Since the 1990s, efficiency and benchmarking analysis has increasingly been used in network utilities research and regulation. A recurrent concern is the effect of environmental factors that are beyond the influence of firms (observable heterogeneity) and factors that are not identifiable (unobserved heterogeneity) on measured cost and quality performance of firms. This paper analyses the effect of geographic and weather factors and unobserved heterogeneity on a set of 128 Norwegian electricity distribution utilities for the 2001-2004 period. We utilize data on almost 100 geographic and weather variables to identify real economic inefficiency while controlling for observable and unobserved heterogeneity. We use the factor analysis technique to reduce the number of environmental factors into few composite variables and to avoid the problem of multicollinearity. We then estimate the established stochastic frontier models of Battese and Coelli (1992; 1995) and the recent true fixed effects models of Greene (2004; 2005) without and with environmental variables. In the former models some composite environmental variables have a significant effect on the performance of utilities. These effects vanish in the true fixed effects models. However, the latter models capture the entire unobserved heterogeneity and therefore show significantly higher average efficiency scores.
Efficiency Quality of service Input distance function Stochastic frontier analysis