Performance estimation when the distribution of inefficiency is unknown

• We compute estimates of inefficiency when its distribution is unknown in Stochastic Frontier Models (SFMs) and Data Envelopment Analysis (DEA). Our procedure based on the Fast Fourier Transform (FFT) utilizes empirical characteristic function. The new techniques perform well Monte Carlo experiments. They deliver reasonable results an application to large U.S. banks. Deconvolution DEA scores with FFT brings closer from SFMs. show how or performance one-sided error component unknown; we do same procedure, which (FFT), function residuals SFMs efficiency DEA. experiments In both cases, deconvolution much SFM.