Liquidity, liquidity risk and stock returns the evidence of viet nam

The common results are a negative relationship between these two variables as investors demand for higher returns to compensate for higher stock volatility.. In other words, we document

1

Liquidity, liquidity risk and Stock Returns – Evidence from Vietnam

Xuan Vinh Vo & Hong Thu Bui

Abstract

The question of whether liquidity is priced is a subject for a huge volume of papers in the asset pricing literature The common results are a negative relationship between these two variables as investors demand for higher returns to compensate for higher stock volatility This paper investigates the relationship between liquidity and stock return in Vietnam by employing an updated dataset of market and financial data of listed companies in

Ho Chi Minh City stock exchange ranging from 2007 to 2012 Our results are proving the reverse In other words, we document a reliable positive relationship between liquidity measures and stock returns and negative relationship between illiquidity measures and stock returns We also confirm the results by controlling for many frequently used factors determining stock returns which are well documented in the literature However, we do not find evidence in supporting the relation between risk associated with fluctuation in liquidity and stock returns

Keywords: liquidity, stock return, Vietnam

GEL Classification Code: G11, G12, G14

2

Liquidity, liquidity risk and Stock Returns – Evidence from Vietnam

1 Introduction

One of the very important primary functions of capital markets is the efficient pricing of asset Sharpe (1964), Lintner (1969) and Black (1972) develop the asset-pricing model and shape the way academics and practitioners think about risk and average returns However, many researchers find evidences that those returns on stocks display only little relation to those of CAPM's projection For example, Banz (1981) demonstrates that firms’ market equities could be another explanatory factor to explain returns in addition to the market’s return When examining the size-return relationship, the writer conjectures that most of the investors are not interested in small firms’ stock due to insufficient information; therefore, it brings higher rewards to shareholders who invested in Another prominent critics of CAPM is Chan et al (1991), in which they find evidence in Japan stock market in support to the role of book-to-market equity in explaining the cross-section average returns Continuing this stream of theory, after investigating the roles of other factors in explaining average stock return, Fama & French (1992) claim that besides the market’s return, size and book-to-market play strong roles and diminish the impact of two other factors, leverage and E/P in explaining average return Furthermore, Fama & French (1993) propose another asset pricing model which augments the CAPM model with excess returns of small caps over big caps and of value stocks over growth stocks

Although being supported by more empirical evidence (Fama & French 1995, 1998), the economic fundamentals of the additional factors to CAPM model are not reasonably clear It is for that reason, several studies do not reach the same agreement (Kim 1995; Kothari et al 1995; Y Peter Chung et al 2006) While Fama & French (1993) document that factor loadings explain stock return, Daniel & Titman (1997) advocate that stock characteristics such as size and book-to-market, not factor loadings, explain stock returns when they provide some significant in-sample evidence that stock returns do not co-move with the differences of returns between neither small size and large size portfolios nor high and low book-to-market portfolios

It is quite reasonable to expect that liquidity is an important variable for asset pricing As defined by Amihud and Mendelson (2008), liquidity is the capacity of the assets that can be traded quickly and at low cost Amihud

& Mendelson (1986) is one of the very first papers that pioneer research on the relation between the liquidity and stock return Using the bid-ask spread as the proxy for measuring liquidity, they point out that investors requires additional liquidity premium for holding illiquid stocks Motivated by this observation, many subsequent studies continue to focus on addressing return-liquidity linkage Following the approach of Amihud and Mendelson(1986) with updated data, Eleswarapu & Reinganum (1993) address the question of this relation when they find that the liquidity premium is a seasonal effect as it is reliably positive during the month of January Nevertheless, Brennan & Subrahmanyam (1996) find strong evidence consistent with the notion of a premium for illiquidity as their findings are stable over time and robust to controlling for Fama and French (1993)’s risk factors

In keeping with the literature, however, Petersen & Fialkowski (1994) and Brennan & Subrahmanyam (1996) criticize that the quoted spread is a poor proxy for liquidity as smaller equity trades are often executed inside

the quoted prices, while larger trades often face prices far inferior to those quoted As such, many alternative proxies for liquidity have been used for further exploring the association between asset returns and liquidity, such as trading volume (Brennan et al 1998), turnover ratio (Chan & Faff 2005; Datar et al 1998), zero return (Lesmond et al 1999), price impact of trading (Breen et al 2002), Amihud’s illiquidity ratio (Amihud 2002), the Pastor & Stambaugh (2003) liquidity measure, and the Liu (2006) liquidity ratio Most of these papers support Amihud and Mendelson(1986)’s finding

In recent study, many researchers have extended this notion by regarding liquidity as an aggregate risk factor instead of as a characteristic of stock in earlier papers For instance, Datar et al (1998), using his proposing liquidity proxy – turnover ratio in investigating the return-liquidity relation, posit that liquidity establishes its roles in explaining asset returns, stronger than the size risk factor and conjecture that the size effect might probably be a reflection of liquidity impact This might partly be due to the fact that institutional investors has boosted the demand for large and liquid stocks and thus diminished the relative performance of small stocks (Gompers & Metrick 2001) Base on the preceding argument, Jacoby et al (2000) re-construct the CAPM model

by re-measuring excess returns over one-period of both stock and market by taking liquidity costs, which are effective spread, into account and prove that the measure of systematic risk must relate the changes in the security’s spread

Pastor & Stambaugh (2003) assess whether market-wide liquidity is priced and conclude that returns of stocks with higher sensitivity to market liquidity exceeds those with lower sensitivity Acharya & Pedersen (2005) present the liquidity-adjusted CAPM model by adding three more risk factors related to liquidity risk, commonality in liquidity of stock and market, return sensitivity to market liquidity and liquidity sensitivity to market return The authors report that their model explains excess returns better than the standard CAPM model does Liu (2006), introducing both a new proxy for liquidity and the liquidity-augmented CAPM model, states that the model explains well the cross-sectional stock returns, particularly in case of the present of well-studied market anomalies which both CAPM and Fama-French three-factor models fail to explain

Questioning on whether the existing asset pricing models with well-documented factors such as market premium, size, book-to-market, coskewness, and Pastor-Stambaugh’s factors capture the characteristic liquidity effect, Nguyen et al (2007) conduct both time-series and cross-section tests on the three-moment CAPM and four-factor model based on Fama–French and Pastor–Stambaugh factors as well as the mix of these two models Their empirical work shows that all of these factors do not subsume the characteristic liquidity premium In other words, characteristic liquidity should play its significant role in explaining stock returns together with other well-known risk factors This view is shared by Keene & Peterson (2007), however, they add that there is

a need for finding more risk factors for capital asset pricing models due to nonzero intercepts

In view of liquidity variability, recent studies have attempted to explore its linkage with asset returns For example, Chordia et al (2001) report a negative association between liquidity volatility and cross-sectional equity returns, in contrast of the intuition of risk-return relationship These findings stand up in the face of various controls for the size, book-to-market ratio, momentum, price level and dividend yield effects However, Pastor & Stambaugh (2003) and Acharya & Pedersen (2005) refute those of Chordia et al (2001) while their findings indicate that investors require higher expected future asset returns to compensate for the

contemporaneous liquidity’s shocks The latter is suppored by Chien & Lustig (2010) as they claim that liquidity risks caused by different business cycles should be rewarded by higher expected return on stocks

As mentioned by Lo and MacKinlay (1990), besides the US market, other markets need to be examined in order to avoid the data-snooping problem The majority of studies of liquidity and liquidity risk associated with asset returns are in US market which is generally recognized as the most liquid market in the world with a small impact of liquidity than those of other markets, especially the emerging markets In recent years, some investigations have been conducted in other equity markets to reinforce return-liquidity relation

Jun et al (2003) report the positive relation between market-wide liquidity and stock returns using data for emerging equity markets However, subsequent studies on the liquidity premium in each market differ from this result A well-known example of this is Loderer & Roth (2005) since their findings are in favor of the liquidity premium theory in the Swiss Exchange Chang et al (2010) provides more significant evidence regarding to the notion using data in Japanese stock market Lam & Tam (2011) contend further that the liquidity augmented Fama-French model explains well stock returns in the Hong Kong stock market In a nutshell, these studies support the notion that illiquidity yield higher stock returns

The conflicting conclusion and division in the literature, particularly the effect of liquidity volatility in security’s expected return, indicate the need for further work This paper firstly contributes to the literature firstly in that way Secondly, most of the empirical research concerning the liquidity–stock return relationship has been focused on US markets (Chang et al 2010) There is not much work done employing emerging market data as the literature investigating this nexus in emerging markets is still very light and this paper considering the case

in Vietnam is another contribution

Our results differ from the large body of the existing literature First, high level of liquidity, especially those related to trading activities, indicates high returns on stock This effect persists after controlling for the well known market-wide determinants such as Fama-French three-factors, momentum factor and liquidity factor as well as well-studied characteristics of stocks The line of reasoning would lead us to believe that newly listed big firms would possibly lead to abnormal trading volume and thus, cause a rise on these stock prices which are supported by the investment preferences towards big firms’ stock (Merton 1987; Miller 1977) Second, when seeking to establish a link between liquidity risk and asset returns, we do not find strong evidence in support the notion that the higher return on stocks should be compensated for suffering high liquidity risk

The remainder of the paper proceeds as follows Section 2 describes our data, key variables and experimental design Section 3 examines the results of the investigation The final section concludes and formulates practical implications

2 Data and variable description

Data are collected from different sources Market data are provided by Ho Chi Minh City stock exchange Financial accounting data are collected and tabulated from financial reports of listed company Our data sample

includes of monthly returns and other firm attributes of non-financial firms listed on Ho Chi Minh stock exchange Our data range from 2009 to 2012

We employ a number of liquidity and illiquidity measures for a thorough investigation Our choice of control variables was based on the existing literature (see, for example, (Chan et al 1991; Chang et al 2010; Chordia

et al 2001)) For each stock, the following variables were calculated each month Liquidity measures and control variables are defined as follows:

RET monthly excess returns risk adjusted using 5 factors, Fama-French three factors, a

momentum, and a liquidity factors, liquidity proxy used is the turnover ratio by Datar et al (1998)

TURN natural logarithm of the average of the share turnover of the prior 3 months where the share

turnover is calculated as the number of shares traded divided by the number of share outstanding (Chang et al 2010)

DVOL natural logarithm of the average of the VND trading volume over the prior 3 months

(Brennan et al 1998)

ILLIQ Amidhud’s illiquidity measure based on the previous 3 months data

ZERO1 proportion of trading days in the past 3 months in which the return is zero (Lesmond et al

1999)

ZERO2 proportion of trading days in the past 3 months in which the trading volume is positive and

the return is zero (Lesmond et al 1999)

V_TURN natural logarithm of the coefficient of variation of share turnover (Chang et al 2010; Chordia

et al 2001)

V_DVOL natural logarithm of the coefficient of variation of VND trading volume (Chang et al 2010;

Chordia et al 2001)

V_ILLIQ natural logarithm of the coefficient of variation of the illiquidity measure of Amihud (2002) SIZE natural logarithm of the market capitalization at the end of the second to the last month

BM natural logarithm of the most recent book value of common shares divided by the market

value in the second to the last month

PRICE natural logarithm of the reciprocal of the closing price in the second to the last month (Chang

et al 2010; Chordia et al 2001)

IDIORISK standard deviation of regression residuals of the factor model based on data month t−24 to

month t−1

DYLD amount of cash dividends for the last fiscal year over the closing price in the second to the

last month (Chang et al 2010; Chordia et al 2001)

ELYD net income before extraordinary items for the last fiscal year over the market capitalization

at the end of the second to the last month (Chang et al 2010; Chordia et al 2001)

CFYLD sum of earnings and depreciation for the last fiscal year over the market capitalization at the

end of the second to the last month (Chang et al 2010; Chordia et al 2001)

RET2_3 cumulative returns over the second through the third prior to the current month (Chang et al

2010; Chordia et al 2001; Jegadeesh & Titman 1993)

RET4_6 cumulative returns over the fourth through sixth prior to the current month (Chang et al

2010; Chordia et al 2001; Jegadeesh & Titman 1993)

RET7_12 cumulative returns over seventh through 12th months prior to the current month (Chang et

al 2010; Chordia et al 2001; Jegadeesh & Titman 1993)

Table 1 shows the summary statistics of data employed in this study

[INSERT TABLE 1 ABOUT HERE]

Following Brennan et al (1998) and (Chordia et al 2001), RET is calculated as followings:

𝑅̃𝑗𝑡 = 𝐸(𝑅̃𝑗𝑡) + ∑𝐿𝑘=1𝛽𝑗𝑘𝑓̃𝑘𝑡 + 𝑒̃𝑗𝑡

where R̃jt is the return on security j at time t, βjk is the factor loading of the security's return on factor k, f ̃kt is the return on factor k at time t, and ẽjt is the error term We begin by estimating each year the factor loadings,

βjk, for all securities that had at least 24 return observations over the prior 60 months The factors we use to measure the excess return are Fama & French (1993) factors, a momentum factors and a liquidity factor The momentum and liquidity factors are calculated as the same way of the Fama-French HML factor, which means that these factors are the differences each month between the average of the returns on the high momentum/liquidity portfolios and the average of the returns on the low momentum/liquidity portfolios The momentum is calculated as the raw return annually, while liquidity is measured as the turnover of every firm through a period of one year Meanwhile, the portfolio foundation procedure is repeated at the end of fiscal years Then, estimated risk-adjusted return on each of stock for each month of the following year is measured

as follow:

𝑅̃𝑗𝑡∗ = 𝑅̃𝑗𝑡− 𝑅𝐹𝑡 − ∑𝐿𝑘=1𝛽̂𝑗𝑘𝐹̃𝑘𝑡

where 𝑅̃𝑗𝑡∗ is the risk-adjusted return, 𝑅𝐹𝑡 is the risk-free rate, and 𝐹̃𝑘𝑡 is the sum of the factor realization and its associated risk premium The next step is to test whether liquidity carry premium after controlling for other firm characteristics:

𝑅̃𝑗𝑡∗ = 𝑐0+ 𝛾𝐿𝑗𝑡+ ∑𝑀𝑘=1𝑐𝑚𝑍𝑚𝑗𝑡 + 𝜀̃𝑗𝑡

where 𝐿𝑗𝑡 is the liquidity of stock j at time t, 𝑍𝑚𝑗𝑡is the characteristic m of stock j at time t, and 𝜀̃𝑗𝑡 is the error term Finally, we also test the liquidity roles in explaining security’s return after controlling for liquidity and other firm characteristics:

𝑅̃𝑗𝑡∗ = 𝑐0+ 𝛾𝐿𝑗𝑡+ 𝛿𝑉𝑗𝑡+ ∑𝑀𝑘=1𝑐𝑚𝑍𝑚𝑗𝑡 + 𝜀̃𝑗𝑡

where 𝑉𝑗𝑡 is the liquidity variability of stock j at time t

3 Results and Discussion of Results

Table 2 reports the coefficients of correlation amongst variables employed in the study There are some particular notes here which are worthy to spell out regarding the correlation amongst liquidity variables Firstly, the relationship between the two liquidity variables TURN and DVOL is quite high (0.7584) and this reflects the ability of substitution between these two liquidity variables Secondly, the correlation between liquidity and illiquidity measures are negative as TURN and DVOL are negatively correlated with ILLIQ, ZERO1 and ZERO2 Thirdly, the correlation between Zero1 and Zero2 is quite high (0.7324) The relationship between Zero measures and liquidity measures are negative

Moreover, the correlation coefficients amongst adjusted stock return and liquidity/illiquidity measures provide some preliminary indications regarding the relationship between stock returns and liquidity Stock returns are positively correlated with the two liquidity measures, TURN (0.0896) and DVOL (0.0857) and negatively correlated with the illiquidity measures, ILLIQ (-0.0315), ZERO1 (-0.0748) and ZERO2(-0.0719)

[INSERT TABLE 2 ABOUT HERE]

Table 2 panel B also shows the correlation coefficients of stock returns and other market and firm characteristics factors Stock returns are negatively correlated with BM, PRICE, while positively correlated with SIZE, IDIORISK, DYLD This observation indicates that the firm with higher return tends to be large, valued, highly priced, risky and paying more dividends As also can be seen in table 2 panel B, DVOL has a positive association with SIZE and negative correlation BM and PRICE, implying that high liquidity level happens in large, valued and highly priced firms The statistics for ZERO1 and ZERO2 show the similar result with that of DVOL Table 3 shows the results of regressions of risk adjusted returns on each of the five alternative measures of liquidity level (TURN, DVOL, ILLIQ, ZERO1, ZERO2) controlling for a wide range of firm specific attributes (SIZE, BM, PRICE, IDIORISK, DYLD, EYLD, CFYLD, RET2_3, RET4_6, RET7-12) In contrast to the majority of the literature, our results show that all of the coefficients of liquidity measures are significantly

affect cross-sectional returns of stocks The coefficients of TURN and DVOL are positive while the other three coefficients are negative In other words, the evidence we present is perfectly inconsistent with the hypothesis that there is a premium for illiquidity

[INSERT TABLE 3 ABOUT HERE]

In order to confirm the robustness of the results, we re-run the regressions in each year in our sample to see whether our results are similar over the year in our analysis We find that in 2009 and 2010, the return-liquidity relations are significant only for liquidity measures (TURN, DVOL), while three illiquidity measures (ILLIQ, ZERO1, ZERO2) do not show significant linkage with stock returns For the year of 2011, the relations between stock return and all of five liquidity/illiquidity proxies are not significant The contrasting view is demonstrated

in 2012 figure which shows that all of the liquidity measures have significant coefficient estimates in explaining returns on stocks

In terms of liquidity variability effect on stock returns, as illustrated in table 8, we do not find it significant as the evidence from a large body of the literature

4 Conclusion

The papers investigate the relationship between liquidity and stock returns in Vietnam Most of the papers in the literature report a negative relationship between these two variables However, we prove the reverse relationship in Vietnamese stock markets As such, our findings cast some doubt in the positive return-illiquidity relation in emerging stock markets since the investors possibly recognize that the illiquid firms are not probably the good performers

The first explanation for this might be the characteristics of Vietnam’s market where participants are small investors are trading more frequently The dominance of these investors means that their preferences towards blue-chip stocks lead to boost the demand for large and liquid stocks, and thus, bring the higher returns on these stocks And further potential reason attributed to individual investors is that portfolio’s decisions of these investors are driven by the high level of trading activities In other words, our findings is supportive in the view stated by Gervais et al (2001) that high trading volume over a period of time could make the stock visible to investors, help stimulate the demand of the shares and push up stock price Second, the rationale behind this finding can be supported on the basis that Vietnamese stock market is the really new market that was introduced

in 2000 Therefore, one important characteristic of the market is that there are newly stocks listed every year

In this situation, one of the realities is that the big firms come to the market and make it visible and tradable to investors, creating races in purchasing and causing a rise on these stocks (Merton 1987; Miller 1977) Another problem in Vietnamese stock market could be the information inefficiency As a consequence, institutional investors do not probably find reliable data for their analysis and investment decision making

Although the main objective of this paper is not to come up with the roles of trade shocks on asset returns, the findings have given us the prediction of this nexus And this finding is consistent with intuition that in emerging markets, which appear vulnerable to stock price manipulation, most of individual investors speculate and follow the trading strategies of institutional and foreign investors

In view of the impact of liquidity risk on equity returns, we do not find this risk-reward relationship in Vietnamese stock market

Table 1: Summary statistics

RET 7653 -0.007 -0.015 0.728 -0.578 0.100 0.940 7.448 7436.401 0.000

BM 7653 0.091 0.128 2.225 -2.539 0.691 -0.321 3.094 133.8942 0.000

Note: RET = monthly excess returns risk adjusted using 5 factors (Fama-French three factors, a momentum, and a liquidity factors); TURN = natural logarithm of the average

of the share turnover of the prior 3 months; DVOL = natural logarithm of the average of the VND trading volume over the prior 3 months; ILLIQ = Amidhud’s illiquidity measure based on the previous 3 months data; ZERO1 = proportion of trading days in the past 3 months in which the return is zero; ZERO2 = proportion of trading days in the past 3 months in which the trading volume is positive and the return is zero; V_TURN = natural logarithm of the coefficients of variation of share turnover based on the past 3-month data; V_DVOL = natural logarithm of the coefficients of variation of VND trading volume based on the past 3-month data; V_ILLIQ = natural logarithm of the coefficients of variation of Amihud’s illiquidity measure based on the past 3-month data; SIZE = natural logarithm of the market capitalization at the end of the second to the last month; BM = natural logarithm of the most recent book value of common shares divided by the market value in the second to the last month; PRICE = natural logarithm

of the reciprocal of the closing price in the second to the last month; IDIORISK = standard deviation of regression residuals of the factor model based on data month t−24 to month t−1; DYLD = amount of cash dividends for the last fiscal year over the closing price in the second to the last month; EYLD = net income before extraordinary items for the last fiscal year over the market capitalization at the end of the second to the last month; CFYLD = sum of earnings and depreciation for the last fiscal year over the market capitalization at the end of the second to the last month; RET2_3, RET4_6, and RET7_12 = cumulative returns over the second through the third, fourth through sixth, and seventh through 12 th months prior to the current month, respectively;

Table 2: Correlation matrix

Panel A: Part I

RET TURN DVOL ILLIQ ZERO1 ZERO2 V_TURN V_DVOL

TURN 0.0896

DVOL 0.0857 0.7584

ILLIQ -0.0315 -0.5787 -0.5921

ZERO1 -0.0748 -0.6135 -0.6239 0.3838

ZERO2 -0.0719 -0.2962 -0.3201 -0.0932 0.7324

V_TURN -0.0294 -0.1516 -0.2205 0.2110 0.1526 -0.0017