Returns Regimes Based Factors in the Cross-Section of Average Stock Returns
Student thesis: Doctoral Thesis
Related Research Unit(s)
The empirical work documented in this thesis augments the FF5 factor pricing model to a six factor pricing model. Dozens of potential candidate factors are constructed using Markov regime switching models (MS) to estimate the states of a stock’s returns and those of a stock market index, and constructs characteristic measures based on them. The factors are constructed as long-short portfolios of stocks. Besides the construction of the characteristic measures, the adopted research methods are standard and widely used in the literature. The factors investigated belong to the technical analysis factor type. In particular, this is the first research to use Markov Switching models in the context of factor construction, and the first research to exploit the concordance of statistical regimes and stochastic dominance as characteristic measures. Two interesting factors have been discovered: one version of the factor based on stochastic dominance is capable of improving the pricing precision of the cross-section of average stock returns, the other version based on concordance of regimes is characterized by a significant large risk premium, making it exploitable for investment strategies. Moreover, the second version has a clear interpretation that links it directly to the industrial sector and financial market cycles. The discovered factors are also robust against the concerns raised recently with respect to empirical factor models literature, in fact: the significant risk premium has been tested using multiple testing p-hacking thresholds; factors have been constructed using breakpoints controlling for microcaps issue; the factors are value-weights to avoid the impact of small stocks; finally the factor is constructed following a true out-of-sample construction procedure.
- asset pricing, factors model, stock returns, Econometrics