Nation writer Rick Perlstein suffered paroxysms last week over my dismissal of the evidence for universal pre-K, which he defended as “Nobel Prize-winning research.” Perlstein is mistaken. Though James Heckman, a leading preschool advocate, is indeed a Nobel laureate, he was awarded the prize for brilliant but unrelated work on statistical methods.

Far from being “Nobel Prize-winning,” the empirical case for universal government pre-K collapses under mild scrutiny. The central claim, as voiced by President Obama in his SOTU speech, is that “every dollar we invest in high-quality early childhood education can save more than seven dollars later on.” This sweeping statement does not in fact refer to the typical  return from federal or state pre-K programs. It refers to the findings from a single intensive 1960s early childhood experiment  that served 58 children in Ypsilanti, Michigan—the High/Scope Perry preschool program. Out of the literally hundreds of preschool studies conducted in the past half-century, the Perry results are not representative and have never been reproduced on a national or even a state level. In fact, an earnest experimental effort to reproduce them for just a few hundred children at eight locations failed despite an annual investment of $32,000 per child, adjusted for inflation—far more than the President currently contemplates spending.

The president’s case for universal government pre-K singles out the unusually large positive effects of one tiny study—sometimes two or three—from scores of others that show little benefit, no benefit, or even significant harm to participating students. That sea of inferior results, moreover, is drawn in large part from …the federally-funded pre-K efforts of the past 47 years. Indeed the largest, best designed, most recent studies of federal pre-K efforts were published by the Obama administration itself: the Head Start Impact Studies. These studies find little or no net lasting benefit to federal pre-K. The Obama administration was apparently so worried about these findings that the most recent study was released on the Friday before Christmas—despite a publication date on its title page of October 2012.

What we have here, in other words, is a monumental act of cherry picking rather than an example of scientifically grounded policymaking.