In a seminal paper using the DEA and SFA, Berger and Humprey (1997) provide a survey of 130 studies covering 21 countries. They conclude that (i) non-parametric measures of efficiency such as DEA yield a mean efficiency statistic slightly lower with a large dispersion while (ii) the parametric measures such as SFA provide a slightly higher efficiency statistics with lower dispersion. For financial firms, the means in that study are respectively 0.72 and 0.84. For a reference to different concepts of efficiency, see Berger and Mester (1997). They concluded that, although different measures add some valuable information, profit efficiency measures are not positively correlated with cost efficiency measures and so suggested support as a reason that this may be due to different type of optimization measures.