Weak Instruments Test in Discrete Choice Models

This paper proposes a test to reliably detect weak instruments in discrete choice models. As a by-product of our testing approach, we generalize the standard “rule-of-thumb” for linear models to discrete choice models. This generalized rule provides insights regarding instrument strength in a host of discrete choice models. A Monte Carlo analysis compares our proposed testing approach against commonly applied weak instruments tests. The results simultaneously demonstrate the good performance of our approach and the fundamental failure (over-rejection of the null of weak identification) of conventionally applied, i.e. linear, weak instrument tests in this context. We compare our testing approach to those commonly applied in the literature within two empirical examples: married women labour force participation, and US food aid and civil conflicts.

Please sign up for meetings below: