Race, Gender, and Risk Perceptions of the Legal Consequences of Drinking and Driving

Abstract

Introduction

This study investigated whether subjective beliefs about driving while intoxicated (DWI) consequences differ by race/gender.

Method

Beliefs affect driving behaviors and views of police/judicial fairness. The researchers compared risk perceptions of DWI using a survey of drinkers in eight cities in four states with actual arrest, conviction, and fine rates from court data in the same cities.

Results

With state arrest data as a benchmark, Black males were overly pessimistic about being stopped, whether or not actual drinking occurred, and attributed higher jail penalties to DWI conviction. That Black males overestimated jail sentences incurred by the general population suggests that they did not attribute higher jail penalties to racial bias. Arrest data did not reveal disparities in judicial outcomes following DWI arrest.

Conclusions

Blacks’ subjective beliefs about DWI consequences may reflect social experiences, which are not jurisdiction- or crime-specific; this is a challenge to policymakers aiming to deter DWI by changing statutes and enforcement.

1. Introduction

Motor vehicle accidents are the leading cause of death among persons under age 24 in the United States and the fifth leading cause of death among those aged 25–44 (Hoyert & Xu, 2012). Driving while intoxicated (DWI) is a leading cause of motor vehicle accidents leading to fatalities (Brewer et al., 1994; Vaughn et al., 2011). Problem drinking has been much less prevalent among females but differences among genders are shrinking (Keyes et al., 2011). For some age groups, alcohol consumption is lower among Blacks than Whites; for all ages, Blacks have lower rates of abstention than Whites (Caetano et al., 1998). Accounting for number of drinking sessions and average number of drinks per session, Birdsall et al. (2012) found no statistical difference between Blacks and Whites in driving “when you’ve had perhaps too much to drink.” Some studies show that Blacks experience higher rates of adverse effects of drinking than predicted from estimated alcohol use (Costanzo et al., 2007; Hasin et al., 2007), but overall, findings are mixed (Keyes et al., 2012). Although treating the underlying addiction is performed to reduce DWI, major policy interventions remain arrest, conviction, and various penalties (Taxman & Piquero 1998; Dula et al., 2007). Such policies may be considered successful if they deter DWI. Yet such policies should be implemented in ways considered to be fair.

In this study, we assessed relationships between race and gender, drinking and driving behaviors, and legal consequences of DWI, both perceived and actual. Rather than just focus on the rates of being stopped by police for DWI, we analyzed the entire chain of events following being stopped to penalties if convicted. Assessing the arrest resolution process allowed us to examine the possibility that disparities at one stage are offset by decisions in the opposite direction at other stages. Perceptions are important to the extent they influence driving behavior and reflect how people view equity in law enforcement and judicial processes. Laws, and practices implementing such laws, are unlikely to affect decisions if people are uninformed about them. Existing studies on risk perceptions are motivated by the premise that deterrence depends on the perceptions of the risk of penalties from offending (e.g., arrest and conviction; Nagin, 1998). While there is some literature on subjective beliefs and risk perceptions of apprehension for criminal activity (Apel, 2012) as well as research on the relationship between other demographic factors, such as age, on risk perceptions and crime (e.g., Hjalmarsson, 2009), there currently is a lack of systematically collected evidence of subjective beliefs as they differ by race.

If population subgroups underestimate the sanction rates, this is a likely source of under-deterrence. However, if the sanction rates are overestimated, this may have three adverse effects. First, if certain population groups believe that laws are enforced inequitably, this may also adversely influence compliance with such laws. If there is overestimation in arrest rates, these beliefs may cause people to view the criminal justice system as inherently unfair, and this may undermine civic participation (Bobo & Gilliam, 1990; Weaver & Lerman, 2010). Second, overestimation and consequently over-deterrence may result in reductions in the rate of crime in question, a desirable outcome, but individuals may substitute other undesirable behaviors for the crime in question. This type of compensatory behavior has been documented in other safety issue contexts (e.g., for seatbelt requirements; Peltzman, 1975). In the context of drinking and driving, having to drive home from a bar or restaurant might cause a restraint on an individual’s drinking. However, in response to overestimation of the strictness of laws, the drinker may be more likely to find a sober person to drive him home, thus removing the restraint on drinking at the bar or restaurant and causing the individual to consume more alcohol than he or she otherwise would.

In this study, we compared risk perceptions to objective data obtained from arrest records in the same eight U.S. cities. Unlike past studies of stops and sentences following conviction, we included risk perceptions of DWI sanctions by race and gender and actual rates of DWI arrest and judicial outcomes following arrest in the same study.

2. Methods

2.1. Data on Risk Perceptions, Alcohol Consumption, and Driving Behaviors

Our survey, the Survey of Alcohol and Driving (SAD), obtained detailed information on alcohol consumption, drinking and drinking and driving behaviors, addiction, substance use other than alcohol, risk perceptions, knowledge of DWI statutes and judicial practices, personality and demographic characteristics, and income.¹ When possible, questionnaire design was guided by questions asked in prior surveys, albeit not all in the same instrument. No prior survey combines detailed questions on subjective beliefs, particularly about legal consequences of drinking and driving with questions about alcohol consumption and its determinants. Wave 1 was conducted using Computer Assisted Telephone Interviews (CATI), and the other two waves were conducted through the web using Computer Assisted Self-Administered Interviews (CASI). Wave 1, which focused on questions with short answers, more easily answered by telephone. Questions about expectations were asked in Wave 2. The use of the web-based surveys in Waves 2 and 3 allowed us to ask questions which were more complex and included a visual display to aid eliciting information from respondents. This study relied on data from Waves 1 and 2. Wave 3 was a shorter follow-up to Wave 2, designed to allow comparison of expectations about future behaviors with actual self-reported behaviors measured a year later. Wave 1 had 1,520 respondents, 4 of whom identified themselves as Hispanic. The four observations were dropped because of an insufficient number of observations for this population sub-group, yielding a net sample of 1,516 observations. Wave 2 had 1,291 respondents. Due to the use of conditional questions, which skips questions based on responses, the number of responses to some questions were considerably lower than the number of observations for the wave.

Notably, SAD was not about race. Questions about race and gender and past driving history were asked in Wave 1; race was not mentioned in the study description used to introduce SAD to potential respondents or in subsequent interviews.

Since the focus was on DWI, SAD excluded persons reporting no alcohol consumption or driving in the past month during the screener interview. We deliberately oversampled persons who consumed large amounts of alcohol and were prone to DWI to allow us to study decision-making processes and behaviors of such individuals in detail.

Battelle Memorial Institute conducted the three-wave survey of drinking and driving on our behalf in eight cities in four states during 2009–2012: Raleigh and Hickory, North Carolina; Philadelphia and Wilkes-Barre, Pennsylvania; Seattle and Yakima, Washington; and Milwaukee and La Crosse, Wisconsin. We drew our sample from a limited number of cities because the SAD asked about DWI laws and motor vehicle insurance specific to the respondent’s location. We selected these cities to represent a broad geographic spread of large and small cities. While no eight-city sample can be nationally representative, the four study states vary in severity of their DWI problems, e.g., per capita consumption of ethanol in gallons in 2007—ranging from 2.0 in North Carolina to 3.0 in Wisconsin (National Institute on Alcohol Abuse and Alcoholism 2009). The four states also differ in their DWI prevention laws, demographic composition, and histories as applied to race with North Carolina, but not the other states, having a history of legal segregation. Differences among the states exist for sentencing and fines. Scheduling criminal penalties has become commonplace, including among the four study states. The state statutes contain minimum or maximum penalties according to a classification system that differs among the states. The scheduled penalties varied substantially among the four study states. For example, North Carolina and Washington had minimum jail terms for first time DWI offenders while Pennsylvania and Wisconsin did not. North Carolina specified a $200 maximum fine for the lowest sentencing level while the other states specified minimums of at least $150 (in Wisconsin).² Arrest per capita population ratios varied from 0.25% in Washington to 0.67% in Wisconsin in 2009.

We also obtained state court data from each state on individual arrests for DWI in 2009. Except for Washington, which did not include jail and fine amounts, the data contained original charges, reduced charges resulting from plea agreements, and information on conviction and sentencing. From these data, we computed objective measures of arrest, conviction, and penalties by race/gender for the same eight cities where SAD was conducted.

2.2. Statistical Analysis: Empirical Specification

2.2.1. Overview

We analyzed three types of dependent variables using SAD data. The first were binary variables for self-reports of a DWI arrest and a citation for speeding 15+ miles per hour (mph) over the speed limit within the three years before Wave 1. We conducted this analysis to determine whether there were differences by race and gender in experiencing legal consequences related to drinking and driving behaviors that might affect risk perceptions of DWI sanctions. We also compared experiences of sample persons to DWI arrest data in the eight cities in which they resided. The second dependent variable type consisted of expected driving behaviors during the year following Wave 2 to determine whether respondents expected to be at greater or lesser risk of being convicted and penalized for DWI. Third, we analyzed expected legal consequences of driving while intoxicated to gauge subjective beliefs about conviction rates and penalty amounts for DWI.

2.2.2. Dependent Variables: Driving History

We assessed three dependent variables for driving history: the number of times a person when she/he was slightly intoxicated (e.g., felt a little tipsy); whether the person had been arrested for a DWI in the last three years; and whether the person was cited for speeding 15+ mph above the speed limit in the last three years.

2.2.3. Dependent Variable: Subjective Beliefs about Being Stopped, Arrested, and Convicted

We defined three dependent variables for being stopped. The SAD asked respondents about being stopped for DWI in two ways. The first question was only asked of persons intending to drink and drive in the next year; the second asked everyone about the percent chance of being stopped for DWI. The first was based on responses to a question, “What is the percent chance you will be pulled over for DWI in the next year? Being pulled over is a routine sobriety check when the officer has suspicion that you might be driving under the influence.” The second and third questions were, respectively, “On a given occasion, when you drive after you have had too much to drink, what is the percent chance you will be pulled over?” and “On an average weekend evening, what percent of drivers on the road who have had to much to drink will be pulled over?” These types of questions may be difficult to answer in a meaningful way. However, based on a question by question assessment of accuracy, answers to some questions, e.g., “what is the percent chance you will be pulled over,” were surprisingly accurate (Sloan et al., forthcoming).

Chances of being stopped after having too much to drink were asked in the third person to assess the Lake Wobegon Effect (i.e., respondents think that they are better than average; Maxwell & Lopus, 1994). People may believe that they are less likely to suffer legal consequences of driving while intoxicated than the average person. Previous studies of deterrence have demonstrated that the perceived risk to oneself exceeds the perceived risk for others (Jensen et al., 1978; Paternoster & Piquero 1995). This difference in risk perceptions has also been reported in other contexts, e. g., for breast cancer (Lipkus et al., 2000). We also defined dependent variables for being convicted if stopped both in the second and third person.

2.2.4. Dependent Variables: Subjective Beliefs about Expected Penalties Conditional on Conviction

We analyzed dependent variables for subjective beliefs of receiving a fine or jail term if convicted during, and amounts of fine and jail time if assessed a fine or sentenced to jail. SAD elicited subjective beliefs of being fined and jail if convicted of DWI. These questions were asked in second person of people expecting to drink and drive in the next year, and in the third person of all persons. Fine amounts and days in jail were conditional on paying some fine or serving some jail time. Questions were also asked in the second and third person; if there is discrimination in law enforcement, we expected that Blacks would state that they would be more likely to be treated harshly by law enforcement officers than the average (i.e., third) person.

2.2.5. Explanatory Variables

The key explanatory variables were for race and gender. We defined four mutually exclusive groups for Black versus non-Hispanic White race and female and male genders. The omitted reference group was non-Hispanic White males. SAD contained too few persons of Hispanic ethnicity to permit separate analysis of such persons. Other explanatory variables were added in steps to gauge the robustness of findings on race and gender as follows.

We defined four mutually exclusive drinker categories. (1) Heavy drinker—if during the past year the person had on average consumed 14+ alcoholic beverages/week for males under age 65 and 7+ beverages for females and for males 65+ and the person did not binge drink. (2) Binge drinker—person consumed 5+ (males) or 4+ (females, males > 65) per occasion but did not average 14+ or 7+ drinks weekly. (3) Heavy binge drinkers satisfied both criteria for binge drinking on 1 occasion and the weekly threshold for heavy drinkers. (4) Other drinkers were all others.

We included demographic variables for age, educational attainment (years), household income, and binary variables for each city. Finally, we included a binary variable for whether the person had been arrested for DWI during the last three years to account for differences in familiarity with law enforcement and criminal justice practices in areas in which they resided.

2.3. Objective Data from State Court Records

We computed objective estimates of legal consequences from 2009 arrest records from each state for the eight cities, separately by Black versus White race and by gender.

3. Results

3.1. Descriptive Statistics

On average, the number of drinking and driving episodes in the last year was 1.12 with a range of 0 to 4 (Table 1). 3.3% of respondents had been arrested or cited for DWI in the last three years. 13.9% had been cited for speeding 15 miles or more over the speed limit. The mean response to a question asking about the percent chance of being pulled over due to a law enforcement officer’s suspicion that the respondent was under the influence during the last year was 0.053. Among respondents who expected to drive at least once in the next year, the mean subjective belief about being stopped by police in the next year was 0.19. When the same question was asked in the third person, the mean subjective belief was substantially lower, 0.099. By contrast, the mean subjective belief about being convicted after being stopped was higher when asked in the third than when asked in the second person. A reason for the difference may be that the question asked in the third person more clearly referred to evidence that the person was drinking and driving at the time of the stop.

Table 1.

Descriptive Statistics

	N	Mean	Std. dev.	Min.	Max.
Dependent Variables
Legal consequences
DD	1,516	1.121	1.471	0	4
Arrested or cited for DWI Cited for speeding 15	1,516	0.033	0.179	0	1
mph > the speed limit	1,516	0.139	0.346	0	1
Subjective beliefs (next year)
Stopped	1,226	0.053	0.119	0	1
Stopped if DD	745	0.189	0.202	0	1
Stopped if DD (3rd)	1,229	0.099	0.137	0	0.95
Convicted if stopped	686	0.401	0.421	0	1
Convicted if stopped (3rd)	1,228	0.607	0.324	0	1
Subjective beliefs about penalties conditional on conviction (next year)
Fine	520	0.898	0.219	0	1
Fine (3rd)	1,224	0.916	0.178	0	1
Fine amount	503	1,036	1,177	0	9,000
Fine amount (3rd)	1,198	960.50	1,168	0	9,000
Jail	517	0.322	0.332	0	1
Jail (3rd)	1,215	0.289	0.280	0	1
Days in jail	460	30.20	68.82	0	720
Days in jail (3rd)	1,214	50.85	170.25	0	3,000
Explanatory Variables
Race/Gender
White male†	1,516	0.437	0.496	0	1
Black male	1,516	0.040	0.195	0	1
White female	1,516	0.441	0.497	0	1
Black female	1,516	0.082	0.275	0	1
Drinker type
Binge drinker	1,516	0.404	0.491	0	1
Heavy drinker	1,516	0.045	0.207	0	1
Heavy binge drinker	1,516	0.239	0.427	0	1
Demographic characteristics
Educational attainment (yrs.)	1,516	15.49	1.986	11	18
Married	1,516	0.460	0.499	0	1
Age	1,516	42.76	12.53	18	82
Income
Income (‘00000s)	1,516	0.764	0.647	0	3.5
Income missing	1,516	0.017	0.130	0	1
Location
Raleigh†	1,516	0.377	0.485	0	1
Hickory	1,516	0.065	0.246	0	1
Philadelphia	1,516	0.113	0.316	0	1
Wilkes-Barre	1,516	0.050	0.218	0	1
Seattle	1,516	0.154	0.361	0	1
Yakima	1,516	0.035	0.184	0	1
Milwaukee	1,516	0.119	0.324	0	1
LaCrosse	1,516	0.088	0.283	0	1

^†omitted reference group in regressions

Conditional on a conviction, respondents believed the rate of being assessed a fine was substantially higher than being sentenced to incarceration. On average, respondents estimated a jail term of about a month when the question was asked in the second person and about a month and a half when asked in the third person.

3.2. Past Driving While Intoxicated and its Legal Consequences

The odds of females reporting driving while intoxicated was far lower than for males engaging in this behavior (Table 2, cols. 1, 2). There was no statistical difference between Black and White males in the odds of drinking and driving episodes in the past year. Heavy and binge drinking was associated with more drinking and driving episodes. Higher educational attainment and being married were associated with fewer episodes. By contrast, Black males in the SAD sample had higher rates of prior arrests for DWI and citations for speeding. Not accounting for other factors, Black males were over 4 times as likely to report an arrest for DWI in the past three years than White males, the omitted reference group, (col. 3, odds ratio (OR)= 4.10; 95% confidence interval (CI): 1.75–9.57). White females were half as likely to have been arrested for DWI than White males (OR=0.49; 95% CI: 0.24–0.98). There was no statistical difference between arrest or citation rates between Black females and White males. Statistical significance on DWI arrests was lost for Black males with the addition of other explanatory variables (col. 4). Relative to White males, Black males reported a higher rate of having been cited for speeding, also in the past three years (col. 5, OR=2.04; 95% CI: 1.13–3.81). The result for citations for Black relative to White males does not meaningfully change with the addition of covariates (col. 6).

Table 2.

Past Driving While intoxicated and Legal Consequences

	Driving While Intoxicated (DD)		Arrested or Cited for DWI		Cited for Speeding > 15 MPH over the Speed Limit
	(1)	(2)†	(3)	(4)†	(5)	(6)†
Black male	1.512 (0.930 – 2.456)	1.197 (0.708 – 2.025)	4.096** (1.753 – 9.570)	2.124 (0.795–5.672)	2.043* (1.095–3.813)	2.164* (1.112–4.211)
White female	0.581** (0.472–0.716)	0.615** (0.490 – 0.773)	0.487* (0.242 – 0.982)	0.519 (0.247 –1.088)	0.882 (0.642–1.210)	0.818 (0.585 – 1.143)
Black female	0.518** (0.356 –0.755)	0.530** (0.345 –0. 815)	1.342 (0.537 – 3.354)	0.967 (0.349 – 2.684)	1.099 (0.643 – 1.876)	0.962 (0.539 – 1.719)
Binge drinker		5.505** (4.054–7.476)		3.850* (1.091 – 13.59)		1.368 (0.917 – 2.041)
Heavy drinker		2.118* (1.177–3.813)		5.155 (0.767 – 34.64)		0.999 (0.401 – 2.487)
Heavy binge drinker		11.93** (8.512–16.72)		5.599** (1.579 – 19.86)		1.019 (0.640 – 1.624)
Educational attainment (yrs.)		0.920** (0.869 – 0.973)		0.776** (0.655 – 0.920)		1.016 (0.935–1.105)
Married		0.619** (0.487 – 0.787)		0.237** (0.085 – 0.659)		0.783 (0.549–1.116)
Age		1.000 (0.990–1.009)		0.990 (0.964–1.017)		0.963** (0.949 – 0.977)
Income		0.977 (0.810–1.178)		0.645 (0.282–1.479)		1.264 (0.989–1.615)
X	1,516	1,516	1,516	1,516	1,516	1,516

Confidence intervals in parentheses;

^**p<0.01,

^*p<0.05;

^†covariates for income missing and city included but not shown

3.3. Subjective Beliefs about Being Stopped and Convicted Next Year

There was no statistical difference between Black and White males in intentions to drink and drive in the next year (results not shown). Black males were more pessimistic in their beliefs about being stopped irrespective of whether or not they actually drank and drove (Table 3, col. 1). Black males also thought the stop rate was higher if they drove while intoxicated (cols. 3, 4). White males thought the stop rate during the next year, if they drove while intoxicated, was 0.15 (col. 3); for Black males the corresponding rate was over twice this (0.15+0.16). This result was robust to the addition of other explanatory variables (col. 4).

Table 3.

Subjective Beliefs about Being Stopped and convicted Next Year

	Stopped		Stopped if DD		Stopped if DD (3rd)		Convicted if Stopped		Convicted if Stopped (3rd)
	(1)	(2)†	(3)	(4)†	(5)	(6)†	(7)	(8)†	(9)	(10)†
Black male	0.043* (0.019)	0.023 (0.019)	0.156** (0.042)	0.140** (0.042)	0.159** (0.022)	0.141** (0.022)	−0.254** (0.095)	−0.218* (0.097)	0.002 (0.052)	−0.009 (0.052)
White female	−0.007 (0.007)	−0.005 (0.007)	0.068** (0.015)	0.061** (0.016)	0.030** (0.008)	0.030** (0.008)	−0.115** (0.033)	−0.090** (0.034)	−0.088** (0.019)	−0.075** (0.020)
Black female	−0.017 (0.013)	−0.024 (0.013)	0.040 (0.030)	0.017 (0.032)	0.065** 0.015)	0.053** (0.016)	−0.289** (0.070)	−0.228** (0.073)	−0.195** (0.035)	−0.172** (0.037)
Prior arrest/citation		0.005 (0.020)		0.003 (0.054)		0.064** (0.024)		−0.023 (0.104)		0.102 (0.056)
Binge drinker		0.023** (0.008)		0.003 (0.021)		−0.012 (0.010)		0.014 (0.045)		0.054* (0.023)
Heavy drinker		0.046** (0.016)		−0.004 (0.039)		0.004 (0.019)		−0.149 (0.089)		0.077 (0.045)
Heavy binge drinker		0.059** (0.010)		−0.004 (0.023)		−0.010 (0.011)		0.113* (0.049)		0.111** (0.026)
Educational attainment (yrs.)		−0.007** (0.002)		−0.012** (0.004)		−0.007** (0.002)		0.007 (0.009)		−0.015** (0.005)
Married		−0.026** (0.008)		−0.036* (0.017)		−0.009 (0.009)		0.032 (0.037)		−0.041* (0.021)
Age		−0.001* (0.000)		−0.001 (0.001)		0.000 (0.000)		0.001 (0.001)		0.000 (0.001)
Income		−0.001 (0.006)		−0.011 (0.014)		−0.014* (0.007)		0.010 (0.029)		0.010 (0.016)
Constant	0.056** (0.005)	0.177** (0.033)	0.152** (0.010)	0.400** (0.074)	0.075** (0.006)	0.175** (0.038)	0.472** (0.022)	0.250 (0.159)	0.663** (0.014)	0.787** (0.090)
N	1,226	1,226	745	745	1,229	1,229	686	686	1,228	1,228
R-squared	0.007	0.102	0.037	0.082	0.052	0.093	0.040	0.071	0.033	0.091

Standard errors in parentheses;

^**p<0.01,

^*p<0.05;

^†covariates for income mission and city included but not shown

When asked in the third person about people in general being stopped conditional on drinking and driving, White males estimated a stop rate of 0.075 (col. 5)—half the rate in response to the question about being stopped themselves (col. 3). By contrast, relative to White males, Black males did not differ in their subjective beliefs about stopped when driving while intoxicated, whether the question was asked in the second or third person (compare cols. 3, 4 with cols. 5, 6). Females, especially Blacks, gave higher estimates of being stopped than White males did (col. 5). Again, the results were robust to changes in specification.

White males thought the chance of being convicted following a stop was nearly half (0.47, col. 7). For Black males, this chance was substantially lower (0.47–0.25). Results for Black and White males hardly changed at all with the addition of other covariates (col. 8). The corresponding estimate for the question asked in the third person is appreciably higher (0.66, col. 9). There was no difference in Black and White males in the estimated likelihood of conviction. The addition of other covariates does not affect this result (col. 10). Yet females, particularly Black females, gave lower estimates of being convicted (cols. 9, 10).

3.4. Subjective Beliefs about Penalties Conditional on Conviction

Rates for White males were similar, whether the question was asked about their own outcomes or those for the general population. They believed that conditional on conviction, a fine was almost certain (Table 4, cols. 1, 3) and that about a third of convictions resulted in jail (cols. 9, 11). For those receiving a jail sentence, they expected the sentence to be about a month, give or take a week (cols. 13,15).

Table 4.

Subjective Beliefs about Expected Penalties Conditional on Conviction

Variables	Fine		Fine (3rd)		Fine Amount		Fine Amount (3rd)		Jail		Jail (3rd)		Days in jail		Days in Jail (3rd)
	(1)	(2)†	(3)	(4)†	(5)	(6)†	(7)	(8)†	(9)	(10)†	(11)	(12)†	(13)	(14)†	(15)	(16)†
Black male	−0.110 (0.057)	−0.105 (0.057)	−0.109** (0.029)	−0.098** (0.030)	−705.2* (309.4)	−563.0 (298.5)	−394.1 (190.9)	−305.3 (186.3)	0.101 (0.088)	0.124 (0.085)	0.099* (0.046)	0.095* (0.045)	65.32** (18.85)	61.62** (18.78)	101.0** (28.92)	91.12** (29.52)
White female	−0.046* (0.020)	−0.050* (0.020)	−0.040* (0.011)	−0.036** (0.011)	−209.3 (109.6)	−172.4 (107.2)	−101.2 (71.40)	−76.32 (70.74)	−0.020 (0.031)	−0.024 (0.030)	−0.002 (0.017)	−0.005 (0.017)	1.424 (6.817)	−1.802 (6.894)	13.70 (10.65)	10.06 (11.05)
Black female	−0.191** (0.048)	−0.169** (0.050)	−0.112** (0.020)	−0.096** (0.021)	−648.0* (264.2)	−413.8 (262.8)	−469.1 (131.1)	−338.0* (132.7)	0.019 (0.074)	0.044 (0.074)	−0.010 (0.031)	0.002 (0.032)	30.94 (16.75)	17.46 (17.32)	29.84 (19.14)	13.51 (20.32)
Prior DWI arrest		0.008 (0.061)		0.002 (0.032)		3.562 (317.1)		261.4 (198.2)		0.157 (0.091)		0.068 (0.048)		−20.57 (19.27)		−31.56 (30.67)
Binge drinker		0.095** (0.029)		0.019 (0.013)		215.8 (154.6)		92.77 (81.99)		0.070 (0.043)		0.017 (0.019)		14.68 (9.857)		−6.847 (12.71)
Heavy drinker		−0.105 (0.056)		−0.002 (0.025)		−455.9 (297.6)		−202.2 (158.4)		−0.020 (0.083)		−0.018 (0.038)		0.918 (18.61)		−3.326 (25.21)
Heavy binge drinker		0.068* (0.030)		0.026 (0.015)		202.1 (162.3)		117.3 (95.10)		0.033 (0.046)		0.031 (0.023)		−7.681 (10.39)		−17.44 (14.80)
Educational attainment (yrs.)		0.003 (0.005)		−0.002 (0.003)		−12.14 (28.16)		−23.57 (18.27)		−0.024** (0.008)		−0.010* (0.004)		−5.340** (1.786)		−2.161 (2.844)
Married		−0.003 (0.022)		−0.015 (0.012)		−220.9 (117.9)		−139.8 (74.47)		−0.083* (0.033)		−0.046* (0.018)		−12.54 (7.608)		−16.62 (11.62)
Age		0.001 (0.001)		0.000 (0.000)		2.700 (4.666)		−5.631 (2.894)		0.002 (0.001)		−0.000 (0.001)		0.336 (0.304)		0.425 (0.448)
Income		−0.027 (0.018)		0.014 (0.009)		206.8* (98.19)		163.1** (60.04)		−0.011 (0.027)		−0.027 (0.014)		−2.387 (6.190)		−11.95 (9.274)
Constant	0.928** (0.013)	0.778** (0.094)	0.946** (0.008)	0.915** (0.050)	1,167** (70.04)	792.3 (498.2)	1,059** (51.52)	1,357** (323.4)	0.326** (0.020)	0.524** (0.140)	0.287** (0.012)	0.440** (0.077)	27.12** (4.391)	97.61** (31.56)	40.47** (7.664)	76.29 (50.27)
N	520	520	1,224	1,224	502	502	1,195	1,195	517	517	1,215	1,215	449	449	1,177	1,177
R-squared	0.039	0.101	0.036	0.051	0.024	0.162	0.013	0.119	0.004	0.138	0.004	0.111	0.033	0.125	0.012	0.036

Standard errors in parentheses.

^**p<0.01,

^*p<0.05,

^†covariates for income missing and city included but not shown.

Black males were more optimistic about fines (cols. 1–8), but more pessimistic about jail than White males (cols. 9–16). The major difference was in days in jail conditional on serving some time in jail (cols. 13–16). While the differences in beliefs about the rates of incarceration between Black and white males are statistically significant in the variants posed in the third person (cols. 11, 12), statistical significance is lacking when asked in the second person (cols. 9,10). Yet the magnitudes of the coefficients are almost identical for questions asked in the second versus the third person. We attribute this difference in statistical significance levels to the larger sample size for responses to the question asked in the third person. Unlike White males, Black males expected to serve 4 months if convicted and sentenced; Black males estimated even a higher time in jail for the general population—almost 5 months (col. 15, 40+101 days). Females did not differ from White males in their estimates of the likelihood of, or time in, jail. Adding explanatory variables did not alter these findings.

Fine amounts were thought to be higher by respondents with higher income (cols. 6, 8). Jail was thought to be less likely by respondents with more education (cols. 10, 12). Moreover, more highly educated respondents believed that they would have lower jail times but this did not apply to others (compare cols. 14, 16). Married persons thought the chances of being incarcerated were lower.

3.5. Objective Arrest Rates and Resolution of Arrest

In the eight study cities in 2009, White males had only a 1 in a 100 chance of being arrested for DWI (0.0088) whether or not they drink or drive or consumed alcohol at all (Table 5, col. 1). This estimate reflects the number of arrests in 2009 relative to the population of individuals ages 18+ in each city according to race and gender. Black males had a higher arrest rate (0.0104), but not statistically different than that for White males. White and Black females had much lower arrest rates than males did. The actual rate of conviction, if arrested, for White males was 0.63 (col. 2). Black males were convicted at a somewhat lower rate (0.59), as were Black females (0.57). Black males and females had a much higher likelihood of paying a fine if convicted than White males and females did (col. 3). Fine amounts, conditional on paying a fine, ranged from $438 for White females to $747 for Black males (col. 4). Convicted White males were jailed at a rate of 30 per 100 convictions, significantly lower than the corresponding estimate for Black males, 43 per 100 (col. 5). Black females had an actual jail rate between that for White and Black males, 0.39. However, White males were sentenced to 151 days in jail on average, conditional on receiving a sentence (col. 6). The mean sentence for Black males was significantly lower, 127 days. Black females were also sentenced to fewer days in jail than White males were, 118 days. The expected number of jail days (probability of jail if convicted x jail days) was slightly higher for Black males, 55 and 45 for White males (expected values not shown in table).

Table 5.

Actual Probabilities of Arrests and Their Disposition

Demographic Group	Arrest rate	Convicted if arrested	Fine if convicted	Fine amount	Jail if convicted	Jail days
	(1)	(2)	(3)	(4)ⱡ	(5)	(6)ⱡ
White male	0.0088 (0.0016)	0.63 (0.0050)	0.37 (0.0064)	576.26 (54.12)	0.30 (0.00060)	150.98 (5.57)
Black male	0.0104 (0.0011)	0.59** (0.0079)	0.77** (0.088)	747.46** (12.43)	0.43** (0.010)	127.14** (5.59)
White female	0.0010** (0.00048)	0.62 (0.0089)	0.33** (0.011)	437.55 (23.58)	0.28 (0.010)	129.20 (7.85)
Black female	0.0020** (0.00031)	0.57** (0.018)	0.65** (0.023)	663.86 (32.09)	0.39** (0.024)	117.64* (13.59)

Standard errors in parentheses:

^**p<0.01,

^*p<0.05 with White males the comparison group;

^ⱡExcludes Washington (Yakima and King counties).

Discussion

Black males, but not Black females, were more pessimistic about being stopped for a DWI, irrespective of whether actual drinking was involved. Black males were relatively pessimistic about going to jail following a conviction for DWI. Black males were neutral or optimistic about some other outcomes (e.g., fines). However, there was no significant difference by Black versus White race in the number of drinking and driving episodes in the past year or in intentions to drink and drive in the next. This is in contrast to the higher rate of previous arrests for DWI among Black male respondents to SAD.

Thus, in contrast to beliefs of Black males about arrests (a significantly higher rate of stops combined with no statistical difference in the subjective belief of arrest conditional on being stopped), there was no statistical difference in the actual rates of arrest. The actual difference in expected jail time following a conviction was somewhat higher for Black than White males. But the difference in subjective beliefs between Black and White males about expected jail time was considerably higher than the difference in corresponding actual values, implying that Black males were overly pessimistic about jail.

That Black males overestimated jail sentences incurred by the general population by even a larger amount suggests that they did not attribute longer sentences to racial bias per se. Black males cited higher rates of incarceration conditional on a DWI conviction, but the difference was only statistically significant when the question was asked in the third person, again suggesting that Black males did not attribute the high incarceration rate to race. These expectations may reflect the higher proportions of Black males who are incarcerated, even controlling for educational attainment (Western, 2006), and greater exposure to criminal justice system outcomes through contact with peers than others. SAD asked respondents if they knew of a person who had been jailed for DWI: 15.7% of all respondents answered affirmatively; 31.0% of Black males did.

An important issue is whether perceptions reflect reality. If reality is reflected, then the system should be changed. If perception of bias exists despite no actual bias, a change in enforcement policy would not be necessary or effective, but a public relations campaign would be helpful in realigning beliefs. We found that sometimes subjective beliefs and actual rates were quite similar. White males had a mean subjective belief of 0.66 for being convicted if stopped versus an actual rate of 0.63 in the same eight cities.

However, some rates of being fined were substantially overestimated and fine amounts, substantially underestimated. Some discrepancies may reflect lack of enforcement of penalties subsequent to sentencing not reflected in arrest data. The arrest data tracked a case from arrest through sentencing, but did not record events post sentencing. Actual jail days were substantially higher than subjective jail days (compare Table 1 values from our survey with actual rates from arrest data in Table 5, col. 6), perhaps due to the offender’s ability to substitute community service for jail post sentencing or to receive a reduction in actual jail time due to completion of a treatment program post sentencing. If so, the state arrest data we used to compute the objective estimates may overstate actual time served.

We documented substantial differences between genders in number of episodes of driving after having had too much to drink and rates of conviction if stopped. The actual arrest rates were at most one fifth of males.

There are previous studies of policing and sentencing practices as they apply to race. There is a substantial literature on racial profiling related to traffic stops and police actions not specifically on stops for suspicion of DWI, but there is very little comparative evidence on case resolution after stops occur, fractions of stops leading to arrests, and prosecution and conviction rates following arrest. Blacks and males are more likely to be stopped and according to self-reports more often not for a legitimate reason (Lundman & Kaufman, 2003; Engel, 2005). However, many stops may not be recorded (Lundman, 2004).

Recent economic research on racial profiling by police has focused on the difference between statistical and taste-based discrimination. Statistical discrimination occurs when law enforcement officers are uncertain about whether a crime was committed but face the problem of how best to allocate scarce resources. Thus, if there are differences in a crime propensity according to readily observable characteristics (e.g., race or gender, it is rational for an agency to allocate more resources to groups; e.g., neighborhoods) for which crime rates are higher (Knowles et al., 2001; Anwar & Fang, 2006; Antonovics & Knight, 2009). One test of statistical discrimination is whether police equalize the rates at which a crime has been committed given that an investigation has been conducted—for example, the fraction of vehicle stops in which contraband is found. Taste-based discrimination involves investigations not based on statistical evidence of differences in crime rates according to observable characteristics (i.e., because an officer dislikes a person “different” from the officer; Persico & Todd, 2008). Antonovics and Knight (2009) found that police search decisions varied with police officer race, which should not occur if searches reflect statistical discrimination, evidence at least broadly consistent with other studies (Donohue & Levitt, 2001; Close & Mason, 2007); Antonovics and Knight found that White officers were more likely to search Black and Hispanic drivers’ vehicles.

There is some empirical evidence that following vehicle stops, minorities are cited and arrested more frequently (Tillyer & Hartley, 2010). Race/ethnicity has been the focus of research on sentencing outcomes. Blacks are sentenced to longer periods of incarceration on average and sentenced to prison rather than jail but the evidence does not pertain specifically to motor vehicle violations (Holleran & Spohn, 2004; Harrington & Spohn, 2007). In our study, we found that jail sentences were lower for Black than for White males, although the rate of jail given conviction was higher for Black males.

Our results show a concern about being stopped by police among Black males, but they saw some offsetting of police behavior by the courts. Although past arrest/citation rates differed, we found no differences in expectations of Black versus White males to engage in behaviors leading to such arrests.

Risk perceptions of sanctions do not appear to be either crime- or jurisdiction-specific, a result consistent with prior evidence (Ross & Voas, 1990). Black males’ overall risk perceptions of stops and jail are likely to have reflected general perceptions rather than perceptions specifically about DWI legal consequences. For this reason, it seems unlikely that overestimation of risk by this population subgroup leads to deterrence of drinking and driving. In fact, in response to a question in our survey about whether or not the respondent limited drinking when s/he expected to drive, Blacks were less likely to limit drinking than were Whites, even though on other measures of alcohol consumption, the two population subgroups were similar.

Our study has several strengths. First, we studied the link between risk perceptions of sanctions by race/gender for a particular offense. Second, we combined subjective and objective data in the same study. Previous studies have been limited to a single outcome (e.g., police or sentencing practices). We analyzed outcomes from stops to DWI penalties. This permitted a nuanced analysis of subjective beliefs about policing and judicial decisions as applied to DWI. Third, we obtained data from eight geographically and demographically diverse cities of different population sizes. Fourth, we examined the effect of question framing that yielded evidence on perceived discrimination by race without asking about race explicitly, which could have led to strategic responses. Fifth, while we focused on race/gender, we included other covariates to assess robustness of findings on race/gender.

We acknowledge several weaknesses. First, we lacked data on fine and jail amounts for Washington. Second, our objective data pertained to judicial outcomes not to the post-judicial outcomes that some survey respondents might have referred to in stating their subjective beliefs. Third, we implicitly assumed that people respond to laws applicable to the jurisdictions in which they reside and that risk perceptions are crime specific. However, risk perceptions may be based on a broader set of experiences. If so, policy makers may not succeed in changing behaviors of persons in their jurisdictions by changing penalties for specific crimes. Research is at a preliminary stage in understanding how subjective beliefs are formed.

In sum, subjective beliefs about driving while intoxicated differ by race and gender. Some subjective beliefs are supported by objective evidence. Differences between subjective beliefs and their objective counterparts are subtler than is typically acknowledged. Differences in beliefs may not be viewed by minorities as reflecting racial bias, but instead reflect their social experiences which may not be crime- or jurisdiction-specific.

Highlights.

• Drinking/driving (DWI) rates were higher for men than women, across races.

• Black men had higher subjective probabilities of stops and jail for DWI than white men.

• Responses imply blacks did not attribute differences in their beliefs to racial bias.

• Overall, actual DWI enforcement and judicial decisions did not differ by race.

• Patterns of misperceptions suggested weak transmission of deterrent signals.

Biographies

Frank Sloan is the J. Alexander McMahon Professor of Health Policy and Management and Professor of Economics at Duke University since 1993. He holds faculty appointments in five departments at Duke, with Economics being his primary appointment. He did his undergraduate work at Oberlin College and received his Ph.D. in economics from Harvard University. Professor Sloan is currently the chair of the IOM committee investigating adjustment factors in Medicare payment. He is also the president elect for the American Society of Health Economists. His current research interests include specialty courts, drug/alcohol use prevention, long-term care, medical malpractice, and cost-effectiveness analyses of medical technologies.

Lindsey Chepke is a Research Associate at Duke University. She obtained her J.D. from Benjamin N. Cardozo School of Law and her B.A. from Wells College.

Dontrell Davis has a masters in Economics, and was a Associate in Research in the Department of Economics at Duke University when this paper was drafted.