Which Factors? International Evidence
MetadataShow full item record
This thesis investigates five major asset-pricing models recently proposed in the literature. Those models are Fama and French’s (2015) five-factor model and its six-factor extension adding a momentum factor (Carhart, 1994), the q-factor model of Hou, Xue, and Zhang (2015) and its five-factor extension (q5) pro-posed by Hou, Mo, Xue, and Zhang (2015), and the four-factor model of Stam-baugh and Yuan (2017). We apply these models to real-world stock market data from 88 developed and emerging markets and compare their performance. Our primary tool to evaluate their performance is the factor spanning test of Huberman and Kandel (1987) performed by Fama and French (2015) and Hou, Mo, Xue, and Zhang (2018), among others. The comparison aims to find a supe-rior model that should be applied when analyzing stock returns. We find that not all the factors inherent in these five asset-pricing models are significant in the international markets. Among the models, the most disap-pointing is that of Stambaugh and Yuan (2017) because none of its two new factors (MGMT and PERF) or even their aggregate (UMO) delivers a reliably positive average premium in the countries or groups of countries considered. For the other models, the factors based on value and performance did very well for almost all countries. For the value factor (HML), we cannot confirm Fama and French’s (2015) assertion that it is redundant in their five-factor model. Further, the value effect remains unexplained by the q and q5 models. The performance of the size factor is affected by the way it is measured. Fama and French’s size factor is significant for only two countries (China and South Korea). Stambaugh–Yuan’s size factor does better than Fama and French’s, as it is significant for Canada, India, Turkey, and across the emerging, non-US, and all markets. However, we reject their claim that eliminating stocks most likely to be mispriced allows delivering a higher size premium. The reason is that Hou, Xue, and Zhang’s (2015) size factor is positive and significant in all countries (excepted France) and for all the groups of countries considered. Overall, the Stambaugh–Yuan four-factor model failed to yield its promise in international markets over the last three decades. None of the Fama–French five-factor and six-factor models or the Hou–Xue–Zhang q-factor model and the Hou–Mo–Xue–Zhang q5 model spans the other models. Hence, we conclude that these new models are about the same as none of them stand out as the best.