The Worst Statistics is No Statistics
June 26th, 2008Mark Chu-Carroll likes to say that the worst math is no math. John Cole brings us an example from CNN. I posted a comment on their article, but it’ll probably get bigger exposure here. (Assuming CNN’s moderators approve it at all.) This was the offending quote:
There were 143 gun-related murders in Washington last year, compared with 135 in 1976, when the handgun ban was enacted.
Needless to say, this is completely useless. It’s two data points with absolutely no context.
My stats here are kind of an ass-pull, but they’re meant more as a demonstration of how fucking easy it would be to improve on the utterly useless, context-less numbers that CNN gave than a real analysis of the effect of the gun ban. I used two web-accessible sources, U.S. Census QuickFacts with historical census data and District of Columbia Crime Rates 1960-2006.
I strongly recommend following the last link in particular to show how completely asinine it is to use only two data points to judge the effect of the handgun ban. Anyway, here’s my comment:
As stated, this is a useless statistic, because it contains no information about either the population of Washington in the two years examined or about the total number of murders (regardless of weapon) in each year. Either one of those things would tell us something about the effect of the ban. The number of gun murders by itself, with no context of the city’s population or overall murder rates, tells us nothing.
The per capita rate of gun death for last year would be 143/572,059 (2000 Census) or about 2.5. The per capita rate of gun death for 1976 would be 135/702,000 (1970 Census) or 1.92. This doesn’t change the implication of the quote, but DOES provide actually useful information to the reader.
Also, there were 188 total murders and 135 gun murders in 1976, for a ratio of 71%. In 2006, there were 169 total murders. (Numbers for 2007 are not available.) Assuming no dramatic change in gun murders between 2006 and 2007, that gives a ratio of 86%. Again, this doesn’t change the implication, but it does provide more useful information about the overall murder/crime rate between the two years. (Murders actually went DOWN overall.)
But even this doesn’t come close to telling a full story, for a number of reasons. For example, there were 235 total murders in 1975, and in fact every year from 1969 to 1975 had more than 200 murders per year. The years 2004-2006, on the other hand, all had less than 200 murders, down from a peak between 1990-1993 which had over 450 murders per year. This clearly indicates that there is something OTHER THAN gun control legislation affecting the total murder rate. In fact, this factor, whatever it is, completely swamps any difference seen in murders between 1976 and 2006, which the article tries to pretend is a valid measure of the effect of gun control legislation in D.C.
Bottom line: quoting random numbers from random years doesn’t tell us anything. So don’t do it unless you’re willing to invest in a complete statistical analysis. (For the record, I haven’t done anything remotely like that here, but I did considerably more than the article did, and I didn’t spend more than 15 minutes doing it.)


