There has been a lot of talk lately about immigration. In particular, about if immigrants cause crime in the cities they go to. Like a lot of controversial topics, this one is based mainly in anecdotes and exaggerations. Here is some data regarding the issue.
In this post I look at the relationship between crime rates and Mexican immigrant population in 102 counties between 1980 and 2010. For a full methodology and the R code used to analyze the data, see below. Please note that this is just a simple analysis and doesn’t indicate a causal relationship.
I used data from two sources for this analysis. Crime data came from the FBI’s Uniform Crime Report (UCR) and Mexican immigration data came from the Census. There were 102 counties that were available in both data sets for both years 1980 and 2010. Figure one shows a scatterplots of the relationship. The blue line shows a line of best fit. The blue line is slightly downward sloping indicating that there is a negative relationship. As Mexican immigrant proportion increases, crime decreases. However this relationship seems fairly mild. The points are also not very close to the blue line, showing that even if the relationship does exist, it’s not a very strong one.
So it looks like there may be a weak negative relationship here. But looking at all counties together may be misleading. With this data, we can look deeper and see if the relationship changes among subcategories of cities (It would also be wise to see how specific crimes change but I’m not doing that here). Lets break the counties up for into the top and bottom half of crime rates in 1980, and then top and bottom thirds.
Here we see something quite different. Mexican immigrants do increase crime. But only in counties they had low crime in 1980. In counties with high crime in 1980, they decrease it. Figure 3 shows crime broken into top and bottom third (the middle third of data is excluded) and shows an even higher increase in crime in low crime counties.
How should we interpret this? First, it is a good example of problems with aggregated data. The story completely changed (and become more interesting) after breaking up the counties into high and low crime. Second, it appears that Mexican immigrants do not cause crime wherever they go. Nor is the presence always leading to a decrease in crime. They seem to be pretty average.
Crime literature (and I do so myself) often talks about crime as if it’s directly tied to the population. If the number of people go up by X, crime should go up by some multiple of X. But that’s not really true. Crime will go up if a bus full of felons comes into town. It won’t go up if that bus is full of nuns. It really depends on who the new population is. If it’s a bunch of average people, then high crime areas with have less crime, low crime areas will have more crime. These graphs (again, not causal) indicate that Mexican immigrants are just average. Crime reverts to average crime rates wherever they go.