A Dynamic Conditionally Heteroscedastic Stochastic Frontier Model

This paper presents a new dynamic ARCH-related conditionally heteroscedastic stochastic frontier model specification where firm and time-specific technical inefficiency is represented by an autoregressive stochastic process in the error components. Monte Carlo results reveal that a one-sided likelihood ratio test of the proposed model has correct small-sample size and has high power for small to medium sized panels. An empirical application is included using a panel of 23 OECD countries over the 26 year period 1965-1990. The estimation results indicate a clear rejection of the standard frontier model and existence of first order dynamic conditionally heteroscedastic technical inefficiency.