<< Attorney General Sessions Speaks at Georgetown | Main | News Scan >>

Quickie Correlations on Crime and Punishment Stats

| 1 Comment
How do changes in sentencing among the various states correlate with changes in their crime rates?  I did some quickie calculations on yesterday's crime stats to find out.

Yesterday, as noted earlier, the FBI released Crime in the United States -- 2016.  This is the official compilation of crimes known to the police nationwide.  Crimes committed but not reported are another issue.  Table 2 reports crimes both as numbers and rate for the states and regions for 2015, 2016, and the percent change.  I extracted the percent change numbers for the states, though the FBI's spreadsheet format didn't make it easy.

The Bureau of Justice Statistics annually reports Prisoners in the United States.  The latest such report, for 2015, gives us the change in the number of prisoners from 2014-2015.  So, with a one-year lag, how do state-by-state prisoner changes correlate with crime changes?

My quickie spreadsheet calculation shows a coefficient of correlation between violent crime rate and prisoner change of -0.27.  For property crime it is -0.31.  If we use numbers of crimes instead of rates, the correlations are a tad stronger.

A negative correlation means that the two variables tend to move in opposite directions.  As number of prisoners goes down, crimes tend to go up.  That is just what persons of sense would expect.

This is not proof, of course.  Correlation is not causation, as we have noted here many times.  I haven't yet done the further analysis required to state the magic "p factor" required by respectable publications to even report a correlation.  Compare this post.  I will do that when I get back to the office and have better tools.  There is also the usual caveat about relying too much on one year's data, and a few others could be thrown in.  Even so, these correlations are strong enough that I thought readers would find them interesting.

1 Comment

Would love to see you run these numbers over, say, a 10 year period. Could be VERY interesting and valuable, especially to see which states have this correlation stronger and weaker. Thanks.

Leave a comment

Monthly Archives