Central Limit Theorems and Inference for Sources of Productivity Change Measured by Nonparametric Malmquist Indices

Malmquist indices are often used to measure productivity changes in dynamic settings and have been widely applied. The indices are typically estimated using data envelopment analysis (DEA) estimators. Malmquist indices are often decomposed into sub-indices that measure the sources of productivity change (e.g., changes in efficiency, technology or other factors). Recently, Kneip et al. (2018) provide new theoretical results enabling inference about productivity change for individual firms as well as average productivity changed measured in terms of geometric means. This paper extends those results to components of productivity change arising from various decompositions of Malmquist indices. New central limit theorems are developed to allow inference about arithmetic means of logarithms of the sub-indices as well as geometric means of (untransformed) sub-indices. The results are quite general and extend to other subindices not explicitly considered in this paper