مقاله سال ۲۰۱۶

It is well documented that stock returns are affected by firm size (Fama and French, 1992), book-to-market (Fama and French, 1993), momentum (Jegadeesh and Titman, 1993; Moskowitz and Grinblatt, 1999; George and Hwang, 2004), volatility (Goyal and Santa-Clara, 2003; Ang et al., 2006) and liquidity (Amihud, 2002; Acharya and Pedersen, 2005; Liu, 2006). However, there is no consensus in the literature as to which one of these characteristics best explains returns. Asness et al. (2013) find that momentum and value characteristics significantly influence asset prices across countries and across asset classes. Liu (2006) contend that the illiquidity characteristic for stock trading discontinuity is able to subsume the book-to-market effect. Foran et al. (2014, 2015) show that the liquidity risk premium is reduced when asset pricing models incorporate the momentum factor. Cotter et al. (2015) find that the idiosyncratic volatility characteristic has been the most important priced factor during the recent UK economic downturn. Fama and French (2015) and Hou et al. (2015) find that the book-to-market and momentum factors become unimportant in explaining returns after controlling for the investment and profitability factors. The investigation of the explanatory power of stock characteristics is not only of empirical value for investors in selecting stocks,1 but is also of theoretical value for academics in understanding the relative importance in priced factors.2

This study employs a novel econometric approach, the semiparametric characteristic-based factor model developed by Connor et al. (2012) (CHL hereafter), to evaluate the relative power of stock characteristics to explain returns on the London Stock Exchange (LSE). The widely adopted approach in multi-factor asset pricing models relies on sorting each variable by a predetermined cut-off rate to construct characteristic portfolios which are used to generate factor returns for a given stock characteristic (Fama and French, 1992, 1993, 1995). When more than three characteristics are included in asset pricing models, the total number of characteristic portfolios will increase substantially causing higher correlations between factor returns due to poor diversification of the portfolios (e.g. Fama and French, 2015). Using the traditional sorting method to evaluate the relative explanatory power of a large number of stock characteristics is therefore empirically challenging. The challenge is even greater in a country with fewer listed stocks than in U.S markets. Lee (2011) reports that around 3000 stocks were listed in U.S markets from 1988 to 2008 while the UK market is the largest European market with only one third of this number of stocks. Given that U.S. markets have the largest number of stocks in the world, it can be argued that the characteristic sorting approach is still able to estimate factors there.3 However, the CHL method which does not rely on characteristic sorting to obtain factors is more suitable for the UK market and other developed and emerging markets which have a smaller number of stocks than in U.S. Thus, the CHL method is an important novel methodology for assessing relevant investment risk in international financial markets outside U.S. Moreover, multinational institutions and investors can use this approach to identify regional factors in order to diversify their portfolios.

Fama and French (2015, p. 19) note that “the most serious problems of asset pricing models are in small stocks”. It is therefore important to construct size portfolios that accurately reflect the difference in market capitalisation between small and large firms. However, when a market contains many small stocks and few large stocks the identification of the size portfolios becomes difficult. With 935 stocks on the LSE main board as in August 2014, the FTSE 100 index including 100 largest UK domestic stocks represents 84% of total market capitalisation, indicating that there are a large number of small stocks in the UK market. A 50% cut-off rate to sort firm size according to Fama and French (1993, 1995) can underestimate the large size portfolio’s market capitalisation leading to a downward bias in the size factor return. One possible solution is to use a finer sorting on firm size such as a 25% or 20% cut-off rate. Since the size portfolios have to interact with other characteristic portfolios, the finer sorting can significantly increase the total number of characteristic portfolios, some of which will not be well diversified in markets with fewer stocks. The CHL approach overcomes this empirical difficulty because it uses stocks’ own market capitalisation to estimate the size factor return. Therefore, the UK market is ideally suited to the application of the CHL method to obtain multiple characteristic factors.