Evaluation of the StoNED method for benchmarking and regulation of Norwegian electricity distribution companies

We evaluate the StoNED methodology for benchmarking and regulation of network companies, and we compare StoNED to the two-stage DEA method currently used by the Norwegian regulator. We find that the estimated values for the skewness parameter in the second stage of the StoNED procedure can be inconsistent with the assumed positive skewness for the inefficiency term. Setting the skewness parameter to an arbitrary value can have significant consequences for the efficiency levels. This effect is partly neutralized by the revenue calibration performed by the NVE, depending on how the calibration is implemented. Our comparison of results from StoNED and two-stage DEA show that the efficiency scores from the two methods are highly correlated, but that the levels can differ significantly. We also interpret the StoNED coefficient estimates and compare them to the corresponding (dual) DEA estimates. Finally, we illustrate the robustness of the efficiency estimates to noise in data, as exemplified by noisy pension costs