Publication

Abstract: We test for a "Ferguson Effect" by studying how police effort responds to different incidents of police violence. We do so using two settings in Minneapolis: (1) George Floyd's murder, and (2) police-involved shootings. We find that following George Floyd's death, arrests and police-initiated calls decreased by 62 and 69 percent, respectively. By comparison, arrests and police-initiated calls decreased by 3 and 1.5 percent following police-involved shootings. We conclude that incidents of police violence generate "de-policing," and the effect is much larger following highly publicized incidents.

Working Papers

Abstract: Recent events of social unrest in the U.S. have heightened concerns about police violence and its consequences. While the literature has mostly focused on the deterring effects of policing, the effects of police use of force on crime and civilian cooperation have not been extensively explored. We use ShotSpotter data and 911 calls to measure the effect of exposure to police shootings in Minneapolis. Since the ShotSpotter data provides an objective measure of the actual level of gun-related crimes, we are able to differentiate between the effect on crime incidence and crime reporting. Exploiting the variation in the timing and the distance to these incidents, we show that while exposure to a police shooting increases gun-related crimes by 3-6 percent, it has no effect on shots reported. Taken together, this implies police shootings reduce civilian crime reports to police by 4-6 percent.


Kleinman Center Policy Digest 

Abstract: Voltage quality issues---such as sags, spikes, and fluctuations---are pervasive across many low- and middle income countries, yet their unobservability has hampered large-scale analysis of their economic importance. This paper uses state-of-the-art measurements to shed light on this novel dimension of publicly provided infrastructure, which underpins most modern industrial, commercial, and residential activity. We characterize voltage quality issues experienced by electricity customers in urban Ghana, quantify the associated economic costs, and estimate the causal impact of a modest improvement in voltage quality as a result of quasi-random electricity grid investments. To do so, we combine minute-by-minute customer-level power measurements with detailed panel surveys of more than 1,500 households and small businesses across Accra. The analyses produce three main findings. First, we document widespread voltage problems: approximately 20% of grid electricity is of poor quality (more than 10% below nominal voltage, 230V), including 30% during peak hours. Second, both residential and business customers purchase expensive equipment to protect against voltage problems and are willing to increase monthly electricity spending by 10% for power supply with half of the existing voltage problems. Third, using a differences-in-differences design, we find that a 5V increase in average voltage---resulting from the construction of additional transformers at selected sites---reduces the frequency of appliance damages at treatment sites relative to control sites, but generates no other economic impacts after 1 year. Longer-term and systemic improvements in electricity usage may require more significant voltage quality improvements through additional infrastructure investments.


Abstract: Racial disparities have been well-documented at every stage of the criminal justice system, as have the roles of police, prosecutors, judges, and juries in driving those disparities. In contrast, little is known about the role of race in the defense of low-income defendants by court-appointed attorneys. This is important since 80 percent of criminal defendants rely on assigned counsel for legal defense. Exploiting the quasi-random assignment of court-appointed attorneys to cases in Travis County, Texas, we test whether defense attorneys secure better deals for same-race defendants using a difference-in-differences approach. Results indicate that while Black and White attorneys are similarly effective at securing dismissals for White defendants, Black attorneys are less effective than White attorneys at securing dismissals for Black defendants. Specifically, Black defendants who are represented by White rather than Black attorneys are 20-22 percent more likely to have their charges dismissed and 17-27 percent less likely to be incarcerated. Moreover, we show that Black defendants who are represented by White attorneys are not more likely to re-offend in the future.


Click here for the AEA poster video! 

Abstract: While there is much desire for holding police accountable for misconduct, there is little evidence on whether the most common accountability system, internal affairs, impacts police behavior in intended or unintended ways. Using data from a large city where there is conditionally random assignment of officers to 911 calls, I employ regression discontinuity and difference-in-differences methods to distinguish the impact of investigations from confounding factors. Results indicate that increased oversight from internal investigations does not change an officer's likelihood of making an arrest or using force.  This is true across different types of allegations, including those that are sustained. Surprisingly, even imposing sanctions after a sustained allegation does not change police behavior, irrespective of the severity of the sanction. This has important policy implications, as it suggests that the current system of internal oversight has no impact on police behavior.


Abstract: In this paper, we study the disparate impact of court-appointed attorney quality on Black defendants. Using the quasi-random assignment of court-appointed attorneys to cases in a large Texas county, we show that Black defendants are disproportionately harmed by low-quality attorneys. We show this in two ways. First, while a one standard deviation increase in attorney quality---measured as the ability to get a case dismissed---increases the likelihood of case dismissal by four percent for Black defendants, it has no effect on White defendants. Second, relative to dismissal rates achieved by the best attorneys in our sample, Black defendants with lesser-quality attorneys are 15% less likely than White defendants to have their cases dismissed.  Additional results indicate that lower-quality attorneys exert less effort when representing Black relative to White defendants, consistent with racial bias driving some or all of the disparate impact.