The NY Times reports on some preliminary results from one of Roland Fryer’s field experiments in which students are rewarded with cash for high AP test scores.

Results from the first year of the A.P. program in New York showed that test scores were flat but that more students were taking the tests, said Edward Rodriguez, the program’s executive director.

Fabio Rojas at Orgtheory.net, interprets this as saying that the incentive didn’t work.

The question is simple: does paying kids improve performance? As I mentioned yesterday, the preliminary evidence is that children are more likely to participate in the test, but they are not more likely to get better grades.

I think he has jumped to his conclusion.  If more students take the test, then we are drawing in the marginal students whose baseline test scores would be lower than average and would bring therefore bring the average down.  Since the average did not go down that means that performance by infra-marginal students (and probably even the marginal students) improved.