Time: Mar. 17 (Thu.), 10:00 – 11:30 (Beijing Time)

Title:  Bayesian Inversion of Demand Systems

About the Speaker:

Zhentong Lu is a Principal Researcher in the Financial Stability Department of the Bank of Canada. Before joining the Bank in 2019, he was an Assistant Professor in the Shanghai University of Finance and Economics, after obtaining a PhD in economics from University of Wisconsin-Madison in 2015. His main research areas are empirical IO, applied econometrics and payment economics. He has published academic papers in Journal of Econometrics, Journal of Law and Economics, China Economic Review, etc, and policy reports on important issues for central banks.

Abstract:

In this paper, I show that the “Bayesian inversion” of demand, which is defined as the posterior distribution of random utilities given realized choices, can replace Berry, Levinsohn, and Pakes (1995)’s “mean utility inversion” and radically simplify the estimation of discrete choice models with aggregate data from many markets. A striking implication is that the logit model, which is often criticized for its oversimplification, yields a consistent and robust estimator for the means of random coefficients (RCs) in a RC logit model. Building on this insight, I propose a simple sequential estimator for the parameters in the mean and variance of random utilities and establish its consistency and asymptotic normality under primitive conditions. The estimator is free from common parametric assumptions on the distribution of random utilities, e.g., normal, and is in closed-form for workhorse models. Results from Monte Carlo experiments and an empirical application support the theoretical findings and demonstrate the effectiveness of the proposed approach.