Parametric aggregation of random coefficient Cobb-Douglasproduction functions : evidence from manufacturing industries

A panel data study of parametric aggregation of a production function is presented. A four-factor Cobb-Douglas function with random and jointly normal coefficients and jointly log-normal inputs is used. Since, if the number of micro units is not too small and certain regularity conditions are met, aggregates expressed as arithmetic means can be associated with expectations, we consider conditions ensuring the existence and stability of relationships between expected inputs and expected output and discuss their properties. Existence conditions for and relationships between higher-order moments are considered. An empirical implementation based on panel data for two manufacturing industries gives decomposition and simulation results for expected output and estimates of the aggregate parameters. Illustrations of approximation procedures and aggregation errors are also given.