EcoSta 2017: Start Registration
View Submission - EcoSta2017
A0179
Title: A new semiparametric approach for nonstandard econometric problems Authors:  Bo Zhou - Tilburg University (Netherlands) [presenting]
Abstract: A new approach is developed to derive efficiency bounds and efficient tests for non-standard problems in semiparametric econometric models where the innovation density is treated as an infinite-dimensional nuisance parameter. The approach is based on an explicit nonparametric modelling of the innovation density and a structural version of the limit experiment. The structural limit experiment exhibits an invariance restriction that is employed to eliminate the nuisance parameter. The associated maximal invariant gives the efficiency bound using the Neyman-Pearson lemma. Moreover, the invariance structure naturally leads to statistical inference procedures. For example, the appearance of Brownian Bridges suggests the use of rank statistics. A simple linear regression model is used as a running example, leading to the same efficiency bounds and tests as the traditional least-favorable parametric submodel approach. We apply our results to two nonstandard econometric problems: testing cointegration rank and testing hypothesis with weak instruments. In both cases, we derive the semiparametric power envelopes and the induced rank-based inference procedures.