Extending the Veil of Darkness Approach: An Examination of Racial Disproportionality in Traffic Stops in Durham, NC

Published date01 December 2017
Date01 December 2017
Subject MatterArticles
Extending the Veil of
Darkness Approach: An
Examination of Racial
Disproportionality in
Traffic Stops in
Durham, NC
Travis A. Taniguchi
, Joshua A. Hendrix
Alison Levin-Rector
, Brian P. Aagaard
Kevin J. Strom
, and Stephanie A. Zimmer
Developed in 2006, the veil of darkness approach is one of the most widely accepted
methods for assessing the impact of driver race on traffic stops. Building on the
original methodology, we innovate in three important ways to enhance the veil of
darkness approach: (a) invoke generalized linear mixed models to account for the
lack of independence among observations in traffic stop data sets, (b) decompose the
relationship between daylight and driver race to consider the role of driver sex, and
(c) assess variability in racial disproportionality across law enforcement units. Nearly
20,000 traffic stops are analyzed for the Durham (NC) Police Department. Results
indicate that more than 10% of the variability in the rate of Black drivers stopped is
accounted for by officer-level factors, racial disproportionality was only for male
drivers, and evidence of disproportionality was found among some units, but no
evidence was found among others.
veil of darkness, traffic stops, racial disproportionality
RTI International, Research Triangle Park, NC, USA
Corresponding Author:
Travis A. Taniguchi, RTI International, 3040 East Cornwallis Road, Research Triangle Park, NC 27709,
Email: taniguchi@rti.org
Police Quarterly
2017, Vol. 20(4) 420–448
!The Author(s) 2017
Reprints and permissions:
DOI: 10.1177/1098611117721665
Racial prof‌iling by law enforcement occurs when authorities target individuals
because of their race or ethnicity rather than their behavior. It is a social issue
that is counter to the constitutional guarantee of equal treatment under the law,
and when it is perceived to exist, police–community relations may suf‌fer. Traf‌f‌ic
stops are not only the most common reason for contact with the police (Bureau
of Justice Statistics, 2016), for many people, traf‌f‌ic stops are likely to be the only
type of interaction they ever have with law enforcement. Thus, the nature of
police contact, its circumstances, and outcomes can have dramatic and possibly
unrivaled inf‌luence on how people view the police (see Smith, Graham, &
Adams, 1991; Worrall, 1999). It is well established that perceptions of unfair
or discriminatory practices by the police have serious implications for commu-
nity cooperation and civil disobedience (Cox & Fitzgerald, 1996; Tyler & Fagan,
2008), support for the police (Tyler & Wakslak, 2004), race relations, crime
reporting, and public safety (see Brown & Benedict, 2002 for a review).
Therefore, racial prof‌iling by police is a critical social issue and an important
research area in the social sciences.
Assessing how much racial prof‌iling by police occurs in the United States is a
complex issue, primarily because it is one of the most dif‌f‌icult social phenomena
to study scientif‌ically. It is an extremely challenging task to prove def‌initively
that the nature or outcome of a traf‌f‌ic stop would have been dif‌ferent if the
driver was of a dif‌ferent race, or that the of‌f‌icer has an explicit or implicit bias
toward people of color that inspires discriminatory police practices. Although a
Gallup survey in 2003 revealed that 59% of Americans consider racial prof‌iling
to be widespread among police stops (Ludwig, 2003; see also Weitzer & Tuch,
2002), this study cannot speak to whether perceptions of racially biased police
practices ref‌lect reality.
While an extensive body of literature exists examining the factors that inf‌lu-
ence police conduct once a community encounter occurs (Riksheim & Chermak,
1993), there is less knowledge of the factors that inf‌luence a police decision to
make a traf‌f‌ic stop in the f‌irst place. Traditionally, the impact of race on traf‌f‌ic
stops has been assessed by assuming that driving patterns can be approximated
using census population estimates (Baumgartner & Epp, 2012a, 2012b) or traf‌f‌ic
collision data (McDevitt & Iwama, 2016) or by conducting traf‌f‌ic surveys to
quantify the race distribution of motorists (Smith et al., 2004). All of these
approaches have serious limitations. Census populations are a poor proxy for
populations at risk for traf‌f‌ic stops, and traf‌f‌ic surveys are not only expensive,
they tend to have limited generalizability because of small sample sizes and
geographic coverage. Use of traf‌f‌ic collision data is a better denominator in
some respects because it more accurately ref‌lects the driving population
(Alpert, Smith, & Dunham, 2004). But this benchmark construction approach
assumes that accidents, and reporting of accidents to the police, do not vary
systematically. There have in fact been numerous attempts to establish an appro-
priate, unbiased benchmark with which to compare police traf‌f‌ic stop data, but
Taniguchi et al. 421

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT