Gender bias in power relationships: evidence from police traffic stops

Abstract

We test for the existence of gender bias in power relationships. Specifically, we examine whether police officers are less likely to issue traffic tickets to men or to women during traffic stops. Whereas the conventional wisdom, which we document with surveys, is that women are less likely to receive tickets, our analysis shows otherwise. Examination of a pooled sample of traffic stops from five locations reveals no gender bias, but does show significant regional variation in the likelihood of citations. Analysis by location shows that women are more likely to receive citations in three of the five locations. Men are more likely to receive citations in the other two locations. To our knowledge, this study is the first to test for gender bias in traffic stops, and clearly refutes the conventional wisdom that police are more lenient towards women.

Acknowledgements

We thank Nicola Persico at the University of Pennsylvania for graciously providing the data. We also thank Kate Antonovics at the University of California, San Diego, and Brian Withrow at Wichita State University for providing documentation for some of the data.

Notes

¹ We have data on the duration of stops in two locations. In Wichita, Kansas, more than two thirds of all stops take 15 minutes or less, while in Bloomington, Illinois, more than three fourths of all stops take less than 15 minutes, with the median stop taking 10 minutes.

² We do not study Prob(Trangress); while it is quite possible that the probability of committing a violation differs between men and women, our focus in this article is on potential gender differences in treatment. Similarly, we do not study Prob(Caught | Transgress), though it is quite possible that this probability also differs by gender. This might happen because of a gender difference in behaviour (for instance, women might spend more time driving in school zones when school is in session, and police might tend to patrol in such areas and enforce the speed limit with particular zeal) or because in some cases the police officer might observe the driver's gender in advance, and this might influence the officer's decision to stop the driver, though in many cases the driver's gender may not be observed by the officer until after the decision to stop is made. This would surely be the case if the driver were stopped at night, an idea that was exploited in the analysis of racial profiling by Grogger and Ridgeway (2006), who refer to it as the ‘veil of darkness hypothesis’.

³ In the smaller classes, we know that every student in attendance on the survey day completed and returned a survey, though in the very large classes it is possible that some students exited the room without completing the survey.

⁴ See Altonji and Blank (1999) for a survey of these and other economic models of discrimination.

⁵ The notion that people prefer to interact with similar others has been central to a large body of theoretical literature in social psychology (e.g. Byrne, 1971; Tajfel and Turner, 1986). This literature attempts to explain the preference for similarity by arguing that compatibility, interpersonal attraction, and identity reinforcement are all increased when people interact with others who are similar, whereas incompatibility, discord and alienation result from interactions with others who are different. The idea that people prefer to interact with similar others has also been recognized in the sociological literature, in particular theories of organizational demography (Pfeffer, 1983).

⁶ Although we do not observe the gender of the police officers in our data, we know that in all five locations for which we have data at least 87% of the police officers were male. Furthermore, in 2003, 89% of police officers nationwide were male, according to the Uniform Crime Reporting (VCR) program data on police employee.

⁷ Although we could find no studies using traffic stop data to compare ticketing rates for men and women, two Department of Justice (DOJ) reports (2001, 2005) compiled from drivers’ self-reporting indicate that women receive a citation 3% less often than men conditional on being stopped by police.

⁸ Though it does not focus on traffic stops, the related analysis of Donohue and Levitt (2001) uses panel data for 122 large US cities to examine the relationship between the racial composition of a city's police force and the racial patterns of arrests. Increases in the number of minority police are associated with significant increases in arrests of whites but have little impact on arrests of nonwhites. Similarly, more white police increase the number of arrests of nonwhites but do not systematically affect the number of white arrests.

⁹ Three recent papers extend the theoretical analysis of Knowles et al. (2001) and conduct additional empirical tests. Persico and Todd (2004) show that the test proposed in Knowles et al. (2001) can also be applied in a more general environment where police officers are heterogeneous in their tastes for discrimination and in their costs of search and motorists are heterogeneous in their benefits and costs from criminal behaviour. Applying their proposed tests to data on police searches of motor vehicles gathered by the Wichita Police Department (WPD), they find evidence consistent with the notion that police choose their search strategies to maximize successful searches and not out of racial bias. Antonovics and Knight (2004) use data from the Boston Police Department (BPD) to show that, consistent with preference-based discrimination, officers are more likely to conduct a search if the officer and driver are of different races. Dharmapala and Ross (2004) re-examine the data used in Knowles et al. (2001), showing that the data are consistent with prejudice against black males, no prejudice, and reverse discrimination, depending on the type of equilibrium that exists.

¹⁰ The model assumes that officers observe a characteristic, θ, which is correlated with the likelihood that the driver is guilty but is unknown to the driver when he decides to carry drugs or other contraband. The same assumption appears in the model of Bjerk (2005).

¹¹ We downloaded these datasets from the website of Nicola Persico at the University of Pennsylvania.

¹² This is equivalent to simply considering the difference in the probability of receiving a ticket for men and women. We estimate probits for consistency with later analyses in which we include control variables.

¹³ To calculate this figure we divide the marginal effect by the corresponding probability that a male driver receives a ticket.

¹⁴ We also tried controlling for the number of passengers in the vehicle, whether the driver was a resident of the state in which he/she was stopped, as well as the year, month and day of the week that the stop occurred. Doing so does not change our results.

¹⁵ Of course, the results we report are identical to those obtained by estimating the model for all traffic stops and interacting a dummy for young drivers with the female dummy.

¹⁶ We created four different versions of this question: (1) Mike and Mr. Smith; (2) Kate and Mr. Smith; (3) Mike and Mrs. Smith and (4) Kate and Mrs. Smith.