Title: Loosening the chains on multi factor asset pricing models: Letting time varying betas speak
Authors: Fabio Calonaci - Queen Mary University of London (United Kingdom) [presenting]
George Kapetanios - Kings College, University of London (United Kingdom)
Richard Baillie - Michigan State University (United States)
Abstract: Estimates of the Fama MacBeth version of the 3 Factor model are derived by implementing some recent, locally smoothed, time varying parameter methods which are non parametric. Importantly, the approach selects the bandwidth on the basis of cross validation of estimating the 3 Factor model. Another novelty of the approach allows computation of time variation in the factor loadings, and avoids heavy theoretical restrictions on the model. Once we make these allowances on the specification of the 3 Factor model, we see there is a new level of explanatory power in the model both in terms of time series and cross sectional variation in the expected returns. There are also further insights into the asset pricing anomalies such as size, value and momentum.