An instrumental variables (IV) identification strategy that exploits statutory class size caps shows significant achievement gains in smaller classes in Italian primary schools. Gains from small classes are driven mainly by schools in Southern Italy, suggesting a substantial return to class size reductions for residents of the Mezzogiorno. In addition to high unemployment and other social problems, however, the Mezzogiorno is distinguished by pervasive manipulation of standardized test scores, a finding revealed in a natural experiment that randomly assigns school monitors. IV estimates also show that small classes increase score manipulation. Dishonest scoring appears to be a consequence of teacher shirking in answer sheet transcription, rather than cheating by either students or teachers. Estimates of a causal model for achievement with two endogenous variables, class size and score manipulation, suggest that the effects of class size on measured achievement are driven entirely by the relationship between class size and manipulation. These findings show how consequential score manipulation can arise even in assessment systems with few NCLB-style accountability concerns.