Proposing a Benchmark Based on Vehicle Collision Data in Racial Profiling Research

• Date: 01 December 2015
• Author: Howard Williams,Brian L. Withrow
• DOI: 10.1177/0734016815591819
• Published date: 01 December 2015
• Subject Matter: Articles

Article
Criminal Justice Review
2015, Vol. 40(4) 449-469
Proposing a Benchmark
ª 2015 Georgia State University
Reprints and permission:
Based on Vehicle Collision
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0734016815591819
cjr.sagepub.com
Data in Racial Profiling
Research
Brian L. Withrow1 and Howard Williams1
Abstract
Estimating the racial and ethnic proportions within a constantly changing population of drivers is
difficult. Commonly called benchmarks, these estimates are the basis upon which researchers
determine the potential for racial profiling. Most benchmarks do not consider the effect driving
frequency has on exposure to routine police supervision. Furthermore, racial profiling research has
a tendency to focus on overrepresentation at the expense of underrepresentation. This research
demonstrates the use of a benchmark based on vehicle collision data. Benchmarks based on these
data are more valid and reliable and enable researchers to consider driving frequency and the
potential for disengagement.
Keywords
police organization/management, law enforcement/security, police culture/accountability, police
processes, evaluation research, other
Introduction
Our inability to devise a universally acceptable method for measuring racial and ethnic proportions
within an ever-changing driving population remains one of the most controversial methodological
challenges in racial profiling research. Commonly called benchmarking, this measurement chal-
lenge is essential to our ability to determine whether law enforcement programs (e.g., traffic stops)
might be devoting too much attention to citizens of certain racial and ethnic groups.
The importance of a valid and reliable benchmark is not limited to the ruminations of research meth-
odologists. Racial profiling studies based on poorly constructed benchmarks cause political and public
relations problems and sometimes result in ill-fated litigation. We offer the following as an example.
During the summer of 2010, the State of Missouri released the 10th iteration of its Annual Vehi-
cle Stops Report. This report was based on police stop data collected in 2009, thus it is referred to as
1 Texas State University, San Marcos, TX, USA
Corresponding Author:
Brian L. Withrow, Texas State University, 601 University, San Marcos, TX 78666, USA.
Email: bw32@txstate.edu

---

450
Criminal Justice Review 40(4)
the 2009 report. Almost immediately, attention focused on the disparity index, arguably the most
controversial statistic in the report. The disparity index is calculated by dividing the proportions
of individuals stopped (reported by the policing agency) by the proportions of individuals available
to be stopped (based on the benchmark estimate of the driving population) for each racial or ethnic
group. If these proportions are equal, then the result will be 1, indicating no disparity for that racial
or ethnic group. If the disparity index is greater than 1, then the attorney general concludes that racial
or ethnic group is overrepresented in stops. Alternatively, if the disparity index is less than 1, then
the attorney general concludes that racial or ethnic group is underrepresented in stops.
The 2009 report indicates that overall ‘‘the disparity indexes for African-American drivers have
increased [in] each of the last five years’’ (Missouri Attorney General, 2010, p. viii). With one
exception (2004), this disparity index has increased each year for the past decade with respect to
African American drivers. The attorney general interprets this to mean that African American driv-
ers are increasingly more likely to be stopped by police officers throughout the state of Missouri.
To estimate the population of individuals available to be stopped in each jurisdiction, the
Missouri attorney general uses the most current estimate of the residential population over the age
of 16 years. In doing so, the attorney general makes no attempt to adjust these estimates to account
for actual driving patterns in and around population centers.1 In short, Missouri’s attorney general
assumes that the population of individuals available to be stopped by the officers in any community
is limited to the individuals who actually reside in that community. This creates a problem for some
communities, particularly those that are populated overwhelmingly by one racial or ethnic group,
located near or adjacent to a large urban area, and responsible for policing a significant transporta-
tion system, such as an interstate highway that introduces a large amount of transient traffic2 into
their community. Here is an example.
In an article appearing in the St. Louis American3 on June 2, 2010, the City of Ladue (a suburb of
St. Louis) was identified as one of the ‘‘10 worst places to drive black [sic] in Missouri’’ (St. Louis
American, 2010, p. 2). The 2009 disparity index for African American drivers in the City of Ladue is
17.11—the highest in the state. Attorney General Chris Koster observes that ‘‘it was more than 1,700
percent more likely a black [sic] driver will be stopped in Ladue based upon the African-American
population of Ladue’’ (pp. 2–3).
As one might imagine, the response to this was quick and loud, particularly from advocacy groups
like the The National Association for the Advancement of Colored People (NAACP). And, it should
be. For a decade, the disparity index in Ladue, MO, has been consistently either the highest or within
the top 10 highest in the state. Taken literally, the numbers suggest that in 2009, the 63 Black residents
of Ladue, MO, were stopped nearly 18 times each because according to the attorney general, only indi-
viduals who actually reside in Ladue are subject to being stopped by the Ladue Police Department.
Ladue is a small, wealthy, suburban community populated nearly exclusively by White residents
next to a large urban center that is populated principally by Black residents. The accusation that the
Ladue Police Department targets Black drivers in order to deter those people from entering their
quiet leafy streets is an easy one to make. But, such an accusation based on the attorney general’s
disparity index alone would be wrong. Interstate highways 170 and 64/40 and state highways 67 and
340 either course through or abut the city. All four highways are within Ladue’s city limits and are
actively patrolled by the Ladue Police Department. Collectively, according to the Missouri Depart-
ment of Transportation, these four major transportation systems handle over 300,000 vehicles each
day, a number of people representing 48 times the population of Ladue, assuming of course that each
vehicle contains only one driver. It is a safe bet that the majority of these drivers are not counted in
the residential population of Ladue.
It would seem that the exclusive use of a single benchmarking method for every community
regardless of their particular enforcement contexts would be rather irresponsible. In the case of
Ladue, it would appear that the use of the residential population to estimate the driving population

---

Withrow and Williams
451
would be inappropriate, given the large amount of transient drivers who enter but do not reside in the
community. Of more concern is the potential that this benchmarking method would fail to reveal an
actual pattern of racial disparity in stops or disengagement.
The purpose of this article is to demonstrate the viability of a benchmarking strategy based on
vehicle collision data. Such benchmarks have been used for many years by traffic engineers and
insurance companies in order to determine the relative risk (RR) of collision involvement among
various classes of drivers and types of vehicles. This strategy would be beneficial to communities
like Ladue that are populated overwhelmingly by one racial or ethnic group, located near or adjacent
to large urban areas, or responsible for policing a significant transportation system that introduces a
large amount of transient traffic into their community. In addition, benchmarks based on vehicle col-
lision data enable researchers to account for driving frequency and to measure the potential for
disengagement.
Literature Review
Information on police stops is rather meaningless unless it can be compared against a valid measure
of individuals, by race and ethnicity who are exposed to police supervision, or more accurately sub-
ject to law enforcement attention. Without a reliable estimate of the driving population and its racial/
ethnic proportions, conclusions of racial prejudice are premature (Engel, Calnon, & Bernard, 2002),
accusations of racial profiling may not be successfully proven (Statistical Assessment Service,
1999), and policy changes may result in inappropriate corrective measures (McMahon, Garner,
Davis, & Kraus, 2003). According to MacDonald (2003), ‘‘Until someone devises an adequately
sophisticated benchmark that takes into account population patterns on the roads, degrees of law
breaking, police deployment patterns, and the nuances of police decision making, stop data are as
meaningless as they are politically explosive’’ (p. 22).
According to Walker (2003), effective benchmark must meet three basic criteria. First, the bench-
mark must be scientifically credible. It should be methodologically sound and able to withstand the
rigors of peer review. Second, the benchmark should have practical utility. It should provide insight
into the findings and illustrate a solution to the problem. And third, the benchmark must have polit-
ical credibility. Racial profiling research never...