In the Fall of 2018, I worked with the researchers at the Stanford Open Policing Project (OPP) in the Stanford Computational Policy Lab, to implement and run the Veil of Darkness Test, as proposed by Grogger & Ridgeway in 2006 using the large OPP dataset. The premise of the test is that the race of a driver is less visible to the police when it is dark outside than when it is light. Therefore, in the presence of racial bias, there should be a greater disparity in the racial breakdown of drivers stopped during daylight hours than at night. There are a number of factors to control for, most critically time of day, location of the stop.
My responsibilities included:
- Finding states and cities for which the data had the necessary covariates for the test, e.g. driver race and stop location
- Developing raw data visualizations to provide context and intuition for regression models
- Implementing the test in R for a large dataset aggregated across cities and states
- Analysing the results of the visualizations and regression models and drawing conclusions about the existence of racial bias and the statistical strengths and limitations of the method
