Testing misspecification of the stochastic frontier model with a random effects panel data model

Conventional time invariant modeling has proven illustrative of a range of firm effects, but has failed to encompass the misspecification problem. With a view to accounting for the temporal correlations of misspecification, we posit a simple time invariant stochastic frontier with reference to a random effects panel data model (REPDM). Our model departs from more traditional models; insofar as the latter's operating assumption is that firm effects are independently distributed in a uniform manner. These theoretical presuppositions are then tested through empirical application to the cost data of IC design companies in Taiwan. Comparative analysis is also performed in relation to the orthodox methodologies of Battese and Coelli (1992), and Greene (2003). Results indicate that misspecification is amenable to testing and the development of a model capable of evaluating the inefficiencies of firms.