Tuesday, April 8, 2008

More Fun with Bad Statistics

As a way of arguing against a free trade deal with Colombia, Matthew Yglesias and Ezra Klein link with approval to this graph posted by the left-of-center Economic Policy Institute. According to Klein, Yglesias, and EPI, the graph shows that union members in Colombia can be murdered with impunity under right-of-center Colombian President Uribe. Indeed, the trio go even further than that, calling all of these murders "assassinations," thereby implying that every time a member of a trade union is murdered in Colombia (one of the world's most dangerous countries in general), the motive for the murder is the prevention of unionization.

Except that there are a whole host of problems with this graph and with the purposes for which it is being used.

First, it terms every murder of a trade union member an "assassination." But as I alluded to above, Colombia has one of the world's highest murder rates, period. The idea that trade unionists might be murdered for the same reasons as anyone else (primarily related to the wonderful War on (some) Drugs) doesn't seem to have crossed Klein, et al's collective minds. Indeed, one commenter on Yglesias' post points out that the murder rate of trade unionists appears to actually be lower than the murder rate for the rest of the Colombian population.

Second, it's obvious that the number of murders of trade unionists peaked in the same year as trial prosecution of such murders peaked during Uribe's reign. As the number of unionists murdered has gone down, so too has the percentage of such murders brought to trial. Obviously, this strange coincidence is not a result of fewer prosecutions encouraging fewer murders. But it does suggest something else. Any cop will tell you that the easiest murders to solve are the ones committed by someone the victim knew, while random killings are far and away the most difficult murders to solve. It strikes me that in the year murder of unionists peaked, it is possible (even likely) that this was a result of a greater number of politically-motivated killings. Since such killings would provide an easy motive and thus an obvious suspect, they would be killings with a higher likelihood of being brought to trial. Meanwhile, random killings would be much more difficult to solve. In addition, in a country with such a high homicide rate, it is pretty easy to imagine a police force that tries to make the more easily solved crimes a priority for prosecution over random acts of violence.

Third, it's worth pointing out (as I did in a comment to Yglesias' post) that the graph shows only the percentage of union murders "brought to trial." It does not show the percentage of union murders actually prosecuted or resulting in convictions of some sort. Nor does it show the average sentence meted out in such cases compared to other murders in Colombia. When a motive is easily found, a case is more likely to be open and shut, and thus it is more likely to result in a guilty plea of some sort (and thus not be brought to trial).

Finally, there is this commentary piece from a week ago in the NYT that debunks these statistics even further, which is well worth a quick read of its own.
***UPDATE***I should have noted that this is not to say that there are no unprosecuted politically motivated murders of trade unionists. What it is to say, however, is that the situation for trade unionists appears to be improving the last few years. It's worth also pointing out that the number of trade union murders is now just a fraction of what it was even ten years ago.
I should also add that even if everything Klein, et al, presume is true, this actually argues in favor of a free trade deal. It's fairly well-established at this point that free trade is the best way of spreading liberal values like free speech, due process, etc. While a free trade deal might be viewed as a "reward" for Colombia, it is a reward that will ultimately result in the situation for trade unionists improving significantly over time.