Robustness of non-parametric measurement of efficiency and risk aversion: A simulation analysis with applications to agricultural banking and Ethiopian farming

Daniel M Settlage, Purdue University

Abstract

In this dissertation a non-parametric method for the measurement of firm efficiency and risk aversion is developed, tested, and applied. This method is a modification of data envelopment analysis (DEA) that allows the user to simultaneously determine the firm's efficiency score while taking the individual firm's aversion to risk into account. This risk-adjusted DEA approach also produces estimates of the Arrow/Pratt measure of risk aversion for each firm in the sample. A Monte Carlo framework is used to examine the performance of this method relative to traditional DEA methods under a diverse array of agent behavior. The accuracy of both the estimated efficiency scores and risk aversion scores from the model are computed and analyzed. The method is then applied empirically to agricultural banking and Ethiopian farming data sets. The Monte Carlo experiment performed in this paper simulates a wide variety of agent behavior. Firms are modeled as risk averse expected utility maximizers using the flexible expo-power utility form. Results from the Monte Carlo experiment indicate the risk-adjusted DEA method provides efficiency estimates that are far more accurate than traditional measures. Standard DEA is shown to drastically overestimate the degree of inefficiency present in a sample of risk averse agents, while this method does not. This method also appears to be a promising alternative to the difficult and time-consuming methods of utility elicitation that have previously been used to estimate agent risk aversion levels. Applying the method to a data set consisting of agricultural banks shows that the risk-adjusted DEA method provides much higher estimates of efficiency in this sample than do cost or standard profit efficiency tests. The risk aversion estimates from the model indicate that, in general, smaller banks are more risk averse than larger banks. The results from the Ethiopian farming application indicate that the risk-adjusted DEA method gives higher efficiency scores than standard DEA. The average profit efficiency level in the sample is 61 percent while the average risk-adjusted profit efficiency score is 86 percent. Based on the results of this application, about one third of the farms in the sample are classified as risk averse.

Degree

Ph.D.

Advisors

Preckel, Purdue University.

Subject Area

Agricultural economics|Banking

COinS