The impact specifications have on efficiency scores in SFA and DEA models: The case of District Courts in Norway

Abstract

Abstract When measuring efficiency, the assumption made about production technology and the accuracy of the efficiency scores are of great importance. The comparison between SFA and DEA has become increasingly popular among authors. The empirical implication model specifications have on the potential for efficiency among the Norwegian District Courts is less studied. This paper examines what implications assumption made about production technology have on efficiency scores. We find that model specification is of great significance for the results of efficiency potential among District Courts in Norway. Different model specifications is tested using parametric Stochastic Front Analysis (SFA)), non-parametric Data Envelopment Analysis (DEA) and Free Disposable Hull (FDH). The mean efficiency potential varies as much as 19 percent points between the most extreme models tested. We find that DEA-models in general return lower efficiency scores, indicating a bigger efficiency potential compare with SFA-models. As expected the FDH-model show less potential of efficiency. Efficiency comparisons across models are based on cross sectional data on 61 district courts in Norway from 2018. Regulators or performance auditors often favour models with large production technology possibilities e.g. constant returns to scale. Under the property of constant returns to scale we find that DEA models are more likely to overestimates the efficiency potential compare to SFA models, also when corrected for bias. To be empirical relevant for policy makers one should include a variety of different advanced benchmarking models. As stakeholders will have different preferences for technology, a one-dimensional approach (either SFA or DEA) is not enough to make an empirical fundament for policy making to improve performance in the public sector.