Asymmetrical Information - Megan McArdle

02.13.13

Can We Really Give Everyone Access to High Quality Preschool?

It sounds great. But can the president's plan survive contact with the real world?

I said in my last post that the president wasn't proposing much that was truly transformative.  But if it works, there was one proposal that truly could have lasting (positive) effects on society: his suggestion of universal, high quality pre-school.  

There is evidence that pre-school makes big differences if it's done right.  Both the Perry Pre-School Project, a groundbreaking experiment conducted in Ypsilanti in the early 1960s, and the Abcedarian experiment performed in North Carolina in the early 1970s, seem to have made substantial improvements in the life outcomes of the kids they served.  They did not turn the children into middle class college graduates, but it did improve school graduation rates and reduce the likelihood of a criminal arrest.  A team lead by economist James Heckman, who is one of the smartest guys around on educational research, estimates that the return on investment for Perry's pre-school program are 7-12%.  

So what's not to love?  It seems like the sort of idea that no one could possibly oppose.  Nonetheless, there are three big questions about the president's proposal that need to be resolved before we move forward:  

1.  Should it be universal?  Liberals like universal programs because they build widespread support for the benefit.  Their customary tagline, when people propose means testing or otherwise narrowly targeting the neediest, is that "a program for the poor is a poor program".  

But in this case, I think universality is the enemy of the president's stated goal, which is to help give poor kids a leg up into the land of opportunity.  First, and most obviously, a universal program will be much more expensive than a program targeted to the 20% of kids who are poor (or the somewhat higher percentage whose families have incomes within 1.5 or 2 times the poverty line.)  But perhaps even more importantly is the way that the middle class portion of the program will siphon off the best resources, which is to say, the best teachers.  Think of the flow of teachers within school districts from poorer to richer schools.  The affluent kids are easier to teach: fewer behavior problems, much better preparation for school, and few of them have trouble getting their homework done because the family's crammed into three rooms with Grandma and Aunt Marie.   So the best teachers flow towards the kids who need them least.  

There's no evidence that I'm aware of that pre-school helps middle class kids; it helps poor kids because it makes up for the stuff that middle class parents do (reading readiness, for example), and poor parents can't or don't.  So if we're going to pass a big expensive new program aimed at helping poor kids with serious deficiencies in their home environment, I want to target it on those kids, to make sure that they get as much benefit as possible.  

2.  Will it be high quality?  The biggest issue with high quality pre-school is simply that we don't know if we can do it on large scale.  Perry had 123 kids from 104 families; the number in Abecedarian was even smaller.  There are huge dangers in extrapolating from studies this small to national programs:  

  • In small samples, it's more likely that the results you get will be random chance rather than a real effect
  • Small programs can be very selective about hiring, marshalling a comparatively large number of great teachers, administrators, and so forth, all of whom are on fire about making this program work.  A national rollout cannot be nearly so choosy.  All the teachers in Perry, for example had bachelor's degrees (this in the 1960s, when that was much less common).  Head Start couldn't even require an associate's degree; the most recent update, from 2007, mandated that 50% of Head Start teachers needed to have a BA in early childhood education or a related field.
  • Large programs are much harder to administer than small programs: less flexible, less accountable.  Think of trying to reschedule dinner plans with one friend--and then compare that to trying to reschedule dinner plans with a dozen.  That will give you some idea of the difficulty of scaling.  
  • People who are being studied sometimes react by making extra investment in whatever is being studied.  People who are gettng a normal government benefit don't.  

You can read more about the problems of scaling here.  The upshot of which is that you should always, always keep scaling problems in mind when you are reading big claims made on the basis of small projects.  And a frightening proportion of the evidence for the benefits of early childhood intervention comes from these two programs, which covered a couple of hundred kids 40-50 years ago.

In fact, we've already tried scaling high-quality pre-school once: that's Head Start. And Head Start just doesn't seem to do much good.  If we're going to argue for pre-school based on the Perry and Abecedarian results, we need to know in advance why Head Start didn't work, and how we're going to avoid those problems this time around.  

3.  How much will it cost?  Perry was really, really expensive.  James Heckman estimates a present-day cost of about $20,000 per child.  And that's pocket change compared to Abecedarian, which cost an amazing $90,000 per student in today's dollars.  

There are about 4 million kids born every year.  Giving all of them the sort of high-quality pre-school that Perry and Abecedarian provided would cost between $80 billion and $360 billion a year.  It would also, at a teacher ratio of 6:1 (what Perry had) require 1.3 million new high quality preschool teachers.  

And that may be an underestimate of the cost, because the price of high-quality teacher labor has gone up.  Abecedarian and Perry were done in an era where very smart, highly educated women simply expected to make less than a less-well-educated man.  They were also restricted in the jobs they could get, so that many more of them ended up in education and nursing.  That made it easier for Perry and Abecedarian to put together a skilled, committed team. Actually attracting similarly high quality, skilled teachers in the requisite numbers today might require enormous salaries, pushing up the cost still further.  

Of course, we can restrict the program to the 20% of kids who are poor.  That would cut our cost to somewhere between $15 billion and $72 billion, and require only 250,000 highly skilled teachers.  Per year.  Double those figures if you want to cover the near-poor kids.  

The Center for American Progress says it would cost less--about $10.5 billion.  But they get there by assuming less-than-universal takeup (probably a safe bet) and also, that the Federal government will only pay for half, with states picking up the rest of the tab.

But surely, you will say, there will be cost savings in a large national program!  It won't cost exactly as much per kid when there are millions as it did to set up the first program for 100 kids.

To some extent, that's certainly true.  Bulk purchases of supplies, and streamlining bureaucratic overhead, can save real money.  But don't overestimate those savings--scaling a program up also means you need more layers of management to coordinate decisions which used to be made on an ad-hoc basis when you were dealing with a single school.  You might also reduce the costs of measurement (pilots spend a lot of time monitoring how things are going), but I wouldn't count on that; regular programs need evaluation too, particularly in education.

And the big costs of running a pre-school are things like buildings and teaching labor.  There's no way to scale the teaching labor--it still takes exactly an hour to spend one hour with a kid, something that hasn't improved since we lived in caves and hunted with rocks.  And adding an extra classroom costs money.  


If the president wants high-quality pre-school, he needs to make the case that we can muster the required resources.  He needs to make the case that we can actually deliver high quality at mass scale. And then he needs to make the case to the American public for spending billions and billions of dollars on poor kids.  

I think that it's possible to make that case.  These kids didn't ask to be born to parents who can't give them what it takes to succeed.  And if do help them, Heckman makes a convincing case that the benefits will be large down the road: less crime, more capable workers, fewer people collecting benefits.  

But the president hasn't done it yet.  This is a big new program, and it needs a big political push, not a side note in a single speech.  


This article has been corrected.  The original figures for the cost of Perry and Abecedarian were overstated by a factor of two.