Abstract:
Regulatory authorities in the European electricity sector use benchmarking techniques to determine the cost-efficient production level for an incentive regulation of distribution system operators (DSOs). With nearly 900 DSOs operating in the German electricity sector, of which 200 subject to incentive regulation, the issue of heterogeneity of DSOs has to be addressed. Using publicly available data of 133 Norwegian DSOs and replicating the model employed by the German regulator (who refuses access to the data), I show its assumption of homogeneous technology cannot be maintained. Quantile regressions (QR) across the cost distribution reveal heterogeneity in the coefficients of the explanatory variables, resulting in biased efficiency scores derived from stochastic frontier analysis. To correct for this heterogeneity in coefficients, I propose a Bayesian estimation of a more flexible SFA with latent classes for selected parameters that reflect variation in technologies. This estimation has better goodness of fit, reduced variance of all coefficients, and higher efficiency scores for nearly all DSOs, compared to the conventional alternative.