Threshold Effects of Technological Regimes for the Stochastic Frontier Model

In the domain of measuring productivity, the stochastic frontier (SF) model has advanced in leaps and bounds and attracted a large amount of academic attention. Inherently, the model utilises input/output combinations of production units as references to construct a frontier for measuring efficiency. Silent curiosity has followed the model for years in regard to whether all the units actually share an identical technology, or more specifically, how many technological groups should be included in the sample observations. This study directly spotlights the perplexity by employing econometric concepts and techniques to make the data talk and to show the due grouping points. Using the computer, communications, and audio and video electronics products manufacturing sectors during 2001 in Taiwan as the sample, the empirical work revealed that the production possibility set of the sample industry should be divided into six regimes, namely micro, micro–small, small, moderate–medium, medium, and large enterprises. For the smallest- and largest-scale firm groups, the size–efficiency nexus is proven to be positive, while for the micro-small firm group, smaller is more efficient than larger. It is also implied that under the appropriate segmentation framework, the size–efficiency nexus has been confirmed as not monotonic in nature and depending on firm size.