Schaffer Online Library of Drug Policy Sign the Resolution
Contents | Feedback | Search
DRCNet Home
| Join DRCNet
DRCNet Library | Schaffer Library | Drugs and Driving 


Factors Influencing Aggregate Indicators of Drinking-Driving in the United States

Robert E. Mann, Reginald G. Smart and Lise Anglin

Addiction Research Foundation, 33 Russell St., Toronto, Ontario, Canada M5S 2S1


The value for prevention purposes of understanding the factors that influence aggregate indicators of drinking-driving and related problem rates in populations is substantial. Often, however, aggregate-level measures of factors of potential interest, such as specific efforts to prevent drinking-driving or other alcohol-related problems, are difficult to obtain or unavailable. This research examined the contributions of several prevention-relevant measures to traffic fatality rates (alcohol-related and total) in U.S. states in 1982 and 1990, as well as to changes between those two years. Using Ordinary Least Squares regression procedures, alcohol-related and total fatality rates were regressed onto per capita consumption of alcohol and rates of drinking driving arrests, alcohol abuse treatment and AA membership for each of the two years. As well, for each measure, relative change between the two years was calculated and similar regression analysis performed. For each year, alcohol-related and total fatality rates demonstrated significant positive relationships with per capita consumption of alcohol and drinking-driving arrest rates, and significant negative relationships with AA membership rates. Changes in alcohol-related and total fatality rates were significantly associated with changes in per capita consumption rates. The importance of these observations for understanding and preventing impaired driving will be discussed.


Understanding of factors influencing aggregate levels of the alcohol-crash problem is limited. Per capita consumption of alcohol and alcohol availability have been shown to be major determinants of many alcohol-related health problems such as cirrhosis mortality rates (Skog, 1980; Norstrom, 1986; Seeley, 1961; Schmidt and Popham, 1978). More recently, research has demonstrated with some consistency important associations of alcohol availability and consumption measures with alcohol-related crash measures (e.g., Mann and Anglin, 1990; Holder and Blose, 1987). Legal measures also may influence alcohol-crash rates. Ross (1973) showed that the introduction of the breathalyser law in Great Britain in 1967 was associated with a substantial, but temporary, reduction in nighttime fatal crashes. Homel (1990) has demonstrated that continued high levels of publicity and enforcement may cause a sustained reduction in alcohol-related crashes. Holder and Wagenaar (1994) also observed that the introduction of mandatory server training in Oregon was associated with reductions in crash rates.

A substantial amount of prevention (educational and rehabilitative) has been undertaken. Public information campaigns, school education, and rehabilitative measures for apprehended drinking drivers have been implemented in varying degrees in different jurisdictions. No studies have yet attempted to determine the effects of these measures at the aggregate level, perhaps because of the difficulty in obtaining suitable data.

Population-level indicators of the drinking driving problem appear to have been decreasing in recent years in many jurisdictions. For example, in Ontario the proportion of fatally injured drivers who had been drinking declined from about 60% in the early 1980's to about 40% by the end of the decade (Wilson and Mann, 1990). Fell (1990) has outlined several possible reasons for these declines, including increased public awareness, stricter enforcement, increase in the legal drinking age, fewer young drivers, declining per capita consumption, and increased prevention and intervention. No research has evaluated the hypothesis that these factors have influenced declining drinking-driving crash rates at the population level.

In this research we address two questions. The first is whether differences in per capita consumption levels, measures of enforcement activity (drinking-driving arrest rate), and measures of prevention and rehabilitation efforts (rates of AA membership and treatment for alcohol abuse) are associated in a meaningful way with differences among U.S. states in total and alcohol-related traffic fatality rates. The second is whether there are any meaningful links between changes in traffic safety measures and changes in per capita consumption, enforcement activity and prevention activity.


The relevant data were obtained from several sources. Figures on Alcoholics Anonymous membership (number of clients) for the 50 states and the District of Columbia for 1983 and 1990 were kindly provided by the AA World Service in New York. The AA Central Office surveys AA groups every 3 years; each group is asked to report the number of members who attend meetings regularly. Population data by state for 1982 and 1990 were obtained from the Statistical Abstract of the United States (1982, 1990). Total traffic fatalities (1982 and 1990) and total estimated alcohol-related fatalities (1982 and 1990) by state were obtained from the Fatal Accident Reporting System (FARS) of the National Highway Traffic Safety Administration (1992). Arrests for driving under the influence for each state were based on the Federal Bureau of Investigation's Uniform Crime Reporting Program. Data for 1986 were obtained from U.S. Department of Justice Reports (U.S. Department of Justice, 1987) while 1992 data were compiled by Mothers Against Drunk Driving (MADD, 1993). Numbers of individuals receiving treatment for an alcohol-related problem by state in 1982 and 1987 were based on the National Drug and Alcoholism Treatment Utilization Surveys (National Institute on Alcohol Abuse and Alcoholism, 1983 and 1988). Per capita consumption of gallons of absolute alcohol per individual aged 14 and over for 1982 and 1990 was obtained from NIAAA Surveillance Report #27 (National Institute on Alcohol Abuse and Alcoholism, 1993).


We regressed (ordinary least squares) total fatality rates and alcohol-related fatality rates onto rates of AA membership, alcoholism treatment, drinking-driving arrests and per capita consumption for 1982 and 1990 separately. For some measures, 1982 or 1990 data were not available (e.g., arrest rates). In these cases, we used the data from the nearest available year. We also looked at the regression of relative changes (i.e. [1990-1982]/1982) in total fatality rates and alcohol-related fatality rates onto relative changes in AA membership, alcoholism treatment, drinking-driving arrests and per capita consumption from 1982 to 1990.

The overall regression equations were significant for all six regression analyses (P<.05). (Only two Tables are presented - Tables 1-2). For both total and alcohol-related fatalities for 1982 and 1990, higher AA membership was associated with lower rates of total and alcohol related fatalities (see Table 1 for analysis of 1982 total fatality data). However, alcohol treatment levels were not significantly associated with total or alcohol-related fatalities in either 1982 or 1990. DWI arrest levels were positively associated with total fatalities in 1982 and 1990 and with alcohol related fatalities in 1990 but not 1982. Per capita alcohol consumption was positively associated with alcohol-related fatalities in 1982 and 1990 and with total fatalities in 1990 (not 1982). Only relative change in per capita alcohol consumption was associated with relative change in total fatalities and alcohol-related fatalities between 1982 and 1990.

Table 1
Regression of Total Fatality Rates onto Rates of A.A. Membership. Alcoholism Treatment, DWI Arrests and Per Capita Consumption: 1982

Variable Zero-Order Correlation Slope (+SE) Standardized Regression Coefficient T
AA Membp. -.35 -.02(.01) -.48 -3.52*
Alc. Trtmt. .06 .01(.01) .10 .70
DWI Arrests .28 .01(.00) -.29 2.24*
Cons. .14 .02(.01) .22 1.60
  Multiple R
R Squared
F(4/45 df)
* p<.05

Table 2
Regression of Changes in Total Fatalities onto Changes in A.A. Membership, Alcoholism Treatment, DWI Arrests and Per Capita Consumption: 1982-1990

Variable Zero-Order Correlation Slope (+SE) Standardized Regression Coefficient T
AA Membp. Changes -.02 .05(.07) .08 .67
Alc. Trtmt. Changes .10 -.01(.03) -.04 -.36
DWI Arrests Changes .09 .01(.02) .06 .53
Cons. Changes .65 1.54(.27) .67 5.78*
  Multiple R
R Squared
F(4/45 df)
* p<.05


Per capita alcohol consumption and enforcement activity were positively associated, and AA membership was negatively associated, with state-by-state variations in alcohol-related fatalities in both 1982 and 1990. AA membership was negatively associated with total fatalities in both years and per capita alcohol consumption was positively associated with total fatalities in 1990 only. Levels of treatment were not significantly associated with either type of fatality in 1982 or 1990. Only changes in per capita alcohol consumption made a significant contribution to changes in alcohol-related and total fatalities.

We expected that alcohol consumption would be positively associated with both cross-sectional and trend differences in alcohol-related fatalities. Several previous studies have shown that per capita alcohol consumption is positively related to alcohol problems such as cirrhosis (Skog, 1980; Norstrom, 1986; Seeley, 1961; Schmidt and de Lint, 1977) and to alcohol-related crash measures (e.g., Mann and Anglin, 1990). While it may seem surprising that alcohol consumption relates to changes in total fatalities, other studies have suggested that, in recent years, alcohol-related crashes have declined more than non-alcohol crashes and thus may account for most of the changes in total crash rates (Beirness et al., 1994). In our data, only declining per capita consumption helped to explain the downward trend in road fatalities. It is therefore important that policy makers consider road safety consequences when forming alcohol policies that may contribute to increasing overall alcohol consumption.

Previously, we have suggested that prevention measures such as education, treatment and A.A. membership may now be having a measurable impact on aggregate indicators of alcohol-related problems such as cirrhosis mortality and morbidity levels (Mann et al., 1988; Mann et al., 1991; Smart and Mann, 1991). The analyses of factors influencing alcohol-related and total fatality rates for 1982 and 1990 provide some support for this suggestion. In all four analyses, A.A. membership rates were significantly and negatively associated with fatality rates. There are several possible mechanisms which might account for this association. Many states have rehabilitative programs for convicted drinking drivers which may involve referral to A.A. (Green et al., 1991; Speiglman, 1994). Thus, A.A. membership (and other rehabilitative measures) are possibly reducing or eliminating drinking (and subsequent drinking-driving crashes) in offenders referred to these programs. Of course, other explanations in which A.A. does not have direct effects on traffic safety cannot be ruled out by the present analyses. Thus, for example, declining traffic fatality rates and increased A.A. membership may both reflect increased health consciousness in the population of a particular area.

The observation that changes in fatality rates between 1982 and 1990 were unrelated to changes in A.A. membership rates moderates any interpretation involving direct effects of A.A. membership rates on fatality rates. If, in fact, higher A.A. membership rates do reduce drinking-driving rates, there are several potential explanations for why this difference in results may have occurred. One possibility is that the effects of consumption (r's = 65 and .52) were of sufficient strength to mask any effects of changes in A.A. membership rates. It may also be that the reduction in traffic crashes involved mainly younger drivers whereas A.A. members are typically in their 40's and 50's. Whether these demographic issues are important would be a matter for further research.

Alcohol abuse treatment levels did not relate to state by state or trend differences in alcohol-related or total fatalities. Earlier research has demonstrated that increased treatment levels did relate to reductions in cirrhosis morbidity rates in Ontario (Mann et al., 1988) and to cirrhosis mortality rates in North Carolina (Holder and Parker, 1992). Again, it may be that fatally injured drivers, especially those in alcohol-related accidents, are often younger than the average alcoholic clinic patient (about age 40-44). It is reasonable to expect that different interventions will have varying impacts at the aggregate level depending on the group most affected. Therefore, demographic factors may moderate the aggregate-level impact of different interventions. For example, most alcohol abuse treatments may not much affect levels of alcohol-related problems most characteristic of young drinkers if their clientele is mainly middle-aged and older.

Our data revealed a positive relationship between enforcement rates and traffic fatality rates. Thus, a beneficial deterrent effect of higher levels of enforcement was not observed; on the contrary, states with higher levels of enforcement had higher fatality rates. Perhaps the simplest interpretation of these findings is that enforcement levels represent a response to drinking-driving and traffic fatality rates. Thus, more drinking drivers on the roads and involved in accidents leads to greater efforts by police to apprehend them. Other research suggests that the type of enforcement activity has a more important deterrent effect than the level of enforcement activity (e.g., Homel, 1990).


We are indebted to Dr. James T. Fell and Bob Williams for their help in obtaining data.


Due to space limitations, references can be obtained from the senior author.