Title Banner - Alcohol and Highway Safety 2001: A Review of the State of Knowledge

4 - Drinking Drivers, Pedestrians 
and Bicyclists


Although this chapter is primarily concerned with drinking drivers, pedestrians, and bicyclists, we also examined some of the literature on drinking by the larger class consisting of all persons of drinking age. An overview of pertinent research is presented in this section.

Apparent Consumption

The National Institute on Alcohol Abuse and Alcoholism (NIAAA) has been compiling data on the apparent per capita consumption12 of ethanol in the U.S. for a number of years. Its latest report (Nephew, Williams, Stinson et al., 1999) shows a long-term decline in total consumption during the period 1981 through 1995, with an indication of a leveling out in 1996 and 1997 (Figure 4-1). In 1981, apparent per capita consumption was 2.76 gallons, but decreased 29% to 2.17 gallons in 1995. This decline appears to be due primarily to a decline in the consumption of spirits: the consumption of wine and beer has exhibited much less change.

Figure 4-1: Apparent Per Capita Consumption of Alcohol 
in the United States, 1935-1997

Figure 4-1: Apparent Per Capita
Consumption of Alcohol in the United States, 1935-1997 

Note that these figures include abstainers in the denominator: per capita consumption capita for the drinking population is considerably higher and varies widely by State. The NIAAA report cited above estimates consumption per drinker to be 3.49 gallons for the lowest State (Iowa), and 9.17 gallons for the highest State (the District of Columbia).

There has been considerable controversy in past years about the relationship of alcohol consumption to alcohol-related crashes. That a relationship exists is certainly plausible: if consumption were reduced to zero, all alcohol-related crashes would be eliminated. And, as can be seen in Figure 4-2 (which combines the data from Figure 2-1 and Figure 3-1), there was a strong correlation between per capita alcohol-related (BAC .10+) crashes and per capita consumption during the 1982-1997 period. However, other factors (e.g., degree of driving after drinking) also influenced the crash rate, so the contribution of consumption alone to alcohol-related crashes cannot be determined from such a simple relationship.

Figure 4-2: Alcohol-Related Crash Fatalities 
per 100,000 Population by per Capita Alcohol Consumption, 
Driver BAC = .10+, 1982-1997

Figure 4-2: Alcohol-Related Crash
Fatalities per 100,000 Population by per Capita Alcohol Consumption, Driver BAC = .10+, 1982-1997

Several studies in the United States have addressed the question of drinking-driving and consumption in considering the effect of taxes on alcoholic beverages. The seminal study was an econometric analysis by Cook (1981), who also reviewed prior studies of the effect of beverage price on consumption. Using historical data on 39 instances of tax changes during 1961-1975, Cook concluded that all highway fatalities could be reduced by increasing the price of distilled liquor and thus decreasing liquor consumption. Our examination of Cook's data indicates a positive correlation between fatal crashes and consumption amounting to a .7% reduction in fatalities for every 1.0% reduction in liquor consumption. However, Cook's regression model did not include other independent variables that might have influenced fatalities.

Skinner (1989) reviewed the Cook study and also performed a separate analysis of differences in 1984 tax rates on distilled liquor, beer, and wine in 31 States. The dependent variable was all highway fatalities per vehicle mile traveled (VMT). Independent variables were: per capita VMT; proportion of urban miles of total VMT; per capita consumption of beer, wine, and distilled liquor; per capita income; per capita homicides; and tax rates per gallon for each beverage type. The effect of consumption on fatalities was not reported, but higher taxes were found not to contribute to lower highway fatalities.

A study by Saffer and Grossman (Saffer and Grossman, 1987) also addressed the problem indirectly in assessing the effect of beer taxes on motor vehicle death rates of young drivers in the 48 contiguous States. The study was reviewed in-depth in Jones and Lacey (1991b). The review found that the study suffered from numerous deficiencies and failed to establish that there is a link between beer taxes and fatality rates. The review concluded that, by not analyzing beer consumption (and beer price) directly, the authors (Saffer and Grossman), took an unwarranted wide leap, bridging the intermediate steps, beer tax/beer price, beer price/beer consumption, and finally, beer consumption/alcohol-related crashes.

We found no recent studies of the effect of alcohol consumption on alcohol-related crashes in the United States, but a study by Voas and Tippetts (1999) of the effectiveness of three alcohol safety laws in the United States included per capita beer consumption in its analysis. The study considered a number of factors related to alcohol-related crashes and found that "states with higher beer consumption had more alcohol-related crashes." (p. 12) However, the analyses leading to this conclusion are not described in sufficient detail to assess them here. This study is discussed further in Chapter 5.

Finally, the 1989 update examined two foreign studies which had found alcohol-related crashes to be positively correlated with alcohol consumption (one in Canada and one in Ireland), and found their conclusions about the relationship crashes and consumption to be questionable13.

Patterns of Consumption

Persons who drink alcohol have been classified by the quantity and frequency of their drinking. Room (2000) recently summarized current developments in characterizing drinking patterns through surveys, concluding that "frequency of drinking at all, and frequency of heavier drinking occasions, are dimensions important both in terms of the social meaning of drinking and of the relation to potential consequences of drinking." A significant percentage of persons of drinking age in the United States are believed to be abstainers who consume no alcohol at all. The actual percentage varies widely among states. Nephew, Williams, Stinson et al. (1999) present data from the Behavioral Risk Factors Surveillance System (BRFSS) on the percentage of abstainers in each State for the period 1986-1997. (The annual BRFSS survey is coordinated by the Centers for Disease Control and Prevention.) In 1997, the percentage of abstainers in the survey ranged from a low of 29.8 percent in Wisconsin to a high of 71.7 percent in Utah. Southern states typically had higher abstinence rates than Northern states, Utah being a notable exception.

Data on abstinence from the 1998 National Household Survey on Drug Abuse (U.S. Department of Health and Human Services, 1999) are generally consistent with State data reported by Nephew and associates, indicating that about 36% of the population of age 12 or more were abstinent during the past year.

Persons who drink have been classified in the traffic safety literature as social drinkers and problem drinkers. The 1978 update defined social drinkers as:

". . . those whose consumption of alcohol is part of their socially defined interactions with family, friends, neighbors and co-workers. . . The health and social functioning of the social drinker are not impaired by this pattern of alcohol consumption. " (p. 55)

This definition appears to be valid today, but attempts have been made to arrive at a more precise definition of problem drinkers than was available circa 197814. As a part of a study conducted in 1990-1993, an expert panel of acknowledged leaders in the field of problem drinking assessment developed a criterion measure for assessing the validity of instruments for the preliminary screening of DWI offenders for alcohol (Lacey, Jones, and Wiliszowski, 1999). The scheme classifies a subject as a problem drinker when that subject exhibits at least one of the following characteristics:

Persons who exhibit two or more of the following characteristics are also categorized as problem drinkers:

The study used these criteria for assessing four screening instruments which were known to be in widespread use and were constructed using sound psychometric procedures. Since the study was mainly concerned with screening DWI offenders, the subjects tested were drawn from two populations of DWI offenders rather than from the general population of drinkers. However, as an additional task, the study validated a combination of the consumption questions from CAGE15 with CAGE, and with the Alcohol Clinical Index (ACI). These questions were suggested by the expert panel as a possible shorthand way of identifying problem drinkers who were also drivers on a telephone questionnaire NHTSA was developing.

The 1997 survey of drinking and driving reported by Balmforth (1998) used the criterion defined by a version of the combined CAGE / ACI items alluded to above to identify problem drinkers in the general population of drinkers. It was concluded that, by the survey's criterion, 17% of all non-abstainers age 16 or older were problem drinkers. Interestingly, Skinner and Holt (1987), authors of the ACI, estimated that some 20% of the drinking age population in North America were problem drinkers, and that abstainers comprised another 15% (Lacey, Jones, and Wiliszowski, 1999).

Combining the results on abstainers from Nephew and associates presented above, and the results on problem drinkers reported by Balmforth, would result in some 5 - 12% of the U.S. drinking age population (depending on the State) being classified today as problem drinkers.

Two other classifications of drinking patterns are more often used in the general scientific literature dealing with drinking and drinking patterns: alcohol abuse and alcohol dependence. These classifications have been published in various editions of the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders, usually abbreviated as DSM-x, where x indicates the edition of the Manual. The criterion assessment instrument developed in the report by Lacey and associates (1999) used several items from DSM-IIIR, the revised edition of DSM-III. In general, DSM-III defines the concept of substance abuse as a pattern of pathological use for at least one month that causes impairment of social or occupational functioning, and defines the concept of dependence as involving either alcohol tolerance or withdrawal, and also the impairment of social or occupational functioning. Nathan (1991) and the U.S. Department of Health and Human Services report (1997) discuss the evolution of the DSM criteria, some of the considerations that led to the development of the current DSM-IV criteria, and some of the differences among various versions of the DSM criteria.

The U.S. Department of Health and Human Services (1997) summarized national survey data prior to 1993, indicating that some 6 - 10% of the respondents were alcohol abusers or alcohol-dependent. These percentages are in the same range found in a telephone survey involving a probability sample of 6,250 respondents in Ontario and Quebec (Cochrane, Goering, and Lancee, 1992). The survey instrument included a psychiatric symptom inventory and questions relating to problem drinking that were adapted from the Diagnostic Interview Schedule. This survey found a rate of 7.1% for problem drinking.

A third classification often used for drinkers is "alcoholic," roughly defined as someone who shows major symptoms of alcohol dependence, including impaired control over drinking, alcohol withdrawal symptoms, and obsessive-compulsive drinking style (Skinner and Holt, 1987). The percentage of alcoholics in the drinking age population is not well known, but was estimated by Skinner and Holt at about 5% of the drinking age population in North America. This percentage is quite close to the percentage of heavy drinkers (persons who consumed five or more drinks on the same occasion on at least five different days in the past month) found in the 1998 National Household Survey on Drug Abuse, namely, 5.9%. Further, the U.S. Department of Health and Human Services (1997) summary of pre-1993 survey data indicates that 3 - 6% of the respondents were alcohol-dependent.

Based on the literature we have examined in this review, we conclude that a precise estimate of the percentage distribution of the drinker types discussed above is not possible. Our own rough estimate of the distribution of drinker types in the U.S. is provided in Table 4-1 below.

Table 4-1: Rough Estimates of the 
Distribution of Four Commonly Used Drinking Types 
in the United Sates at the Millennium
Drinker TypePercent of Drinking
Age Population
Abstainer 35%
Social Drinker 57%
Problem Drinker, Not Alcoholic 4%
Problem Drinker, Alcoholic 4%

The panel convened for the 1981 study of alcohol and public policy (Panel on Alternative Strategies Affecting the Prevention of Alcohol Abuse and Alcoholism, 1981) estimated the alcohol consumption of different drinker types, commenting that it was remarkable how much of the population was completely abstinent or drank very little. The panel also observed that the heaviest-drinking 5% of the of the population accounted for about 50% of total alcohol consumption, and that the heaviest third accounted for over 95% of the total alcohol consumed.

We note that drinking patterns vary widely among different groups of persons. Midanik and Clark (1994) studied the drinking patterns of a number of demographic groups and found that the percentage of males who were current drinkers, weekly drinkers, and drinkers of five or more drinks on at least one occasion at least once a week, were less than the corresponding percentages for women. In addition, the data presented in the study by Midanik and Clark showed that the differences in the percentages between men and women increased with increasing drinking frequency and quantity of drinking. Other studies have found that women are less likely to engage in heavy drinking or to be problem drinkers than are men. The survey reported by Balmforth (1998) found that 74% of problem drinkers were male, and the Canadian survey reported by Cochrane, Goering, and Lancee (1992) found that problem-drinking rates for men are approximately six times greater than for women. Data from the U.S. Department of Health and Human Services report (1997) show that the percentage of males who were alcohol abusers or alcohol dependent was roughly two to three times that of alcohol abusing / dependent females. (The same report gives a lengthy discussion of the drinking patterns of women, concluding, overall, that women drink less than men.)

Age, race, and ethnicity also play a significant role in drinking patterns. With respect to age, data from the1997 survey reported by Balmforth indicate that 21 -29 year and 30 - 45 year age groups have the highest percentage of problem drinkers (by the survey's definition), at roughly 30% for each group. The 46 - 64 group was next at 18%, followed by the 16 -20 group and the 65+ group at 13% and 9%, respectively. (Note that the minimum legal drinking age in all States is 21 years.)

The data from the 1994 Midanik and Clark study cited above show that the 18 - 29 year age group had the largest percentage of heavy drinkers in 1990 (7%), and the 60+ age group had the lowest (1.5%). The other age groups had percentages in the 3 - 4% range. Data from the 1992 National Longitudinal Alcohol Epidemiologic Survey presented by Grant (1994) show that the 18 - 29 age group also had the highest percentage of alcohol abusers (6.5%), and 65+ age group the lowest (0.3%). The percentages of persons classified as alcohol- dependent followed a similar pattern, showing 3.2% of the18 - 29 age group and 0.2% of the 65+ age group to be alcohol-dependent.

Grant (1994) also tabulated data on alcohol abuse and dependence as a function of race, categorized as either Black or non-Black. The data show that the percent of non-Black alcohol abusers is about twice as high as the percent of Black alcohol abusers, but that the percent of alcohol dependence is about the same for the two groups (Figure 4-3).

Figure 4-3: Percent of the Drinking Age Population 
Who Are Alcohol Abusers or Alcohol Dependent by Racial Group

Figure 4-3: Percent of the Drinking
Age Population Who Are Alcohol Abusers or Alcohol Dependent by Racial Group

In a more recent study, Caetano and Clark (1998) examined national trends in alcohol consumption patterns among Whites, Blacks and Hispanics between 1984 and 1995. Data were obtained from two National Alcohol Surveys of U.S. households, the first conducted in 1984 (and used in the above-cited study by Myotonic and Clark), and the second in 1995. The study found that, between 1984 and 1995, the rate of abstention remained stable among Whites but increased among Blacks and Hispanics. It also found that frequent heavy drinking decreased among White men (from 20 percent to 12 percent), but remained stable among Black (15 percent in both surveys) and Hispanic men (17 percent and 18 percent). Frequent heavy drinking was found to have decreased among White women (from 5 percent to 2 percent), but to have remained stable among Black (5 percent in both surveys) and Hispanic women (2 percent and 3 percent). White men and women were two times more likely to be frequent heavy drinkers in 1984 than in 1995. The authors concluded that the reduction in per capita consumption in the U.S. (noted above) is differentially influencing White, Black and Hispanic ethnic groups, and that the stability of rates of frequent heavy drinking places Blacks and Hispanics at a higher risk for problem development than Whites.

The effect of many other demographic variables on drinking patterns is presented in the Myotonic and Clark study, including variables related to socioeconomic status. A recent study by Van Oers, Bongers, Van De Goor et al. (1999) presented results from a survey of 8,000 subjects and noted that lower socioeconomic status (SES) is generally associated with more health problems, and shorter life expectancy. Educational level was used as a surrogate for SES, and the dependent variables were alcohol consumption, alcohol-related problems, and problem drinking. The study found that educational level and abstinence were inversely related in both sexes, and that excessive drinking was more prevalent among men in the group with the lowest educational level. However, it found no significant relationship between educational level and prevalence of excessive drinking in women. No relationship was found between problem drinking prevalence and educational level in either sex. The authors concluded that: there were differences between educational levels with respect to abstinence, differences with respect to excessive drinking were limited, and prevalence of alcohol-related problems in both sexes are higher in lower educational levels after controlling for differences in drinking behavior.

As a final commentary on the effect of demographic variables on drinking patterns, we note the results of an analysis conducted by Midanik and Clark (1995). The analysis involved separate statistical studies of dependence symptoms and social consequences in which demographic variables were used as controls. After all of the demographic variables (11 in all) were taken into account, only younger age (the 18 - 29 age group) was associated with alcohol problems.

There is evidence that age of drinking onset also may be associated with drinking problems later in life. Hingson, Heeren, Levenson et al. (2001) analyzed data from the National Longitudinal Epidemiology Survey to see whether persons who begin drinking at younger ages are more likely to report drunk driving and alcohol-related crash involvement over their lifetime. The study found that, the earlier the age respondents started drinking, the more likely they were to report driving after drinking too much and being in a motor vehicle crash because of their drinking, even after adjusting for current / ever diagnosis of alcohol dependence and other characteristics and behaviors associated with the age respondents started drinking. Even among persons who were never alcohol dependent, those who began drinking in each age group under 21, relative to those starting at age 21 or older, were more likely to report "ever" and "in the past year" being in a crash after drinking too much. Another study of the same data set revealed that this finding also applied in a general sense to unintentional injuries of any kind (Hingson, Heeren, Jamanka et al., 2001).

The above discussion of alcohol abuse is based largely on survey data. However, another means of identifying alcohol abuse, biochemical test abnormality, is receiving increased emphasis in the literature. Several types of such tests, and their value in clinical settings, are described briefly by Rosalki (1999).

As important as drinking patterns are for explaining alcohol problems and dependency, recent research indicates genetic factors may be of equal or greater importance. In fact, the Tenth Special Report to the U. S. Congress on Alcohol and Health (which just became available as this review was being completed) stated:

"Perhaps the single greatest influence on the scope and direction of alcohol research has been the finding that a portion of the vulnerability to alcoholism is genetic. Approximately 50-60 percent of the risk for developing alcoholism is genetic."

Studies leading to this conclusion are discussed in the report (U.S. Department of Health and Human Services, 2000b).

Drinking by college students has been the subject of considerable study in recent years and has generated a fairly sizable amount of literature. Much of the literature deals with so-called binge drinking, now generally defined by researchers as five consecutive drinks for male students and four for females. Wechsler, Molnar, Davenport et al. (1999) used data from the 1993 Harvard School of Public Health College Alcohol Study to describe weekly alcohol consumption and its associated problems among a representative national sample of 17,592 students at 140 colleges. A drink was defined as either a 12 ounce can/bottle of beer, a 4 ounce glass of wine, a 12 ounce bottle/can of wine cooler, or a drink containing 1.25 ounces of liquor. Three categories of drinkers were analyzed: non-binge drinkers, infrequent binge drinkers, and frequent binge drinkers. They found that the median number of drinks consumed per week by all students was 0.7 for those who did not binge drink and 3.7 for those who did so infrequently. Frequent binge drinkers drank a median of 14.5 drinks/week. By these researchers' definitions, nationally, 1 in 5 college students was a frequent binge drinker, and binge drinkers consumed 68% of all the alcohol that students reported drinking. Further, binge drinkers accounted for the majority of reported alcohol-related problems.

A later study by Wechsler and Kuo (2000) of 1999 data found that students themselves defined binge drinking as six drinks in a row for men and five for women, one drink higher than that used by researchers. The students estimated that 35% of all students were binge drinkers by their criterion, and binge drinkers were more likely to overestimate the prevalence of binge drinking. Wechsler, Austin, and Schuckit (1998) examined the validity of the "five/four" criterion, finding that 92% of the college students surveyed who reported five or more alcohol-related problems (e.g., drinking and driving, job problems, and alcohol dependence) in the previous year were identified by the five/four measure as binge drinkers. They concluded that "binge drinking on college campuses is normative, and recognizing this problem and confronting it is an appropriate response."

Brown-Pearson (2000) examined the results of a survey of 2,291 students in an Hispanic-serving university in Texas, and reported a number of interesting findings, as follows:

In another study of college students in Texas, the Texas Commission on Alcohol and Drug Abuse (TCADA), in conjunction with the Public Policy Research Institute (PPRI) of Texas A&M University, conducted a telephone survey of substance use and related behaviors among some 2,420 randomly-selected, full-time undergraduate students in Texas aged 18 to 26 (Kerber and Wallisch, 1999). Though the majority of college students did not misuse alcohol, 29% engaged in binge drinking, and 15% were abusing alcohol at the time of the survey. Being male, over the age of 21, Anglo, Hispanic, and having parents with an annual income of over $60,000 increased the risk for binge drinking. Lifestyle characteristics that increased the odds for binge drinking included receiving low grades, being a fraternity or sorority member, and believing that drinking is a very important part of college life. The misuse of alcohol was associated with several other risky behaviors, including driving while intoxicated, risky sex practices, and problem gambling.

The results of a statewide study in California yielded similar results with respect to the prevalence of college binge drinking (Patrick, Covin, Fulop et al., 1997). It found that 37% of students had binged at least once while drinking, that 25% had driven following consumption of alcohol, and that 32% had ridden in a car with someone who had been drinking.

Dowdall, Crawford, and Wechsler (1998) studied data from a survey of 508 students at women's colleges and 9,624 students at coeducational colleges, finding that women at women's colleges binged less frequently, had fewer alcohol-related problems, experienced fewer negative effects of others' drinking, and were less likely to drink and drive. The researchers hypothesized that self-selection factors at women's colleges may contribute to a healthier environment for women.

Clements (1999) delved more deeply into the drinking problems of 306 college students (74.8% female, 25.2% male). The subjects were administered a number of assessment instruments. Of the total sample, 16% reported that they abstained from alcohol. Of the students who consumed alcohol, the men drank significantly larger quantities than the women, and drank more frequently. White students drank more frequently than Hispanic or African American students. A significantly larger proportion of the men (26%), compared with the women (11%), engaged in binge drinking on a typical day of drinking. Almost one quarter of the sample met the DSM-IV criteria for alcohol abuse or dependence within the past 12 months. The lifetime prevalence for abuse and dependence was an astonishing 35%.

Not all studies have shown such high prevalence of drinking problems among college students. For example, Prince (1999) found that some 9% of 633 students surveyed reported a drinking problem. Seniors reported more problem drinking than any other class. White students reported more problem drinking than the other races, but Hispanic students reported the highest rate of binge drinking.

Data from the 1997 Harvard School of Public Health College Alcohol Study survey also reveal a number of correlates of underage alcohol consumption and related problems (Wechsler, Kuo, Lee et al., 2000). Compared to students aged 21+ years, underage students (< 21 years) engaged in less frequent and less extensive drinking, but consumed more drinks per occasion and had a greater likelihood of drinking in private settings. Correlates of binge drinking overall included residence in a fraternity or sorority, easy access to alcohol, ability to obtain drinks at lower or set prices, and consumption of beer.

Wechsler, Dowdall, Maenner et al. (1998) compared college student binge drinking and related problems in 1993 to those in 1997, again using data from the Harvard School of Public Health College Alcohol Study surveys of students. In 1997, 130 of the original 140 colleges surveyed in 1993 were re-surveyed. The authors found little change in binge drinking. They also found that two out of five students were binge drinkers, that one in five were abstainers, that one in five was a frequent binge drinker, and that four of five residents of fraternities or sororities were binge drinkers. Finally, their data indicated that, among those students who did drink, frequency of drinking, of drunkenness, of drinking to get drunk, and alcohol-related problems, including drinking and driving all increased in 1997 over what they were in 1993.

Finally, Wechsler and associates (2000) compared the results of their most recent (1999) College Alcohol Study by the Harvard School of Public Health with those of the 1993 and 1997 surveys. It was found that 44% of the respondents were binge drinkers in 1999, essentially the same rate as in 1993. Interestingly, both abstention and frequent binge-drinking rates increased significantly, with 19% abstainers and 23% frequent binge drinkers in 1999. Also, binge drinkers still were more likely than other students to experience alcohol-related problems, and students who did not binge drink were at higher risk of the secondhand effects of other students' heavy drinking at those colleges with high binge-drinking rates.

Essentially all of the findings on drinking by college students reviewed here are based on self-reported data, in most instances from surveys completed by the respondents. A program to reduce excessive/binge drinking at the University of North Carolina (UNC), currently being evaluated by NHTSA, measured the actual BACs of students who indicated they were either a binge drinker (by the five/four criterion) or a non-binge drinker. Their survey was conducted to support the development of a program approach based on the nature of college student drinking at UNC, and sought to determine the extent of alcohol use by college students, the amount of drinking-driving by college students, and factors contributing to heavy drinking.

A total of 1,846 randomly selected UNC-Chapel Hill students were interviewed on all nights of the week as the students returned to dorms, fraternity/sorority houses, and off-campus apartments. As a result, BACs were obtained from 1,790 (97%) of the sample, and it was found that only 28% had any measurable alcohol, and only 10% were at .10+. It was also found that:

The self-reported binge drinkers were 49% of those sampled. About 56% of binge drinkers had a zero BAC at the time of the survey compared to 92% of non-binge drinkers, and 18% of binge drinkers and 1% of non-binge drinkers had .10+. The UNC researchers concluded that:

We note again that the research discussed in this section is based largely on self-reported data. The limitations of this kind of data have been discussed elsewhere, including the U.S. Department of Health and Human Services report (1997), and will not be repeated here.


The prior section has discussed some of the characteristics of drinkers in general in the United States. In this section, we examine the characteristics of various groups of drinking drivers. Several categories of variables describing the characteristics are discussed, viz.:

The treatment is an extension of Chapter 2, which was concerned with the overall magnitude of the alcohol-crash problem and discussed the problem in broad terms.

For the most part, these variables are discussed one by one in univariate fashion, but interaction effects are noted in some instances. A comprehensive treatment of interaction effects is not possible here because of the time and space that would be required, and the lack of pertinent literature. The same can be said for multivariate analyses.

Several populations of drinking drivers are examined, including:

Biographical Variables

Age. Prior state of knowledge updates have found that very young drivers and very old drivers were found to drink and drive less frequently, but the very young drivers had a much greater alcohol-crash risk when they did drive. The 1984 update presented data from NHTSA's Fatal Accident Reporting System (subsequently renamed the Fatality Analysis Reporting System or FARS) providing additional confirmation that the youngest drivers tend to have the highest risk of an alcohol-related fatal crash, on the order of 4.5 per 100 million vehicle miles traveled for teen-age drivers, compared to about 1.5 for drivers aged 25-44. Younger drivers were found to be on the road more often during late nighttime hours and on weekends than were older drivers. The 1989 study reported that the most important development since the 1984 update was that the percentage of crashes involving alcohol had declined during the 1980s, and that the share of young drivers in alcohol-related crashes has been disproportionately reduced. The findings of the 1998 review with respect to driver age tended to be consistent with those of prior studies discussed in prior state of knowledge updates, with additional information being provided to confirm and augment prior findings.

Nationwide data on driver age as a factor in alcohol-related fatal crashes are available from the latest FARS report at this writing (U.S. Department of Transportation NHTSA, 1999). These data are plotted in Figure 4-4 and show that the 21-24 year age group had the highest percentage of drivers in alcohol-related fatal crashes, both at .01+ and .10+ (36% and 28%, respectively). The percentages declined steadily from this peak with increasing age, reaching minimums of 3% and 5%, respectively, for the oldest age group, >75 years. Drivers in the under 16 years and 16-20 years age groups (both under the minimum legal drinking age) also had much lower percentages than the 21-24 years age group in both BAC ranges.

Figures 4-4: Drivers in Fatal Crashes by BAC and Age, 1998
Figures 4-4: Drivers in Fatal Crashes
by BAC and Age, 1998

These age groups should not be used in computing the share of fatal alcohol-involved crashes as a function of driver age. This is because the age categories in the figure are not of equal size: the categories for ages 25 through 74 contain 10 years each, while the category for ages 16-20 contains five years, and the category for ages 21-24 contains only four years. We used FARS data for 1998 to calculate the drivers of a given age in years with a given BAC (.01+ and .10+) as a percentage of drivers of all ages in that BAC range. We found that for both BACs, this percentage peaks at age 21 (Figure 4-4), and drops sharply on either side of the peak.

In 1998, the 21-24 years age group contained 6.7% of all licensed drivers (U.S. Department of Transportation FHWA, 1999), compared to the 28% of the same age group of .10+ BAC drivers in fatal crashes. This amounts to an over-representation by a factor of 4.2 (28 / 6.7) of this age group in fatal crashes that are clearly alcohol-related. The degree of over-representation of other age groups is shown in Figure 4-5 for BACs of .01+ and .10+. The sharp peaking of over-representation for the 21-24 years age group is clear. It can also be seen that over-representation is smaller and quite flat for older age groups, but still high for the 16-20 years age group.

Figure 4-5: Overrepresentation of Various Age Groups 
of Drivers in Fatal Alcohol-Related Fatal Crashes, 1998

Figure 4-5: Overrepresentation of
Various Age Groups of Drivers in Fatal Alcohol-Related Fatal Crashes, 1998


Figure 4-6: Drivers of a Given Age in Years 
with or Over a Given BAC as a Percentage of Drivers 
of All Ages with that BAC, Fatal Crashes, 1998

Figure 4-6: Drivers of a Given Age in
Years with or Over a Given BAC as a Percentage of Drivers of All Ages with that BAC, Fatal Crashes, 1998

NHTSA's General Estimates System (GES) provides data on the age of drivers involved in non-fatal crashes which, in the opinion of the investigating officer, involved alcohol. Figure 4-7 shows that the 21-24 years age group had the highest percent of drivers in alcohol-involved injury crashes in 1998 (8%), but that the 25-34 age group had the highest percentage in alcohol-related property damage crashes (5%). Both curves decrease on either side of their peak, but the property damage curve has an increase back to its peak value for the 55-64 age group. The 25-34 years age group accounts for the largest percentage of alcohol-involved non-fatal crashes, but this is probably due to unequal category sizes as indicated in our previous discussion of fatal crashes (Figure 4-8).

Figure 4-7: Drivers in Not Fatal Alcohol-Involved 
Crashes by Age Group, 1998

Figure 4-7: Drivers in Not Fatal
Alcohol-Involved Crashes by Age Group, 1998

Figure 4-8: Drivers in Given Age Groups as a Percentage 
of All Drivers in Alcohol-Involved Non-Fatal Crashes, 1998

Figure 4-8: Drivers in Given Age
Groups as a Percentage of All Drivers in Alcohol-Involved Non-Fatal Crashes, 1998

Studies of drivers injured in traffic crashes and admitted to regional trauma centers in the U.S. and Canada provide information on the age distribution of this special group of drivers (Dischinger and Cowley, 1989; Soderstrom, Dischinger, Ho et al., 1990; Vingilis, Stoduto, Macartney-Filgate et al., 1994). The study by Vingilis and associates is especially interesting, comparing the biographical characteristics of 96 alcohol-negative and alcohol-positive drivers who had been seriously injured in traffic crashes in Ontario, Canada in the 1986-1989 time period, and referred to a regional trauma unit. The mean ages of the two groups were about the same (37.1 years for alcohol-negative group versus 35.0 years for the alcohol-positive group), but the age distribution of the alcohol-positive drivers peaked at a higher age (26-35 years) than that of the alcohol-negative group (18-25 years).

The 1999 FARS report on 1998 fatal crashes includes trends in alcohol-related fatal crashes by age group for the period 1982 through 1998. Figure 4-9 shows drivers at .10+ as a percentage of all fatal-crash involved drivers versus year for various age groups. The trend is clearly down for all of the groups. Figure 4-10 depicts the percentage decrease for each of these groups from 1982 to 1998, indicating that the 16-20 years age group and the 65+ years age group had the largest decreases (~50%), and that the 21-44 years age group had the smallest (~25%-30%). Note that the upward trend to the left of the minimum is reversed at the lowest age group, which also contains the smallest number of drivers in fatal crashes (361 in 1998). (There were a total of 56,543 drivers in fatal crashes in 1998, including 882 of unknown age.)

Figure 4-9: Drivers in Fatal Crashes at a BAC of .10+ 
as a Percentage of All Drivers in Fatal Crashes 
by Age, 1982-1998

Figure 4-9: Drivers in Fatal Crashes
at a BAC of .10+ as a Percentage of All Drivers in Fatal Crashes by Age, 1982-1998

Figure 4-10: Percent Decrease From 1982 to 1998 
in Percent of Drivers with BACs of .10+ 
in Fatal Crashes by Age Group

Figure 4-10: Percent Decrease From
1982 to 1998 in Percent of Drivers with BACs of .10+ in Fatal Crashes by Age Group

As indicated earlier in this report, roadside surveys have measured the BACs of drivers using the roads, but not involved in crashes. The nationwide surveys have been conducted on weekends and during nighttime hours. The age distributions reported for the latest survey (1996) was limited to four groups, <21 years, 21-34 years, 35-44 years, and >45 years (Voas, Wells, Lestina et al., 2000). The 21-34 and the 35-44 groups had an almost identical percentage of drivers at .10+, 3.8% and 3.7%, respectively. Only 0.3% of the youngest age group (<21 years) were at .10+, and the >45 years age group had 1.7% at .10+.

The percentage of non-crash involved drivers at .10+ shows a downward trend for all four age groups since the first survey in 1973 (Figure 4-11). However, only the <21 years age group showed a statistically significant decrease (p <0.05) in percentage at .10+ since the 1986 survey, but that decrease was quite large 88%.

Figure 4-11: Non-Crashed Drivers with a BAC of +.10 
as a Percentage of All Non-Crashed Drivers by Age, 
Nationwide Roadside Surveys 1973-1996

Figure 4-11: Non-Crashed Drivers with
a BAC of +.10 as a Percentage of All Non-Crashed Drivers by Age, Nationwide Roadside Surveys 1973-1996

The surveys cited earlier in this report present self-reported data on drinking driving as a function of driver age (Balmforth, 1998; Townsend, Lane, Dewa et al., 1998). The Balmforth report provides percentages of drivers in various age groups who said they had driven within two hours after drinking any alcohol in the past 30 days, and Townsend and associates give a breakdown of the total number of drivers by age who said they had driven within two hours in the past year after drinking any alcohol. The percentages are shown in Figure 4-12.

Figure 4-12: Age Effects on Driving Within Two Hours 
After Drinking Any Alcohol, Data from Two National Surveys

Figure 4-12: Age Effects on Driving
Within Two Hours After Drinking Any Alcohol, Data from Two National Surveys

The figure shows that the 25 years age group is most likely to drive after drinking within two hours, and that the youngest and the oldest drivers are the least likely. However, the 45 years age group had the highest percentage of all persons who said they drove within two hours after drinking.

Zador and associates (2000) considered the effect of driver age in their analysis of fatal crash risk, but provided relative risk curves only for three age groups, by driver sex. The age groups were 16-20, 21-34, and 35+ years. The relative risk curves for the 16-20 years group are shown in Figure 4-13. The curves for the other two age groups are the same as the curve for 16-20 years females, except at .001-.010 BAC. At .001-.010 BAC the relative risk of the 21-34 and 35+ years age groups for both sexes drops to .18, compared to 1.22 for the females of age 16-20 years.

Figure 4-13: Relative Risk of a Fatal Crash 
for Males and Females Age 16-20, 
Data from Zador and Associates (2000)

Figure 4-13: Relative Risk of a Fatal
Crash for Males and Females Age 16-20, Data from Zador and Associates (2000)

Waller (1998) reviewed research on the effects of alcohol and aging on driving performance and found that, after about age 55, crash risk per mile driven begins to increase, and continues to increase with age at an accelerating rate until it may exceed that of young beginning drivers for drivers in their eighties. Alcohol was found to further increase the crash risk of older drivers.

Sex. Prior state of knowledge updates have found that drinking-drivers are predominately male, but that female drivers nevertheless comprise a significant percentage of drivers in alcohol-related crashes. Further, recent updates have found that the role of female drivers in alcohol-related crashes is increasing.

FARS data for 1998 show that 20% of males involved in fatal crashes had a BAC of .10+, compared to 10% of females (Figure 4-14). However, the percentages for males and females were about the same (4%) at lower BACs.

Figure 4-14: Drivers in Fatal Crashes by BAC and Sex, 1998
Figure 4-14: Drivers in Fatal Crashes
by BAC and Sex, 1998

Males drivers constitute a much larger portion of the alcohol-crash problem than female drivers16. In 1998, about 84% of the drivers at .10+ in fatal crashes were male, but the percentage of female drivers continues to increase slowly, from 13% in 1982 to 16% in 1998 (Figure 4-15).

Figure 4-15: Female Drivers as a Percent of All Drivers 
in Fatal Crashes at BAC=.10+, 1982-1998

Figure 4-15: Female Drivers as a
Percent of All Drivers in Fatal Crashes at BAC=.10+, 1982-1998

Roadside survey data indicate that the percentage of male non-crashed drivers at .10+ has declined since the first survey in 1973, but that the decline decreased since 1986 Figure 4-16. The percentage of female drivers at .10+ declined by about 50% between the first two surveys, but then increased slightly in 1996.

Figure 4-16: Non-Crashed Drivers with a BAC of .10+ 
as a Percentage of All Non-Crashed Drivers by Driver Sex, 
Nationwide Roadside Surveys 1973-1996

Figure 4-16: Non-Crashed Drivers with
a BAC of .10+ as a Percentage of All Non-Crashed Drivers by Driver Sex, Nationwide Roadside Surveys 1973-1996

The survey data reported by Balmforth indicate that 36% of the males and 13% of females had driven within two hours after drinking in the past year. Data from the survey reported by Townsend and associates show that 68% of those who driven within two hours after drinking in the past year were male, compared to 32% who were female.

Prior state of knowledge updates have indicated that females may have a higher alcohol-related crash relative risk than males, but the recent analysis of FARS data and roadside survey data by Zador and associates shows no such effect for drivers of age 21 years and higher. In fact, their data show that young males (i.e., age 16-20 years) have a higher relative risk than young females at all BACs (See Figure 4-13).

Race and Ethnicity. The 1998 update observed that the role of race in the alcohol-crash problem has rarely been addressed in prior research, noting that the 1978 update found some suggestion of a race effect, but could not separate the effect of race from the effect of socio-economic status. Indeed, it has been argued that race / ethnicity is not a valid attribute for describing most human behaviors, since it implies some kind of inborn genetic characteristic rather than a human condition or life event such as education, employment status, income, and drinking patterns. Nevertheless, racial and ethnic groups have been found to differ with respect to many variables that describe their condition, and race / ethnicity are widely used as independent variables in the scientific literature on many disciplines.

FARS does not contain data on driver race and ethnicity, but some States include race as a variable in their crash files. We examined one such State's crash files (Florida) to calculate the number of crash-involved drivers in various racial groups as a percentage of all had-been-drinking (HBD) crash-involved drivers for the period 1993-1998. We found that: drivers classified as "White" accounted for about 80% of the HBD drivers, "Black" drivers for about 10%, "Hispanic" drivers for about 9%, and "Other" drivers for about 1%. These percentages have varied slightly over the period reported, with the percentage of Whites decreasing slowly from 82% to 78%, and the percentages for the non-White groups increasing slowly (Figure 4-17). The Hispanic group showed the largest increase, from 7% in 1993 to 10% in 1998.

Figure 4-17: Had-Been-Drinking Drivers 
as a Percentage of All Drivers in Given Racial/Ethnic Groups 
in Crashes in Florida, 1993-1998

Figure 4-17: Had-Been-Drinking
Drivers as a Percentage of All Drivers in Given Racial/Ethnic Groups in Crashes in Florida, 1993-1998

Whites also had the largest percentage of HBD drivers in crashes of any group, followed by Hispanics, Blacks, and others in that order (Figure 4-18). The percentages declined for all groups over the 1993-1998 period, with Hispanics showing the largest decrease over the period.

Figure 4-18: Drivers in Various Racial/Ethnic Groups 
as a Percentage of All Had-Been-Drinking Drivers 
in Crashes in Florida, 1993-1998

Figure 4-18: Drivers in Various
Racial/Ethnic Groups as a Percentage of All Had-Been-Drinking Drivers in Crashes in Florida, 1993-1998

Further insights on the role of alcohol in crashes have been gained by matching FARS data with death certificate data collected by the National Center for Health Statistics. Voas and associates (2000) recently compiled such data for the period 1990-1994, finding that "Caucasian Americans" accounted for about 72% of alcohol-related fatalities, "African Americans" for about 12%, various Hispanics groups for about 13%, and other ethnic groups (including Native Americans) for about 3%. Note that an alcohol-related fatality was one that occurred in a crash in which a driver, pedestrian, or cyclist had a BAC of .01+. These percentages are quite close to those for the HBD drivers only in Florida, with the exception of the percentage of "Other" ethnic groups (Table 4-2). Native Americans have been found to be highly over-represented in pedestrian crashes (discussed later in this report), which may help account for the larger percent of other ethnic groups in the study by Voas and associates.

Table 4-2: Racial/Ethnic Composition
of Two Alcohol-Related Crash Populations
Florida, HBD Drivers in All Crashes 80% 10%9%1%
FARS / NCHS, Fatalities in Crashes Involving a .01+ Driver, Pedestrian, or Cyclist72%12%13%3%

The study by Voas and Associates also examined fatality rates of different ethnic groups and sub-groups, finding that Native Americans had the highest rates (68%), and that Asian-Pacific Islander Americans had the lowest (28%). The rates for Caucasian Americans and African Americans were nearly the same (~45%). Among Hispanics, Mexican Americans had the highest rates (55%), and Cuban Americans the lowest (37%). Males had higher driver fatality rates than females for all ethnic groups, with smallest differences occurring for Native Americans and Cuban Americans. The 21-40 years age group had the highest alcohol-related driver fatality rates for all ethnic groups, 80% for Native Americans and 70% for Mexican Americans.

Roadside surveys and other surveys provide some recent information on the racial / ethnic characteristics of drivers not necessarily involved in crashes. The 1996 nationwide roadside survey of non-crash involved drivers (Voas, Wells, Lestina et al., 2000) found that about 9% of African Americans had a BAC of .05+ and that about 4% were at .10+. For Hispanics, the percentages were 15% and 8%, respectively. Further, the percentages for African Americans declined from the two prior surveys, while the percentages for Hispanics increased.

The percentages for Whites were not presented, but the odds of a BAC of .05+ for African Americans and for Hispanics relative to the odds for Whites were computed using a logistic regression model that controlled for age, sex, weekend day, and time period (late or early). The results show that, in 1996, the odds for African Americans were about the same as the odds for Whites, but the odds for Hispanics were 1.7 times the odds for Whites. Also, the odds ratio for African Americans / Whites decreased compared to the first two surveys, while the odds ratio for Hispanics / Whites increased.

Quite different results for Blacks and Hispanics compared to Whites were obtained in the 1996 survey reported by Townsend, Lane, Dewa et al. (1998). Here, the measure was "drove within two hours after use of alcohol alone in the past year." The odds ratios were 0.5 and 0.6 for African Americans / Whites and Hispanics / Whites, respectively. We computed these odds ratios directly from the data in the report (Table 3a) which did not control for the effects of other variables such as age and sex.

The survey reported by Balmforth (1998) did not present any results for different ethnic groups. However, a recent study of results from NHTSA's 1997 survey (the same survey summarized by Balmforth), pooled with the results of two prior surveys conducted in 1993 and 1995, examined the ethnicity factor in some depth (Royal, 2000). Again, the primary measure for drinking-driving was "drove within two hours after use of alcohol alone in the past year." The self-reported prevalence for drinking-driving was 28% for Whites, 21% for American Indians / Eskimos, 17% for Hispanics, 16% for Blacks, and 13% for Asians. Also reported was the percentage of crash-involved drivers who had consumed alcohol within two hours prior to the crash. The figure was 3% for White, non-Hispanic drivers, 7% for Black non-Hispanics, and 8% for Hispanics.

No recent studies of the effect of race or ethnicity on the alcohol-crash relative risk were found in this review. However, raw data from the Grand Rapids study (Borkenstein, Crowther, Shumate et al., 1964) suggest that there is little difference in the relative risk of Whites and non-Whites for crashes of all severities (Figure 4-19). The calculated relative risks do not control for any other variables that may have affected relative risk.

Figure 4-19: Relative Risk of a Crash by Race 
at Given BACs - Grand Rapids Study

Figure 4-19: Relative Risk of a Crash
by Race at Given BACs - Grand Rapids Study

Other Biographical Variables. Other biographical characteristics of drinking-drivers that have been studied include marital status, employment status, education, and annual income. Little new knowledge has been gained in recent years on the role of these and other biographical variables in drinking-drivers alcohol-crash involvement, amount of drinking driving, or alcohol crash risk. The 1998 update observed:

"The 1978 update provided considerable data on the effects of other biographical variables on alcohol-crash risk and incidence, but little new data on this subject were reported in the 1984 and 1989 updates." (p. 19)

However, the 1998 update cited a study of seriously injured drivers in Ontario, Canada (Vingilis, Stoduto, Macartney-Filgate et al., 1994) which revealed large differences between employed and unemployed drivers with respect to BAC level -- 20% of employed drivers were alcohol-positive versus 67% of unemployed drivers. This difference is highly significant despite the small sample sizes involved in the study. The 1996 household survey reported by Townsend and associates (1998) found the opposite to be true for persons who "drove within two hours after use of alcohol alone in the past year." Here, 29% of employed drivers reported driving under these conditions, compared to only 21% of unemployed persons.

The Ontario data also indicated significant differences among groups of seriously injured drivers differentiated by income, with the under $15,000 per year group and the $30,000-$65,000 per year group having the lowest percentage of alcohol-positive drivers (about 10%), and the $15,000-$30,000 group having the highest percentage of alcohol-positive drivers (about 44%). This finding was interesting in light of the findings of earlier roadside survey research reported in the 1978 update which showed that percentage of low income on-the-road drivers increased steadily with increasing BAC. Again, the 1996 household survey indicated a different relationship, with self-reported drinking-driving increasing steadily with income, from 11% at incomes of less than $10,000 to 64% at incomes of $75,000 per year.

With respect to education, research reported in prior updates indicated that drivers with a relatively small number of years of formal education were over-represented, both among higher-BAC crash-involved drivers and among higher-BAC drivers who had not crashed. The 1986 National Roadside Breathtesting Survey (Lund and Wolfe, 1989) reinforces the result for non-crash involved drivers, with 4.1% of the drivers with no more than a high school education having a BAC .10+ compared to 2.2% of drivers with a bachelor's degree or more. By contrast, the 1996 household survey found that drinking-driving increased steadily with formal education, from 11% for persons with less than a high school education to 35% for college graduates.

We note that the above household survey results were for persons who reported driving within two hours after any drinking, and so may be biased in favor of persons who drank very little before driving. This suggests totally different relationships with the various biographical than for persons who drank more before driving.

FARS contains no information on biographical variables other than age and sex, nor do state databases on crashes of all types. The report on the 1996 roadside survey also contained no data on "other biographical variables."

Drinking Variables

Literature on drinking variables associated with drinking drivers has most often been concerned with alcoholic beverage preference, quantity and frequency of drinking, prevalence of drinking problems, and drinking location. Many of the studies have dealt with subjects who had been convicted of DWI and referred for alcohol-problem assessment or alcohol-problem treatment.

With respect to alcoholic beverage preference, studies dating back to the 1960s consistently have shown beer to be favored by a wide margin (Jones and Joscelyn, 1978; Jones and Lacey, 1989; Jones and Lacey, 1998a). Also, prior updates suggest that the drinking location preceding illegal drinking-driving was most frequently bars or taverns and other person's homes. The 1998 update noted a study of New York State DWIs by Wieczorek, Miller, and Nochajski (1992b) that found that subjects who drank at more than one location engaged in DWI more often than did subjects who drank at only one location. The multiple-location DWIs had an average of 4.00 self-reported drunk driving events in the past 30 days compared to 2.46 for the single-location DWIs (p= 0.0009). In addition, the multi-location drinkers drove almost twice as far prior to their arrest (an average of 11.5 miles for multi-location versus 6.7 miles for single-location, p<0.0001 ), thereby exposing other drivers and their passengers to a greater crash risk. These two groups also differed significantly on a number of drinking variables, with the multi-location drinkers indicating patterns of heavier drinking and stronger evidence of alcohol problems and alcohol dependency.

The 1998 update also cited a study of Vermont university students by Musty and Perrine (1990) that found that some 70% of those under the legal drinking age reported drinking most frequently at private locations (for example, home, other person's homes, and clubs), and that about 75% of students above the legal drinking age reported drinking at public bars and restaurants.

Another study cited in 1998 update (Lang and Stockwell, 1991) examined the effect of type of drinking location on crash involvement in Australia. This study involved subjects in Perth, Australia who had been arrested for DWI, either as a result of being involved in a crash (n=257) or having failed a roadside sobriety test (n=1 ,909). Two types of drinking locations were considered in the study, "unlicensed" locations (for example, private residences or public places such as parks) and "licensed" locations. The study found that the prior drinking location of the drivers whose DWI arrest occurred after a crash was more likely to be an unlicensed location than it was for drivers whose DWI arrest occurred after some other event or activity such as speeding or reckless driving (p=0.001). Thirteen percent of the drivers who drank in an unlicensed location were arrested after a crash compared to 8% of the drivers who drank in a licensed location. Note that these figures reflect the conditional probability of an arrest given a crash, not the unconditional probability of crash.

Since the 1998 update, additional literature has appeared considering the interaction of drinking location, beverage type, and other factors, including biographical factors. Especially noteworthy is a study by Gruenewald and associates (2000). The study sought to determine whether the association of beer drinking with drinking and driving is due to cultural norms or is an artifact arising from the demographic profile of beer drinkers (young and male), the drinking patterns of this sub-population (frequent and heavy), and the venues in which they prefer to drink (bars and restaurants). The data were obtained from a carefully designed telephone survey in six U.S. communities involving a test sample that included 2,275 drinkers, 985 of whom had driven after drinking.

The study found that frequent consumers were more likely to drink outside the home, preferred beer and spirits to wine, and were more likely than others to drink and drive. Most important, beverage preferences were not directly associated with drinking and driving. Beer drinkers, however, were from the sub-population most likely to drink and drive, heavier drinking younger men, who prefer ro drink at bars and restaurants. The results suggested that the association of beer consumption with drinking-driving arises from the circumstances in which the subpopulation of beer drinkers more commonly find themselves (as a result of their efforts to maximize, within economic constraints, the social and amenity value of drinking), as opposed to any culturally induced disposition beer drinkers may have to drink and drive.

Three recent studies in New Mexico have examined the the role of drinking location in drinking-driving. In the first study, Chang, Lapham , and Barton (1996) compiled the sociodemographic characteristics of 5,154 mostly male DWI offenders) referred for alcohol-related assessment. They found that some groups showed a higher rate of DWI convictions, compared to the adult county population, that is: young, single male; Hispanic and Mexican National; and divorced/separated/widowed (increasing with age). They also found that older, educated or employed offenders reported drinking more in bars or lounges, while younger offenders were more likely to drink in private parties. Further, Hispanic and Mexican National men showed equal likelihood of drinking with friends and relatives in bars or lounges, whereas non-Hispanic White males reported drinking more with friends. Native Americans were associated with higher blood alcohol concentrations (BAC), and with alcohol-related problems.

Lapham, Skipper, Chang et al. (1998) estimated the distance driven between drinking and arrest locations among 3,107 New Mexico offenders convicted of DWI and determined whether the drinking location, the driver's appearance (factors such as race, age, gender), or age of the vehicle accounted for any differences in the estimated distance driven. Among those who were not arrested in the immediate vicinity, the number of miles driven ranged from 0.5 to 18.2, with a mean of 3.4 miles (median= 2.6). Further, persons who drank in a high or medium-high arrest intensity area, those with BACs of .20+, and those drinking at bars, restaurants, or private parties, drove fewer miles compared to other offenders. The authors also found that factors such as age, gender, and vehicle age were unrelated to how far drunk drivers travel before their arrests.

In the third study of New Mexico DWIs, Lewis, Lapham, and Skipper (1998) examined alcohol purchase locations of convicted drunk drivers to determine the characteristics and arrest circumstances of offenders who bought alcohol at a drive-up liquor window compared with those who obtained alcohol elsewhere. The study results revealed that drive-up windows were the preferred place of purchase of package liquor by offenders who bought the alcohol that they drank prior to arrest. The odds of being Hispanic (p<0.0001), a high-risk problem drinker (p<0.01), and drinking in the vehicle prior to arrest (p<0.01) were significantly higher for drive-up window users than for offenders who purchased package liquor elsewhere. The authors concluded that a statistically significant relationship exists between the use of drive-up windows and "certain high-risk drinking behaviors."

Jones-Webb, Toomey, Short et al. (1997) examined relationships among perceived alcohol availability, drinking location, alcohol consumption, and drinking problems. Their subjects were 3,372 adolescent drinkers, ages 16-18, who participated in the authors' Communities Mobilizing for Change on Alcohol Project baseline survey (See page 132). The authors found that perceived alcohol availability was significantly associated with higher levels of alcohol consumption for males. Drinking in a public location, such as a bar, restaurant, or party, was marginally associated with higher levels of alcohol consumption for females.

Greenfield and Rogers (1999) examined effects of measurement on risk curve analysis in an application involving prediction of frequency and indicator measurements of drunk driving with beverage-specific alcohol consumption and risk perception measures. Again, self-reported data were used, this time from a 1995 household survey in which the responses of 1,260 adult drivers who reported any drinking in the prior year were selected for analysis. Relationships between drinking pattern, beverage choice, perception of risks and drinking before driving, and frequency of drunk driving, were investigated. Self-reported drunk driving (occurrence) was measured by a question assessing driving after drinking enough to be in trouble if stopped by the police within the prior 12 months, and those affirming this (n = 191) were asked how many times they did so.

It was found that, after controlling for demographics, heavy beer consumption (p < .01) more than heavy wine (not significant) or liquor / spirits (p < .05) intake was strongly predictive of risk perception. A significant interaction was found between heavy beer consumption and perceived risk (p < .001) in predicting reported frequency of drunk driving. The authors concluded that individuals' underestimation of beer's intoxicating effects, compared to other alcoholic beverage types, helps explain beer's over-representation in drinking-driving violation reports.

The same authors (Rogers and Greenfield, 1999) generalized their findings on beer drinking to "hazardous drinking" in general, which they defined as occasions in which five or more drinks were consumed in a day. Their data came from a survey involving 2,817 respondents who had consumed at least one drink in the previous year. The results showed that, in the U.S., beer accounts for the bulk of alcohol consumed by the heaviest drinkers and that beer also accounts for a disproportionate share of hazardous drinking. Logistic regression analysis revealed that drinkers who consumed beer in a hazardous fashion at least monthly are more likely to be young, male, and unmarried and less likely to be Black than are other drinkers. Hazardous beer consumption was found to be more predictive of alcohol-related problems than hazardous consumption of wine or spirits.

A wide range of epidemiologic and clinical studies have clearly indicated the over-involvement of persons with drinking problems in drinking and driving incidents, including alcohol-related traffic crashes as well as arrests and convictions for DWI. Many of these studies are reviewed in prior updates.

Several recent studies have examined the relationship between BAC at arrest and alcohol problems. Lapham, Chang, Skipper et al. (2000) studied the association between arrest blood alcohol concentration (BAC) and alcohol use disorders among convicted drunk driving offenders in New Mexico. Subjects were interviewed using the Diagnostic Interview Schedule, discussed in Chapter 2. If the age at onset of alcohol disorders was the same as or younger than the age at screening, the person was classified as having alcohol abuse or dependence at the time of screening. Arrest BAC ranged from .01 to .45 (mean = .156). Alcohol dependence at screening was found for 58% of offenders with BAC <.15, 66% of offenders with BAC .15 to .19, and 72% of offenders with BACs of .20 or above (p<.001). The overall accuracy of BAC of .15 or higher and .20 and higher as a screening test for alcohol dependence ranged from .45 to .64. The authors concluded that, although arrest BAC is associated with alcohol use disorders, it provides limited utility as an objective indicator of alcohol dependence.

Bergman, Hubicka, Laurell et al. (2000) assessed 1,600 Swedish drivers suspected of DUI, and a control group of 785 drivers not suspected of DUI using the Alcohol Use Disorders Identification Test (AUDIT) instrument. The authors found that hazardous or harmful alcohol use according to the AUDIT was four times as common among the male, and ten times among the female, suspected DUIs, as compared to the control drivers. Further, more than half (58%) of the suspected DUIs had such drinking problems and 18% had severe problems. Also, almost half (46%) of the suspects with a BAC below the Swedish illegal limit of .02 had such problems. This lead the authors to conclude that BAC level has low sensitivity and specificity as a means for identifying drinking problems in suspected DUIs.

Wieczorek, Miller, and Nochajski (1992a) had arrived at similar conclusions after examining the relationship between BAC at arrest and a number of variables related to drinking patterns and alcohol abuse or dependence17. Their sample consisted of 235 persons convicted of DWI and referred to a drinking driver treatment program in Erie County, New York. BACs at arrest were available for all of these subjects, and only those with a BAC greater than .05 were included in the study. Data were collected through face-to-face interviews, clinical evaluations, and the complete Mortimer-Filkins test. None of the results indicated any significant relationship between BAC at arrest and diagnoses of alcohol problems. Significance levels were 0.40 for all but one of the tabulated analyses, the clinical analysis (p>0.10), which indicated that the high-BAC group (BAC > .15) had slightly fewer non-critical alcohol problems.

However, Lucker and Gold (1995) obtained different results in a study of the association between breath alcohol concentration (BAC) at arrest and problem drinking for a sample of 1,283 male DWI offenders in the U.S. Army. The results indicated a moderate but statistically significant association between BAC at arrest and DSM-III diagnosis. BAC's ability to indicate problem drinking was also compared with the diagnostic ability of three well-known, paper-and-pencil instruments designed for that purpose. BAC performed as well in identifying problems with alcohol as did the MAST, the MacAndrew Scale of the MMPI, and the Vaillant.

Hutchinson (2000) conducted in depth interviews with a sample of 48 male convicted drink drivers, all drinking 50+ units per week. An analysis of their drinking patterns and styles showed that there is a wide range of drinking patterns and styles and no consistency. He found that drinking patterns vary over the years, from occasional to daily drinking, and are altered by major lifestyle changes. Likewise drinking styles had no consistency but there was a concentration on separating drinking from work-related activities. The quantity consumed on any or each drinking occasion bore no relation to either pattern or style of drinking. The author concluded that the only controlling device is the concept of "alcoholic," and that there is no safe or sensible limits as such.

Driving Variables

The driving variables associated with drinking drivers most often examined in the literature have been the time of day and day of the week of drinking-driving, types of vehicles driven, number of drinking driving trips per unit time, and enforcement actions experienced as a result of their drinking driving.

Research overwhelmingly supports the intuitively obvious conclusion that drinking driving and alcohol-related crashes occur most often when most people engage in recreational drinking during the nighttime and on weekends. FARS data for 1998 (U.S. Department of Transportation NHTSA, 1999) provide further confirmation of alcohol-related (BAC > .01) crashes as a predominately nighttime-weekend phenomenon. Figure 4-20 shows that the midnight to 3 a.m. period had the highest percentage of alcohol-related fatal crashes (76%), and that the three-hour periods immediately before and after the midnight to 3 a.m. period also had high percentages (62% and 69%, respectively). By contrast, the morning and early afternoon hours of 6 a.m. to 3 p.m. had the lowest percentages (in the 11% to 15% range), with the percentage starting to rise again in the late afternoon to early evening periods (24% to 45%).

Figure 4-20: Percent of Fatal Crashes that were 
Alcohol-Related by Time Period, FARS Data for 1998

Figure 4-20: Percent of Fatal Crashes
that were Alcohol-Related by Time Period, FARS Data for 1998

The late nighttime to early morning hours also contained the largest percentages of all alcohol-related fatal crashes (Figure 4-21). Nearly 80% of all alcohol-related fatal crashes occurred during the 6 p.m. to 6 a.m. period.

Figure 4-21: Alcohol-Related Fatal Crashes 
in Given Time Periods as a Percent of All 
Alcohol-Related Fatal Crashes, FARS Data for 1998

Figure 4-21: Alcohol-Related Fatal
Crashes in Given Time Periods as a Percent of All 
Alcohol-Related Fatal Crashes, FARS Data for 1998

With respect to day of the week, the FARS report of 1998 data also contains information on the percentage of fatally injured drivers in alcohol-related fatal crashes by time of day (daytime or nighttime), day of week (weekday or weekend day), driver age (<21 years or 21+ years), and number of vehicles involved in the crash. The data for single-vehicle crashes are plotted in Figure 4-22 and show that the highest percentages for both age groups occur during nighttime-weekend hours -- 77% for 21+ years drivers and 57% for <21 years drivers. These percentages fall to 22% and 7%, respectively, during daytime-weekday hours.

Figure 4-22: Percent of Fatally Injured Drivers in 
Single-Vehicle, Alcohol-Related Crashes by Driver Age, 
Time of Day, and Day of Week, FARS Data for 1998

Figure 4-22: Percent of Fatally
Injured Drivers in Single-Vehicle, Alcohol-Related Crashes by Driver Age, Time of Day, and Day of Week, FARS Data for 1998

Wilson and Chen (2000) developed a model for predicting the BAC of nighttime drivers, identifying several biographical variables (age, sex, and educational level) and situational variables (coming from a bar, and a group of passengers of the same sex as the driver) that differentiate nighttime drivers on the basis of their BAC. However, the model left much of the variance unexplained and had a high rate of false positives.

The 1998 update observed that prior updates had not found much useful information on the effect of vehicle type on the alcohol-crash problem, but presented 1996 FARS data on the types of vehicles that were involved in fatal crashes. The data showed that drivers of motorcycles most often had been drinking (42% alcohol-positive and 30% >.10), and drivers of large trucks least often had been drinking (3% alcohol-positive and 1% >.10). Drivers of light trucks were slightly higher than passenger cars for both measures (28% alcohol-positive and 22% > .10). Later data for alcohol-positive drivers in 1998 are shown in Figure 4-23 and indicate small reductions since 1996 for all of the vehicle types shown.

Figure 4-23: Percentage of Alcohol-Positive Drivers 
in Fatal Crashes by Vehicle Type, FARS Data for 1996 and 1998

Figure 4-23: Percentage of
Alcohol-Positive Drivers in Fatal Crashes by Vehicle Type, FARS Data for 1996 and 1998

Using FARS data for 1992, Preusser, Williams, and Ulmer (1995) developed a program for generating computerized "crash reports" for crashes that were fatal to a motorcycle driver. The reports were then analyzed to define types of motorcycle crashes. Five such types accounted for 86% of the crashes: ran-off-road (41%), ran traffic control (18%), oncoming or head-on (11%), and motorcyclist down (7%). Alcohol and excessive speed were commonly associated with motorcycle crash involvement.

A paper by Sun, Kahn, and Swan (1997) reported a test of the hypothesis that a motorcyclist is more likely to be compromised by alcohol than the driver of an automobile. During 1992, they measured the BACs of 40 male drivers of motorcycles and 411 drivers (68% male) of four-wheeled vehicles who were admitted to a university hospital trauma center. They also measured the Glascow Coma Scale (GCS), Revised Trauma Score (RTS), and Injury Severity Score (ISS) for each patient. The mean age was 32 years for the motorcyclists and 35 years for the automobile drivers. Among the motorcyclists, 94 percent wore helmets, and among the automobile drivers, 26 used seat belts. The authors found that one-third of the motorcyclists had a positive BAC, averaging .12, whereas a positive BAC was found in 35 percent of the automobile drivers, averaging .18. The difference in BAC between the two groups was statistically significant (p < 0.05), and no significant differences in drug use were found between the two groups. The authors concluded that among comparable accident victims, motorcycle drivers have lower blood alcohol concentrations than do drivers of automobiles.

Several Australian researchers have studied the role of alcohol in motorcycle crashes. Holubowycz and Mclean (1995) interviewed a sample of 302 male drivers and motorcycle riders admitted to the Royal Adelaide Hospital in Adelaide, South Australia between June 1985 and April 1987. They found that with one exception, the likelihood of having a BAC > .08 did not differ with demographic profile. Also, as BAC increased, there was a significant increase in: various indices of quantity and frequency of drinking; beer being the preferred beverage; percentages drinking alone, in a hotel, in a vehicle and for various less socially acceptable reasons; frequency of drink-driving; likelihood of previous drink-driving suspension; and, more liberal attitudes towards drink-driving. About 25% of those with a BAC of at least .15 were believed to experiencing alcohol-related problems prior to the crash, compared with only a very small proportion of those with lower BACs. Pre-crash drinking was found to be most commonly involved drinking in a hotel, drinking with friends and drinking beer, with no significant differences between BAC groups. The authors concluded that the results suggested that usual drinking and drink-driving patterns, as well as attitudes to drink- driving, become more extreme as the BAC of male crash-involved drivers and riders increases.

Haworth (2000) conducted an interesting case-control study of injured motorcyclists and passengers in Melbourne, Australia using data from 222 crashes and from 1,200 motorcyclists riding through the crash sites at the same time of day and week. He found that a BAC greater than zero was associated with a five-fold increase in the odds of crashing compared to having a BAC of zero. A BAC greater than .05 was associated with about a 40-fold increase in risk. Haworth also found that nighttime crashes were more likely to involve alcohol.

The report on the 1997 National Survey of Drinking and Driving (Balmforth, 1998) contains data on the mean number of drinking-driving trips by drivers who drove within two hours after drinking during (1) the past 30 days and (2) the past year. The data are presented by driver gender and age group and are shown for trips in the last 30 days in Figure 4-24. The mean number of trips for all drinker-drivers (n=964) was 1.7, but the number for males was nearly three times that for females (2.1 versus 0.8). With respect to driver age, drivers in the three middle age groups averaged about 1.7 trips, but the lowest age group had a lower number of trips (1.2), and the highest age group averaged more trips (2.2).

Figure 4-24: Mean Number of Drinking-Driving Trips 
in Past 30 Days by Driver Age and Sex

Figure 4-24: Mean Number of
Drinking-Driving Trips in Past 30 Days by Driver Age and Sex

Although this chapter is primarily concerned with drinking users of the U.S. Highway Transportation System, we also examined some of the literature on drinking by the larger class consisting of all persons of drinking age. Our main interest was alcohol beverage consumption and its relationship to traffic crashes.

Drivers who have been subject to law enforcement actions have been subject to considerable scrutiny by researchers. This is particularly true for drivers who have been convicted of DWI or some related offense such as breath-test refusal. Two obvious reasons are: (1) such drivers clearly have engaged in hazardous drinking-driving behavior and (2) they comprise a convenient group for study as a result of the conditions of their conviction, including assessment of alcohol problems and follow-up during probation.

Much the material examined in our recent review of the scientific literature on repeat DWI offenders (Jones and Lacey, 2000) also dealt with DWI offenders who were not repeat offenders. We reported that FARS data combined with data from the U.S. Department of Justice indicated that drivers with one or more DWIs are over-represented among fatal-crash involved drivers, but that the degree of over-representation had declined from 1.8 in 1988 to 1.4 in 1997. It was estimated that in 1997, roughly 2.2% of all fatal crashes (810) involved a driver with one or more DWI convictions.

However, an analysis of Minnesota data by Ross, Simon, and Cleary et al. (1995) indicates a much higher percentage of repeat DWI offenders in fatal crashes in that state. These researchers defined an administrative license revocation as a prior alcohol-related driving incident, and found that such drivers had an involvement rate of 34% in fatal crashes. The reason for this large discrepancy could be at least in part that the FARS definition of a prior offense is a court conviction, rather than an administrative action. Judicial processes provide more opportunities for a non-conviction (plea bargaining to a lesser offense, for example) than administrative processes, and court records are often less accurate than the records of an administrative agency.

The repeat offenders review cited two studies in California that also involved large data bases. In the first, Peck and Helander (1999) (Peck and Helander, 1999) examined how the mean number of traffic crashes during 1985-1991 in California varied as a function of DWI convictions in the same period. The data are re-plotted below (Figure 4-25).

Figure 4-25: Mean Number of Traffic Crashes 
in California in 1985-1991 by Number of 
DWI Convictions in the Same Period

Figure 4-25: Mean Number of Traffic
Crashes in California in 1985-1991 by Number of DWI Convictions in the Same Period

Drivers who had no DWI convictions in that period had the least number of crashes, and the largest incremental increase in crashes was from no convictions to one conviction (.357 to .628, 76% increase). By contrast, very little percentage increase in crashes occurred for drivers with convictions in the 2-3+ range (5% increase).

In the second California study, Tashima and Helander (1998) reported that out of 17,189 alcohol-involved fatal or injury crashes, 42.5% involved drivers with no DWI priors or alcohol-reckless convictions, and that only about 17% of crashes involved drivers who had been convicted of one or more DWIs occurring prior to the crash. Further, an even smaller percentage (8%) of the 810 alcohol-involved fatal crashes involved drivers who had been convicted of one or more DWIs occurring prior to the crash. The report also included data that showed a steady, linear increase with priors for alcohol-related crashes of about 20% per prior, but a decrease with priors for crashes of all types. Fatal/serious injury crashes remained about the same as a function of priors.

The repeat offenders report concluded that available data from the literature indicate a higher alcohol-crash involvement among repeat offenders than among drivers with no priors or just one prior, but that exactly how much higher nationwide cannot be said with any degree of confidence. The report also concluded that the involvement of repeat offenders in crashes of all types may actually be less than that of first offenders, possibly because sober repeat offenders may drive more carefully than sober first offenders, or may not drive at all because their license was suspended. Finally, the report found that, in terms of sheer number of crashes of all types, both serious and non-serious, persons with no priors at all appear to show the highest involvement in total crashes and in alcohol-related crashes of all degrees of severity.

Perrine, Naud, and von Eye (2000) examined the relationships over time between quantity of alcohol consumed and self-rated intoxication level while driving. The authors used an automated touch-tone, interactive, voice response (IVR) system to collect their data in Vermont which were reported daily over a two-year period. Data collected dealt with questions such as type and amount of alcoholic beverage consumed, drinking location, day of week and time of day of drinking, and perceived level of intoxication. In-depth case studies of two the 33 subjects were discussed in the paper, confirming results from prior roadside surveys in Vermont that bar drinking "becomes once again a major focus of concern regarding drinking and driving." The paper showed that very detailed data regarding drinking and driving patterns can be obtained by the authors' innovative approach.

Our 1998 update observed that prior updates had indicated that drinking drivers had substantially more enforcement actions (including DWI) against them than did other drivers. Several studies published since the 1989 update were reviewed and their findings reported, as follows:

Our own recent review of the scientific literature on repeat DWI offenders (Jones and Lacey, 2000) cited a number of later studies of DWIs, many of which contained data on first offenders as well as repeat offenders. For example, Peck and Helander (1999) cited above provided California data on the mean number of crashes of all types in a seven-year period as a function DWI convictions and moving traffic violations in the same period. The data are plotted in Figure 4-26.

Figure 4-26: Mean Number of Traffic Crashes 
in California in 1985-1991 by Number of DWI Convictions 
and Number of Moving Violations in the Same Period

Figure 4-26: Mean Number of Traffic
Crashes in California in 1985-1991 by Number of DWI Convictions and Number of Moving Violations in the Same Period

The figure shows that, in general, the number of crashes increased with number of DWIs and also with number of moving violations such that, for example, drivers with one DWI and four moving violations had about twice as many crashes as drivers with one DWI and no moving violations.

In the same paper, Peck and Helander presented an interesting summary of the recidivism of first offenders and repeat offenders in California, showing among other things how recidivism rates have varied over time. Their data (Figure 4-27) indicate that the one-year recidivism rates for repeat offenders and first offenders alike decreased in the 1989-1995 period, from nearly 10% to 7% for repeat offenders, and from about 9% to 6% for first offenders. Peck and Helander also listed a number of correlates of recidivism of DWIs in general, and showed how the predicted recidivism of repeat offenders varies with arrest BAC and number of priors.

Figure 4-27: Recidivism Trends in California, 1989-1995
Figure 4-27: Recidivism Trends in
California, 1989-1995

In another study of repeat offenders, Jones, Wiliszowski, and Lacey (1996) examined the effect of prior DWI convictions on the DWI recidivism of 506 repeat DWI offenders assigned to an intensive supervision probation program in Milwaukee County, Wisconsin during 1992 - 1994. It was found that the more priors a subject had, the higher that subject's recidivism at any given time. For example, 28.3% of treatment-group subjects with four priors were predicted to recidivate after one year, compared to 7.8% of such subjects with two priors. A similar relationship between priors and recidivism was found in a study of an individualized sanctioning program in Rockdale County, Georgia (Jones and Lacey, 1998b). Here, the recidivism rate was found to increase by about 8% for each prior DWI.

The biographical and other characteristics of DWIs have been investigated in a number of studies. Many of these have been reviewed in prior updates, the latest being our review of the literature on repeat offenders. There, we presented a table of characteristics of repeat offenders, which we re-produce here for convenience (Table 4-3). The table is pertinent to DWI offenders in general, in light of the finding of multivariate studies that indicate that repeat offenders do not appear to differ much from first offenders (Arstein-Kerslake and Peck, 1985). The study by Arstein-Kerslake and Peck found no first-offender group that was distinguishable from a repeat-offender group, a finding that suggested to Perrine and associates (1989) that "most first offenders are problem drinkers who have simply not yet had their second offense." Marowitz (1998) provides more support for this notion, concluding that "first [DUI] offenders with high BAC levels and prior 2-year traffic convictions are at as high a risk of recidivating as many repeat offenders."

Table 4-3: Summary of Attributes of Repeat Offenders
SexPredominately male, typically over 90%
AgeUsually (~75%) under 40, mean around 35
Marital Status Unmarried 
Education HS or less 
Employment Non-White collar 
BAC .18+ at arrest; higher in fatal crashes
Prior DWIs  Typically 2 or 3, higher for some in treatment programs 
Prior Other Traffic Infractions Several 
Prior Criminal Offenses Yes, more than first offenders, include serious crimes against persons 
Alcohol Problems Often have problems, alcohol dependency common 
Personality & Psychosocial Problems Yes, probably more common and severe than those of first offenders 
Drinking Locations Multiple locations favoring bars; at home; parties. Often plan to drive after drinking 
Final destinations Home 
Beverage Mostly beer, often distilled spirits 
Recidivism ~10%+ per year, increasing with number of prior DWIs 
Implied Consent More than 50% are BAC test refusers 
Sentences Traditional, treatment often 
Reasons for DWI Thought he/she was fit to drive 
Perceived Detection Low for first offenders, increases with priors

Source: (Jones and Lacey, 2000)

In the narrative accompanying the table, we stated that:

"What is known from the recent literature about repeat offenders is summarized in [Table 4-3]. There are few surprises. Repeat offenders are nearly always male, and are typically under age 40, White, low income, unmarried, not college educated, and employed in non-White collar occupations. Their BAC at arrest is typically slightly higher than that of first offenders; they often have alcohol problems; and they commonly suffer from alcohol addiction.

They prefer to drink beer and distilled spirits in bars at multiple locations, thus increasing the probability of their driving while impaired. Because they are such experienced drinkers, they very often believe they are quite capable of driving after drinking and do so knowing that they may be arrested for DWI. Personality and psychosocial problems are common among this group.

By definition, they have prior DWI offenses, usually two or three, but those who have been assigned to treatment programs often have more. But they also have a record of other, often non-major, traffic infractions, an attribute that has been found to be a very powerful predictor of DWI recidivism. In addition, they usually have a record of criminal offenses that include serious crimes against persons as well as against property. When stopped for suspicion of drunk driving, they often refuse to submit to a chemical test for alcohol. When convicted of DWI, they are given traditional sanctions (jail and license suspension), but are also often required to participate in alcohol treatment programs." (Pages 19-20)

Three other recent studies also contain information on the characteristics of DWIs. The study by Chang, Lapham, and Barton (1996) cited above present a detailed compilation of the sociodemographic characteristics of 5,154 mostly male DWI offenders referred for alcohol-related assessment.

Siegal, Falck, Carlson et al. (2000) studied the sociodemographic and psychiatric characteristics of 126 "hardcore" DWI offenders incarcerated in Ohio prisons in 1998-1999. Again, detailed compilations are presented, with the authors concluding that virtually all of the sample could be described as alcohol dependent, that almost two-thirds manifest a concurrent substance abuse disorder, and that the rates of psychiatric illness were many times that which could be found in a general population sample. Educational achievement was low with only a single subject reporting any college experience and 43% indicating less than a high school education. In addition to their multiple arrests / convictions for alcohol and / or drug-related vehicular offenses, 61% of the sample reported arrests for disorderly conduct / public intoxication. The substance abuse problems were found to be long standing: the average age of the sample was nearly 36 years, and the age of the first alcohol dependency symptoms was about 21 years of age, with relatively few subjects having been exposed to identification and early intervention for their alcoholism or substance abuse.

Finally, Siegal and associates found that their hardcore population was, in fact, not a homogeneous group but three distinct groups each exhibiting specific clinical needs:

The third study (Wieczorek, Callahan, and Nochajski, 2000) used the Alcohol Use Inventory (AUI) to identify subgroups of persistent drinking drivers (n=363), where persistent drinking drivers were defined as individuals with at least two drinking and driving convictions. The analysis suggested groupings of persistent drinking drivers based on overall severity of alcohol problems. Interestingly, there were few differences among the groups with respect to demographic factors, the only significant difference being employment status full time employment was much less common among the groups with more severe alcohol problems. By contrast, the groups differed significantly on nearly all of the driving or drinking-driving measures, with traffic violations, alcohol-related crashes, and total DWIs higher among the more severe groups. The higher severity groups also reported more alcohol consumption, with the more severe groups reporting over 20 drinks prior to arrest, and a showing a history of multiple treatment experiences. The higher severity groups also showed more significant psychiatric pathology on ten measurement scales, with the level of psychiatric problems reaching clinical levels.

Motivations for Drinking-Driving. Another class of literature on drinking drivers has dealt with the reasons why persons drink and drive. Wiliszowski, Murphy, Jones et al. (1996) addressed the question directly by asking repeat DWI offenders why they continue to drink and drive after being convicted of DWI. Most of the subjects gave multiple reasons for driving after drinking, the most frequent being that the person thought he or she was "OK to drive." (Table 4-3)

Subjects were also asked if they ever planned not to drink or to drink only a certain amount of alcohol when they knew they would be driving afterward. Twenty-two percent indicated that they planned to drink when they knew that they would be driving afterward, and this percentage increased with increasing number of prior DWIs: six percent of those with one prior planned to drink; 18% of those with two priors planned to drink; and 31% of those with three or more priors planned to drink.

Table 4-4: Reasons Given for Driving After Drinking
by Repeat Offenders
Reason for Driving After DrinkingResponses
Thought he/she was OK to drive32.2%
Just did not think about it21.0%
Lacks control over him/herself after drinking18.6%
No one available to drive for him/her14.4%
Would be OK if careful (to avoid accident/arrest)13.8%

When asked what they thought the likelihood of police detection was before their first offense, almost 44% said they just had not thought about the possibility of being detected and arrested by police before that first offense. The percentage dropped for subsequent offenses to 16.8% with twice as many males giving this response as females.

Finally, the responses indicated that the majority of persons interviewed thought they were intoxicated at the time of an arrest, and more individuals thought they were intoxicated for first and second offenses than for third or higher offenses, but the difference was not statistically significant at the .05 level.

In a study in Pomerania (Germany), Bornewasser and Glitsch (2000) studied the decision making processes of 185 DWIs compared to those of a group of 145 drivers who had never been detected for drunk driving. Each subject was asked to imagine a standard scenario when they had a low BAC (< .05), and one when they had a high BAC (> .11). Descriptive analysis showed that a decision process took place, that the number of both inhibiting and impelling beliefs was reduced at high BACs, and that there was a lack of inhibiting beliefs in the group of DWIs. The DWIs had a strong and significant tendency to neglect social norms and had lower perceived risks of detection and crashing.

Burns and Fiorentino (2000) examined the relationships of drinkers' ratings of their own intoxication and driving impairment in an alcohol experiment with 48 men and women, ages 21-54 years, who were light, moderate and heavy drinkers. The subjects rated their degree of intoxication at BACs of .000 to .125. The authors found that heavy drinkers rated their intoxication levels lower than either moderate or light drinkers, a finding that was said to reflect their acquired tolerance to alcohol effects. Heavy-drinking men generally had lower intoxication ratings than women, but driving ratings between heavy-drinking men and women did not differ.

In a study of drinking practices and attitudes of pub patrons in Israel, Shinar (1995) found a pattern of "alarming ignorance of the effects of drinking, total disregard for the risks of driving when under the influence of alcohol, but coupled with relatively conservative amounts of alcohol consumption." This was seen as particularly troubling, given a trend toward increasing alcohol consumption by Israeli youth.

Turrisi and Jaccard (1991) used psychological theories of judgment and decision making in analyzing drunk driving decisions. Four groups of individuals were examined (a) those having multiple convictions for drunk driving, (b) those having one conviction for drunk driving, (c) those never having been convicted of drunk driving, but who admit to having driven while intoxicated, and (d) those who drive, but claim to not have driven while intoxicated. Cognitions examined included perceived drunkenness relative to legally allowable blood alcohol levels for driving, perceived probabilities for being stopped and arrested and being involved in an automobile accident if driving drunk, and drunk driving tendencies.

Three major findings are of interest. First, there were a number of misinterpretations in the use of "external cues" (number of drinks and time between drinks) of the effects of impairment. These misinterpretations existed across all groups tested, and also applied to the probability of being in a crash after drinking. Second, "situational cues" (weather and driving distance) did not modify judgments about the probability of a crash / arrest. Third, persons who drove after drinking but had not been caught were less cautious about drinking driving than were persons who had been caught. This was true even for persons who knew they were over the illegal limit.

Thurman, Jackson, and Zhao (1993) used factorial surveys to examine the drunk-driving judgments of a national probability sample of 528 non-abstaining adults. They found that key components in decisions to drink and drive included the extent of the driver's behavioral impairment, the availability of drunk-driving alternatives, weather conditions, the number of miles that have to be driven after drinking, the legal consequences of drunk driving (in terms of jail sentences and license revocations), the community response to drunk driving, where drinking occurs, fines that might be issued, the use of traffic roadblocks, and the driver's familiarity with roads that must be driven after a drinking event. Their analyses also indicated that the influence of these factors varied across levels of drinking-driving experience, suggesting that those persons most experienced with drunk driving tend to rate legal sanctions as more important in judgments to drink and drive than those persons who typically refrain from drunk driving.

A study of United Kingdom drivers examined the relationship between subjective perceptions of safe driving and legal driving consumption limits and other factors important in the decision to drive after drinking (Albery and Guppy, 1995). Responses from over 900 drivers established that those who perceived safe consumption levels to be greater than those required to break the law indicated reduced moral commitment to present and possible future countermeasures. These drivers also had previous experience of being breath tested (but not charged with a drink-driving offence), reported comparatively lower estimates of their chances of apprehension and accident involvement when over the illegal limit, showed higher consumption levels on a driving trip and greater self-reported driving while impaired by alcohol. The implications of the findings for the development and delivery of measures to counter drink-driving are discussed.

A recent study involving Albery and Guppy (Guppy, Clay, and Albery, 2000) investigated the self-reported frequency of driving when over the illegal limit of more than 1,400 British drivers. Dependent variables were biographical variables, driver self-perceptions, risk perceptions and reported risk-taking experiences. The respondents were 800 drivers randomly sampled from a national database of licensed drivers, 250 culpable crash-involved drivers identified through police records, and a further 400 drivers sampled locally to the accident group. Just over 20% of the variability in drink-driving frequency was predicted by their model. Higher drink-driving frequency was significantly associated with younger male drivers, those driving fewer miles per week, and those with previous crashes. In addition to the biographical variables, lower drinking-driving risk perceptions and self-perceptions of carelessness and irritability were significantly associated with more frequent drinking-driving.


Prior updates have found a very limited amount of scientific literature on the characteristics of drinking pedestrians and bicyclists. The 1998 update presented 1996 FARS data on the age distribution of intoxicated pedestrians (BAC=.10+) who were killed in traffic crashes, and observed that the distribution was quite similar to that of intoxicated drivers. Trend data on alcohol-related fatally-injured pedestrians by age were also presented, showing very little change in the distribution over time.

The most recent FARS report available for this review (U.S. Department of Transportation NHTSA, 1999) did not provide breakdowns by age group or by sex. We used the FARS database for 1998 available on a CD-ROM to develop some basic information for this report. Figure 4-28 shows that males are about twice as likely as females to have a BAC of .10+ (35% versus 17%). With respect to age, pedfatalities in the 21-44 age groups were the most likely to be alcohol-related, with a fall-off at both sides of the peak age (Figure 4-29).

Figure 4-28: Alcohol-Related Pedestrian Fatalities 
as a Percentage of All Pedestrian Fatalities by BAC and Sex, 
FARS Data for 1998

Figure 4-28: Alcohol-Related
Pedestrian Fatalities as a Percentage of All Pedestrian Fatalities by BAC and Sex, FARS Data for 1998

Figure 4-29: Alcohol-Related Pedestrian Fatalities 
as a Percentage of All Pedestrian Fatalities by BAC and Age, 
FARS Data for 1998

Figure 4-29: Alcohol-Related
Pedestrian Fatalities as a Percentage of All Pedestrian Fatalities by BAC and Age, FARS Data for 1998

These two graphs used data generated by applying NHTSA's imputation method for estimating BACs where no measurements were made (Klein, 1986). The method provides for only two levels of BAC, .01+ and .10+. In a landmark study of the role of alcohol in pedestrian crashes in New Orleans, Blomberg and associates (1979) found that, not only were typical alcohol-impaired pedestrians predominately male and in the mid-age range, but that they also had very high BACs. To get an idea of the percentage of fatally-injured pedestrians at higher BACs in 1998, we once again examined the actual data in 15 states that have high percentages of BAC measurements and have been studied in past examinations of the alcohol-crash problem.

The results with respect to the sex of pedestrians are shown in Figure 4-30. Nearly 30% of the males were at .20+, and 10% were at .30+. The percentage for males was about twice as high as that for females at any given BAC level. With respect to age, the percentages peaked for the 35-44 age group at all BAC levels except .40+, which peaked for the 25-34 age group (Figure 4-31). An astonishing 41% of fatally injured pedestrians of age 35-44 had a BAC of .20+, and 18% were at .30+.

Figure 4-30: Alcohol-Related Pedestrian Fatalities as a Percentage 
of All Pedestrian Fatalities at High BACs by  Sex in 15 States, 
FARS Data for 1998

Figure 4-30: Alcohol-Related
Pedestrian Fatalities as a Percentage of All Pedestrian Fatalities at High BACs by  Sex in 15 States, FARS Data for 1998

Figure 4-31: Alcohol-Related Pedestrian Fatalities as a Percentage 
of All Pedestrian Fatalities at High BACs by Age in 15 States, 
FARS Data for 1998

Figure 4-31: Alcohol-Related
Pedestrian Fatalities as a Percentage of All Pedestrian Fatalities at High BACs by Age in 15 States, FARS Data for 1998

Leaf and Preusser (1997) reviewed the pedestrian-alcohol problem and analyzed FARS data for the years 1983 through 1994. The records of 53,904 pedestrian victims whose BACs (67% of all cases) were measured were used in their analysis of the overall pedestrian-alcohol problem. Since race is not coded in FARS, the FARS data were matched against data in the Centers for Disease Control's Multiple Cause of Death (MCOD) file for the years 1987-1989. As a result, race data on 16,957 fatally injured pedestrians of age 15 years or higher were obtained.

The results for two high BACs, .10+ and .20+, are especially interesting here. For males, 48% had BACs of .10+ and 30% had BACs of .20+. For females, these percentages were 25% and 15%, respectively. These two sets of percentages are quite close to those we obtained for the 15 states in 1998.

There were distinct differences in the age-sex distributions for the racial groups coded in the MCOD file. Overall, Whites (including White Hispanics) had the lowest percentages, and Blacks were moderately higher. Native Americans were the highest, with 84% at .10+ and an astonishing 63% at .20+ (Figure 4-32). Asians and Pacific Islanders (not shown) were very few in number (n=266) and had extremely low incidences of high BACs.

Figure 4-32: Alcohol-Related Pedestrian Fatalities as a Percentage 
of All Pedestrian Fatalities at High BACs by Race, 
Data from FARS and MCOD, 1987-1989

Figure 4-32: Alcohol-Related
Pedestrian Fatalities as a Percentage of All Pedestrian Fatalities at High BACs by Race, Data from FARS and MCOD, 1987-1989

With respect to sex, males were twice as likely as females to have BACs of .10% or higher (47 percent vs. 24 percent). Whites (including Hispanics), male and female, had BACs of .10+ less frequently than Blacks (41 percent vs. 47 percent). Native American males were about twice as likely to have high BAC levels as other males (86% vs. 43%); Native American females were three times as likely to have BAC levels of .10+ as other females (75% vs. 23%). Data for six sites that had information on Hispanics showed that Hispanic males had greater alcohol involvement than did White males (52% of Hispanic male victims had BACs of .10+ vs. 46% of White males), but that Hispanic females had much lower involvement (17% for Hispanic females at .10+ vs. 27% for White females). Again, Blacks had somewhat higher proportions of BACs at .10+ than did Whites.

With respect to age, about 90% of males between 15 and 34 had BACs of .10+, with values dropping only slightly for older Native Americans. Nearly three-fourths (73%) of males 25-34 had a BAC of .20+ or higher; and more than half of Native American males had a BAC of .20+ in every age category. Native American females (based on a small number of cases) showed similar high alcohol involvement. For them, maximum involvement was for ages 25-34, but levels of involvement stayed very high from ages 15 through 54 and possibly beyond. Over all ages, three-fourths of Native American females had a BAC of .10+, and more than half (54 percent) had a BAC of .20+.

For Whites, Blacks, and Hispanics, levels of alcohol involvement were lower, although still very high, and generally similar. Highest BAC values were shown for males ages 25-54, where three out of five had BACs of .10+, and two out of five had BACs of .20+. Close behind were males ages 21-24, who had nearly the same level of involvement at BACs of .10+, but somewhat lower numbers at BACs of .20+. Males ages 15-20 had lower, but still large, levels of alcohol involvement: 38% had BACs of .10+, but only about one in eight (12-15%) had BACs of .20 or higher. Male pedestrians ages 55 and older had still lower levels of alcohol involvement: One-fourth had BACs of .10+, and a relatively high one-sixth had BACs of .20+.

Females showed similar distributions of BAC levels across ages, but the peak was narrower (reached only in the 25-34 age category); dropped off more sharply for younger and older women, and never quite reached the levels shown for males. Black females showed the greatest levels of alcohol involvement, followed by White females, followed by Hispanic females.

The above discussion of age-sex-race interactions is taken from Leaf and Preusser (1997), pages 22, 23, and 25.

Leaf and Preusser also examined the role of several driving-related variables in crashes involving three high-risk racial groups and made comparisons with their White counterparts. The combined FARS / MCOD file for 1987-1989 was used in developing the comparisons. We summarized data from their tabulations in Table 4-5, which indicates the groups and the variables considered, and whether there was any significant or meaningful difference in the percentage of high-BAC crashes between each high-risk group and its White counterpart for a given variable.

Table 4-5: Differences in the Percentage of High-BAC Pedestrian Crashes
Among High-Risk Racial Groups and their White Counterparts
for Several Driving-Related Variables
VariableBlack Adults vs. White AdultsHispanic Males vs. White MalesNative American (NA) Adults vs. White Adults
Time of Day/
Day of Week
No DifferenceHispanic higher on weekday nights and weekend daysNo Difference
LightBlack higher during daylightNo DifferenceNA higher during dark, unlighted
WeatherNo DifferenceNo DifferenceNo Difference
Road ConditionNo DifferenceHispanic more on dry pavementNo Difference
Road ClassBlack higher on local streets, less on arterialsHispanic higher on local streets, less on rural connectorsNA higher on local streets, rural connectors; lower on arterials
Speed LimitBlack higher at lower speed limitsHispanic higher at lower speed limitsNA higher at higher speed limits
LocationNo DifferenceNo DifferenceNo Difference
Pedestrian BehaviorBlack higher for walk in roadwayHispanic higher for improper crossingNA higher for walk in roadway
No. VehiclesNo DifferenceNo DifferenceNo Difference

It is seen that none of the contrasts showed any difference with respect to weather, location, and number of involved vehicles. However, all of the contrasts showed a difference with respect to road class, speed limit, and pedestrian behavior. For road class, all of the racial groups had higher percentages than their White counterparts on local streets, and Native Americans had a higher percentage on rural connectors. For speed limit, Blacks and Hispanics had a higher percentage at lower speed limits, and Native Americans had a higher percentage at higher speed limits. For pedestrian behavior, Blacks and Native Americans had a higher percentage of walk-in-roadway crashes, while Hispanics had a higher percentage of improper-crossing crashes. Finally, two variables were associated with differences for one or two contrasts. For time of day / day of week, Hispanics had higher percentages on weekday nights and weekend days, while for light, Blacks had a higher percentage during daylight, and Native Americans had a higher percentage during dark, unlighted conditions.

As indicated on page 59, Voas and associates (2000) recently complied information on race and ethnicity obtained by matching FARS data with death certificate data collected by the National Center for Health Statistics. The data covered the period 1990-1994, and includes some information on pedestrians in fatal crashes for that more recent period. More racial / ethnic categories are included than in the study by Leaf and Preusser, with data again confirming very high alcohol involvement for Native Americans (74%) and very low involvement for Asian-Pacific Islanders (10%). Blacks, Mexican Americans, and Other Hispanic Americans had slightly higher involvement rates than Caucasian Americans, which were followed by Central and South Americans, Puerto Ricans, and Cubans in that order.

In assembling background data for their development of a countermeasure program for alcohol-involved pedestrian crashes, Blomberg and Cleven (2000) performed extensive case studies of 20 fatally-injured pedestrians with high BACs. From the case studies, they developed a detailed profile of the crash victims. The profiles included personal and residence items, alcohol/drug items, pedestrian habits items, and crash items, totaling 39 items altogether. Profiles on pedestrians who were not victims were also constructed. The authors summarized the results of the case studies, and findings from police crash reports, as follows:

"It is estimated that alcohol use on the part of the pedestrian is involved in approximately 40% of the age 14+ pedestrian crashes in Baltimore. The pedestrian alcohol problem in the city is therefore similar to that in other large cities in the United States.

As in other cities, the problem in Baltimore is largely experienced by a middle-aged male who is walking in dark clothing on weekend nights.

The HBD [had been drinking] pedestrian who gets involved in a crash has a very high BAC--usually more than twice the legal limit for driving.

The pedestrian alcohol problem in Baltimore occurs largely in the center of the city and on a few major city corridors.

Most crashes occur near the victim's home and when the victim is making a relatively short trip (for example, to go to a nearby store for food or cigarettes).

High BAC victims and non-victims who regularly drink to excess come in contact with people who could intervene and possibly prevent a pedestrian crash. In addition to relatives and friends, these include liquor sellers and servers, social service representatives, the police and others.

Both [non-fatally injured] victims and non-victims feel crashes could be avoided if pedestrians drank in moderation and paid better attention to safe pedestrian behavior. They also feel that driver behavior could be improved (particularly driver speeds and knowledge of the cues of an impaired pedestrian), and that engineering improvements (e.g., improved lighting, installation of traffic and pedestrian signals) could make the city's streets safer. The lack of knowledge by the general public of the pedestrian alcohol problem was also noted." (p. 15)

No other recent comprehensive studies of the characteristics of alcohol-impaired pedestrians in the U.S. were located in our search of the literature. However, two state studies in the U.S. should be noted, the first in Arizona using combined FARS / MCOD files (Campos-Outcalt, Prybylski, Watkins et al., 1997), and the second in Florida (Miles-Doan, 1996). The Arizona study contrasted the prevalence of "American Indians" and "non-Indians" in alcohol-related crashes and had findings that were generally consistent with those of Leaf and Preusser. The Florida study found that alcohol impairment, indicators of crash severity, and rural location were correlated with each other and also with the likelihood of serious injury or death. The Florida study also found that, controlling for several environmental, crash, and pedestrian characteristics, being under the influence of alcohol increased the odds of dying over those of a non-fatal injury by a factor of five.

There is evidence that many of the characteristics of alcohol-impaired pedestrians in the U.S. are present among alcohol-impaired pedestrians in some other countries. For example, Holubowycz (1995) found similar age-sex-BAC interactions in South Australia to those in the U.S., although the Australian percentages were somewhat lower than U.S. percentages at the higher BACs.

We found no new studies of pedestrian alcohol-crash risk (as defined in Chapter 2). The study in New Orleans by Blomberg and associates (1979) had found that relative risk of pedestrian crash of any severity increases with BAC, and becomes very high at BACs in excess of .20. (How high depends on the non-crashed group used for comparison, with the most precipitous rise occurring when the crashed and non-crashed groups were matched for age and sex.)

The literature on the characteristics of alcohol-impaired bicyclists is much sparser than that for pedestrians. Again, the latest available FARS report (U.S. Department of Transportation NHTSA, 1999) did not provide breakdowns by age group or by sex, and so we used the FARS database for 1998 available on a CD-ROM to develop information on age and sex for bicyclists. We used NHTSA's imputation method for estimating the BACs of subjects whose BACs were not tested.

We found that 20% of fatally injured male bicyclists had a BAC of .10+, compared to only 8% for females. Impairment (BAC =.10+) peaked for the 45-54 age group for both males (33%) and females (20%), dropping rapidly at higher ages. At ages less than 45, the percentage for males decreased slowly until age 20, and then dropped rapidly. For females, the decrease at ages less than 45 was immediate, increasing again at age 16-20 (Figure 4-33).

Figure 4-33: Percentage of Fatally Injured Bicyclists 
at BAC = .10+ by Age and Sex, Data from FARS, 1998

Figure 4-33: Percentage of Fatally
Injured Bicyclists at BAC = .10+ by Age and Sex, Data from FARS, 1998

In their study comparing fatal and nonfatal bicycling injuries in Maryland, Li and associates (1996) reviewed some of the major studies of the role of alcohol in bicycle crashes prior to 1996. Cited was a study of FARS data for the years 1987-1991 by Li and Baker (1994) which found that 23% of the fatally injured bicyclists aged 15 or older whose BACs were tested had a BAC of .10+. Studies indicating that some 8%-15% of injured bicyclists were alcohol-positive were also cited in the study.

The 1996 study by Li and associates included only 63 fatalities, 52 males and 11 females. Twenty-five percent of the males had a BAC of .10+, compared to 9% of the females. The sample size of bicyclists whose injuries were not fatal was much larger, with a total of 214 males and 39 females. Of the males, 13% were at .10+, compared to 7% of the females. With respect to age distribution, the sample size for fatalities was too small to arrive at any meaningful conclusions, but the distribution for the larger sample of non-fatalities was quite smooth, peaking at about 20% at 30-39 years and was nearly that for ages 20-39 and 40-49. There was a rapid fall-off at the high and low ends of the age spectrum. Li and associates also found that helmet use was far less frequent among fatally or seriously injured bicyclists at .10+ than at lower BACs (6% vs. 31%).


Overall, per capita consumption of alcohol in the U.S. has decreased nearly 30% since 1981, with most of the decrease occurring in the consumption of spirits. And while consumption is strongly correlated with fatal alcohol-related traffic crashes since 1981, it still remains to be shown that, all other possible contributing factors considered, a causal relationship exists between per capita consumption and crashes.

The research shows that most persons of drinking age may be classified as social drinkers, and that more than a third drink no alcohol at all. Problem drinkers and alcoholics together account for less than 10% of the drinking age population. Note that these figures are for the U.S. as a whole; the percentages vary widely by geographic location.

The age 21 to 45 group has the highest prevalence of problem drinking and alcohol-dependency, and the 65+ group the lowest. The 18-29 subgroup has been identified as having a particularly high risk of alcohol dependency and adverse consequences of drinking. In general, women drink less than men and are less likely to have alcohol problems or to be alcohol-dependent.

Race and ethnicity have also been found to play a role in drinking patterns. For example, males identified as "non-Black" have about twice the prevalence of alcohol abusers as males identified as "Black." However, while heavy drinking is believed to be decreasing among Whites, it seems to be remaining stable among Blacks and Hispanics, implying a possible larger share of future drinking-related problems for the latter two groups.

Recent research clearly establishes that there is a drinking problem on college campuses, but there is some question as to the magnitude of that problem. Much research has focused on so-called binge drinking. Estimates of the prevalence of binge drinkers are based on self-reported data, and range from 26% to 49% of all students. And while one study describes binge drinking as normative, these studies suggest that it is not. Nevertheless, binge drinkers account for a disproportionate share of drinking problems among college students, with one study finding that 92% of students with drinking problems were binge drinkers. Estimates of the percentage of binge drinkers with drinking problems ranging from 9% to 26%. Other studies suggest that the prevalence of binge drinking has been relatively stable over the past several years, but that the perception of the prevalence is increasing.

With respect to the alcohol-crash problem, research continues to show that young drivers are more often involved in alcohol-related crashes than any other comparable age group. Alcohol-crash involvement rates, share of the alcohol-crash problem, and alcohol-crash risk all reach their peaks for young drivers, with the largest share of the fatal-crash problem occurring at age 21.

As noted in past updates, a large proportion of the alcohol-crash problem involves young White males. In 1998, 84% of fatal-crash involved drivers with BACs of .10+ were male, and more than 70% were White. However, certain racial / ethnic subgroups have higher involvement rates than other subgroups. Of these, American Indians have the highest rate, and Asian / Pacific Islanders the lowest.

The impact of other biographical variables on drinking and driving is less understood. Crash data suggest that:

It is interesting that self-reported data collected in household surveys have arrived at very different conclusions with respect to employment, income, and education using a different criterion for drinking driving: "drove within two hours after drinking [any amount] in the past 30 days [or year]." A possible explanation for this difference is that the household surveys may be biased in favor of persons who drank very little before driving, suggesting totally different relationships with the various biographical variables than for persons who drank more before driving.

Many studies have found that beer is the preferred beverage of drinking drivers. One study found that frequent consumers were more likely to drink outside the home, preferred beer and spirits to wine, and were more likely than others to drink and drive. These individuals were more likely to underestimate the effects of beer and "hazardous" drinking. The authors of the study suggested that the association of beer consumption with drinking-driving arises from the circumstances in which beer drinkers are often found, rather than some other disposition beer drinkers may have to drink and drive.

A number of other studies have examined the role of drinking location in drinking-driving, finding that heavier drinkers prefer to drink at bars and other persons's homes, and at multiple locations requiring longer driving distances. Younger drivers have been found to prefer drinking at private parties, while older, more educated drivers prefer bars and taverns. Those frequenting "drive-up windows" for acquiring alcoholic beverages have been found more often to have drinking problems and to engage in high-risk behaviors.

It has often been suggested that the high BACs frequently found in drivers arrested for DWI are indicative of a drinking problem, and therefore that high-BAC drivers in crashes are also problem drinkers. Three well-designed recent studies have found that a high BAC at arrest is not in itself indicative of a drinking problem, sending a cautionary note to those concluding without other evidence that heavy drinkers are also problem drinkers. By contrast, a study of a special group of DWI arrestees - male DWI offenders in the U.S. Army - found a significant association between BAC at arrest and drinking problems.

Studies continue to show that drinking-driving is primarily a nighttime, weekend phenomenon. Household surveys indicate that male drivers make three times as many trips within two hours after drinking any amount of alcohol than do females. Using this measure, such drinking drivers as a whole made 1.7 drinking-driving trips in the past 30 days, with the oldest drivers making the most trips and the youngest drivers making the fewest. Motorcycles have the highest rate of alcohol-related fatal crashes, followed by light trucks, passenger cars, and large trucks in that order.

Drivers who have previously been arrested for DWI continue to be the subjects of a large number of alcohol-safety research studies. This is interesting, since FARS data suggest that only a few percent of fatal crashes involve drivers who have recent convictions of DWI. One research issue that has frequently been addressed is the role of number of prior DWI convictions in crashes. Studies in California have found that only 8% of drivers in fatal crashes had one or more DWI offenses on their driver record. However, studies in Minnesota suggest a much higher percentage. In California, crashes of all types actually decreased with number of priors, and in terms of sheer number of alcohol-related crashes, persons with no priors had the highest rate of involvement.

The characteristics of repeat offender DWIs and first offender DWIs have in general been found to be quite similar in many respects, but DWIs with large numbers of priors have been found more often to have long-standing problems of alcohol dependency, and to differ on the severity of their alcohol problems rather than on their demographics. A history of participation in multiple treatment programs is common for these individuals, as well as diagnoses of psychiatric pathology.

A smaller body of literature has addressed the factors that influence one's decision to drive after drinking. Decision theory holds that a rational person will try to maximize the expected gain resulting from a decision, and there is some evidence that drivers attempt to do this in an informal way. Research suggests that experiencing a prior negative event (such as an arrest or a crash) has a positive effect, tending to make a driver less inclined to drive after drinking or to drive more cautiously after drinking. Factors that have a negative effect include a lack of knowledge of the impairing effects of alcohol or a misinterpretation of the cues of impairment, a reduction of inhibitions at higher BACs, a lowered perception of alcohol-crash risk, and a neglect of social norms after drinking. Research suggests that it is not just the impairing effects of alcohol that favors a decision to drive after drinking; some drivers plan to drink knowing they will drive afterward.

We note that past updates reported the findings of several studies that found drinking-driving to be just one of a series of problem behaviors exhibited by certain groups of individuals. This line of research appeared quite promising at the time, but we found no recent studies of this nature in our search of the literature for this update.

The scientific literature on the characteristics of alcohol-impaired pedestrians and bicyclists is far less extensive than that for drivers. What exists indicates that the alcohol-crash problem for pedestrians is, at it is for drivers, predominately a male problem. Very high BACs are common for pedestrians in alcohol-related fatal crashes, especially for those in the 35-44 age group (which is estimated to have 41% at .10+ and 18% at .20+). Alcohol-impaired bicyclists in fatal crashes are also more likely to be male, with the highest percentage of bicyclists at .10+ occurring for those in the 45-54 age group, an older peak age group than that for either drivers or pedestrians.

Locations of pedestrian alcohol-related crashes as a whole are most likely to be near the victim's home or a short distance from the starting point of the trip. Recent research on race and ethnicity indicates that Native Americans have the highest prevalence of alcohol-related pedestrian crashes, roughly three times that of Caucasians at .20+. Blacks and non-Black Hispanics fall somewhere between these two extremes.

Several situational variables (such as road class and speed limit) have been found to distinguish White alcohol-impaired pedestrians in crashes from those of other racial / ethnic groups. In-depth studies of such situational variables are rare in the larger body of literature on drinking-driving: a recent study in New Zealand by Bailey and Bailey (2000) found that such situational factors as adverse weather, the driver falling asleep, and involvement of a truck are important in alcohol-related crashes and observed that some of these factors may be more readily addressed with remedial measures than the alcohol factor.

  1. The term "apparent consumption" is used in the literature to indicate gallons of absolute ethanol consumed per person of drinking age. Drinking age is usually assumed by researchers to be less than the minimum legal drinking age. The NIAAA report used a drinking age of 15 for the years prior to 1970, and a drinking age of 14 for 1970 through 1997.
  2. Critiques of these two studies were also published by Joksch (1989; 1991).
  3. Jacobson (1989) notes the difficulty of arriving at satisfactory definitions of the terms "problem drinker" and "alcoholic" and discusses some of the definitions that had been used.
  4. CAGE (an acronyn for Cut down, Annoyed, Guilty, Eye-Opener) is a screening questionnaire.
  5. Data on non-fatal alcohol-related crashes as a function of driver sex were not presented in the 1999 FARS report.
  6. Reviewed in Jones and Lacey (1998a).