Robust Inference for the Frisch Labor Supply Elasticity

The Frisch labor supply elasticity plays a key role in many policy debates, but its magnitude remains controversial. Many studies use 2SLS regressions of hours changes on wage changes to estimate the Frisch elasticity. But a little appreciated power asymmetry property of 2SLS causes estimates to appear (spuriously) imprecise when they are shifted away from the OLS bias. This makes it difficult for a 2SLS t-test to detect a true positive Frisch elasticity. We illustrate this problem in an application to NLSY97 data. We obtain a Frisch estimate of 0.60 for young men, but the t-test indicates it is insignificant. In contrast, the Anderson-Rubin (AR) test – which avoids the power asymmetry problem – implies the estimate is highly significant. The same power asymmetry issue that afflicts the t-test here will arise in many IV applications. Thus, we argue the AR test should be widely adopted in lieu of the t-test.