Liquidity and size effects on the Johannesburg Stock Exchange (JSE)

ABSTRACT

This study explores the existence of a liquidity premium on the Johannesburg Stock Exchange (JSE) and its potential interaction with the well-documented size effect. It builds on the stream of South African literature that examines liquidity as a standalone factor and adds further weight to its existence. Over the 2000-2015 sample period, this study finds evidence of a significant liquidity effect. Importantly, the liquidity premium is found to be separate from the size effect. Furthermore, the liquidity premium captures a different underlying effect than the value premium. The efficacy of Liu’s (2006) Liquidity-Augmented Capital Asset Pricing Model (CAPM) as a useful asset-pricing model for the cross-section of returns on the JSE is examined and found to perform better than the Fama-French 3-Factor model.

KEYWORDS:

• Asset pricing
• Liquidity factor

JEL CODES:

• G12
• G14

ORCID

James Britten http://orcid.org/0000-0002-1323-1196

Notes

1 Liquidity should be regarded as multidimensional. Liu (2006) considers four dimensions of liquidity: (1) Trading cost; (2) Trading quantity; (3) Price impact; and (4) Trading speed. Liu’s (2006) constructed liquidity measure takes these aspects into account with a focus on the trading speed dimension. This measure forms a core focus of the analysis of this study.

2 An alternative explanation for this is that the JSE was highly concentrated with the top five securities comprising half the market cap of the JSE in total at the time of the van Rensburg and Robertson (2003) study.

3 The liquidity measures that Reisinger and van Heerden (2014) examine are the closing bid-ask spread, turnover, price impact, zeroreturn and number of positive zeroreturn days. Price impact is identical to the Return-to-Volume (RtV) measure used in this study. The number of positive zeroreturn days, which showed the strongest performance as a liquidity factor, is arguably the most similar proxy to Liu’ (2006) LM variable. However, the LM measure captures zero daily trading days and not simply the number of days where a share posts a zero return and can be considered a more complete measure of liquidity.

4 It should be noted that the original methodology makes use of daily data for determining price impact; thus the use of monthly determinants for this variable in this study may affect the power of the variable. The variable does, however, still capture the price impact of the data in that it captures the monthly return per rand value traded of a company and, as such, it is considered a viable liquidity variable for this study.

5 This study uses 12-month LM measure () as it was found to be the most robust (Liu, 2006).

6 Quartiles are used as opposed to quintiles or deciles in this study due to the small amount of stocks traded on the JSE relative to developed markets such as the NYSE.

7 It should be noted at this point that although the IL portfolio performed exceptionally over the period, it is doubtful that this portfolio could ever be replicated in practice. This is because the methodology of this study assumes that all shares can be bought every month in any quantity; however, with a portfolio as illiquid as the IL portfolio it may be difficult, if not impossible, to trade the constituents on a monthly basis and continually rebalance the portfolio.

8 Frazzini and Pedersen (2014) find that low beta shares outperform high beta shares. As there appears to be a relationship between beta and liquidity, for the sake of completeness in unreported results portfolios are sorted according to their pre-ranking betas so as to be sure that the liquidity premium is not the low beta anomaly in disguise. The pre-ranking betas are estimated using a rolling 36-month window excluding the most recent month. Even when shares are ranked according to their betas, liquidity, proxied by LM remains a significant factor.