Tuesday, November 28, 2006

Who Really Cares about Arthur Brooks

Those of you who read this blog regularly (both of you) may remember Arthur Brooks, the guy who claimed that the entire country would soon be hard-core conservative (within fourteen years!) because conservatives were out-breeding liberals. Of course not only is this sheer nonsense because he ignored little things like immigration, it turned out that Brooks' numbers weren't even right.

So when I see that he's got a new book out claiming that conservatives are more charitable than liberals, I'm naturally quite skeptical. My previous experience with Brooks has led me to believe that he is highly disingenuous and not beyond jimmying the numbers to fit his thesis.

So I checked the General Social Survey, one of his sources, to see if the raw data do indeed fit his thesis. What a surprise, they don't. In nearly every case, the GSS data show that liberals contribute more and volunteer more than do conservatives. There are exceptions of course (for example, conservatives donate far more to religious organizations, which do some charitable work but are otherwise just social clubs), and there are many, many cases in which the data is ambiguous. But the general trend is that liberals are more generous than conservatives.

Of course Brooks claims to be correcting for certain things like age and income, and god only knows what else, so I don't doubt that he can pull his desired conclusion out of the data if he tries hard enough. But I'm more interested in what the data actually tell us about charitable giving, not in trying to score points for one ideology over another.

The most salient thing is that there is a huge variance in charitable giving among individuals within any one group. A small number of people give a whole lot, and a whole lot of people give very little. Effectively the same thing is true of volunteering, but even worse. The vast majority of people don't volunteer at all. This makes comparisons among arbitrarily selected groups rather meaningless, because in each group it's only a select few who are doing the giving. To me the central question that Brooks puts forth makes about as much sense as asking who is better at basketball, Americans or Chinese. If you limit the comparison only to the professionals, then Americans are probably better, but it would be meaningless to compare the entire populations of each country because most people either don't play basketball or don't take it very seriously if they do. Additionally, this makes appending percentages (as in, group A volunteers 50% more than group B) very misleading. A tiny number plus 50% is still a tiny number. Such differences would be far more meaningful if volunteering were widespread.

Here's something else that's worth mentioning. All of the data that Brooks number-crunched (a term that in this case may mean more than usual) came from surveys. Surveys are very good at telling us some things, but notoriously unreliable at others. One thing that they are definitely not good at is giving us an accurate gauge of how often people commit behaviors perceived as socially desirable. For example, twice as many people claim to attend church when asked in a survey than actually do attend church. Whether they're fooling themselves or trying to fool the interviewer, people will consistently overestimate how often they do things that they are "supposed" to do. (And they will likewise underestimate how often they do things they're not supposed to do, like drink alcohol.) It's impossible to believe that charitable giving doesn't fit the same pattern. So when Brooks claims that a religious person is 57% more likely than a secularist to help a homeless person, does that mean that he's really more likely to do so, or just more likely to say that he did? Without some method of resolving this issue, Brooks' claims are pretty much meaningless, even if we accept his numbers at face value (which I do not).

Perhaps later I'll look more into this issue and peruse some of the other surveys that Brooks claims to have used (or abused).