A1456
Title: Asset pricing with missing data
Authors: Markus Pelger - Stanford University (United States) [presenting]
Svetlana Bryzgalova - London Business School (United Kingdom)
Martin Lettau - UC Berkeley (United States)
Sven Lerner - Stanford University (United States)
Abstract: The aim is to show how to impute missing information for stock returns and to study the implications for asset pricing relative to the current standard of using only observed or ad-hoc imputed values. Missing data in firm characteristics is a prevalent problem. As firm characteristics are not missing at random, using only observed data for building or evaluating asset pricing model results in biased estimates. We exploit the dependency of firm-characteristics in their time and in the cross-sectional dimension to impute the missing values for a large dimensional cross-section. Our imputed values are a substantial improvement relative to ah-hoc procedures as for example simple cross-sectional averages or past observations. In a large scale empirical analysis we study the asset pricing implications.
