AGGREGATION ISSUES IN THE ESTIMATION OF MALMQUIST PRODUCTIVITY MEASURES

The paper contributes by demonstrating the sensitivity of nonparametric programming productivity measures to the choice of model –time series versus panel models of Malmquist productivity, and to various levels of commodity aggregation compared to the traditional Tornqvist-Theil index approach employing U.S. state-level data from 1960-96. To illustrate the sensitive of nonparametric programming productivity measures, we compare the implicit shadow shares recovered from the dual values of the Malmquist productivity and total factor productivity methods to the observed shares of the Tornqvist-Theil index for U.S level data from 1948-1994.