Monday, September 22, 2014

Disparity vs. profiling

As I mentioned yesterday, I take exception with the ACLU's assertion that racial disparity in traffic stop data indicates"alarmingly high" rates of racial profiling. Disparity and profiling are not the same. While I think the disparity should cause (and has caused) us to step back and take a good look at what's going on, to conclude that racial profiling is causing the disparity without further examination is a logical leap.

Profiling, the act of targeting people of color due to their race, assumes that police officers and agencies, either intentionally or subconsciously, are acting upon racial prejudice and bias. Disparity however, can occur for a variety of reasons. For example, the Crime Commission compares traffic stops to population demographics. But what if census data doesn't reflect who's actually on the road, or what their likelihood is of committing a traffic infraction? A better denominator than population might be the data on licensed drivers, or maybe the age profile of drivers (since we know age is related to traffic violations), or an actual roadside count of motorists.

After lots of thinking about this over the past 12 years, and lots of exploration in our data, I think the best denominator is involvement in traffic collisions. One's likelihood of being stopped, all other things being equal, ought to approximate the race of the 12,915 drivers who were involved in traffic crashes in Lincoln last year. Crash data would automatically account for racial differences in driving behavior, age profile, driving frequency, and mileage. What it would not account for, however, is where the police are actually patrolling, and the racial makeup of the people driving in these areas.

Some racial disparity in traffic stops may be the result of the common police practice of deploying officers into areas where crime and disorder problems are most prominent. We work very hard on this in Lincoln, as does Omaha, and we've gotten much better at it with sophisticated crime analysis and GIS. There are far more officers per square mile in the areas with high crime and disorder, and these also tend to be lower income and considerably more diverse.

This screen shot is from HunchLab 2.0, one of the analytical software products we use to forecast hotspots of crime in the next few hours. The colored blocks are individual cells or groups of cells. Each is a 200 meter square area where the risk of crime is elevated, and the colors denote the dominant crime type. The cluster in the core of the city south of downtown and along 27th Street is persistent, and this is the area of Lincoln where you will find the highest concentration of police officers. It is also among the most racially diverse areas in Lincoln.


As a result of the deployment pattern, the likelihood of police stops in these areas is greater in the areas where there is also a greater percentage of minorities than in the general population. You can question the deployment strategy, but if some of the disparity in stops is emerging from this cause, rather than from police bias, that portion is not racial profiling. If you want to impact this portion of disparity, all the sensitivity and cultural awareness training on earth will fail. Rather, you need to convince the police to deploy differently. We could have a healthy discussion about that, but I think people living and working in Lincoln neighborhoods that are most impacted by crime and disorder generally want more police presence and activity, not less.

The type of activity, however, is also relevant. I believe that problem-solving, prevention, and early intervention, ought to be an important part of the mix, and are less likely to be perceived as racially-motivated than some other tactics, such as intensive use of stop-and-frisk. When the community sees the police working collaboaratively with the neighborhood stakeholders, it is less likely to view such things as arrests, traffic stops, and hot spot patrols as bias-based, and more likely to view these are legitimate efforts to reduce crime and disorder.

Links to the series:

Sunday
Monday
Tuesday
Wednesday
Thursday
Friday

6 comments:

Andy W said...

Some scholarly work by Geoff Alpert in Miami has used crash data as the denominator, see:

Alpert, Geoffrey P., Michael R. Smith, and Roger G. Dunham. "Toward a better benchmark: Assessing the utility of not-at-fault traffic crash data in racial profiling research." Justice Research and Policy 6.1 (2004): 43-70. (You can find a public PDF if you search Google Scholar)

Some of the work I have been involved in (traffic stops in Syracuse, NY) has looked at stops right before and after sunset (referred to as the "Veil of Darkness"). The premise is that when it is dark officers have a diminished capacity to see the race of the driver. See:

Worden, Robert E., Sarah J. McLean, and Andrew P. Wheeler. "Testing for racial profiling with the veil-of-darkness method." Police Quarterly 15.1 (2012): 92-111. Tech report available here.

Tom Casady said...

Andy W.,

Fascinating. I suppose one would expect that if racial disparity was being driven by race-based decisions by police officers, that racial disparity would be greater in the daylight hours than at night, but your research in Syracuse does not support that hypothesis.

You may want to check in on this series as the week unfolds, for some more observations from a wannabe practitioner-researcher.

Anonymous said...

Profiling because of gang tatoos, gangster attire and language used. Does that qualify as racial profiling also?
Gun Nut

Anonymous said...

Night-time traffic stops with PC of no plates or papers, or expired tags - I wonder if anyone has mined stats on race of occupants on those stops, since PC is crystal-clear.

Also, night-time stops with PC of obviously illegal dark tint on the front side windows (easier to judge than illegally dark tint on the rear and rear side windows), where race would be very difficult to determine before making the stop.

Eric M. said...

So maybe I am jumping ahead to a future blog post, but what results do you actually get if you use people in traffic accidents to estimate the demographics of the people at risk of being stopped, as you suggest? Is there a disparity then? (I understand, though, your point that this won't account for differences in police presence. However, you certainly could stratify on zip code or some other geographic unit to try to control for this confounding.)

Tom Casady said...

Andy W,

I did my own amatuer version of your "veil of darkness" research using our 2013 citations, by comparing racial disparity of tickets issued 0800-1700 to those issue 2200-0500. I picked these times because regardless of season, these are always daylight hours and darkness hours in Nebraska.

Racial disparity is greater during hours of darkness than daylight for Hispanic and black motorists in Lincoln--opposite of what you would suspect if race were being used as a basis for the decision to stop. Obviously some other variables to consider, though.