Minneapolis Traffic Stops

To be frank, Minneapolis-St. Paul has had its fair share of incidents with the police, specifically at traffic stops (getting pulled over) (see Hamza Jeylani incident or Jamar Clark incident). Between 2006 and 2012, Minneapolis made $14 million in payouts of police misconduct, and other incidents caused there to be lots of attempts at reforms.

Many of the horrible incidents involved police misconduct involved traffic stops, or pulling cars over for various things. These caused riots, and lined up and surely influenced a lot of the Black Lives Matter movement, which was rather prominent in Minneapolis in the past couple of years. After seeing that a lot of this came from traffic stops, Minneapolis tried enacting reforms, which changed the way officers pulled over and handled traffic stops.

The city of Minneapolis published data on their traffic stops, and I decided to take a data science approach to this issue, to see how these changes and incidents might have affected the traffic stops. Of course the data might be biased due to the city wanting to have a better image (whatever that may be), but we’ll take it at face value since it’s funded by the government.

I found the dataset here, and the first thing I did was cleaned up the dataset, and got it into a format that I liked. Then, I aggregated the overall stops by month, and plotted them out over time.

I realized that this wasn’t super helpful, so I decided to take a rolling average over the course of 12 months, and add that to see how the trends are changing.

The rates since 2001 seem rather sporadic, and possibly decreasing? It’s a little hard to tell what was going on at these peaks and troughs. I decided to run some other tests to see whether these trends relied on seasonal aspects.

From the third plot in there, its pretty easy to see that there are some clear seasonal trends showing up. I decided to plot out the stops by month.

It seemed like the peaks are in March, with drops in September, when students presumably all go back to school.

This is all fine and dandy, but to go back to my so I decided to do some research and find if there were events that might have caused them.

I tried using multiple regression to plot out a prediction for the future, but it got a little messy since there’s so much ‘randomness’. Below I also plotted the residuals.

This doesn’t look the best, the residuals are high even with multiple regression. Since this was pretty off, I instead tried to build a statistical model, using the values and months.

My original model only gave me an R-squared value of .33, so I did some refining. I added some more variables, like adding the month and years, and it spat this back out:

.41 is a little better. We basically built up a statistical model that will take the aspects given in the first dataset, and use them to build a model for what will happen in the future, to try to predict things. but I decided to plot them out to see how it looks.

To be blunt, neither model fits the data extremely well, but I think it’s just a result of the data being so sporadic. The changes and rates of stops seems just so random. I tried figure it out, but I was having a hard time figuring out variables (aside from seasonal changes), so I decided to pull in some outside information: I tried to figure out if events caused the shifts in stops.

I tried the best I could to map out the major events within the police department and city wide, but it was hard for me to see any trends. Perhaps after 9/11 and the officer shooting raised traffic stops, and then each shooting lowered them, but it’s a bit of a stretch. One thing I noticed is that the counts go down quite a bit between 2007 and 2010, which is right around the time of the housing crisis. This hit Minneapolis especially hard, and could have caused people to stop driving so much, due to costs, which would have lowered counts.

Overall, traffic stops were hard to map, hard to understand their frequencies, and hard to predict. Perhaps the rates are random, but regardless it will be difficult to see what will happen over the next couple of years. The only steady factor seems to be the seasonal trends.