K-12 needs an R&D budget

An ARPA for education deserves a second look

Two new papers from Dan Lips make the case for boosting federal spending on educational R&D. In the first, Lips notes that,

In 2022, the United States will spend less than $1 billion on K-12 education R&D initiatives through the Department of Education and National Science Foundation, an amount largely unchanged over the past decade.

Given the US spends on the order of $864 billion a year on elementary and secondary education, this implies a national educational R&D budget of only around 0.1%. There are fast food chains with larger research budgets. Worse still, it’s not clear that the money we do spend on education research is really, well, research. More often, it’s spending on stuff like “STEM education initiatives,” like supporting computer science courses in high school. That may be worthwhile, but it begs the question of whether the educational methods involved are really working as intended. Unfortunately, as Lips explains in his newest report released today, US spending on STEM and computer science education is wanting for greater transparency:

Our analysis of federal STEM education and education R&D programs shows that there is significant federal activity aimed at improving STEM education and national competitiveness. However, consistent with past reviews of these federal initiatives, it remains unclear whether and to what extent these initiatives are advancing national goals.

Any significant boost in federal funding for educational R&D should thus insist on an open and transparent framework for evaluating various interventions. One way to do this would be through an APRA-Ed model, similar to the “advanced research projects agencies” we already have for energy, defense and biomedicine. The virtue of the APRA model is how it not only enables experimentation, but also the rigorous, outcome-oriented evaluations needed to know if an experimental invention actually worked, with the presumption that failing interventions won’t be funded in perpetuity.

The Obama administration proposed an ARPA for education back in 2011, but unfortunately it never materialized, as the greater attention they gave to Race to the Top and Common Core hardened conservatives against further federal expansions into K-12. And not without reason. But insofar as an ARPA model would test interventions in micro, it would at least guard against the potential for faddish, one-size-fits-all preemptions of local authorities. Ideally, it would merely spur the creation of new tools and research that providers could draw on if they so choose, injecting some transparency and innovation into the margins of a system that’s otherwise notoriously resistant to change. It also wouldn’t be expensive. The Obama-era proposal only requested $90 million for the program’s first year — a drop in the bucket relative to how highly leveraged ed research could be if it were truly rigorous.

Indeed, between the rapid rise in remote schooling, the troubling evidence of Covid-era learning loss, and transformative AI-powered tutors on the horizon, the idea of an ARPA for education deserves a serious second look.