Tuesday night, Gov. Martin O'Malley sent out this tweet:
"Here's the chart about violent crime in Baltimore that the Baltimore Sun is unwilling to publish: http://www.governor.maryland.gov/documents/crime_arrests.pdf …"
In case you're wondering what the story is, here's the explanation.
Last week, The Sun ran a story reporting that Governor O'Malley had become increasingly outspoken about his belief that the rise in violent crime in Baltimore this year is the result of the significant decline in the number of arrests in the city during the last few years. Mayor Stephanie Rawlings-Blake was not amused, and we wrote an editorial criticizing his position, which ran on Sunday.
On Monday, an aide to Mr. O'Malley sent us a proposed op-ed from the governor defending his position and asking that we run it along with a chart that he has been using to back up his contention about the relationship between arrests and violent crime.
I raised some questions about whether the chart might be misleading to readers who do not examine it closely -- for one thing, the scale used for violent crime is not the same as the scale for arrests, but the way the two lines are overlaid leads to the incorrect impression that they cross at various points in the last 14 years.
On Tuesday morning, I did some more analysis to try to determine whether there is a correlation between the annual number of arrests and the number of violent crimes. I calculated what is known as a Pearson product-moment correlation coefficient, effectively a measure of the degree to which an increase or decrease in one variable is associated with an increase or decrease in another. (That, of course, doesn't mean the change in one causes the change in the other, but that's another story.)
What I found is that over the 14 years of data the governor refers to in his chart, there is a somewhat strong positive correlation between violent crime and arrests (about 0.61 on a scale of -1 to 1). This means that a reduction in arrests is associated with a reduction in violent crime -- precisely the opposite of the point the governor is trying to make with his chart.
However, the same test of the period of 2000-2003 (the years when arrests were going up) shows a very strong negative correlation, that is, as arrests go up, violent crime goes down. (Pearson coefficient of -0.95). But a test of the period when arrests were going down (2004-2013) shows an even stronger positive correlation (Pearson coefficient of 0.98), meaning that as arrests go down, violent crime goes down.
What that suggested to me was that the comparison of arrests and violent crime in Baltimore during the last 14 years doesn't provide much conclusive evidence at all about whether the two things are related or how. The governor may eventually be proven right -- perhaps we have crossed a threshold of arrests below which the gains in violent crime reduction will be reversed -- but the data in this chart do not seem to conclusively support or contradict his position.
I sent the governor's office an email explaining this analysis in detail. I noted that I am not, of course, a statistician, and I urged the governor's aides to consult with someone -- perhaps a statistics professor -- who could evaluate whether I was right or wrong.
The governor's office did not provide a response, so we ran the op-ed without the chart.
As I said in my email to the governor's office, I'm no expert. If anyone out there reading this is an expert, please let me know: Am I wrong here?
--Andrew A. Green