Lacking data on development and implementation costs for input-saving technologies, we develop a nonparametric procedure to estimate relative differences in marginal technology costs for technical change to be consistent with the induced innovation hypothesis. We apply it to U.S. agriculture.