An Analysis of Police Stops in Rhode Island

by Gursev Pirge

This study is an analysis of police stops in Rhode Island and covers more than 509,000 police stops conducted between the years 2005 and 2015. In this article, I analyzed a data set using Pandas in order to figure out whether race, age, gender, time of the day and other factors were effective in the decision made by the police to make the stop. After some iterations, I decided to analyze the effect of gender and leave the rest of the parameters for future studies.

Considering the effect of gender, there are quite a number of studies, with different results — one study (Racial and gender profiling can affect outcome of traffic stops) claims that gender is important in getting a traffic stop, whereas another study (Gender Bias in Power Relationships: Evidence from Police Traffic Stops) has consistently found gender-based disparities in traffic stops.

Specifically, I tried to examine whether police officers are less likely to issue traffic tickets to men or to women during traffic stops. In the conventional wisdom, it is widely accepted that women are less likely to receive tickets. Then I tried to analyze the results about police search and arrest rates.

The data set is provided by data.world, home to the world’s largest collaborative data community, which is free and open to the public. It’s where people discover data, share analysis, and team up on everything from social bot detection to award-winning data journalism.

Data Cleaning

The first step is to import and clean the data using pandas before exploring the relationships between possible parameters and policing. There will be visual exploratory analysis of the police stops using Pandas and Seaborn.

As always, there are missing values (NaNs), so the number of missing values should be determined before starting a detailed analysis.

There are 509681 rows of data, but some attributes (columns) have a lot of missing values and this will cause problems during the analysis. So, deleting the columns with no useful information for the analysis will be more reasonable.

So, four columns are deleted and will carry on with the remaining 22 rows of data.

When you know that a specific column will be critical to the analysis, and only a small fraction of rows is missing a value in that column, it often makes sense to remove those rows from the data set. The driver gender column will most probably be critical to many of the analyses. Missing rows are only small proportion (around 5 %) of the total, so it will be a good idea to drop those rows from the data set instead of filling them. Now 480584 solid rows of data are left instead of the starting value of 509681.

Now, consider the types of violations and try to find a pattern on this parameter.

Speeding is by far the most prevalent violation and its percentage among all the violations is 55.9 %. The plot below shows the distribution of violations.

Does Gender Have an Effect on the Police Behavior?

Now that the data is almost ready for gender-based analysis, two separate data frames for female and male drivers are prepared and the results for the number of violations grouped by the driver gender are given as:

Now, let us compare the outcomes of police stops due to speeding. When a driver is pulled over for speeding, many people believe that gender has an impact on whether the driver will receive a ticket or a warning. Once again, two separate datasets are prepared for speeding: for male and female drivers.

Interestingly enough, once the driver is stopped for speeding, there is a slightly higher possibility for female drivers to get a ticket when compared to the male drivers (95.32 % for female drivers vs. 94.46 % for male drivers). On the other hand, there may be other effective factors, but percentage of arrest for male drivers is almost three times when compared to female drivers (0.53 % for female drivers vs. 1.58 % for male drivers).

During a traffic stop, the police officer sometimes conducts a search of the vehicle. Calculating the percentage of the stops that result in a full vehicle search will give us some valuable data. Figure below shows that around 3.7 % of the police stops end with a car search.

When we check the effect of gender on the percentage of searches conducted, we get an important result; male drivers’ cars are subject to search by a huge difference (1.88 % for female drivers vs. 4.38 % for male drivers).

The following results combine all types of violations, driver gender and the resulting police search and it is easy from the details to see that, regardless of the type of violation, male drivers are always unlucky when it comes to a possible police search:

Conclusion

In this article, I have assessed the effect of gender on the police stops in Rhode Island. The analysis provides evidence that:

· Once a driver is stopped for speeding, there is a slightly higher possibility for female drivers to get a ticket when compared to the male drivers (95.32 % for female drivers vs. 94.46 % for male drivers).

· Percentage of arrest for male drivers is almost three times when compared to female drivers (0.53 % for female drivers vs. 1.58 % for male drivers).

· Male drivers’ cars are subject to search by a huge difference (1.88 % for female drivers vs. 4.38 % for male drivers).

Thank you for reading. I am planning to publish articles from the same data set in the near future.