“Research shows that one of the best investments we can make in a child’s life is high-quality early education.” So said the president last night. It’s a claim that is fueling the push for public pre-K at all levels of government, but is it true?
In the past I’ve criticized preschool advocates for ignoring experimental evidence, asking the wrong policy questions, and offering dubious excuses. But let’s take a look at what many consider the best case for government preschool: New Jersey. The state’s preschool program serving its “Abbot districts” (low-income areas) boasts certified teachers, small class sizes, and a variety of support services (making it more expensive than a number of universal proposals, such as New York City’s). “Lessons for de Blasio in New Jersey’s Free Pre-K,” claimed a recent headline in the New York Times.
A Rutgers University study mentioned by the Times purports to show that attending New Jersey preschool leads to large academic gains that persist at least until fifth grade. But the study is not a rigorous controlled experiment. It’s a simple matching exercise that’s not adequate for informing major policy decisions.
After controlling for school district, age, ethnicity, gender, home language, and parental socioeconomic status, the Rutgers researchers compared the test scores of fifth-graders who attended the state’s preschool program with those who did not. Does including the controls mean the comparison isolates the impact of preschool attendance? Only if you believe that those control variables account for all the factors that might influence both a child’s academic ability and his likelihood of being enrolled in preschool.
The problem here is self-selection bias, which plagues non-experimental research. Higher-ability children may be more likely to be signed up for preschool, meaning the comparison could be flawed from the start. The test-score advantage that the researchers attribute to preschool may actually reflect preexisting ability differences between the preschool attendees and non-attendees. A simple set of demographic controls can’t account for that, but a randomized experiment can. When such experiments have been conducted on similar preschool programs, they have found no lasting gains.
The Rutgers study isn’t useless, but major policy decisions need to be made with the strongest possible evidence. For policymakers, the standard must be the large-scale, randomized experiment. Yes, experiments are expensive and logistically challenging, but the alternative — rushing into a new, unproven entitlement — could be far more costly.