On the WSJ and the D.C. gun ban

Jeffrey Scott Shapiro, in a recent Wall Street Journal opinion piece, writes about the D.C. gun ban:

The D.C. gun ban, enacted in 1976... had an unintended effect: It emboldened criminals because they knew that law-abiding District residents were unarmed and powerless to defend themselves. Violent crime increased after the law was enacted, with homicides rising to 369 in 1988, from 188 in 1976 when the ban started. By 1993, annual homicides had reached 454.

Later in the article, he writes,

Since the gun ban was struck down, murders in the District have steadily gone down, from 186 in 2008 to 88 in 2012, the lowest number since the law was enacted in 1976. The decline resulted from a variety of factors, but losing the gun ban certainly did not produce the rise in murders that many might have expected.

In written form, numbers and relationships like "369 in 1988", "188 in 1976", "186 in 2008", and "88 in 2002" are not always easy to make sense of. Here's what these specific numbers look like graphically. 

Hmm. The picture give a somewhat different impression than the description in words. That is, when I read, "Since the gun ban was struck down, murders... have steadily gone down," a natural interpretation is that the murder rate started to go down at that point, but clearly something was going on quite a bit earlier. Let's look at more data. Here's the number of murders for all years from 1960 to 2012 in D.C.

The description in words turns out not to match the data very well. "Violent crime increased after the law was enacted, with homicides rising..." The murder rate did rise, but the big change started about ten years after the gun ban was put in place. The description is so vague that in 1985 we could have said the opposite with equal accuracy: "Violent crime decreased after the law was enacted, with homicides falling..."

This raises a problem for Shapiro's explanation: "The D.C. gun ban... emboldened criminals because they knew that law-abiding District residents were unarmed and powerless to defend themselves." This may be true, in the abstract, but the number of murders over time doesn't really show this. That is, we'd have to imagine that emboldened criminals took almost a decade to get serious about committing more murders, but then they became disemboldened about five years later. (And then we'd have to ask, "Should we expect some kind of change ten years after the gun ban was repealed?")

I don't mean to dive into the gun control debate with this post. (My own take is that the start and end of the D.C. gun ban look to have had no effect on the murder rate, though more data and further assumptions could change my view.) Rather, it's to make a more general point: If we bring in data to support a point of view, we need to be careful about doing it, both in describing the data and in explaining the patterns we see.

Update: Gene Spafford asks about violent crime--Shapiro focuses on murders as just one category of violence. Here's what that time series looks like, over the same  years but with a different scale on the vertical axis.

And below I've rescaled the murder data and shown it as a fainter blue line in the background. This makes the murder numbers wrong, but it lets us more easily compare the changes in murder and violent crime over time.

Shapiro's explanation looks more plausible with this data; it only took a couple of years after the gun ban was enacted for the violent crime rate to rise. There is still a good deal going on that's hard to explain. The spike around 1980 is more prominent for violent crime in general than just for murders, but the later spike in the mid-1990s is less prominent. Social datasets are complex--I won't even try to interpret these patterns.