Preschool: I Was Wrong

Kelsey Piper has written an article for Vox: Early Childhood Education Yields Big Benefits – Just Not The Ones You Think.

I had previously followed various studies that showed that preschool does not increase academic skill, academic achievement, or IQ, and concluded that it was useless. In fact, this had become a rallying point of movement for evidence-based social interventions; the continuing popular support for preschool proved that people were morons who didn’t care about science. I don’t think I ever said this aloud, but I believed it in my heart.

I talked to Kelsey about some of the research for her article, and independently came to the same conclusion: despite the earlier studies of achievement being accurate, preschools (including the much-maligned Head Start) do seem to help children in subtler ways that only show up years later. Children who have been to preschool seem to stay in school longer, get better jobs, commit less crime, and require less welfare. The thing most of the early studies were looking for – academic ability – is one of the only things it doesn’t affect.

This suggests that preschool is beneficial not because of the curriculum or because of “teaching young brains how to learn” or anything like that, but for purely social reasons. Kelsey reviews some evidence that it might improve child health, but this doesn’t seem to be the biggest part of the effect. Instead, she thinks that it frees low-income parents from childcare duties, lets them get better jobs (or in the case of mothers, sometimes lets them get a job at all), and improves parents’ human capital, with all the relevant follow-on effects. More speculatively, if the home environment is unusually bad, it gives the child a little while outside the home environment, and socializes them into a “normal” way of life. I’ll discuss a slightly more fleshed-out model of this in an upcoming post.

My only caveat in agreeing with this perspective is that Chetty finds the same effect (no academic gains, but large life-outcome gains years later) from children having good rather than bad elementary school teachers. This doesn’t make sense in the context of freeing up parents’ time to get better jobs, or of getting children out of a bad home environment. It might make sense in terms of socializing them, though I would hate to have to sketch out a model of how that works. But since the teacher data and the Head Start data agree, that gives me more reason to think both are right.

I can’t remember ever making a post about how Head Start was useless, but I definitely thought that, and to learn otherwise is a big update for me. I’ve written before about how when you make an update of that scale, it’s important to publicly admit error before going on to justify yourself or say why you should be excused as basically right in principle or whatever, so let me say it: I was wrong about Head Start.

That having been said, on to the self-justifications and excuses!

1) Head Start seems to work for reasons unrelated to the ones that made people want to do it. Those people were still wrong, and this is still a good example of policy effects being difficult to predict. It seems to have succeeded by coincidence, not because “early childhood education” is a good idea.

2) This probably strengthens rather than weakens the Caplanian case against education, since the studies find that the educational parts of preschool are not useful, and better teachers and curricula do not affect the benefits.

3) This strengthens rather than weakens the case that academic achievement is related primarily to IQ, and that IQ is primarily genetic and difficult to change. An intervention targeted at academic achievement and IQ manages to change everything else except those variables, which remain stubbornly the same. Studies consistently find that IQ is only responsible for about 25% of life outcomes, suggesting that education works on the other 75%.

But on a broader scale, this does lower my confidence in biodeterminism. Preschool is a shared environmental effect; your parents have a big effect on whether or not you go to preschool. Why doesn’t this shared environmental effect show up in studies, which generally find no shared environmental effect matters?

This is the same problem raised by Ozy’s post on lead. We know lead is important. We know it can damage your life outcomes. But we also know lead is related to the shared environment. And we also know studies keep finding the shared environment doesn’t matter. Some studies find the shared environment matters a little, when you make extra-double sure to have very high income inequality in your sample. But other studies find that it doesn’t, and almost all of them find that it doesn’t matter much at the still-high levels of income inequality you get by recruiting a convenience sample. How can this be? We have two really excellent and well-replicated scientific literatures, each proving opposite things. What now?

All I can think of is that maybe shared environment can matter, but is so small in the grand scheme of things that it’s below the threshold where zoomed-out studies of everything can detect it. That would help reconcile the two literature bases. But it doesn’t seem right. The lead effects are huge. The preschool effects, while moderate, suggest that something as minor as “whatever social advantage your family gets from your mother not having to take care of you for part of the day from ages 3 – 5” can have lasting and detectable effects. Surely then we would expect much larger effects from whether your mother is independently wealthy and can do whatever she wants, or whether your family otherwise has the ability to accrue social advantage.

It might also be an effect of what we’re measuring. Although there’s conventional wisdom that shared environment shows little effect in twin studies, there are occasional outliers. For example, studies of crime often find shared environment factors around 15-20%, especially in younger or poorer samples. And some of the studies that found effects from preschool measured crime. These are some inconsistent findings, and 15-20% from everything doesn’t seem consistent with measurable effects from preschool alone, but I’m kind of desperate here.

I guess I will just increase my belief in the studies that suggest shared environment matters a bit more when you limit yourself to non-cognitive factors and include the really poor, and hope that future work confirms this result.

I’ll also increase my political support for programs like these. I think these findings make universal childcare (almost) a no-brainer. They make universal pre-K much more appealing, with the strongest arguments against being inefficiency, eg that universal childcare or basic income are a more effective way of doing the same thing. But given the political realities that make universal pre-K more likely to happen than childcare or basic income, I am now happy to support it.