An empirical investigation of the effectiveness and robustness of asset pricing models in Australia : A thesis submitted in partial fulfilment of the requirements for the Degree of Doctor of Philosophy at Lincoln University

Abstract

With a plethora of asset pricing models, it is hard to determine:  Which model is relatively better than other models in successfully explaining the returns of the capital assets in an equity market? and  if any of the previously documented asset pricing models can effectively perform during pre-crisis, crisis, and post-crisis periods and is robust to test-assets' (portfolios) specifications? The main objective of this research is to find answers to the above mentioned overarching questions in the Australian context, but the inquest into the effectiveness of asset pricing models comes with other challenges, e.g. the aggregational level of returns to analyse, test-portfolios’ style, sorting and level of diversification, and these discrepancies produce inconsistent results. The literature on asset pricing in the Australian context produces such inconsistent results that it is challenging to find a consensus, and the few instances where we find some level of consensus, in terms of superiority of one model over another model, it is hard to say if that superiority sustains across different sub-sets of test assets (portfolios) or different economic conditions (e.g. pre-crisis, during the crisis, and post-crisis periods). This study addresses the issue by examining the representativeness and robustness of the CAPM of Sharpe (1964), the three-factor model of Fama and French (1993), the four-factor model of Carhart (1997) five-factor model of Fama and French (2015) and the momentum augmented six-factor model across 73 varied style portfolios of different diversification levels during the periods of the pre-Dotcom crisis of 2001, during the Dotcom Crisis, the pre-Global Financial crisis of 2008-09, during the Global Financial crisis and the post Global Financial crisis. Following Barillas and Shanken (2016) we conducted the test on various subsets of asset returns to better identify a model’s effectiveness and robustness and, following Lewellen, Nagel, and Shanken (2010), we expanded the set of test portfolios beyond size-B/M portfolios and formed univariate and multi-variate style-based portfolios with a variety of portfolio sorting techniques and levels of diversification. We did not find evidence in support of a single model being effective and robust to the study period, e.g. sample period and sub-sample periods, and they are also affected by the style or characteristics of the portfolios, portfolio formation techniques and level of diversification. However, the three-factor model performed relatively better than other candidate models across single-sorted diversified portfolios. The GRS test of Gibbons, Ross, and Shanken (1989) rejects all candidate models in completely explaining the variations in the returns of less diversified or concentrated portfolios with an overall insignificant alpha. However, comparing the models in relative terms, there is no consensus of the superiority of any single model over other models, the four-factor and the five-factor models appear to perform relatively better with lowest absolute alpha Sharpe ratios for quad-sorted and quintile-sorted portfolios. It appears that from a practical point of view, these models may not be as successful in completely explaining the variations in the returns of poorly diversified or concentrated portfolios. Our results support the finding of Lewellen et al. (2010) that choice of test-portfolios significantly affects the power of asset pricing models and the findings of Lo (2004) that portfolio formation affects the effectiveness of asset pricing models in explaining returns and also provide evidence in support of Kan, Robotti, and Shanken (2013) that the results of the asset pricing tests depend upon the characteristics used in sorting stocks into portfolios.