Measuring performance in nonprofit organizations: A stochastic frontier approach

This study examined the applicability of stochastic frontier analysis (SFA) to measuring performance of nonprofit organizations. Performance evaluation in nonprofits is becoming increasingly important as a result of growth in the nonprofit sector, decentralization of government services, tightening funds for social services, and mounting demands for accountability. However, for-profit performance indicators are generally inappropriate when applied to nonprofits, given their multiplicity of services and programs, their lack of profit motive, and the difficulty of measuring outputs. Stochastic frontier analysis is robust econometric technique that uses regression analysis to estimate a conventional cost or production function, assessing technical efficiency as a measure of organizational performance by estimating a best-practice model.; Using the Battese and Coelli (1995) model, data was analyzed from 990 Tax Returns for all active U.S. nonprofits with revenues over {dollar}25,000 in fiscal year 2003. The computer program FRONTIER Version 4.1, which can accommodate several different cost and production models, was used to calculate in a single step an efficiency score, the inefficiency parameters for each external variable and the log-likelihood function of the model. For each nonprofit organization, the technical-efficiency score and its level of significance were estimated using the explanatory variables Major-12 NTEE classification, age, size, ownership category (i.e., public or private), legal organization type (i.e., corporation, trust, etc.), whether it received government funding, and the level of its officers' compensation. In addition, four model-estimation tests were conducted to measure whether the model selected provided an appropriate representation of nonprofits' inefficiency: the truncated translog model, the Cobb-Douglas model, the half normal translog model, and the truncated translog model with explanatory variables.; Results indicated that the truncated translog production model with explanatory variables is an appropriate specification to measure nonprofits' technical inefficiency; that the simpler production models of Cobb-Douglas and half normal translog could provide similar measurements; and that all explanatory variables had significant effect on technical efficiency scores. The "size" variable was found to have the greatest influence on technical efficiency and a positive effect on the performance of nonprofit organizations. Larger nonprofit organizations across all Major-12 NTEE classifications were found to be more technically efficient.