تاثیرات اندازه زمان متغیر و نقدینگی در بازارهای سهام جنوب آسیا: مطالعه سهام صنعت مرغوب

This paper contrasts the performance of the Capital Asset Pricing Model (CAPM) augmented by size and liquidity factors with its time varying coefficient counterpart, using a unique market universe compiled from constituent stocks of blue chip indices BSE-100 (India), KSE-30 (Pakistan), DSE-20 (Bangladesh) and Dow Jones Titans (Sri Lanka). The evidence suggests that substantial size and liquidity effects are present in all markets with the exception of Sri Lanka. Time varying liquidity beta profiles reveal that the financial sectors of all South Asian markets have been affected by the 2008 financial crisis with exception of Sri Lanka where the market is influenced by the prolonged civil war.

The enhancement of standard pricing models such as the Capital Asset Pricing Model (CAPM) with the addition of factors offering improved explanatory power over the cross section of stock returns has received considerable attention in the literature. While pricing theory states that the cross section of expected stock returns are related to the returns’ sensitivities to state variables, which themselves are linked to investor welfare, there is interest in the nature of the state variables themselves. Fama and French (1993) (henceforth FF) first proposed that variations in size, defined as the valuation differences between value and growth stocks, and variations in accounting book value and market value of stocks are two such candidates for state variables. Furthermore, supplementing the traditional CAPM with two additional returns-based factors representing these state variables provided improvements over the simple market-factor alone. More recently, liquidity has been proposed as a state variable with a range of methods cited for its measurement. Pastor and Stambaugh (2003) found evidence of increased support for a trading volume based liquidity factor augmenting the FF model, while Liu (2006) introduced a new trading speed measure designed to capture both traded turnover as well as frequency of trading as elements of liquidity. Furthermore, Liu (2006) found evidence that the addition of the single liquidity factor alone to the traditional CAPM generated increased explanatory power in excess of either the one factor CAPM or the FF model. However, there is a lack of evidence concerning the benefits of including both the FF size and book-to-market value factors in modelling the cross section of stock returns in Pakistan (Iqbal & Brooks, 2007) and India (Ameer, 2007). The presence of size effects is especially likely in emerging South Asian markets given the considerable dispersion of listings that are commonly from either larger internationally focussed firms or indigenous small and medium enterprises (SMEs), which are often controlled by dominant family groups (Athey & Laumas, 1994 and Manos et al., 2007). There is evidence of major differences in trading activity and liquidity within markets across the South Asian region. Poshokwale & Theobald, 2004 and Karmakar, 2010 cite differences in liquidity across sectors within the large Indian equity market while this is a pervasive issue in Sri Lanka (Elyasiani, Perera, & Puri, 1998), Pakistan (Iqbal & Brooks, 2007) and Bangladesh (Akhtaruddin, 2005). Consequently this study investigates whether size and liquidity effects are priced in these markets. The issue is whether differences in cross sectional expected returns can be better explained by including factors accounting for the differences in aggregate market-wide size and liquidity effects than simply the market factor of the traditional CAPM. Liquidity as a concept is very hard to define largely because its characteristics transcend a number of transactional properties of markets including tightness, depth, resilience (Lesmond, 2005) and information (O'Hara, 2003). The literature has traditionally been limited in only using constructs that capture only one dimension of a multidimensional phenomenon. This typically centres on variants of the bid-ask spread (quoted or effective) in Amihud and Mendelson (1986), the turnover measure of Datar, Naik, and Radcliffe (1998), or measures relating to the price impact arising from traded volume such as Amihud, 2002 and Pastor & Stambaugh, 2003. However, there is very little published research concerning measures capturing the trading speed dimension of liquidity, defined as the ability to transact large quantities quickly with little price impact (Liu, 2006 and Pastor & Stambaugh, 2003). Furthermore, there are serious concerns over the ability of existing one-dimensional constructs to fully capture liquidity risk and their inaccurate estimation of the dimension they are intended to model (Amihud, 2002 and Pastor & Stambaugh, 2003). Equally deficiencies in the application of the bid-ask spread measure have been highlighted in Lee (1993) where evidence reveals that many large trades occur outside the bid-ask spread while many small trades are undertaken within it, leading to potential bias. Further concerns over the application of one-dimensional measures focus on the fact that they are undefined in the presence of extremes of illiquidity, as is frequent in smaller regional markets (Lesmond, 2005). A more recent measure developed by Liu (2006) captures the trading speed dimension of liquidity, defined as the standardized turnover-adjusted number of zero trading volumes over the past 12 months. It is multi-dimensional and captures effects relating to trading speed, trading quantity and trading cost, with an emphasis on trading speed, defined as the continuity of trading and the potential delay in executing an order (Liu, 2006). An additional benefit from this measure arises is its robustness in the presence of significant illiquidity (Liu, 2006), again as is often present in emerging markets (Hearn & Piesse, 2009). The literature concerning the inclusion of liquidity as a priced state variable in a valuation framework is very recent. Pastor and Stambaugh (2003) find strong evidence from US stock data that market-wide liquidity is a priced state variable and that it should be positive. The study applied the innovations of a price impact measure of liquidity to sort stocks within a universe into decile portfolios with the market aggregate premium formed by the difference between returns of the highest and lowest liquidity deciles. The explanatory power from including this fourth fact was established by comparison with the Fama and French (1993) three factor model and the traditional CAPM. Stocks with higher sensitivity to aggregate liquidity stocks compensate investors with higher expected returns. Evidence is also found that small stocks have greater sensitivity to liquidity innovations than large stocks. Pastor and Stambaugh (2003) note that intuitively it could be expected that small and illiquid stocks are those most affected by market aggregate drops in liquidity and this causes investors to flee to assets with higher liquidity. However, their findings also show that size and liquidity are not the sole determinants of liquidity betas. This is reinforced by the argument that stocks with a high liquidity beta are not necessarily illiquid. Investor preferences when there are market aggregate falls in liquidity are also likely to focus on rival bond markets. In order to increase portfolio holdings in bonds investors may seek to sell liquid stocks in order to save on transactions costs. Consequently in this scenario the price reaction to aggregate liquidity changes is stronger for more liquid stocks. Equally, prices of liquid stocks could have greater sensitivity to aggregate liquidity shocks if such stocks are held in greater proportions within the portfolios of liquidity-conscious investors. Thus, Pastor and Stambaugh (2003) find little basis for liquidity betas to bear a simple relation to stock size and liquidity. Liu (2006) builds on this first using a new liquidity construct to estimate stock liquidity and then including this factor within a two factor augmented CAPM. While the additional liquidity factor offers strong performance in explaining the cross section of US stock returns the results contradict earlier findings of Pastor and Stambaugh as the liquidity premium alone incorporates anomalies such as size and the book-to-market effects in Fama and French (1993). The literature relating to South Asian stock exchanges typically focuses on the Indian equity market with some peripheral studies on Pakistan, Bangladesh and Sri Lanka. Research on the Indian market has evolved extensively over the last 15 years with studies relating to market microstructure and information transmission between large and small stocks (see Poshokwale & Theobald, 2004 and Karmakar, 2010) as well as a larger volume of studies focussing on the roles of family groups and control in Indian listed firms. The latter investigates capital structure decisions of Indian firms (see Chakraborty, 2010 and Manos et al., 2007) as well as the contrast between sourcing funds internally as opposed to banks or the stock market (see Athey & Laumas, 1994). The scant literature that focuses on the wider sub-continent is concerned with the dynamic linkages between the Sri Lankan equity market and other South East Asian nations (Elyasiani et al., 1998), corporate disclosure and informational content of stock prices in Bangladesh (Akhtaruddin, 2005) and the application of the CAPM in Pakistan (Iqbal & Brooks, 2007). However, any extension of the CAPM to include liquidity measurement or an application of this to industry portfolios across the wider South Asian regional markets is new. Consequently, the motivation here is to focus on the wider South Asian sub-region. A unique perspective on individual industry sectors is provided that justify the consideration of stocks that are constituent members of blue chip indices as these are most likely to satisfy the asset market integration and informational assumptions inherent in the CAPM and are of most interest to overseas investors. Consequently I make a contribution to the literature in extending the scope of the CAPM (see Sharpe, 1964 and Lintner, 1965 for detailed overview) to include additional size and liquidity effects while broadening the application to a regional level across South Asia. The majority of the valuation literature on pricing models assumes a time invariant relationship in the systemic risk of an asset. However, a separate literature addressing the time varying nature of systemic risk has evolved because of an increasing concern about the violation of assumptions inherent in the linear model, such as normality, identity and independence of stock returns (Grout & Zalewska, 2006). Pettengill et al., 1995 and Ho et al., 2006 studied the relationship between risk and return in “up” and “down” markets while Bekaert and Harvey (1995) undertook a similar study using Markov-switching regressions across a sample of emerging markets to examine differences between periods of integration with world markets and segmentation. More recently Watanabe and Watanabe (2008) incorporate a Markov-switching regime model to account for a time varying liquidity premium across a universe of US stocks. However Brooks, Faff, and McKenzie (1998) used time varying techniques based on the Kalman-filter approach applied Australian industry portfolios and found that these techniques produced improved in and out of sample performances than other econometric techniques. Grout and Zalewska (2006) find that the use of Kalman filter methods is preferable to Markov-switching regressions as it was not necessary to define the exact point of the switch (Grout & Zalewska, 2006). Instead any changes in the time path of betas can be assessed using regression results, which is particularly relevant in modelling liquidity effects in the presence of the fluctuation within emerging markets. Thus, following Brooks et al. (1998), this paper uses time varying techniques and the Kalman filter. The results show that aggregate size and liquidity effects are significant in all these South Asian markets with the exception of Sri Lanka. Similar results are found using the time varying techniques. The more illiquid smaller and less developed markets, namely Bangladesh and Sri Lanka, are reflected in the reduced levels of both significance of factors and overall explanatory power from both the one and three factor CAPM. Evidence from the in-sample profiles of the time varying liquidity betas reveals that while liquidity betas are largely centred on zero and insignificant for blue chip Indian stocks they have decreased considerably since 2002 for top tier Pakistan stocks. The liquidity beta time profile for Bangladesh is high and of variable significance while for Sri Lanka it is high and significant throughout. Given the 2008 global financial crisis the time-varying profiles of liquidity betas of financial stocks in India, Pakistan and Bangladesh show clear and significant upward trends. However, this is not the case in Sri Lanka where consistently high values suggest that a different process governing liquidity is likely in Colombo. Overall these results lend further support for the use of the mean-variance theory in an emerging market context. The paper is structured as follows. Section 2 reviews the institutional features of South Asian equity markets, describes the construction of the liquidity measures and provides descriptive statistics of the data used. Section 3 outlines the two modelling approaches: the size and liquidity augmented CAPM and its time varying parameter equivalent. Section 4 discusses the results and the final section concludes.

This study proposes a size and liquidity augmented capital asset pricing model to explain the cross section of expected returns in South Asian emerging stock markets, an area which has previously been largely excluded from empirical research in finance. The constituent stocks from the blue chip investor-orientated indices were used from the large well developed Indian stock market alongside Pakistan and the smaller less well developed Bangladesh and Sri Lankan markets. The performance of the size and liquidity augmented three-factor CAPM is contrasted with a time varying coefficient equivalent with similar results. While size effects alone drive the returns generating process in Sri Lanka, a combination of size and liquidity effects is relevant in Bangladesh and Pakistan. Size effects are prevalent across Indian industries too but the additional liquidity factor is relevant in only a few industries. The three-factor CAPM provides the best fit across Indian and then to a much lesser extent Pakistan industries while explanatory power is considerably lower for the less developed markets of Bangladesh and Sri Lanka. These differences are further reflected in cost of equity estimates, which are highest in Bangladesh and Sri Lanka before marginally decreasing to levels across Pakistan industries to the lowest levels in Indian industry sectors. The evidence from the time varying profiles of liquidity betas shows that while the financial sectors of India, Pakistan and Bangladesh have all been noticeably affected by the onset of the 2008 global financial crisis and recession, Sri Lanka has been largely unscathed owing primarily to the market being heavily influenced by the prolonged insurgency and civil war. However, while aggregate exposure to liquidity has remained the same across the Indian market it has decreased substantially across the sample period for Pakistan, while increasing steadily for the aggregate Sri Lankan market with a sharp drop recently following the conclusion of hostilities. The exposure of Bangladesh overall to aggregate market illiquidity shows substantial variation over the sample period. These results indicate that there is considerable segmentation between South Asia's stock markets with the smaller less well developed markets of Sri Lanka and Bangladesh as these are subject to very different influences from those prevalent in the largest and most international markets of Pakistan and India.