Revisiting Project Star and the Long-Term Impacts of Educational Interventions

Some big news in educational economics came out this week with a team of economists including Raj Chetty and Emmanuel Saez releasing some pretty groundbreaking results on the long-term impacts of Tennessee's Project Star. Slides from the talk where the results were presented are here. The New York Times has a pretty good overview here. Project Star is pretty well known in economics as it was one of the first big purposely randomized education experiments. Students were randomly placed in classes with different numbers of students to try and tease out whether class size had an effect on educational outcomes ("Star" pithily stands for "Student Teach Achievement Ratio"). It famously found that (a) effects were there in the intuitive sense, i.e., students in smaller classes did better on standardized exams, the metric of choice, but (b) those effects trailed off over time.

What Chetty, Saez and their coauthors do (which I think captures pretty nicely the sort of economics that sends a tingle of cleverness-excitement up my spine) is match students who were in the Star experiment with their tax records to get long term earning data. Two things are particularly cool about this. The first are the results themselves. They find that not only are class size effects important in the long run but teacher quality effects and (even more robustly) class quality effects have very measurable long-term impacts on latter life earnings and college attainment / achievement rates, even when those effects take place in kindergarten. If you want the terrifyingly strong result, check out slides 44 and 49 in the pdf. Moreover, they go back and show that test scores, which everyone agrees are only a proxy for what we actually care about, do a fairly nice job of predicting how people will do later on, providing some validity to programs that measure their impact by how well students do on exams. Whether that should be exciting or troubling is, I think, a pretty deep question.

The second thing that's cool about it is that it sets out to find long-term, indirect, but ultimately very important effects in a very picked-over area of research (Star has been written about a *lot*). The education literature has long been one of the main sources of new applied econometric technique, which is why someone like me whose research has very little to do with education knows about things like Tennessee Star or class size cutoffs in Israel. So the precedent this sets, which is basically "sometimes the effects we care about are 20 years down the line and not in the original area we were looking at" comes in a heavily trafficked area of research with both rich existing precedent and a lot of sway over other areas of applied research. Which obviously makes me happy, since that same precedent can be thought of as applying to a lot of the environmental, health, and public goods questions that we concern ourselves with in sustainable development-related research.

Anyway, if you have a second read the Times article or, even better, flip through the slides (they're pretty clean and approachable and have some great graphs). Then go teach a five year old something new.