Advocates for pre-K education programs often cite the Perry Preschool Study as demonstrating many benefits from early education. It is from this study alone that the frequently mentioned seven-to-one ratio of savings in public expenditures compared to cost was estimated.
The study itself, though, is fundamentally flawed and should carry no weight in public debate over pre-K programs.
It was a small study conducted on only 123 children, all of them black and all from very poor circumstances. It was conducted at a single site in Ypsilanti, Michigan, and was begun in the early 1960s.
The strength of the study was supposed to be its design as a true social experiment, a randomized controlled trial. This means that the participants in the study were randomly assigned either to a control group, which did not participate in the preschool program, or to an experimental group, which did participate. Randomized controlled trials are the best way to show that differences between the two groups are indeed due to the intervention being studied.
Early on, the children in the experimental group (those participating in the preschool) showed improvements in tests of intellectual ability compared to those in the control group, and it was this finding that caused so much optimism. However, those improvements faded away after a few years, as is typical. Fans of the Perry Project have subsequently touted other, softer findings that they claim make up for the lack of actual intellectual progress.
Examples of these soft findings include an increased high school graduation rate for the girls only, students’ reports that they had homework, parents’ reports that they hoped their children would attend college, and lower arrest rate for those who were in the preschool program. Arrest rates for both the experimental and control groups, though, were shockingly high.
These meager results are bad enough, but more damning is the fact that random selection of children into control and experimental groups was not maintained. Through the course of the program those who operated it moved some of the children between the groups. They did this for such reasons as keeping siblings in the same group or accommodating the schedules of working mothers. These moves, though, invalidate the entire study.
When asked to name the worst experiences with random selection in social policy research, analyst Robinson G. Hollister Jr. lists the Perry Preschool Project first. It is worth considering his comments at length. He said:
“In the end, 20 percent of the sample that was used in the subsequent multiyear follow-up evaluation were not randomly assigned. One cannot tell which direction these deviations from pure random assignment may have biased impact estimates, in favor of or against the program; for me, these failures of random assignment greatly undermine one’s ability to take estimates of the ‘impacts’ as sound.
“What is most disturbing to me is that the findings from this study have been the most often cited evidence of the great benefits to individuals and society that flow from early childhood programs.”
Hollister’s colleagues concur, and their message is clear. Except for those whose careers are tied to the program, the Perry Preschool Project has no credibility among social science researchers. It should likewise have none with the public.