A0505
Title: Robust joint estimation of treatment effect via possible dependent instrumental variables
Authors: Yiqi Lin - The Chinese University of Hong Kong (Hong Kong) [presenting]
Qingliang Fan - The Chinese University of Hong Kong (Hong Kong)
Xinyuan Song - Chinese University of Hong Kong (Hong Kong)
Abstract: The instrumental variable (IV) estimation with potentially invalid IVs is extended to allow for weak IVs and scenarios where the majority or plurality rules are difficult to hold or verify. In empirical research, weak IVs are common. A novel estimator, called WIT, is proposed to deal with invalid IVs and improve estimation accuracy under many weak IVs. We show that the WIT estimator works remarkably well under more relaxed identification conditions, which is unachievable in previous literature. Theoretical properties are derived for the proposed estimator. The finite sample property is demonstrated on simulated data and an empirical study concerning the effect of trade on economic growth.
