One of the clever tricks in modern applied economics has been the use of instrumental variables (IVs). Using IVs can help tease out causal relationships, where we might otherwise be left with correlations. For example, Caroline Hoxby’sÂ paper on class size seeks to determine whether smaller classes boost student achievement. Because kidsÂ might be sorted into classes based on their ability, the raw correlation doesn’t tell us anything causal. Instead, HoxbyÂ uses variation from a Connecticut rule that caps class sizes in a school to 25. The results of the IV strategy suggest that below 25, cutting class sizes doesn’t boost student performance.
Of course, not all IV papers are as clever as this, and the literature has spawned plenty of bad IV approaches. To help improve the standard, Michael Murray, my first-ever econometrics lecturer at Harvard, and one of the nicest econometricians you’ll ever meet, has just released a terrific paper on the topic. Entitled The Bad, the Weak and the Ugly, it’s well worth reading for anyone working in the area, or for anyone interested in how economists are doing their darndest to prove causation.