Microeconomics Seminar 2016-2017
joint with the Département de sciences économiques, Université de Montréal
room C-6149 (U. de Montréal, Pavillon Lionel-Groulx, 3150, rue Jean-Brillant)
This paper proposes a nonparametric framework to test and estimate the classic exponential discounting, time separable consumer model in an environment with noisy data caused by the presence of measurement error. Our methodology extends the dynamic revealed preferences framework to make it applicable to survey data where measurement error causes a loss of distributional information that prevents us from applying the revealed preferences tools at the individual level. We combine the strengths of the revealed preferences approach to avoid making arbitrary parametric assumptions on the shape of the utility function, and we use a latent variable integration technique proposed in Schennach (2014) to deal with heterogeneity and measurement error without making strong distributional assumptions. We establish the asymptotic behavior of the estimator and prove the validity of inference using subsampling. Monte Carlo simulations show that an otherwise time-consistent consumer may often be mistakenly taken to be inconsistent in the presence of measurement error by the standard deterministic revealed preferences tests; our proposed methodology does not reject the exponential discounting model in such cases. We find support for exponential discounting behavior in a consumption panel survey for single-individual households, while rejecting the model for the case of couples. The second result is specially interesting, as we establish theoretically, that under measurement error, a version of the collective household exponential discounter model with intra-household preference heterogeneity presented in (2014) (that rationalizes the dataset ignoring measurement error) has the same implications that the standard exponential discounting model.