A chart in Fryer's report shows the effect of incentive programs from city to city. (Click to enlarge)

Despite several spectacular setbacks, Harvard economist Roland Fryer isn’t ready to throw in the towel on incentives to boost student performance.

In recent years, New York City abandoned two different inventives programs that Fryer designed — one for students and another for teachers — after it became clear that the promise of more cash for higher test scores wasn’t paying off.

But Fryer, who last week was awarded a “genius grant” by the MacArthur foundation, has experimented with incentives in other cities and gotten different results. In a report released today, he and a colleague from Harvard University’s EdLabs offer instructions for designing incentives programs and argue that, contrary to what economic theory would predict, programs that reward “inputs” such as reading or completing homework are more effective than those that reward “outcomes” such as test scores, as New York’s program did.

In Houston, students who were paid $4 for each math skill they learned mastered more skills — and they did even better when the prize grew to $6 a skill. In Dallas, students who were paid to read books read more books.

More study is needed to figure out exactly why the Texas students responded to incentives and students in New York City did not, the researchers write. But they hypothesize that New York City students might not understand that comprehending content is key to raising scores. The researchers write:

After each of the ten exams administered in New York City, our qualitative team asked students how they felt about the rewards and what they could do to earn more money on the next test. Every student found the question about how to increase his or her scores difficult to answer. Students answering this question discussed test-taking strategies rather than salient inputs into the education production function or improving their general understanding of a subject area.

Fryer’s report is one of three released today by the Hamilton Project, an economy-focused research initiative of the non-partisan Brookings Institution. A second report, by a team of Columbia University researchers, takes New York City data into account when concluding that changing conditions within schools — such as their start time, grade configuration, or teacher deployment — can boost student achievement just as much or more than changing system-wide policies. And a third proposes that states and cities use separate tests to measure student achievement and teacher effectiveness.