Crime rise; more or the same.

Did crime rise at the same rate in Democrats and Republican cities?

Around 2020, Americans were noticing something: crime was on the rise. Although many objections were put forward to this fact, by September, the noticing had reached presidential proportions, with Trump blaming the rise on Democrats, tweeting:

Donald J. Trump @realDonaldTrump

The Ten Most Dangerous Cities in the U.S. are ALL run by Democrats, and this has gone on for DECADES!

Enter fact-checkers. First, there was a lot of hand-waving,  quoting a warning on the FBI website (i.e., USA TODAY and BBC) that "some entities use the figures to compile rankings of cities and counties. These rough rankings provide no insight into the numerous variables that mold crime". If data regarding the amount of crime committed in each city does not provide insight into its relative ranking with regard to crime rates, it’s not clear what does, the vibes?, experts?

Among the bad-vibes there was some actual data cited, prelimineary data on homicide in Republican and Democrats run cities was avalibale for 2019-2020 and a calculation that spread on twitter was cited on the BBC, NYTimes and others showing:

Together with a later calculation of 51 cities showing a similar change in crime in D and R cities, this was taken as definite proof for the "crime rose the same" position.

Jeff Asher @Crimealytics

Analysis of 51 cities with murder data through at least September shows murder up 35.7% YTD relative to 2019. Big cities tend to overstate national trends in crime, but the national change in murder in 2020 will be historically awful. D cities +36.2% R cities +35.6%

But there are problems. The calculated rise in crime was performed by taking the rate of change in each city and averaging all those rates together. This is problematic since ‘R’ cities are smaller and have a lower homicide rate to begin with, many of them have a very low sample size of homicides for each year. Since the smaller the sample size, the larger the variance, yearly changes in such cities may simply be due to random chance rather than an actual increase in violence.

If I take the height of a million people in a city in 2019 and 2020, changes in the measured average would approximate actual changes in height, but if I take the height of 10 people in 2019 and the height of 19 people in 2020, changes in height would probably be due to random chance, and the same is true for changes in the homicide rate in the Republican city of Lubbock, which experienced 10 homicides in 2019 and 19 in 2020, which would represent a 180% rise in homicides.

This methodology is also open to problems with a ‘floor effect," since homicides cannot fall below 0; a city with 1 homicide per year can experience at most a 100% decline but has no limit on the percentage increase. (This is a more general point, looking at the numbers, I see no city that declined to zero in this analysis, in a wider analysis, this would pose a problem because some do).

Recalculating and comparing

A simple way to account for the problem of higher variance in smaller cities is to simply add together all the homicides that occurred in a given year in ‘R’ and ‘D’ cities respectivly and then to compare the rise from 2019 to 2020. (This data was avalibale in Jeff asher file here).

Here is the data for 66 cities:

Now we see that the total change in the number of homicides was much larger in Democrats cities, since it is ‘current year’ we can also look at the change in homicide between 2020-2021, another period in which rates of homicide increased across the US.

*Data for one ‘independent’ city was not available for 2021, and hence it was removed.

Once again we see a larger change in ‘D’ cities, as a matter of fact the percentage change was close to zero in ‘R’ cities.

So it seems probable that violent crime did not rise at the same rate across the political spectrum of cities; it rose more in ‘Democrat-run cities’.