- Email: [email protected]
DUI recidivism
Addictive Behaviors 28 (2003) 963 – 969
Short Communication
DUI recidivism A comparison of Mexican Americans and whites in a northern California county$ Cheryl J. Cherpitel*, Jason Bond Alcohol Research Group, Public Health Institute, 2000 Hearst Avenue, Berkeley, CA 94709, USA
Abstract Objective: While Hispanic offenders for driving under the influence of alcohol (DUI) in California are more likely to have a history of multiple offenses compared to whites, little is known about characteristics associated with DUI recidivism in either ethnic group. Demographic and DUI conviction characteristics associated with DUI recidivism are analyzed among Mexican American and white DUI offenders in a Northern California county. Method: A sample of 459 primarily Mexican Americans and 490 whites were randomly selected from records supplied by the California Department of Motor Vehicles (DMV) on over 16,000 DUI offenders in the county during a 3-year period. DMV data on DUI convictions in the 1- and 5-year period preceding the identifying DUI offense and the year following the identifying offense (the recidivism conviction) were analyzed. Results: Rates of recidivism were significantly lower for Mexican Americans (5%) compared to whites (11%) who were arrested for DUI but not convicted, but higher for those convicted (12% vs. 5%). While conviction status of the identifying DUI offense was not predictive of recidivism among Mexican Americans, a DUI conviction in the preceding year was significantly predictive. Among white arrestees, receiving a conviction was significantly, and negatively, predictive of recidivism, but among those who were convicted, a previous DUI conviction in the last year was predictive of recidivism. Referral to a DUI treatment program was not a significant predictor of recidivism among those convicted in either ethnic group. Conclusions: The data suggest that conviction for a DUI may not play the same role in the likelihood of subsequent DUI convictions for Mexican Americans as for whites, and this difference may need to be considered in DUI treatment programs. Additional research
$
Presented at the Annual Epidemiology Symposium of the Kettil Bruun Society for Social and Epidemiological Research on Alcohol, Toronto, Canada, May 28 – June 1, 2001. * Corresponding author. Tel.: +1-510-642-5208; fax: +1-510-642-7172. E-mail address: [email protected] (C.J. Cherpitel). 0306-4603/02/$ – see front matter D 2002 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0306-4603(01)00293-3
964
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
on ethnic differences in DUI offenses and recidivism over longer periods of follow-up is needed to determine ethnic-specific intervention and prevention strategies for DUI. D 2002 Elsevier Science Ltd. All rights reserved. Keywords: Alcohol DUI; Mexican Americans
1. Introduction Rates of offense for driving under the influence of alcohol (DUI) have been found to vary by ethnicity in California, with Hispanics overrepresented among DUI offenders (Chang, Lapham, & Barton, 1996; Perrine, Peck, & Fell, 1989). Hispanic DUI offenders have also been found to be more likely to have a history of multiple offenses compared to whites (Perrine & Arce-Quinones, 1994), but little is known about those characteristics which may be associated with DUI recidivism in general, or variables which may be predictive of recidivism in specific ethnic groups. Studies which have examined characteristics associated with DUI recidivism (without respect to ethnicity) have found prior DUI convictions and other traffic-related arrests, younger age, lower education, and heavier drinking all significant predictors (Arstein-Kerslake & Peck, 1985; Ellingstad, 1974; Lapham, Skipper, Hunt, & Chang, 2000; McMillen, Adams, Wells-Parker, Pang, & Anderson, 1992 ; Peck, ArsteinKerslake, & Helander, 1994; Pollack, Didenko, McEachern, & Berger, 1972). Higher DUI offense rates for Hispanics compared to whites, and higher rates of recidivism, may be related to patterns of drinking in this ethnic group, as well as to a lack of familiarity with DUI laws and their enforcement relative to their home country (Caetano & Clark, 2000). The majority of Hispanics in California are Mexican American, and general population surveys of drinking patterns in Mexico have found higher rates of infrequent but heavy drinking leading to intoxication among males than those found among Anglo males (Caetano & Medina Mora, 1986). DUI recidivism is a major public health problem—35% to 40% of all fatally injured drinking drivers are estimated to have had at least one prior DUI offense (Simpson, 1995; Vingilis, Stoduto, Macartney-Filgate, Liban, & Mclellan, 1994). To begin filling this important gap in the literature regarding those variables associated with DUI recidivism, data on demographic characteristics, prior DUI convictions, and recidivism rates are reported on a sample of Mexican American and white DUI offenders in one California county during a 3-year period.
2. Methods The sample consisted of 979 individuals, stratified on ethnicity (white and Mexican American), randomly selected from all those individuals arrested for (but not necessarily convicted of) at least one DUI offense in a Northern California county from 1995 through 1997. Almost all of the Hispanics living in this county are Mexican American. The
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
965
sample of over 16,000 individuals was provided by the California Department of Motor Vehicles (DMV). 2.1. Data from DMV records The DMV provided data for the entire sample of offenders on the following variables: gender, age, ethnicity, previous DUI convictions in the last year and in the last 5 years, and DUI convictions in the year following the identifying arrest. The DMV category of DUI offenses and convictions also included offenses and convictions for hit-and-run and reckless driving offenses. Data were also obtained on whether the offender was referred to a DUI treatment program. About 76% of the identifying DUI arrests resulted in convictions, and of those 78% resulted in referral to a DUI treatment program. 2.2. Data analysis Recidivism was measured by whether or not the individual was subsequently convicted of a DUI offense within a year following the identifying DUI offense. Data were analyzed comparing sample proportions between recidivists and nonrecidivists, separately for Mexican Americans and for whites, and proportions of recidivists between ethnic groups, on demographic and DUI offense characteristics (Table 1). Recidivism rates for Mexican Americans and whites are reported by conviction status of the identifying DUI offense (Table 2), and logistic regression is then used to evaluate the predictive value on recidivism of ethnicity, gender, age, conviction status, program referral, and previous DUI conviction in the last year and last 5 years (Table 3). Age is coded as a categorical variable (18–21=0, 22–
Table 1 Demographic and DUI offense characterization by DUI recidivism (%)
Gender Male Female Age 18 – 21 22 – 29 30 – 39 40+ Conviction Program referral DUI conviction last 5 years DUI conviction last year
Mexican American recidivism
White recidivism
Yes (51)
No (438)
Yes (31)
No (459)
96 4
91 9
77a 23a
81 19
12 41 23 24 90 53 35 24
12 40 32 16 78* 57 25 7*
10 3a 48a 39 55a 45 23 7a
8 29* 32 31 75* 64* 20 4
a P<.05, comparison of proportions of recidivism between Mexican Americans and whites. * P<.05, comparison of proportions of recidivism vs. nonrecidivism within ethnic groups.
966
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
Table 2 Recidivism rates for Mexican Americans and whites by conviction status (%)
% Recidivists
Mexican American recidivism
White recidivism
Arrested (101)
Convicted (388)
Arrested (129)
Convicted (361)
5
12*
11
5*, a
a
P<.05, comparison of proportion of recidivism between Mexican American and whites. * P<.05, comparison of proportion of recidivism within ethnic groups.
29=1, 30–39=3, 40+=4). Variables are entered simultaneously in the model. Odds ratios (OR) and 95% confidence intervals (CI) are reported for variables in the model. Separate models are then examined for Mexican Americans and whites.
3. Results The rate of recidivism was found to be significantly higher for Mexican Americans (10%) compared to whites (6%) ( P<.05) (not shown). Among Mexican Americans the identifying DUI offense was significantly more likely ( P<.05) to result in a conviction (79%) compared to whites (73%), but among those convicted, Mexican Americans were significantly less likely ( P<.01) to be referred to a DUI treatment program (72% vs. 85%) (not shown). Table 1 shows demographic and offense characteristics by DUI recidivism, separately for Mexican Americans and whites. Among Mexican Americans, recidivists were significantly more likely to have the identifying DUI offense result in a conviction compared to nonrecidivists, while the opposite was true for white recidivists. Mexican American recidivists were also over three times more likely to have had a DUI conviction in the year previous to the identifying offense compared to nonrecidivists. Among whites, recidivists were older and were less likely to have been referred to a DUI treatment program compared to nonrecidivists. Mexican American recidivists
Table 3 OR and 95% CI for demographic and offense characteristics on DUI recidivism (n=979) Covariate
OR
CI
Gender (female) Age Ethnicity (white) Conviction Program Conviction—last 5 years Conviction—last year
1.10 1.24 1.66* 1.21 0.62 0.79 3.66**
0.52 – 2.3 0.96 – 1.60 1.01 – 2.71 0.58 – 2.50 0.34 – 1.12 0.39 – 1.60 1.54 – 8.70
* P<.05, c2 statistic with 1 df. ** P<.01, c2 statistic with 1 df.
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
967
compared to white recidivists were more likely to be male, under 30, to have the identifying DUI offense result in a conviction, and to have had a DUI conviction in the previous year. Rates of recidivism varied by conviction status of the identifying offense between Mexican Americans and whites (Table 2). Among those convicted of DUI, Mexican Americans were significantly more likely than whites to be recidivists in the following year, while among those arrested, but not convicted, whites were more likely to be recidivists. Table 3 shows the OR and 95% CI for demographic and DUI offense characteristics on recidivism. Mexican Americans were over one-and-a-half times more likely than whites to be recidivists, and those with a DUI conviction in the previous year were over three times more likely than those without to be recidivists in the following year. Logistic regressions, which included age, gender, conviction status of the identifying DUI offense, and DUI conviction in the preceding year, were then examined separately for Mexican Americans and whites (not shown). Program referral was not included in the models since only those whose identifying offense resulted in a conviction could have been referred to a treatment program. Among Mexican Americans, while conviction status was not found to be a significant predictor of recidivism (OR=2.5; P=.055), those with a DUI conviction in the previous year were three times more likely to be recidivists (OR=3.45; CI=1.62–7.36). Among whites, however, conviction status was a significant predictor of recidivism, with those whose identifying offense resulted in a conviction about a third as likely (OR=0.37; CI=0.17–0.78) to be a recidivist compared to those whose identifying offense did not result in a DUI conviction. Previous conviction in the last year was not a significant predictor of DUI recidivism in this model for whites. Separate regressions, which included program referral, were then examined by ethnicity for those whose identifying offense resulted in a conviction. Program referral was not found to be a significant predictor of recidivism for either Mexican Americans or whites. Among whites, however, those with a DUI conviction in the previous year were three times more likely to be recidivists than those without a previous DUI conviction (OR=3.15; CI=1.44– 6.88), while for Mexican Americans, a DUI conviction in the last year was not significant among those convicted.
4. Discussion Overall, rates of recidivism were significantly higher for Mexican Americans compared to whites, however, among those arrested but not convicted, recidivism rates were higher among whites. Among Mexican Americans, a conviction is not significantly predictive of recidivism at the .05 level, although regression coefficients suggest that a conviction is positively associated with recidivism. Previous convictions were predictive of DUI recidivism among Mexican Americans, but among whites, only for those convicted. These data suggest that a conviction appears to be protective against recidivism among whites, controlling for prior DUI convictions. Whites who are arrested, but not convicted, may believe they can drink while driving without serious repercussions, but once convicted are less likely to have a repeat occurrence within the next year, due to license suspension, jail sentences, or other
968
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
sanctions. Among Mexican Americans, on the other hand, a DUI conviction is not necessarily a detriment to subsequent DUI convictions, and may be related to their experiences of DUI in their home country. Among those convicted, referral to a DUI treatment program was not significantly associated with DUI recidivism for either Mexican Americans or whites. This finding is not surprising since referral to a program does not necessarily mean that the individual either attended or completed the program. Given the lag time between DUI conviction, referral to a treatment program and actual attendance at a program, a 1-year period following the DUI offense may not have provided sufficient opportunity for offenders to enter or complete a program. A limitation to these data is that DUI arrests and convictions are confounded with arrests and convictions for hit-and-run and reckless driving offenses, many of which probably were not alcohol-related, although the proportions of those referred to a DUI treatment program suggest that the majority (78%) of the identifying offenses were for DUI, and not hit-and-run or reckless driving. The identifying DUI offense was significantly more likely to result in a conviction among Mexican Americans compared to whites, and among those convicted, Mexican Americans were significantly less likely ( P<.01) to be referred to a DUI treatment program than whites. Again, since the DUI category also includes other non-DUI offenses, and since all of those convicted of an actual DUI are not referred to treatment, we do not know whether Mexican Americans arrested for a DUI are more likely to be convicted than whites, but less likely to be referred to treatment. Given the opposite association of DUI conviction with recidivism for Mexican Americans compared to whites in these data, and the potential of a DUI treatment program for reducing DUI recidivism, these are important research questions which need to be addressed. Given the size and availability of DMV records, further analysis with larger samples, as well as alternative methods to the classical hypothesis tests and the distribution theory upon which they are based, is warranted. One area of interest is the extent to which models can distinguish individuals’ propensity towards recidivism based on prior information. One way to measure our ability to discriminate between recidivists and nonrecidivists is the c statistic, or the proportion of all possible pairs predicted probabilities from a model where the predictive probability of a subsequent recidivist is higher than that of a nonsubsequent recidivist. The value using logistic regression for our sample was .65, which is not much higher than the value of .5 obtained from assigning recidivists and nonrecidivists predictive probabilities by randomly generating numbers from 0 to 1. Larger samples containing more information on recidivists’ behaviors and characteristics may increase such predictive ability. Another area of concern is statements regarding significance of coefficients from models, and the stability of such statements under different samples from the population. Not only would resampling techniques within the current sample address questions of such stability, but other samples of similar size from the population could be utilized as well in such a study. Although much of the current statistical literature in the alcohol field, and the social sciences in general, relies on model assumptions, the availability of larger samples here opens the door to study the important questions of whether our findings are consistent in the presence of additional data. Despite the limitations of this study, data from this representative sample of all DUI offenders from an entire Northern California county suggest that conviction for a DUI may
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
969
not play the same role in the likelihood of subsequent DUI convictions for Mexican Americans as for whites, and this difference may need to be considered in DUI treatment programs. Additional research on ethnic differences in DUI offenses and recidivism over longer periods of follow-up is needed to determine ethnic-specific intervention and prevention strategies for DUI.
Acknowledgements This study was supported by a National Alcohol Research Center Grant (AA 05595) from the U.S. Institute on Alcohol Abuse and Alcoholism. We would like to acknowledge the California State Department of Motor Vehicles for providing the data for this study.
References Arstein-Kerslake, G. W., & Peck, R. C. (1985). A typological analysis of California DUI offenders and DUI recidivism correlates. Technical Report, National Highway Traffic Safety Administration. Caetano, R., & Clark, C. L. (2000). Hispanics, blacks and whites driving under the influence of alcohol: results from the 1995 National Alcohol Survey. Accident Analysis and Prevention, 32, 57 – 64. Caetano, R., & Medina Mora, M. E. (1986). Immigration, acculturation and alcohol use: a comparison between people of Mexican descent in Mexico and the US. Berkeley, CA: Alcohol Research Group. Chang, I., Lapham, S. C., & Barton, K. J. (1996). Drinking environment and sociodemographic factors among DWI offenders. Journal of Studies on Alcohol, 57, 659 – 669. Ellingstad, V. S. (1974). An analysis of alcohol rehabilitation efforts. Analytical Study 6. Contract No. DOT-HS045-1-061, National Highway Traffic Safety Administration. Lapham, S. C., Skipper, B. J., Hunt, W. C., & Chang, L. (2000). Do risk factors for re-arrest differ for female and male drunk-driving offenders? Alcoholism: Clinical and Experimental Research, 24, 1647 – 1655. McMillen, D. L., Adams, M. S., Wells-Parker, E., Pang, M. G., & Anderson, B. (1992). Personality traits and behaviors of alcohol-impaired drivers: a comparison of first and multiple offenders. Addictive Behaviors, 17, 407 – 414. Peck, R. C., Arstein-Kerslake, G. W., & Helander, C. J. (1994). Psychometric and biographical correlates of drunk-driving recidivism and treatment program compliance. Journal of Studies in Alcohol, 55, 667 – 678. Perrine, M. W., & Arce-Quinones, W. M. (1994). Hispanic and Anglo DUI offenders. Research Society on Alcoholism meeting, Maui, HI, June. Perrine, M. W., Peck, R. C., & Fell, J. C. (1989). Epidemiologic perspectives on drunk driving Surgeon General’s Workshop on Drunk Driving: background papers ( pp. 35 – 76). Washington, DC: U.S. Department of Health and Human Services. Pollack, S., Didenko, O., McEachern, A., & Berger, R. M. (1972). Drinking driver and traffic safety report: final report, vol. 1. Prepared for the National Highway Traffic Safety Administration, Publication No. PB 212-252, National Technical Information Service. Simpson, H. M. (1995). Who is the persistent drinking driver? Part II: Canada and elsewhere. Transportation Research Circular, 437, 21 – 25. Vingilis, E., Stoduto, G., Macartney-Filgate, M. S., Liban, C. B., & Mclellan, B. A. (1994). Psychosocial characteristics of alcohol-involved and nonalcohol-involved seriously injured drivers. Accident: Analysis and Prevention, 26, 195 – 206.
Short Communication
DUI recidivism A comparison of Mexican Americans and whites in a northern California county$ Cheryl J. Cherpitel*, Jason Bond Alcohol Research Group, Public Health Institute, 2000 Hearst Avenue, Berkeley, CA 94709, USA
Abstract Objective: While Hispanic offenders for driving under the influence of alcohol (DUI) in California are more likely to have a history of multiple offenses compared to whites, little is known about characteristics associated with DUI recidivism in either ethnic group. Demographic and DUI conviction characteristics associated with DUI recidivism are analyzed among Mexican American and white DUI offenders in a Northern California county. Method: A sample of 459 primarily Mexican Americans and 490 whites were randomly selected from records supplied by the California Department of Motor Vehicles (DMV) on over 16,000 DUI offenders in the county during a 3-year period. DMV data on DUI convictions in the 1- and 5-year period preceding the identifying DUI offense and the year following the identifying offense (the recidivism conviction) were analyzed. Results: Rates of recidivism were significantly lower for Mexican Americans (5%) compared to whites (11%) who were arrested for DUI but not convicted, but higher for those convicted (12% vs. 5%). While conviction status of the identifying DUI offense was not predictive of recidivism among Mexican Americans, a DUI conviction in the preceding year was significantly predictive. Among white arrestees, receiving a conviction was significantly, and negatively, predictive of recidivism, but among those who were convicted, a previous DUI conviction in the last year was predictive of recidivism. Referral to a DUI treatment program was not a significant predictor of recidivism among those convicted in either ethnic group. Conclusions: The data suggest that conviction for a DUI may not play the same role in the likelihood of subsequent DUI convictions for Mexican Americans as for whites, and this difference may need to be considered in DUI treatment programs. Additional research
$
Presented at the Annual Epidemiology Symposium of the Kettil Bruun Society for Social and Epidemiological Research on Alcohol, Toronto, Canada, May 28 – June 1, 2001. * Corresponding author. Tel.: +1-510-642-5208; fax: +1-510-642-7172. E-mail address: [email protected] (C.J. Cherpitel). 0306-4603/02/$ – see front matter D 2002 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0306-4603(01)00293-3
964
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
on ethnic differences in DUI offenses and recidivism over longer periods of follow-up is needed to determine ethnic-specific intervention and prevention strategies for DUI. D 2002 Elsevier Science Ltd. All rights reserved. Keywords: Alcohol DUI; Mexican Americans
1. Introduction Rates of offense for driving under the influence of alcohol (DUI) have been found to vary by ethnicity in California, with Hispanics overrepresented among DUI offenders (Chang, Lapham, & Barton, 1996; Perrine, Peck, & Fell, 1989). Hispanic DUI offenders have also been found to be more likely to have a history of multiple offenses compared to whites (Perrine & Arce-Quinones, 1994), but little is known about those characteristics which may be associated with DUI recidivism in general, or variables which may be predictive of recidivism in specific ethnic groups. Studies which have examined characteristics associated with DUI recidivism (without respect to ethnicity) have found prior DUI convictions and other traffic-related arrests, younger age, lower education, and heavier drinking all significant predictors (Arstein-Kerslake & Peck, 1985; Ellingstad, 1974; Lapham, Skipper, Hunt, & Chang, 2000; McMillen, Adams, Wells-Parker, Pang, & Anderson, 1992 ; Peck, ArsteinKerslake, & Helander, 1994; Pollack, Didenko, McEachern, & Berger, 1972). Higher DUI offense rates for Hispanics compared to whites, and higher rates of recidivism, may be related to patterns of drinking in this ethnic group, as well as to a lack of familiarity with DUI laws and their enforcement relative to their home country (Caetano & Clark, 2000). The majority of Hispanics in California are Mexican American, and general population surveys of drinking patterns in Mexico have found higher rates of infrequent but heavy drinking leading to intoxication among males than those found among Anglo males (Caetano & Medina Mora, 1986). DUI recidivism is a major public health problem—35% to 40% of all fatally injured drinking drivers are estimated to have had at least one prior DUI offense (Simpson, 1995; Vingilis, Stoduto, Macartney-Filgate, Liban, & Mclellan, 1994). To begin filling this important gap in the literature regarding those variables associated with DUI recidivism, data on demographic characteristics, prior DUI convictions, and recidivism rates are reported on a sample of Mexican American and white DUI offenders in one California county during a 3-year period.
2. Methods The sample consisted of 979 individuals, stratified on ethnicity (white and Mexican American), randomly selected from all those individuals arrested for (but not necessarily convicted of) at least one DUI offense in a Northern California county from 1995 through 1997. Almost all of the Hispanics living in this county are Mexican American. The
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
965
sample of over 16,000 individuals was provided by the California Department of Motor Vehicles (DMV). 2.1. Data from DMV records The DMV provided data for the entire sample of offenders on the following variables: gender, age, ethnicity, previous DUI convictions in the last year and in the last 5 years, and DUI convictions in the year following the identifying arrest. The DMV category of DUI offenses and convictions also included offenses and convictions for hit-and-run and reckless driving offenses. Data were also obtained on whether the offender was referred to a DUI treatment program. About 76% of the identifying DUI arrests resulted in convictions, and of those 78% resulted in referral to a DUI treatment program. 2.2. Data analysis Recidivism was measured by whether or not the individual was subsequently convicted of a DUI offense within a year following the identifying DUI offense. Data were analyzed comparing sample proportions between recidivists and nonrecidivists, separately for Mexican Americans and for whites, and proportions of recidivists between ethnic groups, on demographic and DUI offense characteristics (Table 1). Recidivism rates for Mexican Americans and whites are reported by conviction status of the identifying DUI offense (Table 2), and logistic regression is then used to evaluate the predictive value on recidivism of ethnicity, gender, age, conviction status, program referral, and previous DUI conviction in the last year and last 5 years (Table 3). Age is coded as a categorical variable (18–21=0, 22–
Table 1 Demographic and DUI offense characterization by DUI recidivism (%)
Gender Male Female Age 18 – 21 22 – 29 30 – 39 40+ Conviction Program referral DUI conviction last 5 years DUI conviction last year
Mexican American recidivism
White recidivism
Yes (51)
No (438)
Yes (31)
No (459)
96 4
91 9
77a 23a
81 19
12 41 23 24 90 53 35 24
12 40 32 16 78* 57 25 7*
10 3a 48a 39 55a 45 23 7a
8 29* 32 31 75* 64* 20 4
a P<.05, comparison of proportions of recidivism between Mexican Americans and whites. * P<.05, comparison of proportions of recidivism vs. nonrecidivism within ethnic groups.
966
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
Table 2 Recidivism rates for Mexican Americans and whites by conviction status (%)
% Recidivists
Mexican American recidivism
White recidivism
Arrested (101)
Convicted (388)
Arrested (129)
Convicted (361)
5
12*
11
5*, a
a
P<.05, comparison of proportion of recidivism between Mexican American and whites. * P<.05, comparison of proportion of recidivism within ethnic groups.
29=1, 30–39=3, 40+=4). Variables are entered simultaneously in the model. Odds ratios (OR) and 95% confidence intervals (CI) are reported for variables in the model. Separate models are then examined for Mexican Americans and whites.
3. Results The rate of recidivism was found to be significantly higher for Mexican Americans (10%) compared to whites (6%) ( P<.05) (not shown). Among Mexican Americans the identifying DUI offense was significantly more likely ( P<.05) to result in a conviction (79%) compared to whites (73%), but among those convicted, Mexican Americans were significantly less likely ( P<.01) to be referred to a DUI treatment program (72% vs. 85%) (not shown). Table 1 shows demographic and offense characteristics by DUI recidivism, separately for Mexican Americans and whites. Among Mexican Americans, recidivists were significantly more likely to have the identifying DUI offense result in a conviction compared to nonrecidivists, while the opposite was true for white recidivists. Mexican American recidivists were also over three times more likely to have had a DUI conviction in the year previous to the identifying offense compared to nonrecidivists. Among whites, recidivists were older and were less likely to have been referred to a DUI treatment program compared to nonrecidivists. Mexican American recidivists
Table 3 OR and 95% CI for demographic and offense characteristics on DUI recidivism (n=979) Covariate
OR
CI
Gender (female) Age Ethnicity (white) Conviction Program Conviction—last 5 years Conviction—last year
1.10 1.24 1.66* 1.21 0.62 0.79 3.66**
0.52 – 2.3 0.96 – 1.60 1.01 – 2.71 0.58 – 2.50 0.34 – 1.12 0.39 – 1.60 1.54 – 8.70
* P<.05, c2 statistic with 1 df. ** P<.01, c2 statistic with 1 df.
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
967
compared to white recidivists were more likely to be male, under 30, to have the identifying DUI offense result in a conviction, and to have had a DUI conviction in the previous year. Rates of recidivism varied by conviction status of the identifying offense between Mexican Americans and whites (Table 2). Among those convicted of DUI, Mexican Americans were significantly more likely than whites to be recidivists in the following year, while among those arrested, but not convicted, whites were more likely to be recidivists. Table 3 shows the OR and 95% CI for demographic and DUI offense characteristics on recidivism. Mexican Americans were over one-and-a-half times more likely than whites to be recidivists, and those with a DUI conviction in the previous year were over three times more likely than those without to be recidivists in the following year. Logistic regressions, which included age, gender, conviction status of the identifying DUI offense, and DUI conviction in the preceding year, were then examined separately for Mexican Americans and whites (not shown). Program referral was not included in the models since only those whose identifying offense resulted in a conviction could have been referred to a treatment program. Among Mexican Americans, while conviction status was not found to be a significant predictor of recidivism (OR=2.5; P=.055), those with a DUI conviction in the previous year were three times more likely to be recidivists (OR=3.45; CI=1.62–7.36). Among whites, however, conviction status was a significant predictor of recidivism, with those whose identifying offense resulted in a conviction about a third as likely (OR=0.37; CI=0.17–0.78) to be a recidivist compared to those whose identifying offense did not result in a DUI conviction. Previous conviction in the last year was not a significant predictor of DUI recidivism in this model for whites. Separate regressions, which included program referral, were then examined by ethnicity for those whose identifying offense resulted in a conviction. Program referral was not found to be a significant predictor of recidivism for either Mexican Americans or whites. Among whites, however, those with a DUI conviction in the previous year were three times more likely to be recidivists than those without a previous DUI conviction (OR=3.15; CI=1.44– 6.88), while for Mexican Americans, a DUI conviction in the last year was not significant among those convicted.
4. Discussion Overall, rates of recidivism were significantly higher for Mexican Americans compared to whites, however, among those arrested but not convicted, recidivism rates were higher among whites. Among Mexican Americans, a conviction is not significantly predictive of recidivism at the .05 level, although regression coefficients suggest that a conviction is positively associated with recidivism. Previous convictions were predictive of DUI recidivism among Mexican Americans, but among whites, only for those convicted. These data suggest that a conviction appears to be protective against recidivism among whites, controlling for prior DUI convictions. Whites who are arrested, but not convicted, may believe they can drink while driving without serious repercussions, but once convicted are less likely to have a repeat occurrence within the next year, due to license suspension, jail sentences, or other
968
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
sanctions. Among Mexican Americans, on the other hand, a DUI conviction is not necessarily a detriment to subsequent DUI convictions, and may be related to their experiences of DUI in their home country. Among those convicted, referral to a DUI treatment program was not significantly associated with DUI recidivism for either Mexican Americans or whites. This finding is not surprising since referral to a program does not necessarily mean that the individual either attended or completed the program. Given the lag time between DUI conviction, referral to a treatment program and actual attendance at a program, a 1-year period following the DUI offense may not have provided sufficient opportunity for offenders to enter or complete a program. A limitation to these data is that DUI arrests and convictions are confounded with arrests and convictions for hit-and-run and reckless driving offenses, many of which probably were not alcohol-related, although the proportions of those referred to a DUI treatment program suggest that the majority (78%) of the identifying offenses were for DUI, and not hit-and-run or reckless driving. The identifying DUI offense was significantly more likely to result in a conviction among Mexican Americans compared to whites, and among those convicted, Mexican Americans were significantly less likely ( P<.01) to be referred to a DUI treatment program than whites. Again, since the DUI category also includes other non-DUI offenses, and since all of those convicted of an actual DUI are not referred to treatment, we do not know whether Mexican Americans arrested for a DUI are more likely to be convicted than whites, but less likely to be referred to treatment. Given the opposite association of DUI conviction with recidivism for Mexican Americans compared to whites in these data, and the potential of a DUI treatment program for reducing DUI recidivism, these are important research questions which need to be addressed. Given the size and availability of DMV records, further analysis with larger samples, as well as alternative methods to the classical hypothesis tests and the distribution theory upon which they are based, is warranted. One area of interest is the extent to which models can distinguish individuals’ propensity towards recidivism based on prior information. One way to measure our ability to discriminate between recidivists and nonrecidivists is the c statistic, or the proportion of all possible pairs predicted probabilities from a model where the predictive probability of a subsequent recidivist is higher than that of a nonsubsequent recidivist. The value using logistic regression for our sample was .65, which is not much higher than the value of .5 obtained from assigning recidivists and nonrecidivists predictive probabilities by randomly generating numbers from 0 to 1. Larger samples containing more information on recidivists’ behaviors and characteristics may increase such predictive ability. Another area of concern is statements regarding significance of coefficients from models, and the stability of such statements under different samples from the population. Not only would resampling techniques within the current sample address questions of such stability, but other samples of similar size from the population could be utilized as well in such a study. Although much of the current statistical literature in the alcohol field, and the social sciences in general, relies on model assumptions, the availability of larger samples here opens the door to study the important questions of whether our findings are consistent in the presence of additional data. Despite the limitations of this study, data from this representative sample of all DUI offenders from an entire Northern California county suggest that conviction for a DUI may
C.J. Cherpitel, J. Bond / Addictive Behaviors 28 (2003) 963 – 969
969
not play the same role in the likelihood of subsequent DUI convictions for Mexican Americans as for whites, and this difference may need to be considered in DUI treatment programs. Additional research on ethnic differences in DUI offenses and recidivism over longer periods of follow-up is needed to determine ethnic-specific intervention and prevention strategies for DUI.
Acknowledgements This study was supported by a National Alcohol Research Center Grant (AA 05595) from the U.S. Institute on Alcohol Abuse and Alcoholism. We would like to acknowledge the California State Department of Motor Vehicles for providing the data for this study.
References Arstein-Kerslake, G. W., & Peck, R. C. (1985). A typological analysis of California DUI offenders and DUI recidivism correlates. Technical Report, National Highway Traffic Safety Administration. Caetano, R., & Clark, C. L. (2000). Hispanics, blacks and whites driving under the influence of alcohol: results from the 1995 National Alcohol Survey. Accident Analysis and Prevention, 32, 57 – 64. Caetano, R., & Medina Mora, M. E. (1986). Immigration, acculturation and alcohol use: a comparison between people of Mexican descent in Mexico and the US. Berkeley, CA: Alcohol Research Group. Chang, I., Lapham, S. C., & Barton, K. J. (1996). Drinking environment and sociodemographic factors among DWI offenders. Journal of Studies on Alcohol, 57, 659 – 669. Ellingstad, V. S. (1974). An analysis of alcohol rehabilitation efforts. Analytical Study 6. Contract No. DOT-HS045-1-061, National Highway Traffic Safety Administration. Lapham, S. C., Skipper, B. J., Hunt, W. C., & Chang, L. (2000). Do risk factors for re-arrest differ for female and male drunk-driving offenders? Alcoholism: Clinical and Experimental Research, 24, 1647 – 1655. McMillen, D. L., Adams, M. S., Wells-Parker, E., Pang, M. G., & Anderson, B. (1992). Personality traits and behaviors of alcohol-impaired drivers: a comparison of first and multiple offenders. Addictive Behaviors, 17, 407 – 414. Peck, R. C., Arstein-Kerslake, G. W., & Helander, C. J. (1994). Psychometric and biographical correlates of drunk-driving recidivism and treatment program compliance. Journal of Studies in Alcohol, 55, 667 – 678. Perrine, M. W., & Arce-Quinones, W. M. (1994). Hispanic and Anglo DUI offenders. Research Society on Alcoholism meeting, Maui, HI, June. Perrine, M. W., Peck, R. C., & Fell, J. C. (1989). Epidemiologic perspectives on drunk driving Surgeon General’s Workshop on Drunk Driving: background papers ( pp. 35 – 76). Washington, DC: U.S. Department of Health and Human Services. Pollack, S., Didenko, O., McEachern, A., & Berger, R. M. (1972). Drinking driver and traffic safety report: final report, vol. 1. Prepared for the National Highway Traffic Safety Administration, Publication No. PB 212-252, National Technical Information Service. Simpson, H. M. (1995). Who is the persistent drinking driver? Part II: Canada and elsewhere. Transportation Research Circular, 437, 21 – 25. Vingilis, E., Stoduto, G., Macartney-Filgate, M. S., Liban, C. B., & Mclellan, B. A. (1994). Psychosocial characteristics of alcohol-involved and nonalcohol-involved seriously injured drivers. Accident: Analysis and Prevention, 26, 195 – 206.