Home > Events > 2014 Seminars & Colloquia > JOHN LIECHTY - Penn State University

JOHN LIECHTY - Penn State University

Main Content

Which Firm Characteristics Predict Stock Returns and When? A Hierarchical Bayesian Variable Selection Approach
16 October 2014 from 4:00 PM to 5:00 PM
201 Thomas Bldg.
Add event to calendar

A large number of firm level characteristics -- {\em e.g.} the Fama French predictors size, momentum and book-to-market -- have been found to predict stock returns.  Increasingly there is evidence that the impact of these predictors changes over time and that these dynamics relate to a wide range of macro level variables.  Most firm characteristics and macro variables have been tested one at a time or in small combinations, leaving researchers unclear as to which really matter.  We address this by introducing a hierarchical, variable selection model where inclusion and impact of firm characteristics vary over time and are driven by subset of macro variables.  Analysis of a range of proposed predictors provides a parsimonious view of what drives stock returns and offers insights into ongoing debates in the literature.  We find that almost all firm level characteristics matter, but that they do not matter all of the time, {\em e.g.}  momentum is rarely included (on the order of 6\% of the time) and when included it's slope is relatively small.  With regards to macro predictors, we find that impact is primarily driven by dynamics in the credit markets and that inclusion is primarily driven by dynamics in the business cycle, with firm characteristics more likely to be included during periods of market expansion; conversely we find that aggregate sentiment and psychological measures provide no additional explanatory power.  Interestingly these results are stronger for large firms than for small firms.  We also discuss implications of our findings for asset management and suggest areas of future research.

Navigation for this Section