The generalized method of moments (GMM) is an important estimation procedure in many areas of economics and finance, and it is well known that this estimation is highly sensitive to the presence of outliers and influential observations. Case-deletion diagnostic has been studied in GMM estimation; however, it is surprised that local influence analysis is under explored. To this end, a local influence method is proposed to assess the effect of minor perturbation on GMM estimation. The local diagnostic measures of GMM estimators under the perturbations of empirical distribution and moment condition are derived to study the joint influence of observations. The obtained results are applied to efficient instrumental variable estimation and dynamic panel data model. Two real data sets are used for illustration, and a simulation study is conducted to examine the effectiveness of the proposed methodology. The advantage of local influence method is analyzed in detail through comparison with the case-deletion method.
Read full abstract