: Maine, 88Michigan, 99 Berk, R.A., and M. Br Berry, R.E., Jr., and J.l Blane, H.T., 13 Blane, H.T., and L.E. 10 Blood-alcohol concentn: 11, 13-14, 51

Index Aaron, P., and D. Musto, 1 Abrupt-permanent-impact model, 128 Alabama, 22 Alberta, 21 Alcohol, benefits of, 105-107 Alcohol, Drug Abuse, and Mental Health Administration, 18 Alcohol availability, 31-33, 103-104 Alcohol consumption and drinking age, 27-30, 80-95 Alcohol Epidemiologic Data System, 22 Alcohol marketing practices, 31 Alcohol sales, 29-30; data, 49-50; research analysis, 80-95. See also Draft-beer sales; Package-beer sales Alcoholism, 17-18 American Businessmen's Research Foundation, 29, 106 Auto-Regressive Integrated Moving Average (ARIMA) model, 50, 125-133 Bacon, S.D., 18, 106 Bako, G., W.C. MacKenzie, and E.S.O. Smith, 21 Barsby, S.L., and G.L. Marshall, 29 Beauchamp, D.E., 17 Beer prices: Maine, 88-89, 99; Michigan, 99 Berk, R.A., and M. Brewer, 123 Berry, R.E., Jr., and J.P. Boland, 1 Blane, H. T., 13 Blane, H.T., and L.E. Hewitt, 9, 10 Blood-alcohol concentration (BAC), 11, 13-14, 51

Bonnie, R.J., 31 Borg, T.M., et al., 127 Borkenstein, R.F., et al., 11, 12 Bowen, B.D., and M.R. Kagay, 3 Box, G.E.P., and D.A. Pierce, 128 Box, G.E.P., and G.C. Tiao, 50, 125 Box, G.E.P., and G.M. Jenkins, 50, 51, 125, 126, 128, 131, 133 Box-Jenkins intervention analysis, 50-51, 113, 133 Box-Tiao intervention analysis, 50 Branch, B., 81 Brissett, D., 106 Brown, D.B., and S. Maghsoodloo, 22 Bruun, K., et al., 18 Burkett, S.R., and W.T. Carrithers, 98 Burritt, B.E., A. Moghrabi, and J.S. Matthias, 127 Cahalan, D., and I.H. Cisin, 106 California, 12 Cameron, T., 11 Campbell, D.T., 109 Campbell, D.T., and J.C. Stanley, 37, 113 Canada, 3, 4, 21, 30 Carsten, 0., 12 Cavan, S., 106 Chu, B., and G.E. Nunn, 127 Clark, N.H., 1 Connecticut, 110 Construct validity, 119-121 Convergent validity, 120 Cook, P.J., 108 Cook, P.J., and G. Tauchen, 22, 102 147

148

! f

£

I

,

Index


Cook, T.D., and D.T. Campbell, 24,37, 113, 114, 119-120, 121, 123 Cross-border purchase of alcohol, 86, 88-89, 100, 101-102 Cucchiaro, S., J. Ferreira, Jr., and A. Sicherman, 19

Filkins, L.D., and J.D. Flora, 12, 25 Flora, J.D., L.D. Filkins, and C.P. Compton, 20, 42, 51 Flora, J.D., et al., 12 Frankel, B.G., and P.C. Whitehead, 18

Dart, O.K., 127 Data-analysis methods, 50-51, 125-133 Data collection, 41-50 Detroit Free Press, 86 Differentiated-drinking policy, 107 Distilled Spirits Council of the United States, 3, 89 District of Columbia, 5 Douglass, R.L., L.D. Filkins, and F.A. Clark, 42, 80 Douglass, R.L., and J.A. Freedman, 20, 30, 36, 80 Douglass, R.L., et al., 13, 20, 99 Draft-beer sales: Michigan, 83-85, 86,91, 100; U.S., 90-91, 100 Drinking age: and alcohol consumption, 27-30, 80-95; and high-risk category, 111; and highway accidents, 18-26, 53-80 Drinking-age laws: constitutionality of, 104-105; enforcement of, 49-50, 95-98; nomenclature, 7 Driving under the influence of liquor (DUlL) arrests: Maine, 99; Massachusetts, 23-24; Michigan, 59-61 Dunn, O.J., and V.A. Clark, 114

Gallup polls, 105, 106 Gasoline shortages and price increases, 118 Gender, 103, 110 Gottman, J.M., 125 Grand Rapids Press, 96 Grandfather clauses, 26, 28, 38 Guy, Ralph, Jr., 14, 104

Evaluation apprehension, 120 External validity, 121-123 Farris, R., T.B. Malone, and H. Lilliefore, 12, 13 Fatalities, 102, 110

Hammond, R.L., 29 Harford, T.C., D.A. Parker, and L. Light, 18 Haugh, L.D., and G.E.P. Box, 133 Hibbs, D.A., Jr., 125 Highway accidents: and drinking age, 18-26, 53-80; fatalities, 102, 110; unreported, 79-80. See also Injury-damage crashes; Propertydamage crashes Hingson, R., et al., 24, 28, 101 Historical events, contemporary, 115-116 History-treatment interaction, 122 Homogeneous nonstationary behavior, 131 Hurst, P.M., 13 Illinois, 2, 5, 19, 21-22, 25 Illinois Department of Transportation, 21 Indiana, 19 Indicators, single/multiple, 120 Injury-damage crashes: Maine, 6669, 78; Michigan, 56-58, 77; New York, 70-72; Pennsylvania, 74, 75

i

Institute for Research Safety, 11 Instrumentation, 116 Internal validity, 115-] Intervention, diffusior Intervention analysis, 133 Iowa, 2

I

Jessor, R., and S.L. J Johnston, L.D., 93 Johnston, L.D., J.G. and P.M. O'Malley Jones, R.K., and K.B. 11, 78

I

Kahane, C.J., 127 Keller, M., 14 Kendall, M., 42, 51, 1 Klein, D., 13 Klein, T., 25 Koch, G.G., 22

I

Lau, H-H., 81 Liability, legal, 111 Lillis, R., et al., 101 Liu, L-M., 128 Ljung, G., and G.E.P Louisiana, 20 Lynn, C., 22

t

I

j

I

----l.

MacAndrew, C., and 106 Maddox, G.L., and B 105, 106 Maine, 5-7, 37, 101; a 49, 86-93, 94; beer crash data, 43, 45-4 77-78, 79, 80; cross chase of alcohol, drinking-age-Iaw en 49-50, 97-98; and f( design, 38-40; resea

:raffic Accidents

Index

L.D. Filkins, and C.P. 20, 42, 51 et al., 12 ., and P.C. White-

Institute for Research in Public Safety, 11 Instrumentation, 116 Internal validity, 115-119 Intervention, diffusion of, 119 Intervention analysis, 50-51, 125133 Iowa, 2

105, 106 rtages and price in8 110 ({., 125 s Press, 96 clauses, 26, 28, 38 Jr., 14, 104

Jessor, R., and S.L. Jessor, 107 Johnston, L.D., 93 Johnston, L.D., J.G. Bachman, and P.M. O'Malley, 10 Jones, R.K., and K.B. Joscelyn, 11, 78

, and J.D. Flora, 12,

29 :., D.A. Parker, and 8 and G.E.P. Box, 133 Jr., 125 dents: and drinking 53-80; fatalities, 102, orted, 79-80. See also lage crashes; Propertylshes et al., 24, 28, 101 nts, contemporary,

nent interaction, 122 , nonstationary be13

19,21-22, 25 tment of Transporta-

19le/multiple, 120 : crashes: Maine, 66higan, 56-58, 77; New ; Pennsylvania, 74, 75

Kahane, C.J., 127 Keller, M., 14 Kendall, M., 42, 51, 133 Klein, D., 13 Klein, T., 25 Koch, G.G., 22 Lau, H-H., 81 Liability, legal, 111 Lillis, R., et al., 101 Liu, L-M., 128 Ljung, G., and G.E.P. Box, 128 Louisiana, 20 Lynn, C., 22 MacAndrew, C., and R. Edgerton, 106 Maddox, G.L., and B.C. McCall, 105, 106 Maine, 5-7, 37, 101; alcohol sales, 49, 86-93, 94; beer prices, 88-89; crash data, 43, 45-46, 64-69, 76, 77-78, 79, 80; cross-border purchase of alcohol, 88-90, 100; drinking-age-Iaw enforcement, 49-50, 97-98; and research design, 38-40; research on legal-

149

drinking age and highway crashes in, 20, 24, 25, 26 Maisto, S.A., and J.V. Rachal, 28, 31, 33 Makela, K., 13 Mandatory-deposit law, 99; Maine, 88, 89-90, 93, 99, 118; Michigan, 81-83, 85, 118 Manitoba, 21 Maryland, 2, 5 Massachusetts, 19,23-24,27,28, 36, 101, 102, 110 Maturation, 116 Maxwell, D.M., 25 McCleary, R., and R.A. Hay, 51, 125 McFadden, M., and H. Wechsler, 27 Medicine in the Public Interest, 18 Merton, R.K., 107 Merton, R.K., and A.S. Kitt, 107 Michigan, 5-7, 37, 53, 101, 102, 104, 105, 110; alcohol sales, 49, 81-86, 92, 93, 94; crash data, 42,43,44-45, 56-63, 76, 77, 79, 80; cross-border purchase of alcohol, 86; drinking-age-Iaw enforcement, 50, 96-97; and research design, 38-40; research on legal-drinking age and alcohol consumption in, 27, 29, 30, 36; research on legaldrinking age and highway crashes, 19, 20, 25, 26 Michigan Council on Alcohol Problems, 3 Michigan State Legislature, 81 Michigan State Police, 12 Minnesota, 19 Mono-method bias, 120 Mono-operation bias, 119-120 Montana, 2 Moore, M.H., and D.R. Gerstein, 18

,

,I

Index

II'

111

1\ I1 :1

Mortality, 117 Moser, Joy, 18 Mosher, J.F., 1,7,31,111 Multiple-test analysis, 114 Naor, E.M., and R.D. Nashold, 19 National Academy of Sciences, 18 National Highway Traffic Safety Administration, 26, 127 National Institute on Alcohol Abuse and Alcoholism, 10, 18, 103 National Safety Council, 1, 10, 12 National Transportation Safety Board, 110 Nelson, C.R., 131 New Brunswick, 21 New Hampshire, 110; beer prices, 88-89; beer sales, 49, 92, 93, 100; legal-drinking age, 90, 100 New Jersey, 4 New York City, 48, 51 New York State, 24, 110; crash data, 42, 43, 46-48, 69-73; Moreland Commission, 18; and research design, 38, 39 Newbold, P., and C.W.J. Granger, 51 Noise model, 125, 132 O'Day, J., D.J. Minahan, and D. Golomb, 127 Oklahoma, 2, 110 Ontario, 5, 19, 20, 26, 28, 29, 79 Organization for Economic Cooperation and Development, 12, 24 Ornstein, S.L, 81 Package-beer sales: Michigan, 8183, 85-86 Pelz, D.C., and S.H. Schuman, 13 Pennsylvania, 20, 110; crash data, 43, 48-49, 73-76; cross-border

Index

150 purchase of alcohol, 101-102; and research design, 38, 39 Perrine, M.W., J.A. Waller, and L.S. Harris, 12, 13 Plaut, T., 14 Police reports, 42-43, 118-119 Popham, R.E., W. Schmidt, and J. deLint, 81, 108 Preusser, D.F., J.F. Oates, Jr., and M.S. Orban, 12, 13 Property-damage crashes: Maine, 64-66,78; Michigan, 54-55,77; New York, 69-70; Pennsylvania, 74,75

Setting-treatment in Shattuck, D., and P 21 Single-vehicle crash, Single-vehicle nightt (SVNM) crash in Smart, R.G., 30, 36 Smart, R.G., and J Smart, R.G., and V 27,29 South Dakota, 2 Statistical-conclusio 113-115 Statistical inference,

Random error, 114-115, 125 Reactive effects, 120 Regression effect, 116-117 Research design, 37-41 Research design validity, 113-123 Residual white-noise, 125 4,110 Rooney, J.F., and S.M. Schwartz, 28 Roy, M.B., and E. Greenblatt, 23 Ryan, W., 15

Temporary-impact I Texas, 20 Toronto, 27, 29 Transfer function II 128, 132, 133 Treat, J.R., 11 Trends, long-term, Type I error probat

United States: beer U.S. Constitution: . Amendment, 105

Saskatchewan, 5, 21 Schifrin, L.G., C.E. Hartsog, and D.H. Brand, 1 Schmidt, W., and A. Kornaczewski, 20 School principals, 29 Seasonal cycles, 50 Seila, A.F., M.A. Entsminger, and C.Z. Silva, 127 Selection, 117 Selection-history interaction, 118 Selection-instrumentation interaction, 118 Selection-maturation interaction, 117-118 Selection-treatment interaction, 121

j

- - - - - - - - - --

150 llcohol, 101-102; design, 38, 39 J.A Waller, and 12, 13 118-119 W. Schmidt, and ,108 J.F. Oates, Jr., Jan, 12, 13 e crashes: Maine, chigan, 54-55, 77; )-70; Pennsylvania,

Setting-treatment interaction, 122 Shattuck, D., and P.C. Whitehead, 21 Single-vehicle crash, defined, 44 Single-vehicle nighttime male (SVNM) crash indicator, 42-43 Smart, R.G., 30, 36 Smart, R.G., and J. Finley, 30 Smart, R.G., and W. Schmidt, 27,29 South Dakota, 2 Statistical-conclusion validity, 113-115 Statistical inference, 123 Temporary-impact model, 128 Texas, 20 Toronto, 27, 29 Transfer function models, 125, 128, 132, 133 Treat, J .R., 11 Trends, long-term, 50 Type I error probability, 113, 114

E. Greenblatt, 23

United States: beer sales, 49, 90-91 U.S. Constitution: Twenty-first Amendment, 105; Twenty-sixth

Hartsog, and 1 I A. Kornaczewski, ;,29 50 ,. Entsminger, and 7 interaction, 118 lentation interlion interaction, nt interaction, 121

151

Index

114-115, 125 120 t, 116-117 , 37-41 validity, 113-123 loise, 125 , 110 ld S.M. Schwartz,

,21

-----------

Amendment, 2 U.S. Department of Commerce, 93 Vermont, 20, 110 Verway, D.L, 99 Vigderhous, G., 51, 133 Vingilis, E., and R.G. Smart, 26, 28,29 Virginia, 2, 5, 22 Voas, R.B., and J. Moulden, 24 Wagenaar, A.C., 7, 25, 102 Waller, J.A., et al., 1, 12, 21, 78 West Virginia, 2 Whitehead, P.C., 21, 79, 102 Whitehead, P.C., et al., 20 Williams, AF., et al., 19, 25, 101 Wine sales, 83, 87 Wiorkowski, J.J., and R.F. Heckard, 127 Wisconsin, 2, 19 Wolfe, AC., 13, 97 Wolfe, A.C., and M.M. Chapman, 27 Works, D.A., 3 Zylman, R., 3, 4, 13, 23

Alexander C. Wagena Michigan Transportat the University of Micl1 with the Transportat alcohol policy, highw< Journal of Safety Re, Public Health Policy, the editorial board of Wagenaar is active in annual-meeting progr; Problems of the Amel

i !1

I

j

>I

1 About the Author Alexander C. Wagenaar is on the senior research staff of the University of Michigan Transportation Research Institute. He received the Ph.D. from the University of Michigan School of Public Health and has been affiliated with the Transportation Research Institute since 1977. His articles on alcohol policy, highway safety, and injury prevention have appeared in the Journal of Safety Research, Journal of Studies on Alcohol, Journal of Public Health Policy, and Alcohol Health and Research World. He is on the editorial board of Abstracts and Reviews on Alcohol and Driving. Dr. Wagenaar is active in several professional organizations and is currently the annual-meeting program chairman for the Forum on Alcohol and Drug Problems of the American Public Health Association.

I

I

Ii II J

I

,

I.