Specification Testing of Production Frontier Function in Stochastic Frontier Model

Parametric production frontier function has been commonly employed in stochas-tic frontier model but there was no proper test statistic for its plausibility. To fill into this gap, this paper develops two test statistics to test for a hypothesized parametric production frontier function based on local smoothing and global smoothing, respectively. We then pro-pose the residual-based wild bootstrap approach to compute the p-values of our proposed test statistics. Our proposed test statistics are robust to heteroscedasticity. Simulation studies are carried out to examine the infinite sample performance of the sizes and powers of the test statistics.