Veil of Darkness and Investigating Disproportionate Impact in Policing: When Researchers Disagree

Published date01 March 2021
Date01 March 2021
Subject MatterArticles
Veil of Darkness
and Investigating
Impact in Policing:
When Researchers
Michele Stacey
Heidi S. Bonner
Disproportionate impact in policing has long been a concern for researchers and
practitioners alike, with much of the focus on traffic stops. While there are many
methods used to determine disproportionality in traffic stops, the veil of darkness
(VOD) approach has increasingly become one of the most popular frameworks.
Although there is consensus on certain aspects of the method, researchers utilizing
VOD disagree on the appropriate sampling strategies. This research examines the
original VOD method and three different sample restrictions proposed within the
VOD literature to demonstrate the effect each has on the conclusions drawn. The
results indicate that there is variation in the estimates of disproportionality depend-
ing on the sampling strategy used. As such, researchers using the VOD method must
be cautious in their sampling decisions in mid-size jurisdictions due to the impact
these such choices have on the conclusions drawn about disproportionate impact.
veil of darkness, disproportionate treatment, benchmark, traffic stops, law
enforcement, police
Department of Criminal Justice, Thomas Harriot College of Arts and Sciences, East Carolina University
Corresponding Author:
Michele Stacey, Department of Criminal Justice, Thomas Harriot College of Arts and Sciences, East
Carolina University, 243 Rivers Building, Mailstop 155, Greenville, NC 27858, United States.
Police Quarterly
2021, Vol. 24(1) 55–73
!The Author(s) 2020
Article reuse guidelines:
DOI: 10.1177/1098611120932905
Early work on disproportionate impact was developed with the intent to inves-
tigate disparity in policing outcomes. These preliminary studies of racial dispar-
ity started in the 1960s (Black, 1971; Black & Reiss, 1970; LaFave, 1965; Reiss,
1971), and research on racial prof‌iling (centered on traff‌ic and pedestrian stops
because they held the greatest potential for perceptions of [or actual] racial bias,
Fridell et al., 2001) was introduced in the late 1990s (Engel & Calnon, 2004;
Smith & Petrocelli, 2001; Smith et al., 2004, 2017; Withrow, 2006). More recent-
ly, following several high-prof‌ile uses of lethal force, there has been a renewed
focus on disproportionality in enforcement decisions which culminated in the
creation of the President’s Task Force on 21st Century Policing (2015; see also
Smith et al., 2017).
More recent inquiries into disparate impact have utilized the veil of darkness
(VOD) approach (Grogger & Ridgeway, 2006). While VOD is seen as a marked
improvement to other forms of benchmarks, there is little consensus on the
sampling strategies to employ when using this method. The current inquiry
utilizes traff‌ic stop data from a mid-sized police department in the southeastern
United States to conduct a series of analyses that represent points of disagree-
ment regarding sampling in the literature about the VOD method in order to
highlight how these variations affect the results reported.
Measuring Disproportionality
Most inquiries into disproportionality require a benchmark, or a baseline com-
parison group. A benchmark is an established standard against which data can
be measured (International Association of Chiefs of Police, 2006), and quality
benchmarks should indicate both the race proportion if there is no dispropor-
tionate impact as well as who is at risk (Fridell, 2005a, 2005b). However, devel-
oping appropriate benchmarks for traff‌ic stops was not easy. Early analyses of
racial prof‌iling used Census data as a benchmark which did not accurately
represent people at risk of law enforcement intervention (Baumgartner et al.,
2018; Engel & Calnon, 2004; Farrell et al., 2004; Fridell, 2004; Tillyer et al.,
2010; Worden et al., 2012). Use of weighted Census data (Novak, 2004; Rojek
et al., 2004; Smith & Petrocelli, 2001; Walker, 2001), DMV data, or blind
enforcement methods (e.g., red light cameras) were preferable to original
Census data but also not without limitations (Fridell, 2005a; Smith et al.,
2017; Walker, 2001; Withrow, 2006).
Scholars have also proposed to compare stops to accident data, as those
involved in accidents are a group that is less likely to be statistically biased in
racial/ethnic compositions (Alpert et al., 2004; International Association of
Chiefs of Police, 2006; McDevitt & Iwama, 2016). However, this also assumed
that there was no variation in accidents, or how they are reported (Taniguchi
56 Police Quarterly 24(1)

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