The Australia Asset-Pricing Debate. Durand et al (2016)

The authors want to explore which of the commonly used models best explains the cross section of returns of the Australian stock market.  The commonly used models in the US are the Fama-French 3-factor model, which includes market, size, and value factors.  Another model is the Fama-French-Carhart model, which adds momentum.  They also want to explore how liquidity and the US market influence the returns of the Australian stock market.

First, by using the 4-factor model, they find that the Australian value and momentum factors significantly explain the returns; however, none of the US factors are statistically significant in explaining the returns.  Although the authors do find a high correlation between the US factors and the AUS factors; most notably, the market factor of the US is 0.72 correlated with the AUS market factors.

To dig deeper into the influence of the US market, the authors create another model that includes the 4-factor model for the US and the 4-factor model for the AUS (orthogonalized) into a single model.  They then compare the 3-factor, AUS 4-factor, combined model, and combined+liquidity factor model.  They find that the alphas of each of the models is statistically unlikely to be zero; therefore, we could conclude that none of the models' factors fully explain the cross section of returns in the Australian stock market.  Although the R^2 of the combined model is 0.76 and provides the best fit when compared to the other models.

Next, the authors explored the factor coefficients and t-statistics in each of the 25 portfolios (created in line with the Fama-French method) for the combined model.  They find that approximately half of the portfolios have statistically significant alphas; in particular, the largest capitalization and highest book-to-market portfolios have insignificant alphas and might pass the Merton zero-alpha test.  The AUS market-beta is significant for all portfolios, the AUS size beta is significant for most portfolios, the AUS value factor is significant for only the high book-to-market portfolios, and the AUS momentum does not explain returns across many of the portfolios.  The US market, size, and value factors significantly explain the returns of most of the portfolios, although the US momentum factor does not. 

Next, the authors added a liquidity factor to the combined model.  They find that the coefficient on the liquidity factor was not statistically significant for most of the portfolios.

Next, they perform a Wald test for the US factors, which measures whether each of the factors is jointly significant in each of the 25 portfolios.  They find that all the US factors (i.e., market, size, value, and momentum) significantly explain the returns of the AUS stock market; however, they find that the liquidity factor does not add much explanatory power to the model.

Finally, the authors sorted the portfolios by the Brailsford portfolio/factor construction method (which is a bit different than the Fama-French method used before); they find the US 4-factor model is the only one with an insignificant GRS, so it seems to be the best at explaining the returns of the AUS stock market.

Durand, R. B., Limkriangkrai, M., Chai, D., & Gallagher, D. (2016). The Australian asset-pricing debate. Accounting & Finance, 56(2), 393–421