RESEARCH

Working Papers

“Job Market Paper – Race and Experimental Design: How Respondents may Read Context into a Neutrally Framed Scenario”

Abstract: In the field of Law and Economics, experiments often use “neutral” language to avoid biasing respondents against certain actions such as stealing. However, a ``neutral'' frame for one group may not be the same ``neutral'' for another. In an online experimental study, we test the effect of framing on a two-player ``stealing game'' and compare the effect across samples of Black and White respondents. The behaviors of Black respondents in a neutral frame aligned much closer to those taken in a less serious ``casino game'' than in a more serious ``police game'' where actions were explicitly mentioned as ``stealing''. This was in contrast to their White peers, who had distinct reactions to all frames. This suggests that past experiments may overstate the tastes for crime of Black respondents. How much a respondent trusted the police was also relevant. Ultimately, we caution that academics doing experimental work should consider whether experimental norms are overly tailored to certain groups. A better approach may be to embrace context.

“Social Identity and Preferences: A Replication” with Nathan Fiala

Abstract: In a highly cited paper, Benjamin et al. (2010) tested the main contention of self-categorization theory: That increasing the salience of an individual’s racial identity will cause them to align their preferences closer to what is expected from them as a member of that racial group. This was done through an experiment that assessed the effect of ethnically charged survey questions on the time and risk preferences of Asian American respondents. Primed Asian respondents were more patient than those in the control group with no effects observed for primed White respondents. However, this experiment was done on a sample of students from very prestigious and even Ivy League universities. We attempt to replicate this study due to concerns that this sample is not properly representative of the general public. Since this was conducted during the COVID-19 pandemic, these experiments were conducted remotely. While we do find the effects of variables such as larger stakes to be similar to what was reported in the original study, we fail to replicate the main result of the paper that Asian Americans chose more patient options and had lower Log-interest rates after treatment. On a sample of the general population, treatment actually increased Log-interest rates for Asian respondents. When we restricted this sample to only students, the sign of the effect was in line with BCS. However, the effect was not significant. This may suggest the results of BCS were indeed due to it using a non-representative sample of convenience. 

“Non-homogenous Force and Factors that can Obscure Racial Bias in its Application by Police” with Mike Shor

Abstract: A number of studies have suggested that strategic compliance can obscure racial bias as inferring racial bias directly from observed use-of-force data ignores the action (and agency) of the suspect. This is important for the contentious debate regarding whether the evidence of racial bias in use of “high-level” force such as officer weapons use is weaker than that of “lower-level'' uses of force that cause less bodily harm. Based on these insights, we develop a strategic interaction model where suspects invest in a level of compliance with police based on their expectation of race-specific use of force, which is itself separated into distinct levels of force. We not only develop a theoretical proof that shows the observed rate of force can obscure disparities in likelihood of force, but that the multi-level nature of force can itself affect the information that can be garnered about officer's sensitivity to race from empirical data. Ignoring strategic compliance may lead researchers to significantly understate evidence for racial bias, and our theoretical results caution against rejecting the hypothesis of racial bias without ensuring the examples of force observed are truly comparable.