Research Summary
Research Summary
Overcoming Large-N, Small-T Issues in Asset Pricing Tests
Description
The large-N, small-T (i.e. large cross-section, short time series) nature of our asset data presents serious estimation problems for empirical asset pricing. In response, the literature tests asset pricing models against 10-25 test assets or portfolios. A valid asset pricing model must hold for every possible test asset, yet we use only a few. Moreover, the restriction to only a few test assets makes parameter estimates terribly imprecise. I show how to implement a new estimator (see: "An Unlimited Moments GMM Estimator" below) that delivers both highly precise estimates and allows the use of the entire universe of test assets. Replication exercises using the new methodology overturn the results of several prominent papers.