Modeling leakage in two-stage DEA models: An application to US mutual fund families

This paper proposes a two-stage DEA model with leakage variables at stage 1 for assessing relative performance of decision making units. We refer to the output variables at the first stage that leave the two-stage system without entering the second stage as leakage variables. In addition to the leakage variables, the proposed model can handle multiple input and output variables at both stages and multiple intermediate variables. The concept of leakage variable adds a new dimension to two-stage DEA modeling. The applicability of the proposed model is demonstrated by assessing the performance of a sample of the US mutual fund families over the period 1999–2008 with operational management and portfolio management processes as the two stages of mutual fund operation. We consider total cash flow to investors (TCF) as the leakage variable. The results reveal that, over the sample period, modeling TCF increases discriminatory power of overall performance considerably. Moreover, we find consistent evidence over the sample period that small fund families are more likely to perform better than large fund families. This is not observed when TCF is not modeled as a leakage variable.