An examination of how alcohol enforcement strategies by sheriff and police agencies are associated with alcohol-impaired-driving fatal traffic crashes

Abstract

Objective.

Over 10,000 people die in alcohol-impaired-driving traffic crashes every year in the U.S. Approximately half of alcohol-impaired drivers report their last drink was at a bar or restaurant, and most bars and restaurants serve alcohol to patrons who are already intoxicated, known as overservice. Law enforcement agencies use various strategies to address alcohol-impaired driving and overservice but research on the effectiveness of these strategies is limited. Our objective was to assess whether law enforcement efforts focusing on alcohol-impaired driving and alcohol overservice were associated with alcohol-impaired-driving fatal traffic crashes.

Methods.

We conducted a survey of police and sheriff agencies in 1,082 communities across the U.S. in 2010 regarding their alcohol enforcement practices. We assessed whether the agency conducted: (1) alcohol overservice enforcement and (2) alcohol-impaired driving enforcement (sobriety checkpoints, saturation patrols, open container, overall alcohol-impaired driving enforcement). From the Fatality Analysis Reporting System (2009–2013), we obtained counts of alcohol-impaired-driving fatal traffic crashes (at least one driver had blood alcohol content ≥ 0.08) within the agency’s jurisdiction boundary and within a 10-mile buffer. Using multi-level regression, we assessed whether each enforcement type was associated with alcohol-impaired-driving fatal crashes (per 100,000 population). For both the jurisdiction boundary and 10-mile buffer, we ran stratified models based on community/agency type: (1) small town/rural police; (2) urban/suburban police and (3) sheriffs.

Results.

In jurisdiction boundary models, urban/suburban communities where police conducted overservice enforcement (compared to those that did not) had fewer alcohol-impaired-driving fatal crashes (5.0 vs. 6.6; p=0.01). For the 10-mile buffer, small town/rural communities where police agencies conducted overservice enforcement (compared to those that did not) had fewer alcohol-impaired-driving fatal crashes (16.9 vs. 21.2; p=0.01); we found similar results for small town/rural communities where police used saturation patrols (18.7 vs. 22.1; p=0.05) and had overall high alcohol-impaired driving enforcement (18.7 vs. 22.1; p=0.05). The direction and the size of the effects for other types of enforcement and agencies were similar, but not statistically significant.

Conclusions.

Alcohol enforcement strategies among police agencies in small town/rural communities may be particularly effective in reducing alcohol-impaired fatal traffic crashes. Results varied by enforcement, agency and community type.

INTRODUCTION

Significant progress has been made in reducing alcohol-impaired driving, yet alcohol-impaired-driving traffic crashes remain a leading cause of death and injury in the United States (U.S.). In 2019, there were more than 10,000 deaths from alcohol-impaired-driving crashes (at least one driver had blood alcohol level [BAC] ≥ 0.08), which represented 29% of all traffic-related deaths for the year (National Center for Statistics and Analysis, 2020). We need to identify the best strategies for continuing to address this public health problem.

One approach to reducing alcohol-impaired-driving traffic crash deaths is to focus directly on reducing alcohol-impaired driving through enforcing a policy that prohibits driving a motor vehicle with BAC of 0.08 or higher. All states in the U.S. have established a BAC limit of 0.08 except Utah, which has a stricter limit of 0.05. Enforcement strategies can increase the effectiveness of BAC laws. Sobriety checkpoints, saturation patrols, enforcement of laws that prohibit open-containers of alcohol in motor vehicles are practices used by police, highway patrol and sheriff agencies to reduce alcohol-impaired driving. Sobriety checkpoints are set up by law enforcement agents along a roadside to stop all vehicles or a certain portion of vehicles to identify impaired drivers (CDC, 2020). Saturation patrols involve a large number of law enforcement agents patrolling a specific area to enhance alcohol-impaired driving enforcement (CDC, 2020). Enforcement of open container laws involve identifying motor vehicles where drivers and/or passengers have open containers of alcoholic beverages.

Sobriety checkpoints are an effective strategy for enforcing BAC laws to reduce alcohol-impaired driving (Bergen et al., 2014; Shults et al., 2001). Saturation patrols have been less evaluated but look promising for preventing impaired driving (Fell et al., 2008; Lenk et al, 2016). Enforcement of open-container laws has limited evaluation (Lenk et al., 2016). Enforcement of impaired-driving policies can be conducted by state agencies (e.g., highway patrol) or through local law enforcement agencies (e.g., city police agencies, sheriff agencies, etc.). All three of these enforcement strategies focusing on alcohol-impaired driving and crashes have not been evaluated across different types of communities or when used in combination with other strategies.

In addition to focusing on alcohol-impaired driving behavior directly, strategies that address alcohol-impaired driving behavior by targeting alcohol consumption may also help reduce alcohol-impaired-driving traffic crashes. Binge drinking (5 or more drinks for men/4 or more drinks for women on one occasion) is associated with alcohol-impaired driving (Naimi et al., 2003; 2009; Quinlan et al., 2005). One strategy to reduce binge drinking is to reduce the sales of alcohol to obviously intoxicated customers at on-premise licensed alcohol establishments (e.g., bars, restaurants). This is particularly important given that bars and restaurants are the source of last drink for approximately half of drivers stopped for alcohol-impaired driving (Naimi et al., 2009). Serving alcohol to obviously intoxicated customers (i.e., overservice) is illegal in all but two states (Mosher et al., 2009); however, studies have found that the likelihood of serving alcohol to an obviously intoxicated customer is high (Toomey et al., 2016).

Law enforcement agencies can enforce overservice laws through efforts such as random inspections and observations at bars and restaurants. A systematic literature review conducted in 2010 as part of the Community Guide concluded that due to a lack of studies, overservice enforcement efforts lacked sufficient evidence to be a recommended strategy to reduce alcohol consumption and related harms (https://www.thecommunityguide.org/findings/alcohol-excessive-consumption-overservice-law-enforcement-initiatives), supporting the need for additional studies in this area. The three studies that have been published since this review suggest overservice enforcement may be effective in reducing binge drinking at bars/restaurants and alcohol-impaired-driving crashes, but study limitations include small samples (Fell et al., 2017), lack of specification of enforcement strategies used (Tomedi et al., 2018) or lack of clarity if effects are due to enforcement or other parts of the intervention (George et al., 2019). While these studies suggest overservice might be effective, more research in a diverse sample of communities is needed to assess community-level effects of overservice enforcement.

In this study, we assessed whether local law enforcement efforts focusing on alcohol-impaired driving and overservice of alcohol w associated with the rate of alcohol-impaired-driving fatal traffic crashes. We hypothesize that communities where agencies report conducting these types of enforcement will have lower rates of alcohol-impaired-driving fatal traffic crashes, and these associations may vary by type of agency and community. Results of the study can inform community enforcement practices.

MATERIALS AND METHODS

We conducted a two-year study using a cross-sectional design. We used secondary data collected in 2010 as part of a survey of a national sample of law enforcement agencies and data from a national database of fatal traffic crashes.

Data Sources

Local law enforcement survey.

Measures of enforcement practices came from a 2010 survey of local law enforcement agencies (see Lenk et al., 2014 for a full description of the survey methods). For this survey, we sampled agencies across the U.S. based on number of agencies per state, the proportion of agencies that were county sheriff versus municipal police, and agency size (n=1,504 agencies). We also added the municipal police agency from the three largest cities in each state if they were not already included in our sample (n=127). Surveys were completed via telephone or online. The final sample in our previous study was 1,082 (66% response rate).

Fatality Analysis Reporting System (FARS).

We obtained counts of fatal traffic crashes in each community from FARS, a database maintained by the National Highway Traffic Safety Administration (NHTSA) containing a census of all motor vehicle crashes in the U.S. that result in a fatality within 30 days of the crash. Although the FARS dataset has limitations (Subramanian, 2002; Briggs et al., 2005), it is the best data source available for measuring fatalities related to alcohol-impaired-driving crashes.

U.S. Census.

We used data from the 2010 U.S. Census for community population and to define types of communities (e.g., urban, rural).

Measures

Independent variables.

From the law enforcement survey, we constructed measures for two main types of enforcement conducted within the past year: (1) alcohol-impaired driving enforcement, and (2) enforcement targeting overservice of alcohol at licensed establishments (e.g., bars, restaurants). For alcohol-impaired driving enforcement, we used data pertaining to three specific enforcement strategies: sobriety checkpoints, saturation patrols, and enforcement of open container laws; for each strategy, we assessed if an agency used the strategy (yes/no). For overservice enforcement, we measured whether an agency conducted any overservice enforcement (yes/no). We also considered an index sum variable of whether an agency used each specific overservice enforcement strategy assessed in the survey: random inspections, walk-throughs, observations for overservice, “last call” enforcement, blood alcohol concentration (BAC) testing, and pseudo-intoxicated purchase attempts; however this sum was highly correlated (r=.94) with the dichotomous (yes/no) variable so we used the dichotomous variable only. Similarly, for sobriety checkpoints and saturation patrols we considered variables measuring the frequency of using the strategy in past year (once, 2–3 times, 4–5 times, or 6 or more times) but these variables were highly correlated (r=.88) with the dichotomous (yes/no) variables so used the dichotomous variables in analysis.

In addition, we created an overall measure of alcohol-impaired driving enforcement based on a measure from a previous study using latent class analysis (Erickson et al., 2015). This analysis grouped agencies based on patterns of whether they conducted sobriety checkpoints, saturation patrols, and enforcement of open container laws as well as on how often in the past year the agency collaborated or communicated with local media outlets to promote their alcohol-impaired driving enforcement efforts (never, once or twice, 3 or more times). The model estimated three classes—Class 1: Uniformly low enforcement; Class 2: Primarily used saturation patrols and media communication; and Class 3: Uniformly high enforcement. For these analyses, we dichotomized this variable to “high” (Class 3) vs “low” (the Class 1 and 2).

Outcome and control variables.

Our outcome variable was the number of alcohol-impaired-driving fatal crashes from 2009 through 2013. An alcohol-impaired-driving crash was defined as at least one of the drivers in the crash had a BAC equal to or above 0.08, which was the legal limit for all 50 states during this time period. We also created a variable measuring the total number of crashes in each jurisdiction where no drivers had BAC ≥ 0.08 in order to control for the variation in number of crashes across jurisdictions/buffers. Because BAC test results are not known for all drivers involved in fatal crashes, NHTSA uses multiple imputation to estimate the BAC of the driver at the time of the crash. Missing data can result for a number of reasons; the most frequent is that drivers are not always tested for alcohol. The imputation model is based on characteristics of the crash including: crash factors (time of day, day of week, type of crash, and relation to roadway [on or off]); vehicle factors (vehicle type and role in the crash); person factors (age, sex, restraint use, and previous driving violations); and the subjective assessment of the police officer at the scene of the crash as to whether alcohol was involved. The imputation model was developed at the national level using all available known data and applied to each individual driver with missing or unknown BAC test results.

The crash data from FARS had either a geographic location (i.e. latitude and longitude) or the Geographic Locator Codes. ArcGIS Desktop 10.5.1 was used to prepare the 152,000 crash locations for analysis. The law enforcement agencies were also mapped using Census boundary files from NHGIS (https://www.nhgis.org). We created measures using two types of buffers (i.e., geographic areas). The first measure was the count of crashes within the boundaries of each agency’s jurisdiction. Using the Create Drive-Time Areas tool in ArcGIS Online, we also created a measure of the count of crashes within 10 miles around the centroid of each agency’s service area (i.e. city or county boundary). This buffer was then spatially joined to the FARS crash data, allowing for the summation of crashes within the buffer. In urban communities, service area buffers overlapped with each other, thus crashes could be joined to multiple service areas.

Stratification variable.

Given our preliminary analysis demonstrated differences in crash rates by type of community and agency, we conducted stratified analyses. We constructed four categories of community type (urban suburban, small town, rural) using U.S. Census designations. Urbanized areas of 50,000 or more people that are a core city (i.e., the city shares its name with the urbanized area name) were designated as urban, and all other cities within that same urbanized area were labeled as suburban. Urban clusters of at least 2,500 and less than 50,000 people were designated as small towns. Rural communities were all other areas not included within an urban area. Cities that shared both an urbanized area and rural area designation (n=296) were designated as urban. There were 16 agencies that could not be matched with a community type due to a lack of geospatial information, so our final sample was 1,066 agencies. The community type of designations were considered along with the type of law enforcement agency in the community (police vs. sheriff) to create a three-level stratification variable: (1) small town/rural police (n=476), (2) urban/suburban police (n=344) and (3) sheriff/constables (n=246).

Analysis

We first assessed descriptive statistics for all variables. We then estimated 30 separate models regressing number of fatal crashes on enforcement; one for each of the five enforcement variables (sobriety checkpoints, saturation patrols, enforcement of open container laws, overservice enforcement, enforcement latent class) across the two geographic areas (jurisdiction boundary and 10-mile buffer) and by community/agency type (small town/rural, urban/suburban, and sheriff). Negative binomial models were used to accommodate the large number of zero fatal crash counts in some models. The natural log of total population of the jurisdiction or buffer was used as the offset (from 2010 U.S. Census) and the natural log of crashes with BAC <0.08 was a control variable (included in the model as a covariate). Models used the FARS imputed dataset (n=10 imputations) and accounted for agencies clustering within state (Stata MI command). Significance was considered at p≤ 0.05. Analyses were done using SAS/Stat Version 9.4 (Cary, NC) and Stata 15.1 (College Station, TX).

RESULTS

Descriptive results are presented in Table 1. The most common enforcement strategy reported by agencies in all three stratifications was saturation patrols and the least common was enforcement of overservice laws. A higher percentage of municipal police in urban/suburban communities (compared to the other two stratifications: police in small town/rural communities and sheriff agencies) reported doing overservice and saturation patrols and had high overall alcohol-impaired driving enforcement. Rates of alcohol-impaired-driving fatal traffic crashes were highest among communities with sheriff/constable agencies when considering the jurisdiction boundaries, and were highest among urban/suburban communities with municipal police for the 10-mile buffer area.

Table 1.

Descriptive statistics (n=1,066)

	Municipal police		Sheriff/Constables
	Small town/rural n=476	Urban/suburban n=344	n=246
Enforcement strategies	Percentage of agencies
Overservice	23	27	15
Sobriety checkpoints	35	51	52
Saturation patrols	60	77	64
Open container	44	43	46
Overall impaired-driving enforcement: high	65	81	69
Jurisdiction boundary		Mean (SD)
Total population	6950 (26531)	136299 (534204)	108455 (255417)
Fatal traffic crashes
BAC ≥ 0.08	0.59 (3.1)	10.6 (36.5)	13.7 (29.1)
BAC < 0.08	0.09 (0.5)	1.5 (5.4)	2.2 (4.1)
10-mile buffer		Mean (SD)
Total population	43983 (113846)	486483 (601548)	61610 (177518)
Fatal traffic crashes
BAC ≥ 0.08	5.7 (9.1)	35.9 (35.2)	6.9 (14.5)
BAC < 0.08	0.9 (1.5)	5.8 (6.0)	1.2 (2.5)

BAC=blood alcohol content; SD=standard deviation

Results of regression models using the agency jurisdiction boundary are presented in Table 2 and results from the 10-mile buffer are in Table 3. Using the jurisdiction boundary, urban/suburban communities where police agencies reported conducting overservice enforcement (compared to those that did not) had fewer alcohol-impaired-driving fatal crashes per 100,000 population (15.0 vs. 6.6; p=0.01). No other results were statistically significant but most were in the same direction (i.e., communities where agencies reported conducting a particular type of enforcement had lower crash rates compared to communities where agencies did not conduct that type of enforcement).

Table 2.

Enforcement actions by alcohol-impaired-driving crashes (BAC ≥ 0.08) stratified by type of agency/community: Jurisdiction boundary (n=1,066)

	Municipal police				Sheriff/Constables
	Small town/rural n=476		Urban/suburban n=344		All n=246
Enforcement type	crashes/100K population	p	crashes/100K population	p	crashes/100K population	p
Overservice		0.29		0.01		0.06
Yes	5.5		5.0		17.5
No	6.7		6.6		23.0
Sobriety checkpoints		0.36		0.25		0.93
Yes	7.7		5.6		22.5
No	6.3		6.8		22.9
Saturation patrols		0.35		0.23		0.16
Yes	6.1		5.9		20.3
No	7.4		7.1		25.3
Open container		0.74		0.09		0.94
Yes	6.3		6.6		21.5
No	6.0		5.5		21.7
Overall impaired-driving enforcement		0.38		0.42		0.30
High	6.0		5.9		21.1
Low	7.3		6.8		25.0

Note: Negative binomial regression models with log of population as offset and log of BAC <0.08 crashes as covariate; estimates are margins (i.e. means); bold type indicates significant p<.05

Table 3.

Enforcement actions by alcohol-impaired-driving crashes (BAC ≥ 0.08) stratified by type of agency/community: 10-mile buffer (n=1,066)

	Municipal police				Sheriff/Constables
	Small town/rural n=476		Urban/suburban n=344		All n=246
Enforcement type	crashes/100K population	p	crashes/100K population	p	crashes/100K population	p
Overservice		0.01		0.11		0.40
Yes	16.9		8.2		22.2
No	21.2		9.2		25.9
Sobriety checkpoints		0.14		0.15		0.31
Yes	21.9		8.3		29.5
No	18.9		9.7		23.7
Saturation patrols		0.02		0.29		0.33
Yes	18.3		8.7		24.0
No	22.1		9.7		28.3
Open container		0.23		0.49		0.26
Yes	20.5		9.0		24.1
No	19.0		8.6		28.1
Overall impaired-driving enforcement		0.05		0.19		0.68
High	18.7		8.6		25.1
Low	22.1		9.8		27.1

Note: Negative binomial regression models with log of population as offset and log of BAC <0.08 crashes as covariate; estimates are margins (i.e. means); bold type indicates significant p<.05

For the 10-mile buffer, small town/rural communities where police agencies conducted overservice enforcement (compared to those that did not) had fewer alcohol-impaired-driving fatal crashes per 100,000 population (16.9 vs. 21.2; p=0.01); we found similar results for small town/rural communities where police used saturation patrols (18.7 vs. 22.1; p=0.05) and had overall high (vs. low) alcohol-impaired driving enforcement (18.7 vs. 22.1; p=0.05). As with the jurisdiction boundary, results for other types of enforcement were largely similar in terms of direction but were not statistically significant.

DISCUSSION

Across different types of communities and law enforcement agencies, we found only a few statistically significant associations between local law enforcement efforts and rates of alcohol-impaired-driving fatal traffic crashes (at least one driver with BAC ≥ 0.08). All statistically significant associations were as we hypothesized with conducing enforcement actions associated with fewer alcohol-impaired driving crashes, but results varied by the type of community, agency, and enforcement strategy as well as the geographic buffer used in the analysis.

When using a 10-mile buffer from the centroid of the jurisdiction, lower rates of alcohol-impaired-driving crashes were found in small town/rural communities where police agencies reported conducting overservice enforcement, implementing saturation patrols and having overall high levels of impaired-driving enforcement. When considering jurisdiction boundaries, lower rates of alcohol-impaired-driving crashes were found in urban/suburban communities where municipal police agencies conducted overservice enforcement. No associations were found between enforcement and crashes among sheriff and constables in any of our models. Perhaps the areas served by sheriff and constable agencies are not adequately captured by using a jurisdiction boundaries of the town or county they serve or by a 10-mile buffer around these jurisdictions. These results highlight the importance of stratifying analyses by type of agencies and communities and considering different geographic boundaries when assessing how enforcement may affect outcomes such as traffic crashes.

Across different type of communities, reported overservice enforcement by local law enforcement agencies was most consistently associated with fewer fatal traffic crashes. These findings suggest that overservice enforcement may be a tool that communities can use to drive down rates of alcohol-impaired driving fatalities. In this study we assessed whether an agency reported conducting past year enforcement of overservice of alcohol but did not assess specific types of overservice enforcement. There are several types of overservice enforcement that agencies may use to target overservice of alcohol at licensed establishments (Lenk et al., 2014), such as conducting undercover observations of alcohol establishments to detect cases of illegal overservice of alcohol. Other agencies may be conducting place of last drink (i.e., POLD) investigations to determine where people consumed alcohol prior to being stopped for an alcohol-impaired-driving incident (e.g., driving while impaired, assault). Currently, there is limited information about the effectiveness of these different types of overservice enforcement strategies.

We assessed whether overservice and impaired driving enforcement were conducted but not how frequently these actions were conducted. In previous studies we found that although agencies tended to conduct sobriety checkpoints and saturation patrols fairly regularly, agencies that conducted overservice enforcement did so less than once per month (Erickson et al., 2015; Lenk et al., 2014). This is consistent with findings that many managers of alcohol establishments reported a low perceived certainty of being cited by law enforcement for service to intoxicated patrons (Lenk et al., 2018). A study of compliance checks to prevent alcohol sales to underage customers (i.e., youthful appearing persons attempt to buy alcohol under supervision of law enforcement) found that this strategy was initially effective in reducing sales to underage youth but effects of this enforcement strategy decayed within three months (Wagenaar et al., 2005). These findings led to a recommendation to conduct compliance checks more than once or twice per year. Similar research is needed to determine how frequently overservice enforcement needs to be conducted.

Many of the relationships between alcohol-related enforcement and traffic crashes were not statistically significant, however most were in the hypothesized direction: conducting alcohol-related enforcement was associated with fewer alcohol-impaired-driving driving crashes. An exception was, enforcement of open containers of alcohol in motor vehicles among municipal police agencies in urban/suburban and small town/rural communities (for both jurisdiction and the 10-mile buffer models). Very little is known about specific strategies that are used for open container enforcement and more research is needed to understand this type of enforcement.

A main limitation of this study is the cross-sectional study design, preventing us from assessing causality. However, local alcohol enforcement data are rare and we were able to use a unique secondary data source available at one time point to assess community-level overservice and alcohol-impaired driving enforcement strategies. Future studies should use a longitudinal design to examine how use of specific alcohol enforcement strategies affect alcohol-impaired-driving traffic crashes over time. In addition, by controlling for fatal crashes where BAC < 0.08, we accounted for many important factors that may affect the relationships between alcohol enforcement and traffic crashes, but other factors that we did not include may also be important such as impaired driving laws and alcohol consumption levels. Another limitation is that there is little guidance in terms of what is an appropriate geographic area for evaluating effects of enforcement on alcohol-impaired driving outcomes. Driving behavior could be affected by enforcement practices of the targeted community – or by enforcement practices of surrounding communities. Another limitation is the number of models estimated and the possibility of significant findings due to chance. With 30 models we would expect at least 1 or 2 significant effects just due to chance.

The two buffers we used—the boundaries of the law enforcement agencies’ jurisdictions and a 10-mile driving buffer from the center of the agencies’ jurisdictions—each have limitations. For example, alcohol-impaired-driving traffic crashes resulting from overservice of alcohol may not occur within the jurisdiction boundaries—especially fatal crashes that usually occur at higher speeds. This could lead to under estimate of effects of overservice enforcement because the enforcement actions within a jurisdiction may reduce crashes both inside and outside the jurisdiction. A 10-mile driving buffer may result in overlapping buffers across communities; which could also lead to inaccurate estimates. These issues demonstrate the importance of further research to determine the best geographic boundaries for assessing effectiveness of efforts by law enforcement agencies. We also assessed a 15-minute driving buffer measured from the center of the law enforcement agencies’ jurisdictions. We were concerned that this buffer was not ideal across a mix of densely populated urban communities and very sparsely populated rural areas. Given the concerns about this buffer, we did not present the results of these analyses; however, the findings were very similar to the findings we presented in this paper for the other buffers.

Another limitation is that enforcement strategies were assessed through self-report by one officer at each agency. It is possible that the participants may not know all enforcement activities that are happening in their agency, potentially leading to an under estimate of enforcement. On the other hand, participants may have given socially desirable responses about enforcement practices, potentially leading to an over estimate of enforcement. We tried to mitigate these biases by asking to survey the officer most knowledgeable about alcohol enforcement for their agency and by assuring respondents that individual answers would only be presented in aggregate. Furthermore, there currently is no other way to collect these type of enforcement data.

Despite these limitations, this study still makes an important contribution to the research literature. There is little research focusing on effects of community-level alcohol-impaired driving or overservice enforcement strategies. We assessed enforcement strategies reported by communities among randomly selected large and small communities across the U.S. Findings suggest that overservice enforcement and saturation patrols should be further assessed as an important additional enforcement tool as communities continue to address alcohol-impaired driving.