Conditional Moment Models under Semi-Strong Identification

We consider models defined by conditional moment restrictions under semi-strong identification. Identification strength is directly defined through the conditional mo- ments that flatten as the sample size increases. The framework allows for different iden- tification strengths across parameter’s components. We propose a minimum distance estimator that is robust to semi-strong identification and does not rely on the choice of a user-chosen parameter, such as the number of instruments or any other smoothing parameter. Our method yields consistent and asymptotically normal estimators of each parameter’s components. Heteroskedasticity-robust inference is possible through Wald testing without prior knowledge of the identification pattern. In simulations, we find that our estimator is competitive with alternative estimators based on many instruments. In particular, it is well-centered with better coverage rates for confidence intervals.