McFreedom

I just finished Joel Best's Damned Lies and Statistics. It's a pretty enjoyable overview of how bad statistics get created and passed along, and it's a short, fast read. His coverage of the main types of statistics errors you see in politics is a useful introduction to thinking critically about statistics, but the real entertainment in the book are his anecdotes of his research into mutant, or just plain bad from the beginning statistics.

For example, he relates how gay activists claim that 1350 gay teens commit suicide each year - representing 30% of all teen suicides. Since the rate of homosexuality in the population is much lower than this, gay activists have tried to use this as a rallying point to try to illustrate how difficult life is for homosexual teens, that it drives them to commit suicide in such large numbers.

Best's first point is - how in the world would you figure this out to the begin with? First of all, getting suicide numbers is hard, to begin with, because many suicides are classified as homocides or unknown cause deaths. Secondly, how would you learn the sexual orientation of the deceased? The coroner doesn't put it on his report; the police don't put it in their report. In fact, the activists themselves claim that this includes a large number of teens still in the closet - which any study, by definition, would miss.

In fact, that number was created by activists. Using the widely-discredicted Kinsey study figure of a 10% rate of homosexuality in the population (a number still widely claimed by gay activists despite the fact that every study since has shown a number in the range of 3% - 6%) as a starting point, they then took the number of teen suicides in the country annually (4500), and then multiplied that by 3 - since another "widely known" stat is that homosexuals commit suicide at 2-3 times the rate in the normal population. Of course, that is in itself another stat which is very suspect, since it's very difficult to study for the same reasons the teen rate is difficult to study. Regardless, this chain of bad statistics leads up to 4500 * .1 * 3 = 1350. Having then derived these numbers from unrelated, uncausal sources, activists then try to claim that something should be "done" about the problem of homosexual teen suicide - but one has to wonder how we would know if such policies were effective, since the percentages are derived from populations that are not teens to begin with?

More simply, he gives the case of anorexia nervosa. This became a popular topic in the press some years ago, with many alarmist stories being written. Someone at one point estimated that 150,000 women anually exhibit symptoms of anorexia nervosa. This statistic was pretty suspect to begin with, as it was an estimate, and undoubtedly is the largest number the originator - an advocate of greater intervention for anorexia nervosa - thought he could get away with. However, someone at some point claimed that 150,000 American women die of anorexia nervosa every year. This statistic was much more impressive, and once it had the imprint of approval (by being stated on television and used in the papers and books), it took on a life of its own. Never mind, of course, that this would make anorexia nervosa the second-most common killer of women in the US. The real figure looks to be something more like 70.

I've found that, once you start thinking critically about things like this, you start having a "nose" for when a stat (or story - these spread and mutate in ways very similar to urban legends) is wrong. This book is a good place to start, and he is absolutely right when he reminds us to apply that critical thinking to all the numbers we see, not just the ones whose politics we disagree with.