| 讲座简介： | This paper proposes a novel time-varying model averaging approach for generalized method of moments (GMM) to capture structural changes in economics and finance. Unlike existing literature, our approach allows for potential misspecification of moment conditions while permitting time-varying parameters and weights. The asymptotic optimality and convergence rate of the selected weights are derived and the consistency of the proposed averaging estimator is obtained. Moreover, we prove that if one or more candidate models contain correctly specified moment conditions, our method will asymptotically assign all the weights to them with probability approaching 1. Simulation studies demonstrate the superiority of our approach over competing methods. Applying our time-varying model averaging to stochastic discount factor models for pricing U.S. equity returns reduces model uncertainty by assigning time-varying weights to various instrumental variable-based moment conditions, constructing an investment strategy. Furthermore, the proposed method yields more profitable investment performance than other existing model selection techniques. |
