How the 'veil of darkness' sheds light on racial profiling June 15, 2016

Andrew Ba Tran
Data Editor

The proportion of minority drivers pulled over by police in Connecticut is greater during daylight hours when officers can more easily identify race, according to an analysis of Connecticut's traffic stop data.

Traditionally, statistical methods of looking for signs of racial profiling by police compare stops to population, however researchers at Central Connecticut State University's Institute for Municipal & Regional Policy have added another level of analysis, looking specifically at what they call the "veil of darkness."

The veil of darkness method looks at stops around the time of sunrise and sunset and assumes an officer can better observe a driver's skin color when the sun is out. The CCSU researchers limited their analysis to stops that happened during dawn and dusk inter-twilight hours, the time span before and after the sun rises or sets on a given day. This helps normalize the effects of a variable driving population.

For example, 6 p.m. during the winter is dark and 6 p.m. during the summer is light, but in general the driving population should be similar at that time. The proportion of stops during the two seasons should theoretically be constant, researchers reasoned.

Statewide, researchers found evidence that minority drivers are stopped more often during daylight hours, particularly for Black or Hispanic drivers during dawn and dusk inter-twilight hours.

Six municipal police departments and one state police troop represent the only jurisdictions that had a statistically significant disparity in either black or Hispanic motorists alone: Ansonia, Bloomfield, New Milford, Norwalk, West Hartford, Wethersfield, and State Police Troop H.

This evidence of racial or ethnic disparity indicates the possible existence of disparate treatment at the department level.

Researchers' model included considering factors such as race, day of the week, visibility, and stops that indicated stops that were more likely to be conducted during day time. The model created a coefficient score for each department, ranging from insignificant to significant.

Trend CT did a simpler analysis than CCSU researchers (nerd alert: Theirs involved a generalized linear modeling, synthetic control analysis, and KPT hit rate adjustment).

Instead we compared the percent of minorities stopped during full daylight versus night to see if there was a disparity at an aggregate level. We determined and counted all stops conducted during daylight and stops in the dark and left out stops in the civil twilight period.

According to our analysis, departments pulled over minorities at a higher rate during daylight than darkness in all but four towns.