The validity of instrumental variables (IV) regression models depends crucially on fundamentally untestable exclusion restrictions. Typically exclusion restrictions are assumed to hold exactly in the relevant population, yet in many empirical applications there are reasonable prior grounds to...

This paper identifies endogenous social effects in mathematics test performance for eighth graders in rural Bangladesh using information on arsenic contamination of water wells at home as an instrument. In other words, the identification relies on variation in test scores among peers owing to...