A Stanford research study about police bias in traffic stops is getting increasing buzz.

Professor Sharad Goel along with graduate students Camelia Simoiu and Sam Corbett-Davies analyzed data from 9.5 million traffic stops from 287 police departments in North Carolina in their article Testing for Racial Discriminationin Police Searches of Motor Vehicles (pdf).

Sharad Goel from 5harad.com

Sharad Goel from 5harad.com

For various reasons they narrowed the data down to 4.5 million stops from the 100 largest departments and excluded the state patrol.

We find that nearly every department applies a lower standard of evidence when searching
blacks and Hispanics than when searching whites and Asians, suggestive of racial bias in
search decisions.


The discrimination was not uniform, finding “discrimination against blacks in 57 of the top 100 departments … [and] ambiguous evidence in 42 departments [with] only one department [showing] to an absence of discrimination against black drivers.”

Overall though the evidence was overwhelming:

In nearly every one of the 100 departments we consider, we find that black and Hispanic drivers are subject to a lower search threshold than whites, suggestive of discrimination against these groups. In many departments, we find the disparity is quite large, with the threshold for searching minorities 10 or even 20 percentage points lower than for searching whites.

Putting their numbers in another way, the article says:

Across the six years in our dataset, this analysis suggests that over 30,000 searches of black drivers would not have been conducted had officers uniformly applied the white threshold, making up one-third of all searches of black drivers.

The researchers softened the blow a little:

We cannot, however, definitively conclude that the disparities we see stem from racial bias. For example, officers might instead be applying lower search thresholds to those from lower socio-economic backgrounds, a demographic that is disproportionately black and Hispanic. At the very least, though, our results indicate that there are concerning irregularities in the search decisions we study.

They’re not done. The team has accumulated data from many other states and more analysis will be coming soon.