We show that observed choices in discounting experiments are consistent with roughly one-half of the subjects using exponential discounting and one-half using quasi-hyperbolic discounting. We characterize the latent data generating process using a mixture model which allows different subjects to...