Investigating New Multifactor Models with a Conditional Dual-Beta : Can a Conditional Dual-Beta in the Market Factor add Explanatory Value in New Multifactor Models? A study of the Swedish Stock Market between 2003 and 2015

Abstract: This thesis investigates pricing-performance of two recently developed multifactor asset-pricing models with the implementation of dual-betas dependent upon prevailing market-conditions. The models included in the study are the Fama and French five-factor model and the Q-factor model by Hou, Xue and Zhang. We test the models on cross-sectional Swedish stock-market data between 2003 and 2015 from the Large-, Mid- and Small Cap-lists and their respective precursors. The models are tested in their ability to explain portfolios sorted on firm beta-values, on a twelve-year period as well as a six-year period characterized by changing market directions and high market volatility.  In our study, we support the presence of changing risk-return relationship in up and down market states by estimating separate market betas with the risk-free rate as threshold. However, we do not find the isolated and volatile period to give rise to a larger difference in the up and down market betas. We consistently find the models to have a decreasing explanatory power on the portfolios of firms with lower beta values. We also find the largest difference in the up and down market betas occurring in the low beta portfolios, suggesting that this is causing measurement problems in the models. While making the models conditional, the measurement problem with the static beta seems to be reduced for the portfolios where the difference between up and down betas differ most. In the applied context, we conclude the conditional dual beta adds explanatory power in the models when the market beta differs in up and down market states.  The insights of this thesis support the method of making the market-beta conditional as suggested by Pettengill, Sundaram & Mathur (Pettengill, et al., 1995), in new multifactor models.