Mark Zuckerberg and Bill Gates have a mild obsession with public charters, schools that focus on core academic subjects, such as science, economics, or government. Now, a massive new study suggests that injecting the successful bits of charter schools into regular schools might help minorities in math -- but not in reading.

It turns out that Zuckerberg's $100 million donation to New Jersey public schools might just help get more minorities on the path to lucrative tech jobs.

Texas gave Harvard researcher Roland G. Fryer an impressive amount of authority to overhaul their worst schools in 2010. He increased instruction time 21 percent, replaced nearly every principal and a one-third of the teaching staff, ran all schools on "data-driven" instruction, and emphasized high expectations for every student.

"We have shown that this particular set of interventions can generate gains in math in both elementary and secondary schools, but that it generated small to no effects in reading," he concludes [PDF].

The results are both incredibly depressing and moderately hopeful. On the one hand, at least something works. All of the experimentation funded by generous Silicon Valley donors proves that we can help minorities in STEM.

On the other hand, the results aren't great. The entire overhaul closes the gap between white and black students by less than half. Schools, it seems, have limited ability to overcome poverty.

For stat nerds, the study found that math test scores improved about 0.18 standard deviations, a measure of how far from the average a student ranks (1 standard deviation would move a student from the 50th percentile to the 67th). One estimate of the average standard deviation gap between whites and blacks is about 0.8, but it fluctuates widely.

Gates and Zuckerberg's commitment to charter schools has been highly controversial, especially since past evidence on such experiments have been mixed. This shows that charter schools are beginning to discover ways of changing the education system at scale.

There's hope, but still a long way to go.