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


DWI Offenders and Alcohol-Related Crashes

William F. Wieczorek

Research Institute on Addictions, 1021 Main Street, Buffalo, New York 14203, USA


Despite the clear linkage between alcohol impairment and crashes, most drinking and driving episodes do not result in a crash. This fact suggests that there may be differences between DWI offenders who have crashes and those who don't. A better understanding of these differences could be useful for prevention and treatment programs directed at reducing DWI and crashes. Therefore, this study examined differences in behavioural and psychological factors between DWI offenders who never had an alcohol involved crash and DWI offenders who had at least two alcohol-related crashes. The sample consisted of 646 convicted DWI offenders: about 11% had two or more crashes. In univariate analyses, the crash involved offenders were significantly different from the no crash group on a large number of measures that suggested lower socioeconomic status, more driving violations, more severe alcohol dependence, more frequent drinking and driving, substantial psychiatric problems, lower self esteem, and greater driving aggression. Multivariate discriminant function analysis found that psychiatric symptoms, driving violations, and arrests for crimes other than DWI were significantly related to crashes even when the frequency of drinking and driving was controlled in the analysis. The results indicate that more intensive interventions specific to improving driving skills are needed. Also, psychiatric problems may require attention if crashes are to be minimised.


Despite the clear linkage between alcohol impairment and motor vehicle crashes (US DOT 1989, Zador 1991), most drinking and driving episodes do not result in a crash. This fact suggests that there may be differences between drinking drivers involved in alcohol-related crashes and those who have not had an alcohol-related crash.

Previous researchers (Simspon, 1985; Waller, 1989) recognized the need to evaluate a variety of human factors that may be involved in alcohol-related crashes. Donovan & Marlatt (1982) identified subgroups of drinking drivers based on driving-related attitudes (e.g., tension-reduction driving, competitive speed, driving-related aggression) and other personality measures (e.g., hostility, control orientation, depression). Donovan et al. (1983) found that a number of personal characteristics including depression, anger, thrill seeking, and driving to deal with emotions were associated with a higher risk for crashes. Donovan and coworkers (1985) went on to suggest that drinking drivers are a subgroup of a larger population of high risk drivers.

Recent research (McLellan et al., 1993; Vingilis et al., 1994) examined psychosocial differences between drinking and non-drinking drivers seriously injured in a crash, and found that the drinking drivers had a higher level of alcohol-involvement in their lives. Comparisons of drinking drivers with and without a history of alcohol-related crashes are lacking in the literature.

The current study examined differences in psychosocial characteristics between DWI offenders (persons convicted of any drinking and driving offense) who never had an alcohol-related crash and those who had multiple (two or more) alcohol-related crashes. The study focused on the difference between the groups with no crashes and multiple crashes to provide as strong a distinction as possible. This is appropriate because little research has been done in this area and any differences between the groups may have important implications for prevention, treatment, and policy.


The sample for this study was drawn from the 656 DWI offenders who were recruited from several sources (i.e., Drinking Driver Programs n=145, Buffalo City Court records n=200, Erie County Probation Department n=311) to ensure inclusion of the entire spectrum of DWI offenders. The subjects were recruited from September 1991 through July 1994 in Erie County, New York.

After signing an informed consent, each subject was interviewed by a trained research assistant in a private setting at the Research Institute on Addictions. The interview assessed a number of topics including sociodemographic information, lifetime DSM-III-R alcohol dependence criteria, alcohol-related problems in the last year, alcohol consumption in the 30 days prior to the last DWI arrest, drinking and driving in the 30 days prior to the last arrest, driving history (crashes and driving violations), criminal history other than DWI, attitudes towards driving (Donovan, 1980), self-esteem (Rosenberg, 1965), locus of control (Levinson, 1981), sensation seeking form V (Zuckerman, 1979), and SCL-90-R psychiatric symptomatology raw scores (Derogatis, 1983).

Univariate comparisons utilized analysis of variance (ANOVA) for continuous variables and the Chi-square statistic for categorical measures. Discriminant function analysis (DFA) was the multivariate technique used in a step-wise fashion to select the combination of variables that best differentiates the two groups. The step-wise DFA used a p-level of .10 for entry of variables because the analyses are preliminary and even marginally significant findings could be important topics for future research.


Of the total sample of DWI offenders, 12% had two or more alcohol-related crashes. The results focus on the 545 DWI offenders who either had no alcohol-related crashes (467), or had multiple alcohol-related crashes (78). There were no significant differences between these two groups based on gender, race/ethnicity, and age. The group with no crashes had significantly higher household income ($29,666 vs $22,784, F=5.25, p=.02) and education (12.9 years vs 12.2 years, F=5.90, p=.02) than the group with multiple alcohol crashes. The crash-involved group was more likely to have arrests for crimes other than DWI (80% vs 51%, Chi-square=21.97, p<.0001).

Table 1 shows the results based on the drinking-related measures. Clearly, the group with multiple alcohol-related crashes has a more severe problem with alcohol than the group with no crashes. Compared to those without any alcohol crashes, DWI offenders with multiple alcohol crashes had more alcohol dependence criteria, more alcohol-related problems, greater alcohol consumption, and a higher probability of qualifying for a lifetime diagnosis of alcohol dependence.

Table 1
Drinking Measures by Number of Alcohol-Related Crashes

Variable None
Mean (ąSD)
Two or More
Mean (ąSD)
F Significance
% Alcohol Dependent 77% 96% 15.08* .0001
# of Dependence Criteria 4.87 (2.68) 6.88 (2.17) 39.86 <.0001
Total Alcohol Problems last yr 9.51 (23.42) 24.32 (48.41) 18.27 <.0001
Drinks/day in 30 days prior to last DWI arrest 4.76 (6.70) 9.98 (13.18) 28.40 <.0001
*Chi-square statistic

Comparisons of the two crash groups on drinking-driving measures and driving measures are found in Table 2. The multiple alcohol crash group was more involved in drinking-driving behaviors as shown by their larger number of DWI arrests, larger percentage of repeat DWI offenders, and more frequent drinking and driving. More traffic violations were associated with the multiple alcohol crash group. The pattern of driving-related attitudes showed no clear differences between the crash groups. Only driving aggression and competitive speed were greater in the multiple alcohol crash group.

Table 2
Driving and Drinking-Driving Measures by Number of Alcohol-Related Crashes

Variable None
Mean (ąSD)
Two or More
Mean (ąSD)
F Significance
Total DWIs 1.95 (1.33) 3.50 (2.98) 60.01 <.0001
% Repeat DWI Offenders 48% 90% 46 .05* <.0001
Lifetime Traffic Violations (excludes crashes & DWIs) 5.38 (8.02) 10.32 (13.56) 18.27 <.0001
Times Drinking-Driving in 30 days prior to last DWI arrest 7.68 (8.74) 14.32 (11.99) 34.28 <.0001
Driving Aggression 4.05 (2.89) 4.84 (3.39) 4.65 .03
Competitive Speed 1.66 (2.07) 2.15 (2.11) 3.84 .05
Tension Reduction Driving 3.03 (1.83) 3.43 (1.86) 3.33 NS
Driving Inhibition 1.15 (1.18) 1.00 (1.02) 1.12 NS
*Chi-square statistic

Table 3 shows the differences between the two alcohol crash groups on personality measures. DWI offenders with no alcohol crashes had higher self-esteem scores than the crash involved offenders. No differences were apparent between the two groups on locus of control (Internal Control, Chance Control, Powerful Others). The only significant difference on sensation seeking between the groups was found for Thrill and Adventure Seeking. The difference on this scale was in the opposite direction than expected: DWI offenders without any alcohol crashes scored higher than the offenders with multiple alcohol crashes.

Table 3
Personality Measures by Number of Alcohol-Related Crashes

Variable None
Mean (ąSD)
Two or More
Mean (ąSD)
F Significance
Self-esteem 3.13 (.48) 2.90 (.55) 15.37 .0001
Internal Control 38.33 (5.94) 37.09 (5.92) 2.91 NS
Chance Control 21.04 (9.05) 23.14 (9.78) 3.53 NS
Powerful Others 18.93 (9.13) 19.75 (10.77) .50 NS
Sensation Seeking Total 19.43 (6.35) 19.47 (5.96) .01 NS
Thrill & Adventure Seeking 6.5 (2.75) 5.7 (2.87) 5.97 .02
Experience Seeking 5.37 (1.78) 5.47 (1.89) .21 NS
Disinhibition 4.75 (2.44) 5.15 (2.53) 1.78 NS
Boredom Susceptibility 2.77 (1.97) 3.14 (2.17) 2.32 NS

Table 4 shows the psychiatric symptomatology for the two groups. For all nine subscales of the SCL-90-R, the group with multiple alcohol crashes reported significantly more symptoms than the group without any alcohol crashes. The Global Severity Index is a measure of the overall level of psychiatric severity, which also indicated significantly greater psychiatric distress in the multiple alcohol crash group.

Table 4
Psychiatric Symptomatology (SCL-90-R) by Number of Alcohol-Related Crashes

Variable None
Mean (ąSD)
Two or More
Mean (ąSD)
F Significance
Somatization 4.81 (5.76) 8.41 (8.13) 22.92 <.0001
Obsessive-Compulsive 6.27 (5.74) 10.59 (8.43) 32.55 <.0001
Interpersonal Sensitivity 5.51 (5.48) 8.86 (7.22) 22.54 <.0001
Depression 8.36 (8.09) 13.97 (11.09) 28.58 <.0001
Anxiety 4.21 (5.37) 7.77 (7.91) 25.10 <.0001
Hostility 2.95 (3.77) 4.64 (5.18) 11.96 .0006
Phobic Anxiety 1.53 (2.81) 3.54 (4.48) 27.85 <.0001
Paranoid Ideation 4.03 (4.10) 6.18 (5.31) 16.74 <.0001
Psychoticism 4.14 (5.22) 7.24 (7.43) 20.49 <.0001
Global Severity Index .52 (.48) .88 (.72) 32.06 <.0001

The discriminant function that differentiates between the DWI offenders with no alcohol crashes and those with multiple alcohol crashes was highly significant (Chi-square=72.5, p<.0001). Although the discriminant function analysis (DFA) selected a number of significant variables, the amount of variance explained by the function is about 14% (see Table 5).

Table 5
Discriminant Analysis of DWI Offenders with None or Multiple Alcohol Crashes

Predictor Standardized Coefficient Significance of F-to-remove Wilks' Lambda Association with Alcohol Crashes
Times Drinking-Driving in 30 days prior to last DWI .45 .0003 .932 More drinking and driving
Global Severity Index .50 <.0001 .890 More psychiatric symptoms
Criminal History .36 .004 .869 Non DWI arrests
Traffic Violations .27 .04 .863 More violations
Thrill & Adventure Seeking -.24 .05 .856 Less interest in thrills

The variable with the largest influence on the discriminant function is the Global Severity Index. The amount of drinking and driving in the 30 days before the last DWI arrest had the next largest role. Having an arrest for crimes other than DWI was the next most influential in the DFA. Traffic violations were directly associated with the multiple alcohol crash group in the DFA. Similar to the univariate analysis, higher scores on Thrill and Adventure Seeking were associated with the no alcohol crash group. The DFA indicated that when the frequency of drinking and driving was controlled for, psychiatric symptoms, traffic violations, criminal history, and a sensation seeking scale still significantly differentiated the two groups.


The univariate findings support the concept that alcohol problems, personality (e.g., hostility), driving attitudes (competitive speed, driving aggression), and bad driving (violations) are characteristics associated with multiple alcohol-related crashes. The current study also identified psychiatric symptomatology and criminal history as risk factors for alcohol-related crashes.

The finding for Thrill and Adventure Seeking was counter to expectations in that the group with no crashes scored higher than the alcohol crash group. The finding may be associated with differences in education, income, and social norms between the two groups. This scale asks about prosocial forms of adventure seeking such as mountain climbing, flying, and water skiing, which are attractive to persons with a higher socioeconomic status and no criminal history other than DWI.

The DFA reinforces the univariate findings by showing that even when drinking and driving is controlled in the analysis, psychiatric severity, traffic violations, criminal history, and a sensation seeking scale still significantly differentiate the two groups. Driving attitudes and personality do not make significant contributions in DFA. The modest amount of variance explained by the DFA suggests that other factors (e.g., environmental and contextual factors) have a substantial role in alcohol-related crashes by DWI offenders.

These results have a number of policy and research implications. The results indicate that alcohol treatment, interventions directed at driving skills, and treatment of psychiatric problems are important for reducing alcohol-related crashes by DWI offenders. The spectrum of problems suggests that efforts are needed to match DWI offenders' characteristics with the most appropriate set of interventions. Future research should examine the association between alcohol-related crashes and psychiatric symptoms. Psychiatric severity could affect the risk of a crash by the symptoms resulting in reduced attention to the driving task, a direct effect of psychiatric medications on driving skills, an interaction between psychiatric medication and alcohol, or a combination of these factors.


This research was supported by the National Institute on Alcohol Abuse and Alcoholism.

References will be supplied upon request to the author.