Police Stops and Naïve Denominators

In their 2025 article published in Crime Science, Jerry Ratcliffe and Shelley Hyland critically examine how reported racial disparities in police stop data can be misleading. They argue that much of the misrepresentation stems from the widespread use of an inappropriate baseline, or “denominator,” when calculating stop rates.

Typically, analysts compare the distribution of police stops across different racial groups to the distribution in the total city population. Unfortunately, this method does not account for differences regarding who is actually more or less likely to encounter police stops. In doing so, this approach ignores important factors such as the uneven geographic distribution of crime and variations in police deployment. According to the authors, by relying on this “naïve denominator,” studies risk overstating racial disparities. Ratcliffe and Hyland’s work challenges researchers, policymakers, and the public to reconsider how police stop data should be interpreted to paint a more accurate and nuanced picture of racial bias in law enforcement.

Introduction

The common practice of comparing police stops to the citywide population (e.g., “Black people make up X% of the population but Y% of stops”) may seem intuitive, but it fails to reflect where policing actually happens. Crime isn’t spread evenly across cities; it’s highly concentrated in specific areas known as hot spots, where police presence and patrols are intensified. Using citywide census data to evaluate racial bias in stops ignores this reality.

This mismatch between who is exposed to police presence and who actually lives in the city is what the authors call a “naïve denominator.” In other words, failing to account for context, such as where crime occurs, who spends time in those areas, or how police are assigned, can make for misleading calculations on racial bias.

Methods

To demonstrate the problem, Ratcliffe and Hyland analyzed over 133,000 vehicle and pedestrian stops made by Philadelphia police in 2022. The raw numbers show that Black people made up 70.9% of stops, while White people accounted for 14.1%. When compared to census population data, it gives the appearance that Black drivers were 4.4 times more likely to be stopped than White drivers. However, authors contend that using population as a benchmark doesn’t tell the full story.

To show how the choice of denominator impacts perceived disparities, the authors introduce nine different benchmarks for evaluating stop rates. Each benchmark represents a different way of estimating who is “exposed” to police presence, ranging from total city population (Benchmark 1) to more refined options like crime suspects (Benchmark 7) or arrestees for serious violent crimes (Benchmark 8).

These alternatives account for things like where 911 calls occur, the concentration of serious violent crime, how many officers respond to a crime, and how much time they spend on scene. Some benchmarks reflect police demand (e.g., areas with more emergency calls or high concentrations of violent crime). For example, when the distribution of stops is evaluated against the distribution of calls for service, it reflects a more realistic picture of who is likely to encounter police.

Results: The Benchmark Changes Everything

The results of the study showed that perceived racial disparities vary depending on the chosen denominator.

When using citywide population as a benchmark, Black people appear 344% more likely to be stopped in vehicles than White people. This is the highest disparity across all benchmarks and also the least accurate, according to the authors.

When using public uses suspect descriptions from public crime reports (via NIBRS) as a benchmark, the disparity drops dramatically, with Black individuals being just 8.5% more likely to be stopped.

When using violent crime suspect data as a benchmark, the disparity reverses, such that White people are now more likely to be stopped.

Conclusion

In the end, this work is a reminder that the context in which we interpret statistics can dramatically shift our understanding. The authors argue that benchmarking against where crime actually happens and who is involved, rather than just who lives in the city, is essential for honest conversations about racial disparity.