Measuring cost efficiency in presence of heteroskedasticity: The case of the banking industry in Taiwan

We measure cost efficiency of commercial banks in Taiwan using annual data for the 1996-2011 period. Following Caudill et al. (1995), and Caudill and Ford (1993), we develop a generalized dynamic model that accounts for heteroskedasticity in the cost inefficiency component of the error term in the stochastic frontier (SF) model and adopt a single-step approach (e.g., Battese and Coelli, 1995; Wang and Schmidt, 2002) to circumvent the problems associated with the two-step estimation procedure used by prior studies. We show that the results based on the heteroskedastic stochastic frontier (HSF) model combined with the single-step estimation approach sharply contrast with that of the traditional 'two-step' approach. These are important findings because they point to the limitations of previous studies on cost efficiency that use a two-step estimation approach. Our findings suggest that asset size, loan quality, and bank equity-to-asset ratio have significant negative impact on cost efficiency of banks. Government-owned banks in Taiwan are more likely to be cost efficient than privately owned banks. Macroeconomic environmental factors such as money supply and foreign exchange reserve ratio are likely to influence the cost efficiency of banking industry in Taiwan.