Become a Patreon!


Excerpted From: Rohit Asirvatham and Michael D. Frakes, Are Constitutional Rights Enough? An Empirical Assessment of Racial Bias in Police Stops, 116 Northwestern University Law Review 1481 (2022) (155 Footnotes) (Full Document)


AsirvathamAndFrakesBy July 6, 2016, getting pulled over was routine for Philando Castile. The Black thirty-two-year-old school cafeteria worker had already been stopped at least fifty-two times in the Minneapolis-St. Paul area for reasons including not wearing a seat belt, speeding, and driving at night with an unlit license plate. This time, the officer told Castile he was being pulled over for driving with a broken taillight. But perhaps that was not entirely true. Before stopping him, Officer Yanez had radioed a colleague that Castile fit the description of a robbery suspect, citing Castile's "wide-set nose." Nonetheless, when Officer Yanez walked up to Castile's car, he said, "the reason I pulled you over [is] your brake lights are out." Minutes later, Philando Castile was dead, shot by Officer Yanez.

In his final minutes, Walter Scott, a fifty-year-old Black man, heard something similar to what Castile had: "The reason for the stop is your brake light is out." Sam DuBose heard: "You don't have a front license plate on your car." Sandra Bland heard: "You failed to signal your lane change."

Traffic stops open the door to violence. They are one of the most common entry points for contact between civilians and the police. And the harms that can accompany a traffic stop encompass far more than physical violence. When an individual is stopped, her day is put on hold by the officer's authority as an arm of the state. When the officer walks up and looks inside the car's window, her privacy is invaded. Even the most routine stop can cause apprehension or fear, for some. And a resulting ticket or fine can have devastating effects on the driver. For example, drivers who cannot afford to pay the fine often lose their licenses. As a result, those who need a license to work will lose their jobs. And that in turn makes it harder for them to pay their fines and have their licenses reinstated. The stakes for ensuring evenhanded traffic enforcement are high. But traffic enforcement is not evenhanded. The evidence is clear: Black drivers are more likely to be stopped than white drivers, even when accounting for the benchmarking problem.

A critical question is how to reduce this racial bias. We must figure out which avenues of attack are likely to be fruitful and which are likely to fail. A dominant view in the literature is that a robust Fourth Amendment right against pretextual stops is one such promising avenue. Our overarching goal in this Article is to assess the merits of this conventional wisdom. Ultimately, our analysis demonstrates that this dominant view--while understandable at first blush--simply lacks empirical support.

To be sure, figuring out how to regulate traffic stops presents a knotty puzzle. Given the sheer number of traffic rules, most people commit some sort of technical violation almost every time they drive. Most would agree it is not a good idea to stop people for literally every traffic violation. That would be incredibly costly, without a sufficient corresponding marginal increase in safety. But most would also agree that it is dangerous to give drivers carte blanche to violate traffic laws. That too would be costly--in terms of safety. So the dilemma is that while we do want officers to stop some people who violate a traffic law, we do not want an officer's decision about whether to stop a given driver to be based on bad reasons. Two such bad reasons are (1) an unsubstantiated hunch of criminal activity and (2) race. Allowing for either of these reasons to serve as the basis for a stop could increase the likelihood of racial bias in stops because a hunch might be influenced by the suspect's race. Thus, the puzzle is how to protect drivers from being stopped on those bad grounds when, because everyone commits traffic violations, those bad grounds will almost always appear alongside a permissible reason for a stop. How do we know whether a driver was stopped because he was Black or because he was driving 7 miles per hour over the speed limit?

The Supreme Court could have said that pretextual stops--stops in which there was a traffic violation but an impermissible reason was the real motivation behind the stop--violate the Fourth Amendment. Instead, in Whren v. United States, the Court held that as long as there is a traffic violation, the stop does not violate the Fourth Amendment, regardless of the officer's real motivations. Understandably, the decision was met with a flood of scholarly criticism taking the position that, by failing to provide a Fourth Amendment right against pretextual stops, the Court exacerbated the problem of racial bias in traffic stops. And many argue that, given the scope that officers have to exhibit racial bias when permitted to conduct pretextual stops, a Fourth Amendment rule against pretextual stops would work to reduce that racial bias.

Nonetheless, there may be reasons to doubt this widely held view--that is, reasons to doubt whether constitutional prohibitions of pretextual stops would actually reduce racial disparities in traffic-stop rates in practice. First, we suggest that the means of enforcing these constitutional protections--i.e., the exclusionary rule--may insufficiently deter officers in the first place. After all, to the extent an officer's decision to initiate a traffic stop is heavily driven by factors other than the remote possibility that any evidence obtained during a pretextual stop will be suppressed, it is unlikely that a constitutional-rights-based approach will meaningfully reduce racial disparities in traffic-stop rates. Second, and relatedly, the evidentiary barriers inherent in proving pretext may fundamentally limit the degree to which a stricter standard would impact case outcomes and thus regulate officer behavior.

Altogether, despite the near consensus in the literature for addressing racial disparities in traffic-stop rates through a constitutional-rights-based approach, the theoretical case for that approach is simply ambiguous. Given this ambiguity, this inquiry comes down to an empirical exercise. And it is this exercise that is at the heart of our Article. We ask the question: Would a Fourth Amendment right against pretextual stops reduce racial disparities in the initiation of traffic stops? Our data suggest: no.

Until recently, empirically assessing this question was very difficult. At the time of the Whren decision, we lacked good data on police stops. And even as data improved, having only post-Whren data did not tell us much because there was nothing to which to compare it. Two Washington Supreme Court decisions changed that. Disagreeing with Whren, the Washington Supreme Court first interpreted its state constitution's Fourth Amendment analogue as prohibiting pretextual traffic stops. But thirteen years later, in 2012, it changed course and lowered that standard, sanctioning previously prohibited stops. As a result, we have a natural experiment through which to analyze the effect of a decision like Whren on racial bias in traffic stops.

We analyze data from over 7 million traffic stops in Washington (along with nearly 2.5 million traffic stops in Colorado in our control analyses) and explore how racial disparities in those stops evolved in the period around the 2012 change in constitutional law in Washington. Our analysis employs a number of empirical techniques to distinguish the effect of this change in the law from a range of both observable and unobservable factors that may likewise impact racial disparities in the rate of traffic stops. One particular challenge of this nature is posed by the fact that the change in the constitutional treatment of pretextual stops in the State of Washington that forms the basis for this natural experiment coincided to the month with Washington's legalization of recreational marijuana. To confront this challenge, we employ a novel approach that takes advantage of the fact that Colorado likewise legalized recreational marijuana within days of Washington but did not alter its Fourth Amendment-like treatment of pretextual stops. These facts allow us to incorporate a comparison of Colorado and Washington to untangle the effects of the two coinciding changes.

Taking a broad array of approaches, our analysis ultimately finds no robust evidence that the legal change in Washington had an effect on racial disparities in traffic stops. If anything, our findings suggest that Washington experienced a trend towards greater and greater racial disparities in traffic stops that predated the relevant change in constitutional rights, with the 2012 Washington Supreme Court decision that lowered the constitutional standard for pretextual stops doing nothing to further exacerbate the preexisting problem. Accordingly, this evidence contradicts the dominant assumption that Whren is a significant contributor to racial bias in traffic stops. And it thus cautions against looking to the Fourth Amendment for a solution to this critical problem. In its place, we propose that policymakers consider solutions that will more effectively deter officer behavior, including extraconstitutional approaches that discipline officer behavior based on their aggregate set of stops, as opposed to the choices that they make in individual cases (in which pretext is fundamentally hard to prove).

We acknowledge that we are not the first to analyze this Washington data and the relevant change in Washington law. Professors Stephen Rushin and Griffin Edwards, in an article in the Stanford Law Review, also used the Washington traffic-stop data to examine the impact of the resulting change in constitutional rights on racial bias in traffic stops. But they concluded that the data indicate that doctrines like Whren's "contribute to a statistically significant increase in racial profiling of minority drivers." As previewed above, our findings suggest no such effect.

When presenting our findings, we discuss several areas of departure between our analysis and that of Rushin and Edwards. Primarily, we build on their work to implement a novel exercise designed to confront a fundamental challenge facing this investigation: the need to separate the effect of the change in Washington's constitutional treatment of pretextual stops from its contemporaneous legalization of recreational marijuana. For these purposes, we take advantage of the fact that Colorado legalized recreational marijuana within days of Washington but did not alter its treatment of pretextual stops, allowing us to estimate a triple-differences specification that effectively uses Colorado as a control in our analysis.

Second, our analyses build on Rushin and Edwards's with respect to another key challenge facing this exercise: the fact that driver race is not recorded for 27% of the stops in the dataset, an amount greater than the proportion of stops with non-white driver race recorded. One of Rushin and Edwards's key approaches assumes that all drivers with missing race codes are non-white. We set forth two empirical markers that question the validity of this assumption. However, we find their alternative approach-- dropping stops with missing race--less problematic. We nonetheless expand on that approach in several ways; primarily, we set forth imputation and other exercises designed to diagnose the extent of any possible bias that might arise from the missing-race data. These exercises suggest no concerns over any such bias.

Third--and perhaps speaking more specifically to why our findings do not align with those of Rushin and Edwards--we depart from their analysis in certain functional-form assumptions that we make in constructing our respective difference-in-difference and triple-differences analyses. Our reading of their approach is that they test for changes in absolute counts of traffic stops for non-white drivers before and after the Washington law change and compare those changes with observed changes in absolute counts of stops for white drivers before and after the reform. Given substantial baseline differences in population counts across races and thus in stop counts across races--especially in large counties--we elect instead to focus on changes in log counts of traffic stops. This allows us to make comparisons across races in the proportional (not absolute) responses in traffic stops to the Washington law change. This provides a more flexible approach that facilitates a comparison that is not sensitive to the scale of the outcome variable (and thus the baseline differences) and that arguably captures the relationship of interest. We demonstrate that Rushin and Edwards's results are rather specific to the functional form they select and are sensitive to this and related modifications.

Our analysis proceeds as follows. In Part I, we first lay out the legal backdrop for our inquiry. Then, we present and question the dominant position in the literature. We argue that a Fourth Amendment-type right against pretextual stops would do little to reduce racial disparities in traffic stops. In Part II, we introduce and analyze the Washington legal change that serves as the basis for our natural experiment. In Parts III and IV, we explain our natural experiment and results. In Part V, we discuss the implications of our results. First, we suggest that the exclusionary rule is an ineffective deterrent in the context of traffic stops. And second, we very briefly sketch out avenues for reducing racial bias in traffic stops that are worth exploring in the alternative.

[. . .]

A dominant view among scholars of the Fourth Amendment is that a permissive stance towards pretextual traffic stops contributes to greater disparities in the rates of traffic stops across race. We suggest, however, that there may be conceptual grounds to doubt this conventional view and stress the weaknesses in the deterrent channel linking the substantive constitutional law and the officer decision-making process. Ultimately, whether or not the imposition of a Fourth Amendment-like prohibition against pretextual stops will reduce the racial disparities observed in such stops is an empirical question. Drawing on a natural experiment based on a change in the constitutional treatment of pretextual stops in the State of Washington and employing a range of empirical techniques, we fail to find evidence in support of the conventional wisdom. To the extent that policymakers wish to address the undeniable problem of racial bias in the rates of traffic stops, our evidence cautions against placing all of our hopes in a constitutional-rights-based approach. This is not to say that such rights are unimportant. They simply are not enough to attack this critical issue. We propose that policymakers consider certain extraconstitutional approaches, including those better designed to bolster a deterrent channel or those that attempt to remove officer discretion from traffic regulation altogether.

[. . .]

We now address one final empirical concern. We have conducted this analysis thus far using data from the Washington State Patrol. However, one may be concerned that we investigated the impact of a change in the law bearing on pretextual stops in a setting in which one might have predicted a weaker impact in the first place. After all, even in the pre-Arreola period when pretextual stops were prohibited in Washington, it would perhaps be hard to challenge the decision of a state trooper since much of their job entails regulating traffic. Perhaps the change in the law would be expected to more meaningfully change the outcome in cases involving other types of officers-- i.e., officers involved less heavily in enforcing traffic laws and perhaps primarily focused on other crimes.

The challenge here is largely one of data. We do have data on traffic stops recorded by the Seattle Police Department--i.e., on a jurisdictional unit other than the Washington State Patrol and one that may be relatively less focused on regulating traffic. However, with these data, we do not possess information on the race of the driver. Nonetheless, to at least shed light on whether the experiences of the Washington State Patrol may generalize to officers more generally, we conduct a simple exercise in which we observe the trend over time in the differential stop rate between drivers stopped in the south and east precincts of Seattle--precincts with notably higher nonwhite populations (especially, larger Black populations)--and drivers stopped in the remaining precincts (note that we find similar results when just using the south precinct among these two given that its non-white population is the largest). While this does not necessarily focus specifically on the differential trend in the rate of stops of non-white drivers relative to white drivers, it focuses on comparing two groups of drivers, one of which is more likely to be non-white. If Arreola did increase racial bias in stops, one might expect to observe an aggregate effect that would be picked up with this approach.

As demonstrated by the following graph, however, we do not find evidence suggesting an increase in racial bias in stops by the Seattle Police Department that arose from Arreola. If anything, we see a trend prior to Arreola towards more stops in those precincts with higher non-white populations relative to the other precincts, and a possible reversal of this trend thereafter, inconsistent with what one would expect if Arreola exacerbated racial bias in stops.

Rohit Asirvatham is a law clerk to Judge David R. Stras, United States Court of Appeals for the Eighth Circuit.

Michael Frakes is the A. Kenneth Pye Professor of Law and Professor of Economics at Duke University and a Research Associate at the National Bureau of Economic Research.

Become a Patreon!