Incorporating contextual information in public sector efficiency analyses: A comparative study of NSW local government

Abstract

Using the planning and regulatory function of 173 NSW local governments, several approaches for incorporating contextual or non-discretionary inputs in data envelopment analysis (DEA) are compared. Non-discretionary inputs (or factors beyond managerial control) in this context include the population growth rate and distribution, the level of development and non-residential building activity, and the proportion of the population from a non-English speaking background. The approaches selected to incorporate these variables include discretionary inputs only, non-discretionary and discretionary inputs treated alike and differently, categorical inputs, "adjusted" DEA, and "endogenous" DEA. The results indicate that the efficiency scores of the five approaches that incorporated non-discretionary factors were significantly positively correlated. However, it was also established that the distributions of the efficiency scores and the number of councils assessed as perfectly technically efficient in the six approaches also varied significantly across the sample.