Defending Empirical Evidence
Matt Yglesias isn't impressed with my post this morning showing that child tax subsidies don't have much impact on fertility:
I buy it, but on another level I don't buy it at all. This is just a chart showing that we've had sweeping waves of social and economic trends over the decades that totally swamp tweaks in the tax code. It's true that you can put together a two-variable chart with appropriately-scaled axes to make it appear silly to say that the tax code is having an influence on fertility rates, but really the chart tells us nothing. We know that some people have children, and that different people have different numbers of children. We know that people exercise some level of conscious choice about this. And we know that having children is costly in both financial and non-financial ways. People also find it rewarding. But the costs are real and extra money to defray those costs should, at the margin, encourage people to have more children.
A few points:
- In fairness, the study itself is a lot more than a "two-variable chart with appropriately-scaled axes." That just happens to be the only part of the study that I included in my post.
- In a sense, though, I agree with Matt: economists are endlessly clever at finding ways to prove that nothing ever has any effect. Design your model right and control for enough variables and pretty much anything can wash out if you really put your mind to it. These things should always be taken with a grain of salt until they get confirmed using a bunch of different approaches.
- On the specific issue of child tax subsidies, of course there's a lot of underlying stuff going on here. And unquestionably, a tax subsidy almost has to have some positive effect on fertility. But the size of the effect is really, really important. Far more important than the mere Econ 101 statement that people react to incentives at the margin. Sure they do. But if the incentive effect is so small that it's swamped by everything else — which is what this study seems to show — then for all practical purposes there's no effect. Alternatively, sometimes there are counteracting incentives that no one has thought about. The only way to find out is to dig into the evidence.
Contra Matt, empirical evidence is not "one of the most overrated things in policy debates." It needs to be treated carefully, and it shouldn't overwhelm common sense. But sometimes common sense is wrong, and sometime incentive effects, no matter how theoretically compelling, are small enough that they don't really matter in the real world. That seems to be the case here.
In a nutshell: size matters. If I have one takeaway that I wish everyone would tattoo on their foreheads, that's it. As usual, then: more evidence, please!
POSTSCRIPT: As always, it's worth being conscious of your own confirmation biases. My intuition, for example, is that tax subsidies are unlikely to have much impact on decisions to have children. The benefits aren't big enough, people don't understand them very well, and other reasons for having (or not having) children are overwhelmingly more important. So naturally when I see a study that confirms this, I'm likely to believe it. Reihan Salam, who supports pro-natal policies in general, and Matt, who has more faith in theoretical constructs than I do, are more likely to be skeptical. Caveat emptor.