It is widely accepted that the presence of some of the firm’s attributes or characteristics attracting premiums in terms of average returns is pervasive and not restricted to a few individual markets. However, the way to derive these premiums by sorting firms based on their characteristics that are known associated with share returns is not without controversy. This chapter takes a different approach by adopting a novel Multi Self-Organising Maps to cluster shares first and then identify fundamental factors afterwards. It finds that firm’s size and book-to-market ratio attributes do have explanatory power over share average returns. There is also lack of evidence for other factors in explaining the share average returns.
Top1. Introduction
In recent years, it has become increasingly common to sort firms based on their attributes or characteristics, such as size, book-to-market ratio (B/M) and dividends yield (yields), to form empirical asset pricing portfolios. Investigators then use them to examine whether asset pricing models can explain the dispersion in firm’s share returns. For instance, after Fama and French (Fama and French (1993, 1995, 1996)) show that their three-factor model can explain more than 90% of the returns of these portfolios and that the unexplained portion of returns is economically small, subsequently, a large number of asset-pricing tests have used portfolios sorted on both size and B/M. Other firm characteristics used to form portfolios include return standard deviation, profit margin, liquidity, relative past return performance, profit margin, earnings-price ratio, and firm industry (Campbell et al (2008), Chava and Jarrow (2004), Fama and French (1988) and Zmijewski (1984)).
Forming portfolios based on characteristics has some advantages. For instance, it can generate a large dispersion in returns. However, it subsequently presents a challenge to any asset pricing model and has sparked a debate about whether this practice is appropriate. Lo and MacKinlay (1990) point out that the ad-hoc nature of the sorting technique - sorting on characteristics that are known to be correlated with returns generates a data snooping bias. Berk (2000) suggests that the characteristics that researchers rely on to share sorting may be mechanically linked to share returns. In addition, Conrad et al. (2003) demonstrate that the increasing popularity for researchers to sort stocks on multiple characteristics, and consequently to form larger number of portfolios, exacerbates the data-snooping bias. Such practice results in the return dispersion weakening or even disappearing in out-of-sample tests because the relation between returns and the characteristic is not robust over time.
Apart from the data-snooping concern, Daniel and Titman (2005) suggest that firm characteristics, such as B/M, serve as a “catch-all” factor. These characteristics capture the differences in the sensitivities of stocks’ returns to a number of different fundamental factors. Consequently, factors based on firms’ characteristics will be bound to have an explanatory power over share returns in the empirical testings using characteristics sorted portfolios. However, these testings are unable to tell us whether other factors, perhaps more fundamental ones can be equally important variables in explaining share returns. Stein (1996) suggests that “the return differentials associated with the book-to-market ratio and other predictive variables be thought of as compensation for fundamental risk. While there seems to be fairly widespread agreement that variables such as book-to-market do indeed have predictive content, it is much less clear that this reflects anything to do with risk.” Therefore, one cannot rely on the prevailing sorting practice in which shares sorted according to firms’ characteristics are used to identify fundamental factors in empirical asset pricing. It is primarily this criticism that motivates this study.
In this paper, we propose a novel computational approach to identify fundamental factors that capture the sensitivities of stocks’ returns. Such approach is well motivated economically, and more importantly, alleviates some of the problems inherent in the prevailing asset pricing practice.
This study makes the following contributions to the finance literature. Firstly, it overcomes the ad-hoc nature of sorting shares into portfolios based on their known characteristics. Secondly, it validates the well known fundamentals, such as size and B/M, in predicting share returns.
The remainder of the paper proceeds as follows. Section 2 describes the proposed methodology to be used in the study. Subsequently, Section 3 delineates the data. Section 4 reports the modelling findings. Section 5 proposes future research direction and Section 6 concludes.