Explaining the cross-section of share returns in South Africa using macroeconomic factor models.

Abstract

Understanding asset prices is critical for the decision-making of many; from professional and individual investors, who seek to earn the highest possible return from their investments, to governments and corporates evaluating investment and consumption choices. Given that the behaviour of asset prices may differ across countries, especially across varying levels of development, applying knowledge of the determinants of asset prices from one country to another may not be appropriate. Asset pricing models can typically be grouped into one of two categories – portfolio or macroeconomic. The principle focus of this study is on models which fall under the latter grouping, where little research has been conducted on the South African market. These models are concerned with identifying the true risk factors which drive share returns, in contrast to portfolio-based models, which simply measure risk as the sensitivity of a share’s returns to portfolios of securities. The consumption-based capital asset pricing model (CAPM), which links consumption to investor behaviour in their demand for securities, provides the foundation for the majority of the macroeconomic models. Labour income and household wealth are seen as two critical measures that are linked to the consumption decisions of investors and several models which have incorporated these two factors are evaluated in this study. In particular, those of Lettau and Ludvigson (2001b), Piazzesi, Schneider, and Tuzel (2003, 2007), Lustig and van Nieuwerburgh (2005), Santos and Veronesi (2006) and Yogo (2006) are examined to assess their ability to explain the size and value anomalies on the Johannesburg Stock Exchange (JSE). The results are compared to several portfolio-based models including the CAPM, the conditional CAPM and the Fama and French (1993) three-factor model. The models are tested over the period June 1990 to April 2013 using a comprehensive sample of JSE-listed shares based on the Fama and MacBeth (1973) and generalised method of moments methods. The study finds that many of the macroeconomic models are less successful in explaining returns of South African shares compared to the developed markets which have been examined internationally. However, there is weak evidence to suggest that returns are correlated with factors which capture how investors’ returns vary with labour income, housing wealth and consumption. In particular, value shares earn higher returns than growth shares partly to compensate investors for greater risk in the macroeconomy where risk is captured by the interaction of consumption, asset wealth and labour income, while small shares are more sensitive to shocks in housing scarcity thus partially accounting for their higher returns compared to larger shares. The results of this study are analysed in conjunction with the international evidence so as to consider possible reasons for the weaker results obtained and the implications for understanding the factors that drive assets prices are reviewed. Finally, suggestions for future research are provided.