From that Lant Pritchett book:
There has been massive success in putting butts in seats—enrollments and grade completion have expanded massively around the globe—even in very poor countries. However, the accumulating evidence from around the globe is that this has not lead to nearly the degree of learning as had been hoped. Students show very little mastery of even the basics, much less the ability to go beyond the rote and actually understand concepts, much much less the ability to use their conceptual mastery applied to novel practical problems, much much much less for even having laid the foundation for “new basic skills” that are not substitutes for, but meant to build off of, the old basic skills.
So what to do? Pritchett then looks at the evidence on returns to educational "inputs" - books, uniforms, teachers, and compares what massive increases in inputs would do, compared to the kind of improvements needed. (The benchmark is one student standard deviation (SSD)).
So inputs don't work either.
Update: I forgot to add this Freakonomics link to a new study finding that providing school uniforms in the US improves attendance, but not grades.