The Italian water and sanitation services regulator, ARERA, has adopted a parametric model for assessing operators’ efficiency and set efficiency improvement targets accordingly. This approach is based on a Stochastic Frontier Approach; deviations of actual costs from the estimated standard cost are interpreted as signals of inefficiency and determine how demanding the efficiency improvement target will be in the next regulatory period. We critically assess the approach adopted by ARERA, exploiting an original dataset fed by the original data of regulatory accounts that water operators transmit to ARERA for biennial tariff reviews. This data is not publicly available and represents a novel contribution since it allows statistically significant estimates (more than 200 observations) from a detailed breakdown of information concerning costs and performance levels. Our analysis applies various alternative estimation models, completing the more traditional SFA based on the Pitt-Lee approach adopted by ARERA, with more sophisticated models aimed at overcoming the problems caused by multicollinearity and data heterogeneity. Finally, we estimate a non-parametric model based on Data Envelopment Analysis (DEA). This comparison leads to some surprising results: different models generate similar overall results, but the resulting ranking is different, and the explanatory potential is limited to dimensional variables, while quality indicators and contextual elements seem more challenging to capture. We conclude that
parametric models still seem far from the possibility of supporting the regulatory process, primarily when the industry is characterised by high heterogeneity of company size and operational environments. At the same time, the DEA is a promising, more flexible, and informative approach, but still unable to support regulatory decisions.
