Bryan Caplan  

Sacerdote's Dog That Didn't Bark

John Stossel Tonight... Krugman on Pop Internationalis...
I keep thinking about Bruce Sacerdote's Korean adoption study.  I've read every twin and adoption paper I could find about parental influence on kids' educational outcomes.  Sacerdote's is the best of the bunch - the cleanest study, the clearest presentation, and the most comprehensive analysis.

You may recall that Sacerdote found small effects of maternal education and family size on how far kids get in school.  As I summarized before:
[E]very 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%.
OK, but why does maternal education and family size matter, however slight the effect?  Many economists will be tempted to say that families with more education and fewer kids have more financial resources to invest in each child.

It is striking, then, that Sacerdote controls for both family income and income in the family's zip code.  Neither of these income variables has a statistically significant effect on kids' educational success.  In fact, the sign is wrong three times out of four.  Despite their best efforts, in short, it looks like rich families fail to give their kids' an edge in school.  It doesn't even look like richer parents' residence in better school districts pays off. 

On reflection, income - both family and neighborhood - is the dog that didn't bark.  It prompts us to ask, "Why would maternal education and small families matter after controlling for their purely financial effects?"  Maybe it really is parental attention; maybe it's some kid of sub-cultural osmosis.  The big news, in my view, continues to be that these effects are small.  But why do you think they matter at all?

Comments and Sharing

COMMENTS (7 to date)
Greg writes:

Where has Occam's Razor gone? Parental attention, expectations, or whatever you want to call it seems like the obvious, logical hypothesis.

Look at sports. I don't know if any studies have been done on childhood sports attainment, but I would bet that parental involvement in sports when they were children is a huge predictor of their children's involvement and maybe achievement. It's the same with anything.

I'm not impressed by the results of these educational and parenting studies, for the most part, because we haven't figured out how to measure the most important variables.

David R. Henderson writes:

Do you mean 2.3 percent or 2.3 percentage points. If the latter, what is the baseline?

RL writes:


The obvious first question to someone completely ignorant of the literature is "why maternal education, not paternal"? The natural inclination is to assume it has to do with a larger maternal role in child-rearing. If so, does the data allow distinction between the impact of stay-at-home moms and mothers who work outside the home?

Dr. T writes:

I haven't read the study, but I have serious doubts about its conclusions. These types of "fishing expedition" studies, because they look at so many possible factors and so many outcome variables, always come up with a few correlations that appear to be statistically significant. (This is how epidemiologists in the early 1990s mistakenly concluded that living near high voltage power lines caused childhood leukemia.)

In this twin study, the reported effects are small, which to me indicates that they might be statistical anomalies. Even if the correlations are true, years of maternal education may be irrelevant because they may be a proxy for other factors that altered the educational attainments of the adoptees. Here's an example alternate hypothesis: Better-educated mothers got higher-paying jobs. Their children received high-quality day care (by a private sitter or a well-run center). The early prolonged exposure to other adults helped the children learn and increased their likelihood of getting more schooling. Thus, in this alternate hypothesis, the mothers' educations were unimportant: it was the extra adult exposure at an early age that mattered. My alternate hypothesis can also explain the 'number of children' effect: after multiple children, high-quality day care was no longer an option, and educational attainments of affected children were reduced.

My completely invented alternate hypothesis is just as likely as the direct correlates stated in the paper, which is another reason why I pay little attention to these kinds of studies.

Gary writes:

Bryan, your theorizing on this matter would be vastly improved by reading Robert Trivers's Parent-Offspring Conflict paper.

I can't find an ungated version on the web, but would be happy to send a copy to anyone who emails me at garymarkhov AT

nicole writes:

On reflection, income - both family and neighborhood - is the dog that didn't bark. It prompts us to ask, "Why would maternal education and small families matter after controlling for their purely financial effects?"

I can't read the article, and don't know whether there is a standard way of doing this, but...does it control just for household income, or for per capita income? You say "family income," and if that's the case, I don't see why family income should be very relevant at all. More children means less money per kid, even if you still have just as much, or more, total.

Walt French writes:

Taking your and the authors' summaries at face value, I question why anybody is surprised at income's irrelevance in predicting adoptees' educational attainment.

Adopting families are screened for various success criteria to ensure that adoptees are not subjected to failing families, so far left tails of income or dysfunctionality are excluded. Extreme right tails would be minimal, or even zero, cases in the sample. Likely, few truly "rural" cases due to the mechanics of international adoption, means that the sample is relatively homogeneous in other ways.

In all, for whatever importance the study presents to sociologists, it does not address some of America's most pressing problems of built-in discrimination or other "market failures" that prevent our country from getting the best outcomes based on the raw material of our country's youth.

However, it DOES suggest that income is a poor proxy for a family's propensity, rather than ability to nurture its children. One might expect that "successful" parents tend to have higher income, but it seems that parents' (OK, maternal, although we expect very high correlations between parents) education is a better proxy for parents' emphasis on children's development. Income differences, might have zero or negative marginal impacts if they represent direction AWAY from the family and towards work. (I speculate, on the assumption that the authors are too professional to do so.)

This might be a wake-up call for the most dogmatic of conservatives, showing that for the middle class, income is not predictive of how well a family nurtures its children. Likewise, an emphasis on availability of private education, voucher assistance, and other cherished dogma might not have much traction with any but the wealthiest.

But those of us on the other end of the political spectrum won't be shocked, nor dissuaded from our beliefs that we subject our children to a dramatically unproductive lottery that wastes talent and stifles our children so that a few can promote a self-serving call for "the American Dream."

Comments for this entry have been closed
Return to top