Stock Selection Using Data Envelopment Analysis

Data Envelopment Analysis is a nonparametric algorithm for ranking the ability of decision making units (DMUs) to produce output given a set of inputs, i.e. the DMUs relative efficiencies. Data Envelopment Analysis uses the methods of linear programming to assign each DMU an efficiency score between zero and one by solving the linear dual of the fractional optimization of a ratio of virtual outputs to virtual inputs. DEA uses variable weights to present each DMU in the best light possible to avoid biases caused by prior selection of weights. It has been used to measure the efficiency of such varied classes of DMUs as hospitals, schools, banks, government services, and software programs. In our paper we describe our implementation of DEA in Matlab and explore the effectiveness of using DEA efficiency scores to pick a stock portfolio, based on prior work by Abad, Thore, and Laffarga.