Adjective-Coefficient Disconnect in Sacerdote
By Bryan Caplan
I’m now writing chapter 3 of Selfish Reasons to Have More Kids, entitled “Que Será,
Será: The Case for Guilt-Free Parenting.” It’s basically a parents’ eye guide to behavioral genetics. As a result, I’m now reading and/or re-reading a bunch of twin and adoption studies. For the most part, Steven Pinker’s summary of this literature is apt:
A handy summary of the three laws [of behavioral genetics] is this: Genes 50 percent, Shared Environment 0 percent, Unique Environment 50 percent (or if you want to be charitable, Genes 40-50 percent, Shared Environment 0-10 percent, Unique Environment 50 percent).
However, some prominent economists working in this area seem to take a very un-Pinkerian position. The most notable, perhaps, is Bruce Sacerdote, who has done impressive work on Korean adoptees. His 2007 QJE piece finds “large effects on adoptees’ education, income, and health from assignment to parents with more education and from assignment to smaller families.” Indeed, he adds that “One reasonable hypothesis to explain the results in Table VI [which displays his main findings] is that the quantity and quality of parental attention matters a tremendous amount for the adoptees’ outcomes.”
How can Sacerdote’s results be so different from those in mainstream behavioral genetics? The answer, as far as I can tell, is that the difference is not Sacerdote’s coefficients, but the adjectives he uses to describe them.
What, for example, does Sacerdote count as a “large effect” of parental education and family size on children’s educational attainment? Let’s see. According to his Table VI, every extra year of maternal education raises kids’ years of education by .097 years – about a month. Every additional child in the family reduces kids’ years of education by .129 years – about a month and a half.
What about the probability a kid finishes college? Every year of maternal education raises it by 2.3%; each kid in the family reduces it by 2.6%.
What about income? This is an important question to ask – for most economists, the whole point of increasing educational attainment is to increase income! But Sacerdote finds no statistically significant effect of years of maternal education on income (the point estimate is 0.3%). The only aspect of family that affects income seems to be the number of kids – each of which reduces it by 4.4%. If you’re finicky, you’ll also note that he shows us results for nine independent variables, and only one is statistically significant at the 5% level.
The biggest effect by far in Sacerdote’s table: If your mother drinks, you’re 18.8 percentage-points more likely to drink. A big deal? Not really, since the question asks whether you drink alcohol at all – not whether you drink to excess.
Don’t get me wrong. Sacerdote’s work is important and careful. My objection is merely to his adjective inflation. Based on my understanding of ordinary English, his effects are not “large,” and they certainly aren’t consistent with a “tremendous” effect of the quantity and quality of parental attention.
Of course, I won’t claim to be the final arbiter of ordinary English. What do you think?