Education

Recognizing our ignorance gives us a starting point for designing a human capital program for America. While the ultimate point of a human capital program is to educate children, the key feature in structuring that program must be that it also educates policy-makers, principals, and parents about education policy.

The genius of Google is that it is a self-teaching program where information from Web visits on the Google toolbar makes the Google search engine more effective. Google is a tool that makes us smarter, but it also makes itself smarter along the way. A Marshall Plan for education should start by acknowledging that we don't know all the answers and that we need to learn quickly if we are going to radically improve this nation's schools.

The need to educate ourselves about education means that the program should start with some good ideas and then charge forward with the great knowledge - generating a trinity of experimentation, randomization, and evaluation. Experimentation means trying out bold ideas, such as providing incentives for children and parents or paying large bonuses to successful teachers or using Web-based learning in ways that many of us over 40 will never understand. Sticking close to the status quo will generate less new information than something more radical.

None of these new approaches needs to be imposed on unwilling schools - it is enough if a small percentage of schools are willing to experiment. If a few schools are willing to participate, then randomization should determine which districts will get the new approach and which districts will sit it out. Throwing dice to allocate education reforms across schools may seem unfair, but randomization is critical for the evaluation that will make the human capital program self-educating. Without randomization, we won't know whether it is the new approach that is responsible for good results or whether those results reflected something else about the school.

Of course, what works in Brockton may not work in Brookline. As successful approaches get brought to scale, ongoing evaluation is needed to learn whether something that worked in an initial sample of schools will work everywhere.

The great difficulty in public education reform is that the education of a child depends significantly on people who aren't on the government's payroll, such as parents, peers, and the children themselves.

Some researchers, such as my colleague Roland Fryer, are using experimental approaches to figure out how to get more out of parents and students. In New York, Washington, and Chicago, Fryer has been conducting experiments where children get direct rewards - such as cash or cellphone minutes - for doing well on tests.