Universal preschool has become the ginseng of American politics, a sort of broad-spectrum nostrum that will cure almost anything that ails you. Inequality, male-female pay gaps, crime, poverty — just apply some early childhood programs, and watch those maladies fade. Expect to hear this a lot from Hillary Clinton in the coming presidential race.
And what kind of a crotchety, sour, greedy old columnist could be opposed to such a wonderful idea? I’m so glad you asked.
In truth, I am not opposed to early childhood education programs. I am opposed to blind boosterism of such programs, the kind that confidently predicts marvelous results from thin empirical evidence, and briskly proceeds to demand huge sums be spent accordingly. There are three big problems with this:
• The empirical evidence is shakier than many boosters suggest. The possible benefits of these programs can be divided into two baskets: cognitive benefits (improvement in academic skills and performance), and noncognitive benefits (improvement in such things as social skills, hyperactivity, gratification delay, and so forth). The evidence of cognitive benefits is underwhelming; they appear, then tend to fade out as the children leave the program and proceed through our nation’s school systems. A lot of hope has therefore been poured into noncognitive benefits. Some early programs seem to show long-term improvements in things like graduation rates, employment and criminal activity. However, many of these programs were very small, which raises the possibility that we’re dealing with small samples plus publication bias, rather than something that actually works. In general, in social science, you tend to see that the larger the sample and the better designed the study, the less remarkable the effects. And this is definitely what you see with early childhood programs. Perry, Abecedarian, the Chicago Child-Parent Center — these are inspiring projects. They’re not nearly large enough to base a national program on.
• Small programs are hard to scale to the national level. Let’s say, however, that these programs are actually exciting, effective programs, rather than outliers (my guess is that this is correct). That doesn’t mean we can replicate them for millions of American children. Indeed, Head Start was our attempt to replicate them, and it didn’t. There are lots of reasons that it’s hard to do things at scale. It’s difficult to maintain the passion and intensity of a carefully selected cadre of dedicated researchers. Politics often alters the program to suit the wishes of entrenched existing constituencies. People in the initial program may respond to the fact that they are being studied in ways that improve your results. Coordination problems increase bureaucracy while decreasing flexibility and efficiency.
• Don’t assume away the possibility of negative results. The argument for taking ginseng is that even if it doesn’t actually cure hemorrhoids, it probably won’t make them worse. This is also often the argument for things like early childhood programs. The baseline is zero, so at worst we’re spending a lot of money on something that doesn’t make kids better off (but does make their parents’ lives easier, and boost female labor force participation), while at best we get a great outcome like higher employment and lower crime. Say that at small scale, you have really high-quality early childhood education — but at large scale, the education may not be that great, because of all the aforementioned difficulties in growing the program. It’s possible that the small project is better for kids than the various child-care arrangements that parents make now, and the large project is worse. And indeed, that’s what a new study out of Quebec seems to suggest. In the 1990s, the province instituted an inexpensive universal child-care program. The program doesn’t seem to have produced much in the way of cognitive benefits, and its noncognitive benefits were actually negative — that is, kids exposed to the program (those who lived in Quebec) were more likely to have various problems than control groups in other provinces. Now, this is, as I always caution, Just One Study. It’s a pretty convincing study, of a pretty large group. But it’s still a single study, which means that we should not rush to say that universal child care is a bad idea, or even that cheap, badly designed universal child care is a bad idea. What we should rush to say, however, is that the background assumption about such programs — that at worst they’re a waste of money for zero results — cannot be safely held. We have to assume some possibility that our early childhood program will actually be worse for the kids than the status quo is.
I would like to see us experiment more with these programs. But the key word here is “experiment.” Which is to say we should:
• Try more programs, with well-designed randomized controlled studies to test their effectiveness.
• Take the programs that seem to work and scale them up to a larger group, with well-designed randomized controlled studies to test their effectiveness.
• Rinse and repeat until we have either a) A well-designed early childhood benefit (either universal, or targeted, depending on what the evidence suggests is cost-effective); or b) a robust conclusion that early childhood benefits don’t do much good, at least within feasible limits of the U.S. workforce and public purse.
That would be the sane, sensible way to go about constructing policy in an important area.
But politically, how insane! Voters don’t want to hear about a decade or two of carefully planned research to help shape solid policy choices; they want to hear promises of immediate solutions to an immediate problem. That’s not a great way to make policy. But it’s a pretty good way to get elected.