Education

Let’s be humble on what we know about education

August 16, 2018

Greg Forster, Ph.D.

From ramping up pre-K programs to imposing uniform standards and curricula, there are a lot of people who want to give centralized authorities the power to improve public education. These approaches assume that central authorities actually know how to improve education. In fact, sound education policy begins by accepting just how much we don’t know—and, in fact, how much we can’t know and never will. 

This summer, the Arnold Foundation reported that a major study in Tennessee found negative results for state-funded pre-K education. The study used random-assignment methods, the gold standard for scientific research. Kids in the program did better than the control group in the first year, but later on their outcomes went down. By third grade, the students were actually worse off for having “benefited” from the pre-K program.  

This is not the first failure for pre-K. As I wrote in my recent OCPA policy brief, the empirical evidence on pre-K is mixed—and the higher the methodological quality of a study, the worse its findings tend to be for pre-K. And as Jay Greene noted, the evidence is even worse for the federally funded early development program Head Start.  

Accepting that we simply don’t know how to provide good education at the pre-K level is a hard pill to swallow. I know, because 10 years ago, I had to swallow it. I was among those excited by the Reading First program, which was based on solid research showing that early intervention for young children behind in reading seemed to be highly effective. But the federal government spent $6 billion on Reading First and got no results. We simply didn’t know how to implement the early intervention insights at scale.

The problem is not simply that we haven’t done enough research. Taking an education reform to scale involves trying to generalize the content of education across millions of unique children. Centralized command-and-control systems simply can’t have enough information about every individual child to know how to apply educational insights effectively to every child.

Centralized command-and-control systems simply can’t have enough information about every individual child to know how to apply educational insights effectively to every child.

That story got told over and over again in the heady education reform days of the early 2000s. Lots of big new things were being tried to improve public education, and they were sensible things to try: merit pay, revised standards, high-stakes testing. Some states managed to have success in some cases—Massachusetts’ standards reforms and Florida’s mix of school choice, exit exams, and incentives to raise test scores across demographic groups are notable examples. But the overall story was failure. We just couldn’t take these good ideas to scale.

The one exception is school choice. Unlike the other flops of the 2000s, which are mostly gone now, it has continued to grow and grow. And it has succeeded far more often than it has failed. Even in those very few places where it has failed, like Louisiana, we know why. We can avoid making the same mistakes in policy design again.

Why was school choice the only winner? Because it takes our ignorance seriously. It doesn’t try to generalize the content of education across millions of unique children. It empowers parents—who know their own children better than anyone else, and love their own children in a way no one else can—to control how each individual child’s education gets improved.

We don’t know much about how to educate millions of children. But one of the things we do know is how to make education policy as if we didn’t know much. Let’s put the people who do know—the parents—in charge.