Estimating multiproduct costs when some outputs are not produced

Pooling diversified and specialized firm data to analyze multiproduct cost technologies raises two issues in applied research: (1) a functional form must be specified that accommodates zero outputs, and (2) assumptions must be made regarding the structure of the multiproduct technology when some outputs are not produced. This article introduces a methodology to estimate the translog multiproduct cost function in the presence of zero outputs. The method adds flexibility to allow for and test competing structural assumptions. The added flexibility can improve measurement of the global properties of multiproduct cost structures. An application to a cross section of U.S. railway firms demonstrates this key advantage. Copyright Springer-Verlag 2003