Friday, August 25, 2006

More on Brooks...

In the continuing the saga of Arthur Brooks' bizarre arithmetic (critiqued here, results of correspondence here and here), there remains the issue of how Brooks came up with the numbers that he relied upon for his simulations. These numbers were supposedly drawn from the General Social Survey (GSS), a publication that I did not have access to and hence wasn't able check. Recall, he's starting with the assumption that self-indentified conservatives have more children than liberals, and then concludes that this will spell doom for the Democratic Party:

According to the 2004 General Social Survey, if you picked 100 unrelated politically liberal adults at random, you would find that they had, between them, 147 children. If you picked 100 conservatives, you would find 208 kids. That's a "fertility gap" of 41%.

These numbers didn't look right to me. Even the silly Phillip Longman came up with a gap of only 11%. But since I couldn't check the numbers, I just assumed for the sake of argument that they were correct and then went on to describe why Brooks' conclusion make no sense.

But Half Sigma does have access to the GSS, and he took a look to see who was really having more children. Guess what? Democrats are having more kids than Republicans. It's not a big difference (and honestly, I can't see why it would be), but it's still the opposite of what Brooks is claiming.

Let me add a couple of caveats: I didn't check Half Sigma's numbers in part because I'm not sure if the chart he lists is his own creation or comes straight from the GSS. So if they're off, go yell at him and not me. Secondly, Brooks lists results for liberal vs. conservative (with independents curiously left out) and not Democrat vs. Republican. Not all people who self-identify as conservative vote Republican, and not all people who think they're liberal vote Democrat. (Of course, surveys have shown that a large fraction of people who self-identify one way or the other don't actually fit the label they've given themselves anyway.) So maybe that's where the discrepancy lies. But Brooks doesn't make this distinction himself. For the purposes of his article, liberal=Democrat.

But what really doesn't bode well for Brooks is the fact that ideological self-identification has remained virtually unchanged for 30 years. How does he explain that? Either 1) the idea that conservatives outbreed liberals is wrong, 2) the idea that children reliably adopt the ideology of their parents is wrong, or 3) the whole thing with conservatives outbreeding liberals is such a recent phenomenon that it hasn't had time to show up in the numbers. The only one of these possibilities that doesn't flatly refute Brooks is #3, and it doesn't make much sense.

Update: I just now noticed that Half Sigma has another post about the fertility of Bush voters vs. Gore voters. This time, it is Bush voters who have more children. The seeming descrepancy between the previous results can be explained by people crossing party lines, given the fact that registered Democrats outnumber registered Republicans by a large fraction, yet Gore and Bush recieved nearly the same number of votes.

The important thing is, the difference is miniscule. Gore voters have an average of 1.92 kids and Bush voters have an average of 2.03, which is about 6% larger. That's means there's a 49/51 split. Recall, Brooks is claiming a 41% difference resulting in a 41/59 split. Now that's a discrepancy.