Steinglass writes:

What’s curious is that both of these initiatives seem to be several years past the point when they were the most convincing in intellectual terms, on the basis of theories and evidence. School choice was an idea that had a lot of proponents in the 1990s, but with well over a decade of large-scale pilots for various implementations, it doesn’t seem to be showing any results. And you have former top proponents like Diane Ravitch actually turning against charter schools.

Many others have subjected Diane Ravitch’s arguments to close scrutiny. The fact that Ravitch has said that charter schools aren’t a panacea — a claim that every charter school advocate I’ve ever encountered would heartily endorse — doesn’t strike me as dispositive.

Anyone even passingly familiar with the history of educational RFTs and their limitations would know that Steinglass’s characterization is problematic. Randomized field trials in education are difficult to devise, and the number of large-scale pilots for various implementations is small, particularly when compared to the number and quality of experiments that preceded the welfare reform efforts of the mid-1990s. We do have a handful of natural experiments involving lotteries. These experiments face a number of limitations, including faulty record-keeping, a failure to properly separate treatment and control groups, and much else besides. But of the big randomized lottery experiments, we have seen nontrivial gains for African American students. We actually don’t have much data for non-black students, in large part because of ferocious resistance to further experimentation.

Because these experiments have yielded nontrivial gains without an increase in resources, I’m hard-pressed to see why we shouldn’t field more experiments, ideally well-designed RFTs. The idea that this is a settled issue is … interesting.

Steinglass goes on to suggest that the case for defined-contribution social insurance programs rests on strong stock market performance. This is another interesting idea. Note that many defined-contribution social insurance proposals involve investing in Treasury-backed assets and not the stock market. And of course many libertarians and conservatives, myself included, back notional defined contribution schemes, like the system used in Sweden, that would allow the benefit growth formula to track broader economic growth, etc., thus creating a more sustainable entitlement.

I do hope that Wilkinson replies to Steinglass.