You have to be careful about coming to fast conclusions based on these statistics. I can imagine these numbers being sucked up by various ill-informed political commentators to mislead the public on what police departments are doing. (Regardless, I love this work and would love to see it done nationwide!)<p>-It could be that some cops are assigned to densely white areas, while others must work in mostly black areas. This does not necessitate racism on behalf of the officers or management. (Perhaps it does in the overall local culture, as indicated by the self-segregation? To police an entire area, you must distribute resources to various portions of that area.) Racist cops might even <i>prefer</i> the dominant race of the area they work in, yet <i>still</i> mostly cite that dominant race, simply because there are so few disliked-race targets to look for.<p>-Certain violations cited to predominantly one race may simply indicate a <i>reality</i> in which people of that race break that law more often. This does not necessitate racist targeting on behalf of the officers. At that point, maybe a discussion about how seriously this law needs to be enforced should be had. Window tint? Dime bag of weed? Is that <i>really</i> such a big deal? (Ideally, these laws would just be removed. But politically this seems harder than using officer discretion.)
I worked in this area for the state of MD at one of the foremost traffic records outfits in the US and have done this work previously, there are a few considerations you are leaving out which comes as a result of knowing these data.<p>You want to never look at things at the citation level, you should group it into stops. One driver can get many citations per stop, which confounds your findings..grouping can be done by making an ID of DL#+Date of stop. If time is included, put that on ONLY if it has the same time for all citation issued in the stop (like it is here in MD) - You could end here, or you could add an indication Y/N for whether a citation was issued. This way, you are looking at stops where at least 1 citation was issued, instead of 3 counts for a race off of one stop where three tickets were issued.<p>Also, the racial categories are based upon the DMV database, and what options they have permitted. The confounder here is racially ambiguous persons may be misclassified, while obviously black people etc will reliably be called such.<p>Additionally, you want to use something like the National Household Travel Survey to get an idea of the racial background of DRIVERS in the area you are examining. It is quite possible that an area has more drivers of a certain race. Note this still will not be perfect due to out of jurisdiction drivers.<p>Please contact me with any questions. Very interesting stuff, I just want to make sure you do it right.....if you hope to make any changes at the govt level, these considerations will be widely known in the traffic records sector.
I plan do do this for many more counties. I'm sure it isn't a coincidence how difficult many are to scrape. This data often feels purposefully obfuscated. It only seems right to me that everyone should see local data aggregated like this. If police intent do use data to track the population, we should be able to track them as well.
The only objection I can think of that an honest police department might have against this sort of analysis taking place is that some figures might be taken out of context or not adequately explained, such as officers’ numbers who patrol neighborhoods that are predominantly one ethnicity.<p>If I believe in my work, I don’t mind it scrutinized, and would welcome the analysis, given that the analyst is not excessively biased.
This seems like a promising way of detecting bias. However, I would be surprised if this type of analysis was permitted to continue without objections from police organisations.
It would be nice if the actual racial distribution of the population of the county is added for context. For a random visitor from the internets the lack of context hinders the ability to understand what implications the data has.<p>When 90% of the population is black the conclusions are different from when the population is 90% white.
Redacting data in your results is very close to the same as not publishing your results. Please do not redact critical information that allows people to validate or otherwise cross-check your results.