This paper examines the properties of the X-inefficiencies in U.S. bank holding companies derived from both stochastic and linear programming frontiers. This examination allows the robustness of results across methods to be compared. While we find that calculated programming inefficiency scores are two to three times larger than those estimated using a stochastic frontier, the patterns of the scores across banks and time are similar, and there is a relatively high correlation of the rankings of banks' efficiencies under the two methods. However, when we examine the "informativeness" of the efficiency measured by the two different techniques, we find some large differences. We find evidence that the stochastic frontier scores are more closely related to risk-taking behavior, managerial competence, and bank stock returns. Based on these findings, we conclude that while both methods produce informative efficiency scores, for this data set decision makers should put more weight on the stochastic frontier efficiency estimates.