Mutual fund liquidity timing ability in the higher moment framework

Research in International Business and Finance 51 (2020) 101105

Contents lists available at ScienceDirect

Research in International Business and Finance journal homepage: www.elsevier.com/locate/ribaf

Full length Article

Mutual fund liquidity timing ability in the higher moment framework⋆

T

Woraphon Wattanatorna, Chaiyuth Padungsaksawasdia, , Pornchai Chunhachindaa, Sarayut Nathaphanb ⁎

a b

Thammasat University, Department of Finance, Thammasat Business School, Bangkok, 10200, Thailand Mahidol University International College, Mahidol University, Nakhon Pathom, 73170, Thailand

ARTICLE INFO

ABSTRACT

JEL classification: G11 G12 G21 G23

Using mutual fund data in Thailand, this study shows that fund managers can time the marketwide liquidity in the higher moment framework. High-performing fund managers demonstrate significantly positive liquidity timing ability, while low-performing fund managers do not. Thus, high-performing fund managers increase (decrease) the funds' exposure to the market during a high (low) market liquidity period, while low-performing fund managers do not show the liquidity timing ability. Moreover, only top-performing bank-related mutual funds possess the liquidity timing ability, supporting the information advantage hypothesis. Nonbank-related funds do not possess the liquidity timing ability at both the aggregate and portfolio levels. Several robustness tests confirm the findings.

Keywords: Liquidity timing Mutual fund performance Market timing Bank-related mutual fund Higher moment liquidity timing

1. Introduction Aragon and Strahan (2012) and Cao et al. (2013b) document that liquidity risk plays an essential role in portfolio management because mutual fund managers are responsible for managing portfolio liquidity to meet daily investors’ operations, especially unexpected large redemptions.1 Furthermore, unlike market return, which is difficult to forecast, market liquidity is more persistent and allows fund managers to forecast more accurately (Cao et al., 2013b). Although studies on mutual fund performance are extensive,

We thank the editor (John W. Goodell), anonymous referee, Bart Frijns, Jairaj Gupta, Maria E. de Boyrie, Russ Wermers, Anutchanat Jaroenjitrkam, Anya Khanthavit, Pantisa Pavabutr, and Seksak Jumreornvong for many helpful and insightful comments. We also thank the participants and discussants at the 2015 Academy of International Business Southeast Asia Regional Conference in Penang, Malaysia, the 2016 Asian Finance Association Conference, Bangkok, Thailand, the 2016 Auckland Finance Meeting, Auckland, New Zealand, the 29th Australasian Finance and Banking Conference, Sydney, Australia, and the Research Finance Workshop, Department of Finance, Thammasat Business School. We are grateful to the Business Research Center (BRC), Thammasat Business School for financial support. ⁎ Corresponding author. E-mail addresses: [email protected] (W. Wattanatorn), [email protected] (C. Padungsaksawasdi), [email protected] (P. Chunhachinda), [email protected] (S. Nathaphan). 1 Prior literature (Acharya and Pedersen, 2005; Amihud, 2002; Amihud and Mendelson, 1986; Holmström and Tirole, 2001, and Pástor and Stambaugh, 2003) conjectures that liquidity risk is an important market risk factor. ⋆

https://doi.org/10.1016/j.ribaf.2019.101105 Received 20 December 2018; Received in revised form 9 September 2019; Accepted 10 September 2019 Available online 14 September 2019 0275-5319/ © 2019 Elsevier B.V. All rights reserved.

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

few studies have focused on mutual fund liquidity timing ability, especially in a developing economy.2 For example, Cao et al. (2013a,b) show that both hedge funds and mutual funds can time market-wide liquidity, which the managers would expose to the market more (less) during a high (low) liquidity period. To the best of our knowledge, no study has been reported on the liquidity timing ability of mutual funds in an emerging equity market, although liquidity risk plays an important role (Brown et al., 2008; Hearn, 2010, and Lam and Tam, 2011). Additionally, stock return distributions in emerging markets are more non-normal than in developed markets (Adcock and Shutes, 2005; Bae et al., 2006; Bekaert et al., 1998; Canela and Collazo, 2007, and Harvey et al., 2010).3 Thus, a mean-variance approach is not sufficient to characterize return and associated risk in a high-volatile environment, such as in emerging markets, and a higher moment framework must be considered (Samuelson, 1970). Moreover, seminal works (Chunhachinda et al., 1997; Kraus and Litzenberger, 1976; Harvey and Siddique, 2000; Moreno and Rodríguez, 2009) have shown that coskewness is priced and plays a significant role in portfolio allocation. Our paper introduces a liquidity timing model in the higher moment to fill the gap in the literature. Another important trend of studies in mutual fund performance is the bank-mutual fund relationship. Gómez-Bezares and Przychodzen (2018) document that bank-related funds in the U.S. managed assets of more than $566 billion in 2014, accounting for 26% of the mutual fund industry. The size of bank-related mutual funds cannot be neglected. Additionally, Bolton et al. (2007) propose a theory to show that the competition among bank conglomerates persuades the businesses to collect and manage their clients’ information. Consequently, bank-related fund managers and nonbank-related fund managers possess different information to manage their portfolios. Bank-related fund managers are in a better position to analyze the information, which is particularly essential when the potential cost of false investment is high. In summary, bank-related fund managers and nonbank-related fund managers have different constraints to manage their portfolios as follows. First, bank-related funds have more new investment flows because of a lower searching cost (Nathaphan and Chunhachinda, 2012 and Sirri and Tufano, 1998) and, hence, possess more liquidity than nonbank-related funds. Second, the relationship between the bank and its affiliated mutual funds can cause an agency problem, which affects bank-related fund investment constraints and investment outcomes. For example, Mehran and Stulz (2007) and Hao and Yan (2012) document that a bank might encourage its funds to support the clients’ IPO stocks to hopefully win future contracts in other lines of the bank businesses, potentially making the funds misallocate their invested portfolios and lose benefits of diversification. This study contributes to prior literature in the following several aspects. First, to the best of our knowledge, this study provides the first evidence of liquidity timing ability in the mutual fund industry in a developing economy. Our pioneering findings are supported by liquidity risk in emerging markets showing a higher risk premium than in developed markets (Brown et al., 2008; Hearn, 2010, and Lam and Tam, 2011) and that incorporating the liquidity factor in financial literature is supported by Benson et al. (2015). Additionally, Kearney (2012) shows that emerging markets illustrate significant growth in terms of the economies and proportion of savings, subsequently warranting attention in the global market. Importantly, this paper fills the gap in the studies of liquidity timing ability of mutual funds. Second, Moreno and Rodríguez (2009) find that coskewness is a significant factor to explain risk-adjusted returns in the mutual fund industry.4 Considering this finding, we incorporate the coskewness risk factor into our liquidity timing ability models, a feature that is not existent in prior literature and is superior to the mean-variance approach. Third, evidence on the abilities of mutual fund managers remains mixed,5 in which contradicting results in existing studies are largely driven by differences in methodology, sample (country), and period of study. Additionally, Bekaert and Harvey (1997) show that returns in emerging markets are more predictable than those in developed markets. We propose our models by controlling the market timing ability as well as the volatility timing ability of mutual fund managers to truly examine the role of liquidity timing ability at both the aggregate and portfolio levels. Finally, based on the bank-mutual fund relationship, our findings support the information advantage hypothesis that bank-related funds occupy superior information than nonbank-related funds. We provide new findings of the liquidity timing ability between bank- and nonbank-related mutual funds that are neglected in prior literature. Our finding suggests that bank-related funds capture dynamic patterns in the market better than nonbank-related funds. As one of fastest growing emerging markets, we focus on the mutual fund market in Thailand for the following reasons. First, Thailand is one of the important emerging markets in South East Asia and has exhibited a rapid economic expansion over the last decade. In addition to the economic growth, the Thai mutual fund industry has impressively expanded at an average of 27% per

2 All existing studies are examined in developed markets. See Bodson et al. (2013) and Karstanje et al. (2013) for the U.S. market and Foran and O’Sullivan (2017) for the U.K market. 3 In a bird's eye view, this is because emerging markets are characterized by an incomplete market structure, political and economic uncertainties, weak regularities, as well as low-quality auditing systems and a more likelihood of structural changes—for example, regulatory changes, financial market liberalization, political crises, and other shocks. 4 It is consistent with prior research (Adcock and Shutes, 2005; Bae et al., 2006; Bekaert et al., 1998; Canela and Collazo, 2007, and Galagedera and Brooks, 2007), especially in emerging markets, suggesting that higher moment return characteristics significantly affect portfolio management and portfolio allocation. 5 For example, Kon (1983) finds no ability of mutual fund managers, while Chang and Lewellen (1984) and Kon and Jen (1979) show that mutual fund managers have both market timing ability and stock selectivity. Tchamyou and Asongu (2017) support the volatility timing and market timing abilities in the U.S. equity and allocation funds. On the other side, Comer et al. (2008) find no evidence of the market timing ability in global asset allocation funds. This finding is consistent with Bauer et al. (2006) finding no evidence of the market timing ability in the New Zealand mutual funds and Romacho and Cortez (2006) showing no evidence of the market timing ability and selectivity in the Portuguese mutual funds. However, Jiang et al. (2007) and Bollen and Busse (2001) strongly support the market timing ability in the U.S. market.

2

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

year.6 Second, because businesses in Thailand are dominated by debt financing through bank loans (Prommin et al., 2014), this unique characteristic allows us to study a relationship between the bank and its affiliated mutual funds. According to the information advantage hypothesis, bank-related funds gain superior information in that banks can share their clients’ loan information with their affiliated mutual funds. Therefore, this study employing the sample in Thailand sheds a new light on the difference in liquidity timing ability between bank- and nonbank-related mutual funds. Overall, our proposed liquidity timing ability model works well at both the aggregate and portfolio levels. The liquidity timing ability factor is statistically significant. Adjusted R2 values are generally high. Low-performing funds show no liquidity timing ability, while high-performing funds show the statistically significantly positive liquidity timing coefficient, demonstrating that the fund managers generate profits from timing the market-wide liquidity. Even in the higher moment environment, the liquidity timing factor remains significant in both the aggregate and portfolio levels, with most findings remaining unchanged. The coskewness risk factor was statistically priced, showing that returns are more predictable in emerging markets (Bekaert and Harvey, 1997). Interestingly, we find that only bank-related mutual funds show the market-wide liquidity timing ability, a finding that is consistent with the information advantage hypothesis. The fund managers utilize superior information obtained from their affiliated banks to time the market-wide liquidity. Our results are robust to several proxies of illiquidity measure, subperiods of study, and different methods of variable estimation, respectively. The remainder of the study is organized as follows. Section 2 provides a brief overview of liquidity measures that are widely employed in emerging markets. Section 3 discusses an important aspect of the higher moment characteristic in emerging markets. Section 4 provides the background and trend of the mutual fund industry in Thailand. Sections 5 and 6 illustrate data and methodology used in the study. Sections 7 presents and discusses empirical results. Section 8 is robustness checks. The last section concludes and summarizes the study. 2. Liquidity measurement In this section, we provide an overview of the major liquidity measurement employed in finance studies. Motivated by Benson et al.’s (2015) study on the essence of incorporating liquidity into finance research, liquidity is a crucial component for portfolio management. A success of major trading strategies (i.e., size, value, and momentum) depends on the role of liquidity.7 8 Additionally, the importance of liquidity as a risk factor in asset pricing is well documented in academic literature (Acharya and Pedersen, 2005; Amihud, 2002, and Pástor and Stambaugh, 2003).Brown et al. (2008); Hearn (2010); Lam and Tam (2011), and Jun et al. (2003) show that emerging markets have lower liquidity than developed markets, supporting the important role of liquidity as an additional risk factor. Because liquidity can be measured from different dimensions (for example, numbers of trading assets, trading volume, and asset turnover), we summarize the pertinent literature using major illiquidity measures, focusing on emerging market studies below. Lesmond (2005) studies the liquidity risk factor in 31 emerging markets using low frequency illiquidity measures and finds that Amihud’s (2002) illiquidity measure and the zero illiquidity measure are the best performers within country analysis. Bekaert et al. (2007) find that the zero return day is frequently observed and fairly persists in emerging markets. Furthermore, their findings show that the zero return measure is highly correlated with the bid-ask spread, a finding that is consistent with that of Lesmond (2005). In summary, several studies (for example, Bekaert et al., 2007, and Kang and Zhang, 2014) support Lesmond’s (2005) findings that the Amihud measure is the best illiquidity measure of low-frequency data.9 From the aforementioned literature, in this study, we adopt widely acceptable illiquidity measures employed in emerging markets, namely, the Amihud’s illiquidity measure, adjusted-illiquidity measure, zero return day measure, and Roll’s effective spread measure. Hence, we briefly introduce each measure below. First, one of the most classical illiquidity estimations is the Amihud's (2002) illiquidity measure of stock i in month t (ILLIQi, t ) presented as

ILLIQi, t =

T |Rti, d | 1 T d = 1 Vti, d

(1)

where is the return of the stock i on day d in month t, T is the total number of trading days in month t, and is the trading volume in million baht of stock i on day d in month t. The larger is the Amihud measure, the lesser is the stock liquidity. Second, Kang and Zhang (2014) suggest that the Amihud measure is undefined under a zero trading volume day. Although zero trading volume days are unusual in developed markets, they are more common in emerging markets. Several studies have found that the zero trading volume day possesses a high correlation with the bid-ask spread (Lesmond, 2005 and Bekaert et al., 2007). Therefore, Kang and Zhang introduce a new illiquidity measure, the adjusted-illiquidity measure, which improves the efficiency of the Amihud’s (2002) illiquidity measure by considering the effect of zero trading volume days as follows.

Vti, d

Rti, d

6

Sources: Morningstar Direct database, as of April 2015. Amihud and Mendelson (1986) document that market-observed expected returns are an increasing concave function of the bid-ask spread. Later, numerous studies (for example, Acharya and Pedersen, 2005; Amihud, 2002, and Pastor and Stambaugh (2003) suggest that liquidity is a relevant factor in asset pricing. 8 Bettman et al. (2010); Li et al. (2014), and Docherty et al. (2013) support the effect of liquidity to size, value, and momentum trading strategies. 9 Although several studies employ Pastor and Stambaugh’s (2003) illiquidity measure, we argue that it is not a good liquidity estimation for emerging markets because of a potential large number of zero return days. The calculation is undefined in zero return days. 7

3

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

AdjILLIQi, t = ln

Zerovoli, t =

T |Rti, d | 1 T d = 1 Vti, d

× (1 + Zerovoli, t )

(2)

number of days with zero trading volume in month t T

(3)

where Zerovoli, t is the ratio of the number of zero trading volume days in month t to the number of total trading days in month t (T) and ln is natural logarithm. Third, Lesmond et al. (1999) argue that the zero illiquidity captures asset liquidity in terms of adverse selection. Informed investors will trade only when the profit from their private information is larger than the transaction costs. For example, if an asset is less liquid, a large transaction cost would discourage informed investors to trade, consequently no private information is revealed. Therefore, a less liquid asset with large transaction costs is less frequently traded than a high liquid asset, potentially showing a large number of zero return days. The zero return day measure is presented below.

zeroi, t =

number of days with zero returns in month t T

(4)

where zeroi, t is the zero illiquidity measure for the stock i in month t. The higher is the number of zero return days, the higher is the market illiquidity. Finally, we employ the Roll’s (1984) spread. The effective spread measures the liquidity of an asset driven by transaction costs, given symmetric information. Transaction costs show a negative relationship with a serial correlation of price changes. An effective spread leads to a negative covariance of price changes in the following period. A wider spread leads to higher transaction costs. Thus, the wider is the effective spread, the lesser is the liquidity.

Roll s Effective spreadi, t = 2 where

(5)

covi, t

covi is the first-order serial covariance of price changes between two consecutive periods.

3. Higher moment risk factor Skewness is an important factor in explaining security return and portfolio return (Jean, 1971, 1973 and Kraus and Litzenberger, 1976).10 Seminal works in portfolio management (Adcock and Shutes, 2005; Bae et al., 2006; Bekaert et al., 1998; Canela and Collazo, 2007, and Galagedera and Brooks, 2007) support the important role of skewness in investors’ decision making. A negative relationship between skewness and expected stock return is also widely documented (Kon, 1984; Kraus and Litzenberger, 1976; Mills, 1995, and Peiró, 1999), allowing investors to trade skewness risk to maximize expected returns (Chunhachinda et al., 1997). Levy and Sarnat (1972) support the importance of higher moments as a risk factor for mutual fund investors. Furthermore, Harvey and Siddique (2000) show that coskewness (nondiversifiable skewness) is priced in asset returns.11 Therefore, it is logical to incorporate the skewness risk factor into the model. We follow Harvey and Siddique (2000) to form the coskewness factor as follows.

Si =

E( E(

2 i, t + 1 m, t + 1)

2 2 i, t + 1 ) E ( m, t + 1)

(6) 12

2 2 i, t + 1 is the residuals from the regression of the returns on stock i on market excess returns. m, t + 1 is the market residuals between the market excess returns and their mean return. We employ returns over the first 60 months in our sample period to calculate i2, t + 1 and 2 m, t + 1. Next, we calculate the coskewness factor (Si ) and rank it monthly. In each month, we form three value-weighted portfolios; S represents the portfolio comprising 30% of the most negative coskewness stocks; S+ represents the portfolio comprising 30% of the most positive coskewness stocks; and S 0 represents the portfolio comprising 40% of the remaining coskewness stocks.13 The coskewness risk factor (CSK) is defined as the difference between S and risk-free return (rf ).

4. Why Thailand? Research on the effect of bank-mutual fund relationship is scant. Most evidence has been devoted to developed markets.14 Because of an increasing important role of emerging markets in the global financial system, this study aims to provide such evidence on the 10 Positive first moment and negative second moment properties imply sufficient conditions for the positive skewness preference for risk-averse investors. 11 Doan et al. (2010); Harvey et al. (2010); Harvey and Siddique (2000); Kostakis et al. (2012); Moreno and Rodríguez (2009), and Smith (2007) show a positive relationship between the coskewness risk factor and expected stock returns. 12 The error term is obtained from t+ 1 = rt + 1 at + 1 t + 1 (rm, t + 1) . See Harvey and Siddique (2000). 13 To confirm that our results are not driven by the procedure of coskewness risk factor formation, in unreported tables, we calculate S from 5%, 10%, 15%, 25%, and 30% of the most negative coskewness stocks, respectively. The results remain unchanged and are available upon request. 14 See, for example, in the U.S. markets, Massa and Rehman (2008); Mehran and Stulz (2007); Hao and Yan (2012); Berzins et al. (2013), and Ritter and Zhang (2007) find that bank-related funds outperform nonbank-related funds because of superior information.

4

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

bank-mutual fund relationship. We select the mutual fund industry in Thailand as our interest for several reasons. First, Thailand is one of the most important emerging markets in South East Asia and has illustrated a rapid economic expansion. A cumulative average growth rate (CAGR) of the GDP in Thailand is 7.38% from 2000 to 2017 compared with 4.99% of the world average CAGR.15 Second, businesses in Thailand are dominated by debt financing through bank loans (Prommin et al., 2014) and the Thai mutual fund industry is influenced by bank-related mutual funds (Charoenrook and Pavabutr, 2017). This allows us to investigate whether bank-related funds exploit the bank’s superior information as proposed by the information advantage hypothesis. This is because banks can readily access to their clients’ information through other lines of bank’s activities. Banks can potentially pass through information advantage to their related mutual funds.16 The information availability of mutual funds affects the investment decisions of investors. Consequently, bank-related funds that have lower information searching cost attract a larger positive fund flow than nonbank-related funds,17 making them have less liquidity constraint. In summary, we question the effect of the bank-mutual fund relationship on the mutual fund performance and mutual fund liquidity timing ability. To the best of our knowledge, prior literature does not address this issue in emerging markets. The mutual fund sector in Thailand was formally founded in 1975 by the collaboration between the government of Thailand and International Finance Corporation. From 1975 to 1992, Mutual Fund Public Co., Ltd. was the only operating asset management company (AMC) with 22 funds, comprising 12 domestic funds and 10 international funds. In 1992, the Security and Exchange Act BE2535 (AD, 1992) allowed the subsidiaries of commercial banks and other financial institutions to operate in the mutual fund industry. As of December 2017, there are 26 AMCs, consisting of 12 bank-related AMCs (1152 funds) and 14 nonbank-related AMCs (370 funds).18 The mutual fund industry in Thailand demonstrates an impressive growth from 2000 to 2017. The size of asset under management (AUM) increases approximately 14.30% per year although the Thai GDP grows approximately 6.18% per year. At the end of 2017, the Thai mutual funds occupy the asset 45 times larger than the Thai national GDP or 40 times larger than the market size of the Stock Exchange of Thailand.19 The premise confirms an increasing popularity among people in investing through a financial intermediary. 5. Data and the classification of mutual funds Stock prices, the one-year government bond index, and the stock market index are from DataStream. The monthly asset under management (AUM), net asset value, and annual reported net expense ratio are from the Morningstar Direct database. Our study period is from January 2000 to May 2018. Only 339 active domestic equity mutual funds we consider in this study, comprising 226 bank-related funds and 113 nonbank-related mutual funds.20 We follow the matching method suggested by Hao and Yan (2012) and Berzins et al. (2013) to categorize bank-related funds and nonbank-related funds as follows. First, we hand-collect the names of commercial banks in Thailand provided by Bankscope. Second, we match the name of the AMC with the name of the bank to identify bank-related AMC. We simply infer that the mutual funds operating under the bank-related AMC are bank-related mutual funds. For the AMCs where the names do not match the name of commercial banks, we manually classify them using information at the mutual fund’s website. We check any statement that implicitly and explicitly shows a relationship between the mutual fund and bank.21 6. Methodology We start our analysis by applying the liquidity timing model suggested by Cao et al. (2013a). From a Taylor series expansion,22 market beta is a linear relationship with the market liquidity timing ability as shown below. 15

Source: World Bank Information advantage hypothesis argues that bank-related funds possess superior information. They potentially utilize this superiority to enhance mutual fund performances. Banks can share clients’ information obtained from the banks’ activities with their affiliated mutual funds. In summary, bank-related funds gain informational advantages in several dimensions. First, they can obtain information at a cheaper cost. Second, they can access unpublished information available only at their banks, such as lending information (Massa and Rehman, 2008; Mehran and Stulz, 2007; Hao and Yan, 2012, and Berzins et al., 2013). Finally, they have the privilege of receiving IPO allocation when their associated bank is an IPO underwriter (Ritter and Zhang, 2007). 17 Sirri and Tufano (1998) show that the higher is the searching cost, the smaller is the fund flow. This finding is consistent with Frye (2001), who finds that bank-related funds have a lower searching cost than nonbank-related funds because corporations conduct business through banking. Thus, a low searching cost of bank-related funds attracts a larger fund flow. 18 Sources: Morningstar Direct database, as of May 2018. 19 Sources: Thomson Reuter Datastream, as of May 2018. 20 At the end of May 2018, 751 equity funds exist. We exclude international funds, funds of funds, index funds, trigger funds, bond funds, and money market funds. 21 For example, the name of “Bualuang AMC” does not match with the bank name. However, the statements provided on the AMC’s website show that it is an affiliated company of Bangkok bank. Therefore, we classify it as a bank-related AMC. Name lists of commercial banks, bank-related domestic equity funds, and nonbank-related domestic equity funds are available upon request. 22 Prior literature (Admati et al., 1986, and Ferson and Schadt, 1996) in market timing models suggests that market beta is a linear function of fund managers’ expectations on market returns. Busses (1999) and Cao et al. (2013a,b) further show the market beta as a linear function of volatility timing and excess liquidity timing, respectively. 16

5

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

mp

=

0mp

+

mt (Lmt

L¯m )

(7)

where mt is the systematic liquidity risk, Lmt is the market liquidity in month t, and L¯m is the rolling mean of the previous 60-month market liquidity.23 Because of the important role of liquidity in asset pricing and portfolio management (Benson et al., 2015), we follow the liquidity timing ability model suggested by Cao et al. (2013a) as follows.

rpt =

p

+

0mp rmt

+

smb SMBt

+

hml HMLt

+

mom MOMt

+

mt (Lmt

L¯m ) rmt +

(8)

pt

th

where rpt is the p portfolio return, rmt is the excess market portfolio return, and SMB, HML, and MOM are the mimic portfolio returns that capture the different effects of size, value, and momentum, respectively. mt measures the liquidity timing ability of the portfolio’s fund managers. A positive mt illustrates that fund managers can foresee the market-wide liquidity. Thus, they increase (decrease) the funds’ exposure to the market during a high (low) market-wide liquidity period. On the other hand, a negative mt illustrates that fund managers wrongly forecast the market-wide liquidity. Thus, they increase (decrease) the funds’ exposure to the market during a low (high) market-wide liquidity period. is the error term. As discussed in prior sections, we add the coskewness risk factor (CSK ) to reflect the nature of the equity market in emerging markets as presented below.

rpt =

pt

+

0mp rmt

+

smb SMBt

+

hml HMLt

+

mom MOMt

+

csk CSK t

+

mt

(Lmt

L¯m ) rmt +

pt

(9)

7. Empirical results 7.1. Summary statistics Panels A, B, and C of Table 1 present basic statistics for the entire sample, bank-related (BR) mutual funds, nonbank-related (NBR) mutual funds, risk factors, illiquidity measures, and correlation, respectively. On average, the mean return and standard deviation of BR and NBR funds, as shown in Panel A, are similar. The skewness values of both BR funds and NBR funds are negative showing a non-normal distribution. More than half of the individual fund returns do not show a normality distribution.24 Preliminary evidence supports an important role of a higher moment in portfolio returns, a finding that is consistent with those of Bollen and Busse (2001) in the U.S. and Moreno and Rodríguez (2009) in Spain. Thus, the inclusion of a higher moment risk factor is necessary. Furthermore, the AUM of BR funds is 3.23 times larger than that of NBR funds, emphasizing the dominant role of commercial banks in the mutual fund industry. A monthly average return (1.399%) of the equity mutual funds portfolio, as shown in Panel A, is higher than that of the stock market return (0.911%), as shown in Panel B. All risk factors (except HML) are positive. A positive value of the illiquidity measure supports a potential effect of the liquidity premium in the Thai stock market. Panel C shows the Pearson correlation between the returns of risk factors. In general, all risk factors are correlated in the same direction. However, we find that a multicollinearity problem is not a concern.25 7.2. Liquidity timing in mutual funds We start our analysis by testing the liquidity timing ability model as shown in Eq. (8). Table 2 presents the liquidity timing ability of the mutual funds for the full sample and decile portfolios. Overall, the results show strong evidence to support the market liquidity timing ability, with a positive significant relationship to value-weighted mutual fund portfolio excess returns. This shows that mutual fund managers can time the market-wide liquidity. Adjusted R2 values are large, confirming that the model is well specified. The size and value premium trading strategies are generally prevalent.26 Additionally, it is interesting to note that the momentum is negative but insignificant, showing that the contrarian trading strategy might work well. Next, we classify mutual fund portfolios into deciles; decile 1 represents the lowest fund performance, and decile 10 is the highest. We form each portfolio based on 12-month lag returns and rebalance them monthly. For brevity, we present the results on deciles 1 (P1), 5 (P5), 10 (P10), and 10-1 (P10-1).27 Overall, the market liquidity timing factor ( mt ) is priced at the portfolio level. Bottom-performing funds (decile 1) do not show market liquidity timing ability, with an insignificantly positive coefficient. Fund managers do not possess market liquidity timing ability. However, the remaining mutual fund portfolios show significantly positive market liquidity timing, suggesting that they successfully time the market by increasing (decreasing) market exposure during a high (low) market liquidity. On average, adjusted R2 values are still large (approximately 95%). The coefficient of liquidity timing ability is larger for top-performing portfolio (0.022) than that for bottom-performing portfolio (0.013). In decile 1 (P1), mutual fund managers do not have an ability to adjust their portfolios’ beta according to the liquidity timing ability. By contrast, managers show the liquidity timing ability in the top-performing 23

See further detail in Cao et al. (2013a,b). In an unreported table, we reject the null hypothesis of a normal distribution in 192 individual fund returns. The results are available upon request. 25 The VIF results are available upon request. 26 However, the value premium is less statistically significant than the size effect. A possible explanation is that value stocks are partially explained by the size effect (Lambert and Hübner, 2014). 27 The results of the other decile portfolios are available upon request. 24

6

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

Table 1 Descriptive statistics. Panel A: Mutual funds

Full Sample

R¯ CV Skewness Kurtosis Age (Year) ¯ (Million THB ) AUM ¯ (Million USD) AUM ¯ Fee

BR Fund

R¯ CV Skewness Kurtosis Age (Year) ¯ (Million THB ) AUM ¯ (Million USD) AUM ¯ Fee

NBR Fund

R¯ CV Skewness Kurtosis Age (Year) ¯ (Million THB ) AUM ¯ (Million USD) AUM ¯ Fee

Mean

P25

P50

P75

1.399% 5.838% 4.174 −0.180 9.745 9.745 2312.67 70.08 1.774 Mean 1.396% 6.72% 0.208 −0.387 9.620 9.061 2807.13 85.06 1.734 Mean 1.400% 6.86% 4.904 −0.247 9.592 11.739 870.497 26.38 1.875

−2.990% 1.666% −0.557 −1.128 2.757 1.750 90.50 2.74 1.425 P25 −2.683% 1.074% −2.498 −0.971 2.600 1.250 141.46 4.29 1.456 P25 −3.054% 2.932% −0.960 −1.426 3.288 4.208 53.535 1.62 1.410

0.312% 3.787% 12.150 −0.506 4.115 10.500 397.47 12.04 1.845 P50 0.338% 3.142% 0.108 −0.449 3.776 7.917 662.67 20.08 1.810 P50 0.483% 4.750% 9.838 −0.750 5.048 13.000 141.584 4.29 1.940

3.010% 5.527% 1.836 0.000 6.769 14.417 1,985.36 60.16 2.130 P75 2.809% 5.191% 0.541 0.023 6.142 13.917 2,849.79 86.36 2.095 P75 3.013% 5.865% 1.947 −0.286 8.128 17.750 399.384 12.10 2.130

Panel B: Risk factor Variable

Mean

Std. Dev.

Min

Max

Skewness

Kurtosis

Rm SMB HML MOM CSK AdjILLIQ

0.911% 0.203% −0.012% 5.617% 1.072% 0.415

7.509% 0.922% 1.018% 1.069% 6.056% 1.062

−36.329% −2.842% −5.455% 3.854% −34.704% −1.705

30.802% 3.874% 3.244% 9.593% 14.162% 3.053

−0.664 0.771 −1.399 0.845 −1.663 −0.089

9.471 5.276 10.510 3.718 10.703 2.550

Panel C: Pearson correlation

Rm SMB HML MOM CSK

Rm

SMB

HML

MOM

CSK

1.000 −0.184* 0.106 0.075 0.167

1.000 −0.256** −0.062 −0.109

1.000 0.099 0.105

1.000 0.020

1.000

Panels A, B, and C provide a summary of descriptive statistics for the variables employed in the study. R¯ and are the arithmetic monthly mean ¯ (Million THB [USD]) , and Fee ¯ are the cross-sectional return and standard deviation, respectively. CV is the coefficient of variation. Age, AUM average fund age, average of assets under management in Thai baht (the United States dollar), and fund annual net expense ratio, respectively. An average of the exchange rate is 33.00 Baht/USD. BR and NBR are the bank-related mutual funds and nonbank-related mutual funds, respectively. Rm is the monthly average market return in excess of a one-year government bond. SMB, HML, and MOM are the mimic portfolio returns accounting for the size, value, and momentum, respectively. CSK is the coskewness risk factor. AdjILLIQ is the Amihud-adjusted illiquidity measure. The null hypothesis of the correlation test is = 0 . ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

portfolio (P10). Ceteris paribus, the mutual fund portfolio’s market beta increases by 0.0234% (1.062 × 0.022), representing a change in one standard deviation of the adjusted-illiquidity measure. The liquidity timing ability seems to diminish the power of the size premium and value premium, in which most are slightly statistically significant. We infer that mutual funds managers rely heavily on timing market-wide liquidity. However, the contrarian 7

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

Table 2 Liquidity timing ability model. Full 0mp

smb hml

mom mt pt

Adj-R2

P1

P5

P10

P10-P1

0.902***

0.853***

0.952***

0.931***

0.077**

(37.26) 0.620*

(39.94) 0.745***

(27.59) 0.789

(22.81) 1.588**

(2.30) 0.824

(1.69) −0.082

(2.34) 0.048

(1.50) −0.080

(2.28) −0.401***

(1.59) −0.441***

(1.71) 0.469*

(-0.98) 0.075*** (5.35) 0.008* (1.86) 96.90

(2.93) 0.468**

(1.37) 0.604

(0.50) 0.013 (1.03) −0.006

(-0.90) 0.015*** (6.25) 0.006

(-1.40) 96.30

(1.49) 96.20

(2.39) 1.133**

(-2.99) 0.022*** (7.35) 0.031*** (4.63) 90.60

(1.43) 0.666

(-4.58) 0.021*** (8.28) 0.037*** (6.60) 23.90

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels. The illiquidity measure in this study is the Amihudadjusted illiquidity. mt is the estimated coefficient of the liquidity timing ability. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worstperforming funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (best-performing funds), and difference between the bestperforming and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. tvalues obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

effect plays a crucial role at the portfolio level, especially for the uppermost decile. Low-performing funds follow the momentum strategy (although insignificant) and have no liquidity timing ability and no abnormal return. By contrast, high-performing funds strongly perform the contrarian strategy and have liquidity timing ability and positive abnormal returns. The monthly abnormal returns for mutual fund portfolios are approximately 3.10% for the best-performing portfolio (decile 10) and -0.60% for the worstperforming portfolio (decile 1). 7.3. Liquidity timing in the higher moment framework Prior literature (Harvey and Siddique (2000); Doan et al. (2010); Harvey et al. (2010); Kostakis et al. (2012); Moreno and Rodríguez (2009), and Smith (2007)) finds that coskewness as a nondiversifiable risk is priced, documenting a positive relationship between the coskewness risk factor and expected stock returns. Thus, we must consider the effect of coskewness, which is usually present in emerging markets to detect the abilities of mutual fund managers. Under the higher moment framework, the coskewness risk factor ( CSK ) is significantly positively priced as presented in Table 3. Table 3 Liquidity timing ability in the higher moment framework. Full

P1

P5

P10

P10-P1

0mp

0.898***

0.849***

0.920***

0.888***

0.039

smb

(40.23) 0.602*

(43.89) 0.732***

(32.28) 0.580

(25.55) 1.279**

(1.25) 0.547

(1.74) −0.082

(2.28) 0.048

(1.15) −0.012

(2.13) −0.298*

(1.23) −0.346***

hml

mom csk mt pt

Adj-R2

(1.72) 0.457*

(-1.18) 0.036*** (3.08) 0.078*** (5.69) 0.007** (2.02) 96.10

(2.90) 0.460**

(1.01) 0.468

(0.55) 0.026

(-0.13) 0.036**

(1.55) 0.016 (1.16) −0.007

(2.24) 0.088*** (4.24) 0.003

(-1.59) 96.40

(0.56) 96.10

(2.12) 0.978**

(-1.86) 0.054*** (3.15) 0.145*** (7.95) 0.025*** (2.92) 90.80

(0.92) 0.518

(-3.02) 0.028

(1.06) 0.129*** (6.18) 0.032*** (4.74) 22.80

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels in the higher moment framework. The illiquidity measure in this study is the Amihud-adjusted illiquidity. mt is the estimated coefficient of the liquidity timing ability. csk is the estimated coefficient of the coskewness risk factor. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worst-performing funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (best-performing funds), and difference between the best-performing funds and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. t-values obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. 8

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

This finding confirms the important role of the higher moment in developing markets. In general, the results in the higher moment framework are consistent with our prior results, as shown in Table 2. The market liquidity timing factor ( mt ) remains significantly positive, confirming that mutual fund managers successfully forecast market liquidity in the higher moment environment. Overall, the liquidity timing ability factor in the portfolio level is positively significant in all decile funds, except for the worstperforming portfolio (decile 1). Interestingly, the alphas in low-performing portfolios (deciles 1–5) in the higher moment seem to be slightly worse than those in Table 2. Additionally, adjusted R2 values in each model slightly change compared with the previous results in Table 2. In summary, we conclude that mutual fund managers can time the market-wide liquidity under the higher moment framework. 7.4. Market timing ability, market volatility timing ability, and market liquidity timing ability In this section, we incorporate market timing ability and volatility timing ability into the model because these abilities are of the main interest in prior research. Thus, the finding on liquidity timing ability would not be biased. We define the market timing ability factor suggested by Treynor and Mazuy (1966) and the market volatility timing ability factor suggested by Busse (1999). Thus, the model is presented as

rpt =

pt

+

0mp rmt

+

smb SMBt

+

hml HMLt

+

mom MOMt

+

csk CSK t

+

mt

(Lmt

L¯m ) rmt +

2 mkt rmt

+

vol (Vmt

V¯m ) rmt +

pt

(10) where Vmt is the market volatility in the month t estimated from GARCH(1,1) specification and V¯m is the time-series average of Vmt . For consistency, we use the same period of the rolling window of [t-60,t-1] to calculate the time-series average of Vmt as the calculation of ¯ m . mkt and vol measure the market timing ability and volatility timing ability, respectively. the time-series average of CSK Table 4 shows the results for the effects of market timing ability and volatility timing ability. We find that the top-performing fund managers (P10) possess both positive liquidity timing ability and market timing ability, but we do not find these abilities in the worstperforming fund managers. Additionally, the top-performing fund managers can time the market return correctly, a finding that is consistent with those of Jiang et al. (2007) and Bollen and Busse (2001). Furthermore, the CSK risk factor plays an important role for both the top- and bottom-performing portfolios. In summary, top-performing fund managers possess the liquidity timing ability, even controlling for the traditional two-timing abilities and higher moment risk, while the bottom-performing fund managers do not. 7.5. Bank-related mutual funds and nonbank-related mutual funds In this section, we focus on the liquidity timing ability in the higher moment framework between two groups of mutual funds Table 4 Market timing ability, volatility timing ability, and liquidity timing ability.

0mp

smb hml

mom csk vol

mkt mt pt

Adj-R2

Full

P1

P5

P10

P10-P1

0.928***

0.860***

0.972***

0.913***

0.053

(32.84) 0.819

(34.07) 0.604***

(28.08) 1.010

(20.39) 1.368

(1.12) 0.764

(1.41) −0.170***

(1.57) 0.027

(1.41) −0.159**

(1.55) −0.339**

(1.16) −0.367**

(1.51) 0.553

(-2.86) 0.026**

(2.13) −0.042*** (-3.09) 0.460*** (5.86) 0.014 (0.90) 0.011*** (3.12) 97.30

(2.58) 0.326

(1.54) 0.677

(0.29) 0.026*

(-2.17) 0.021

(1.67) 0.071*

(1.15) −0.090***

(1.81) 0.011

(-6.43) 0.782***

(0.05) 0.045 (1.28) −0.006

(8.72) −0.029 (-1.35) 0.008**

(-1.34) 96.70

(2.07) 96.70

(1.63) 0.971

(-2.15) 0.045*** (2.72) −0.017 (-0.41) 0.316*

(1.96) 0.106*** (2.79) 0.027*** (3.28) 90.90

(0.97) 0.645

(-2.14) 0.019

(0.69) −0.088 (-1.51) 0.305

(0.86) 0.061 (1.04) 0.032*** (3.64) 23.80

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels when considering all well-known timing abilities of mutual fund managers. The illiquidity measure in this study is the Amihud-adjusted illiquidity. csk is the estimated coefficient of the coskewness risk factor. mt , mkt , and vol are the estimated coefficients of the liquidity timing ability, market timing ability, and volatility timing ability, respectively. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worst-performing funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (best-performing funds), and difference between the best-performing and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. t-values obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. 9

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

Table 5 Bank-related and nonbank-related mutual funds. Bank-related Mutual Funds

Nonbank-related Mutual Funds

0mp

Full 0.916***

P1 0.918***

P5 0.909***

P10 0.892***

P10-P1 −0.026

Full 0.914***

P1 0.934***

P5 0.926***

P10 0.857***

P10-P1 −0.077**

smb

(59.21) 0.982***

(35.55) 1.142*

(51.75) 0.883**

(25.87) 1.492***

(-0.63) 0.350

(59.42) 1.323***

(27.16) 1.219**

(56.74) 1.721***

(24.56) 2.048***

(-2.09) 0.829

(1.44) −0.029

(0.98) 0.146

hml

mom csk vol

mkt mt pt

Adj-R2

(3.49) 0.319

(-0.17) 0.200***

(1.91) 0.485

(2.17) 0.183

(3.09) 0.243

(0.45) −0.242

(4.84) 0.520**

(2.30) 0.469

(4.20) 0.786**

(5.19) 0.799**

(1.35) 0.330

(0.37) 0.181***

(-0.15) 0.229***

(-1.84) 0.294***

(-2.05) 0.113

(-0.25) 0.203***

(-0.44) 0.243***

(-0.20) 0.183***

(-2.52) 0.324***

(-0.97) 0.081

(9.39) 0.009

(2.98) 0.070*

(3.16) 0.024* (1.91) −0.003

(-2.81) −0.099** (-2.57) −0.031

(0.23) 0.226***

(-0.32) 96.80

(1.65) −0.425***

(-1.41) 88.00

(1.07) −0.029

(11.98) 0.016 (0.57) 0.116

(1.29) 0.021 (0.91) −0.003 (-0.26) 96.80

(0.64) −0.467*

(7.50) −0.041* (1.85) 0.442*** (4.82) 0.094** (2.47) 0.032** (2.37) 91.70

(-0.33) −0.613**

(1.52) −0.110** (2.35) 0.560*** (4.46) 0.193*** (4.40) 0.062*** (3.90) 13.30

(2.28) −0.044

(7.19) 0.004

(0.16) 0.246*** (4.45) 0.007 (1.39) −0.002 (-0.23) 96.80

(1.13) −0.160

(2.03) −0.039

(2.12) −0.519**

(0.56) −0.359

(6.40) 0.033

(4.45) −0.028

(7.09) −0.015

(1.42) −0.048

(-1.23) −0.062 (-1.19) −0.011

(2.84) −0.016 (-1.01) −0.005

(4.20) 0.029 (0.55) 0.036***

(4.90) 0.091* (1.74) 0.047**

(0.49) −0.185

(-0.55) 89.50

(-0.86) 0.185***

(-0.45) 95.20

(-0.33) 0.365***

(3.51) 90.00

(-0.89) 0.436***

(2.46) 14.70

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels when considering all important abilities of mutual fund managers. The illiquidity measure in this study is the Amihud-adjusted illiquidity. csk is the estimated coefficient of the coskewness risk factor. mt , mkt , and vol are the estimated coefficients of the liquidity timing ability, market timing ability, and volatility timing ability, respectively. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worst-performing funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (bestperforming funds), and difference between the best-performing and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. t-values obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

based on their relationship with a commercial bank. This line of research is mostly neglected in prior studies. Massa and Rehman (2008); Mehran and Stulz (2007); Hao and Yan (2012), and Berzins et al. (2013) argue that bank-related funds have information advantage over nonbank-related funds. They can enhance their mutual fund performances by using the superior information obtained from their affiliated banks. The funds can potentially boot their performances by strategically investing in IPO stocks when their affiliated banks are underwriters (Ritter and Zhang, 2007). Additionally, bank-related funds have less liquidity constraint because they can attract a larger fund flow than nonbank-related funds. Table 5 shows the results of the bank-related (BR) mutual funds and nonbank-related (NBR) mutual funds in Eq. (10). Obviously, only BR mutual funds can time market-wide liquidity, supporting the information advantage hypothesis that BR funds have superior information than NBR funds. Overall, both BR funds and NBR funds show significantly negative alphas but significantly positive market timing ability ( mkt ). NBR funds have lower negative alphas than BR funds. This finding is consistent with prior studies on the performance of the bank-mutual fund relationship. A negative relationship between the fund performances and coskewness risk factor is present in both groups. Adjusted-R2 is relatively higher for BR funds than NBR funds. The striking results show that the NBR funds do not show the market-wide liquidity timing ability at both the aggregate and portfolio levels. Although the BR funds possess the liquidity timing ability at the aggregate level at 10% statistical significance, the evidence at the portfolio levels demonstrates clearer evidence. Top-performing BR funds show a significantly positive liquidity timing ability, while bottom-performing BR funds show the opposite. BR fund managers can analyze their superior information to time market-wide liquidity better than NBR funds. The information advantage theory explains this phenomenon. A relationship between mutual fund and its associated commercial bank is a key determinant of the market-wide liquidity timing ability. 8. Robustness tests In this section, we perform several robustness tests. First, to ensure that our results are not sensitive to the rolling windows estimation, we construct L¯m and V¯m using the windows of 48 months, 36 months, 24 months, and 12 months, respectively. In general, shorter rolling windows estimations do not affect the results, as shown in Table 6, for the 48-month window period.28 Second, we use other popular illiquidity measures—i.e., the Amihud’s (2002) illiquidity measure, zero return day measure, and Roll’s (1984) effective spread measure—to calculate the liquidity timing factor. In general, the results are not driven by the measurement of marketwide liquidity. For example, Table 7 shows the results of the Amihud illiquidity measure, which are almost identical to those of the

28

The remaining results are available upon request. 10

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

Table 6 A robustness test: 48-month rolling window estimation.

0mp

smb hml

mom csk vol

mkt mt pt

Adj-R2

Full

P1

P5

P10

P10-P1

0.936***

0.867***

0.972***

0.944***

0.076**

(30.01) 0.911*

(28.21) 0.726**

(25.26) 0.974

(18.62) 1.765**

(2.14) 1.039

(1.51) −0.216***

(1.58) 0.006

(1.38) −0.175**

(1.69) −0.469***

(1.33) −0.475***

(1.65) 0.585

(-3.32) 0.026**

(2.11) −0.057*** (-4.37) 0.498*** (7.71) 0.001* (1.69) 0.014*** (3.60) 97.30

(2.29) 0.361

(1.39) 0.671

(0.05) 0.021

(-1.99) 0.026

(1.29) 0.030

(1.46) −0.066***

(1.07) 0.185**

(-5.80) 0.658***

(2.06) 0.000 (0.03) −0.005

(8.31) 0.001 (0.75) 0.010**

(-0.92) 96.70

(2.04) 96.70

(2.11) 1.097*

(-4.21) 0.036***

(2.92) −0.120*** (-3.89) 0.694*** (6.07) 0.003* (1.85) 0.034*** (5.46) 90.60

(1.41) 0.736

(-3.59) 0.015

(0.68) −0.150*** (-4.28) 0.509*** (4.03) 0.003 (1.45) 0.039*** (4.99) 23.70

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels when considering all well-known timing abilities of mutual fund managers. The illiquidity measure in this study is the Amihud-adjusted illiquidity. csk is the estimated coefficient of the coskewness risk factor. mt , mkt , and vol are the estimated coefficients of the liquidity timing ability, market timing ability, and volatility timing ability, respectively. The length of the estimation window for the liquidity timing ability and volatility timing ability is 48 months. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worst-performing funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (best-performing funds), and difference between the best-performing and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. t-values obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Table 7 A robustness test: Amihud’s (2002) illiquidity measure. Full

P1

P5

P10

P10-P1

0mp

0.930***

0.867***

0.966***

0.932***

0.065

smb

(30.50) 0.866*

(32.12) 0.705***

(26.68) 0.992

(21.63) 1.508*

(1.31) 0.803

(1.59) −0.176***

(1.66) 0.005

(1.37) −0.139*

(1.54) −0.405***

(1.32) −0.409**

hml

mom csk vol

mkt mt pt

Adj-R2

(1.67) 0.565

(-2.86) 0.024*

(1.75) −0.064* (-1.90) 0.556*** (3.31) −0.001 (-0.28) 0.011*** (3.04) 97.30

(2.72) 0.357*

(1.49) 0.667

(0.05) 0.023

(-1.73) 0.019

(1.07) 0.048

(0.96) −0.119***

(0.68) 0.093

(-4.34) 0.945***

(0.22) 0.003 (0.25) −0.005

(6.12) −0.010** (-2.14) 0.007

(-1.16) 95.80

(1.61) 94.60

(1.70) 1.025

(-3.45) 0.047*** (2.58) 0.020

(0.29) 0.055*

(1.83) 0.023* (1.80) 0.031*** (5.25) 87.50

(1.10) 0.668

(-2.44) 0.024

(0.74) −0.028 (-0.25) −0.038

(-0.05) 0.020 (0.98) 0.035*** (4.04) 22.80

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels when considering all well-known timing abilities of mutual fund managers. The illiquidity measure in this study is Amihud’s (2002) illiquidity. csk is the estimated coefficient of the coskewness risk factor. mt , mkt , and vol are the estimated coefficients of the liquidity timing ability, market timing ability, and volatility timing ability, respectively. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worst-performing funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (best-performing funds), and difference between the best-performing and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. t-values obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

11

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

Table 8 A robustness test: impact of financial crisis. Full 0mp

smb hml

mom csk vol

mkt mt pt

Adj-R2

P1

P5

P10

P10-P1

0.971***

0.921***

0.979***

0.974***

0.054

(70.80) 0.908**

(29.77) 0.778**

(40.81) 1.033***

(24.63) 1.505**

(1.03) 0.728

(1.94) −0.033

(1.24) 0.013

(2.14) −0.011

(1.47) −0.499***

(0.40) −0.512

(2.49) 0.489*

(-0.27) 0.199*** (2.76) −0.016 (-0.59) 0.038

(1.23) −0.001 (-0.45) −0.000 (-0.02) 97.80

(2.21) 0.445

(2.72) 0.534**

(0.04) 0.224*

(-0.09) 0.189***

(1.98) −0.019

(2.89) −0.031

(-0.42) 0.192***

(-0.83) 0.025

(3.81) −0.010** (-2.06) −0.022

(0.66) −0.002 (-0.60) −0.003

(-1.33) 91.40

(-0.34) 97.60

(2.27) 0.679

(-3.24) 0.292** (2.03) −0.016 (-0.47) 0.009

(0.15) 0.000 (0.07) 0.039*** (5.49) 93.40

(1.04) 0.234

(-1.45) 0.068* (0.97) 0.003

(0.06) −0.184** (-2.47) 0.010** (2.27) 0.061*** (3.40) 13.80

This table demonstrates the liquidity timing ability model at the aggregate and portfolio levels when considering all well-known timing abilities of mutual fund managers and excluding the financial crisis period (July to December 2008). The illiquidity measure in this study is the Amihudadjusted illiquidity. csk is the estimated coefficient of the coskewness risk factor. mt , mkt , and vol are the estimated coefficients of the liquidity timing ability, market timing ability, and volatility timing ability, respectively. The length of the estimation window for the liquidity timing ability and volatility timing ability is 48 months. P1, P5, P10, and P10-P1 are the 1st decile portfolio (worst-performing funds), 5th decile portfolio (average-performing funds), 10th decile portfolio (best-performing funds), and difference between the best-performing and worst-performing funds, respectively. p is the portfolio abnormal return. Adjusted R squared (Adj-R2) values are in percentages. t-values obtained by the Newey and West (1987) procedure are shown in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

adjusted-illiquidity measure.29 Finally, the market liquidity is obviously drained during the crisis period. As suggested by Cao et al. (2013b), we exclude the data between July to December 2008, and the results shown in Table 8 are largely similar to the those of the full sample period. 9. Conclusion and summary Recent studies have shown the important role of the liquidity timing ability of mutual funds; however, they are limited in developed markets. In this study, we propose a new liquidity timing ability model in the higher moment framework. Our study contributes to prior literature at least by fourfold. First, our proposed liquidity timing model allows for the coskewness risk factor, which is negligible in prior literature. To the best of our knowledge, this is the first paper to provide evidence in the selected developing country (Thailand). Our evidence suggests that mutual funds change their portfolios' exposure to the market according to the market-wide liquidity. Second, we select four low-frequency illiquidity measures, namely, the Amihud’s (2002) illiquidity measure, adjusted-illiquidity measure, zero return day measure, and Roll’s (1984) effective spread measure, which are widely used in studies of liquidity in emerging markets. The adjusted-illiquidity measure is best to capture the market-wide illiquidity in our setting. Third, at the portfolio level, high-performing funds, but not low-performing funds, show positive liquidity timing ability. They increase (decrease) the funds' exposure to the market during a high (low) market liquidity period and time the market-wide liquidity. Finally, only bank-related mutual funds possess liquidity timing ability, supporting the information advantage hypothesis that they utilize superior information obtained from their affiliated banks. Our results are robust given several checks. Future studies should examine the liquidity timing ability in different funds’ objectives because prior literature notes significantly different performances. References Acharya, V.V., Pedersen, L.H., 2005. Asset pricing with liquidity risk. J. Financ. Econ. 77 (2), 375–410. Adcock, C.J., Shutes, K., 2005. An analysis of skewness and skewness persistence in three emerging markets. Emerg. Mark. Rev. 6 (4), 396–418. Admati, A.R., Bhattacharya, S., Pfleiderer, P., Ross, S.A., 1986. On timing and selectivity. J. Finance 41 (3), 715–730. Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series effects. J. Financ. Mark. 5 (1), 31–56. Amihud, Y., Mendelson, H., 1986. Asset pricing and the bid-ask spread. J. Financ. Econ. 17 (2), 223–249. Aragon, G., Strahan, P.E., 2012. Hedge funds as liquidity providers. J. Financ. Econ. 103 (3), 570–587. Bae, K.H., Lim, C., Wei, K.C.J., 2006. Corporate governance and conditional skewness in the world’s stock markets. J. Bus. 79 (6), 2999–3028. Bauer, R., Otten, R., Rad, A.T., 2006. New Zealand mutual funds: measuring performance and persistence in performance. Account. Financ. 46 (3), 347–363.

29

The results of the other illiquidity measures are available upon request. 12

Research in International Business and Finance 51 (2020) 101105

W. Wattanatorn, et al.

Bekaert, G., Erb, C.B., Harvey, C.R., Viskanta, T.E., 1998. Distributional characteristics of emerging market returns and asset allocation. J. Portf. Manage. 24 (2), 102–116. Bekaert, G., Harvey, C.R., 1997. Emerging equity market volatility. J. Financ. Econ. 43 (1), 29–77. Bekaert, G., Harvey, C.R., Lundblad, C., 2007. Liquidity and expected returns: lessons from emerging markets. Rev. Financ. Stud. 20 (6), 1783–1831. Benson, K., Faff, R., Smith, T., 2015. Injection liquidity into liquidity research. Pacific-basin Financ. J. 35 (Part B), 533–540. Berzins, J., Liu, C.H., Trzcinka, C., 2013. Asset management and investment banking. J. Financ. Econ. 110 (1), 215–231. Bettman, J.L., Sault, S.J., von Reibnitz, A.H., 2010. The impact of liquidity and transaction costs on the 52-week high momentum strategy in Australia. Aust. J. Manag. 35 (3), 227–244. Bodson, L., Cavenaile, L., Sougné, D., 2013. A global approach to mutual funds market timing ability. J. Empir. Finance 20, 96–101. Bollen, N.P., Busse, J.A., 2001. On the timing ability of mutual fund managers. J. Finance 56 (3), 1075–1094. Bolton, P., Freixas, X., Shapiro, J., 2007. Conflicts of interest, information provision, and competition in the financial services industry. J. Financ. Econ. 85 (2), 297–330. Brown, S., Rhee, S.G., Zhang, L., 2008. The return to value in Asian stock markets. Emerg. Mark. Rev. 9 (3), 194–205. Busse, J.A., 1999. Volatility timing in mutual funds: evidence from daily returns. Rev. Financ. Stud. 12 (5), 1009–1041. Canela, M., Collazo, E.P., 2007. Portfolio selection with skewness in emerging market industries. Emerg. Mark. Rev. 8 (3), 230–250. Cao, C., Chen, Y., Liang, B., Lo, A.W., 2013a. Can hedge funds time market liquidity? J. Financ. Econ. 109 (2), 493–516. Cao, C., Simin, T.T., Wang, Y., 2013b. Do mutual fund managers time market liquidity? J. Financ. Mark. 16 (2), 279–307. Chang, E.C., Lewellen, W.G., 1984. Market timing and mutual fund investment performance. J. Bus. 57–72. Charoenrook, A., Pavabutr, P., 2017. A window into Thai mutual fund managers’ perception and decision making process. Rev. Pacific Basin Financ. Mark. Policies 20 (3), 1750020. Chunhachinda, P., Dandapani, K., Hamid, S., Prakash, A.J., 1997. Portfolio selection and skewness: Evidence from international stock markets. J. Bank. Financ. 21 (2), 143–167. Comer, G., Larrymore, N., Rodriguez, J., 2008. Measuring the value of active fund management: the case of hybrid mutual funds. Manag. Financ. 35 (1), 63–77. Doan, P., Lin, C.-T., Zurbruegg, R., 2010. Pricing assets with higher moments: evidence from the Australian and us stock markets. J. Int. Financ. Mark. Inst. Money 20 (1), 51–67. Docherty, P., Chan, H., Easton, S., 2013. Australian evidence on the implementation of the size and value premia. Account. Financ. 53 (2), 367–391. Ferson, W.E., Schadt, R.W., 1996. Measuring fund strategy and performance in changing economic conditions. J. Finance 51 (2), 425–461. Foran, J., O’Sullivan, N., 2017. Mutual fund skill in timing market volatility and liquidity. Int. J. Finan. Econ. 22, 257–273. Frye, M.B., 2001. The performance of bank‐managed mutual funds. J. Financ. Res. 24 (3), 419–442. Galagedera, D.U.A., Brooks, R.D., 2007. Is co-skewness a better measure of risk in the downside than downside beta?: Evidence in emerging market data. J. Multinatl. Financ. Manag. 17 (3), 214–230. Gómez-Bezares, F., Przychodzen, W., 2018. Bank-affiliated mutual fund managers’ trading patterns of parent banks’ stocks: international evidence. J. Behav. Financ. 19 (2), 199–208. Hao, Q., Yan, X., 2012. The performance of investment bank-affiliated mutual funds: Conflicts of interest or informational advantage? J. Financ. Quant. Anal. 47 (3), 537–565. Harvey, C.R., Liechty, J., Liechty, M.W., Müeller, P., 2010. Portfolio selection with higher moments. Quant. Finance 10 (5), 469–485. Harvey, C.R., Siddique, A., 2000. Conditional skewness in asset pricing tests. J. Finance 1263–1295. Hearn, B., 2010. Time varying size and liquidity effects in South Asian equity markets: a study of blue-chip industry stocks. Int. Rev. Financ. Anal. 19 (4), 242–257. Holmström, B., Tirole, J., 2001. LAPM: a liquidity‐based asset pricing model. J. Finance 56 (5), 1837–1867. Jean, W.H., 1971. The extension of portfolio analysis to three or more parameters. J. Financ. Quant. Anal. 6 (1), 505–515. Jean, W.H., 1973. More on multidimensional portfolio analysis. J. Financ. Quant. Anal. 8 (3), 475–490. Jiang, G.J., Yao, T., Yu, T., 2007. Do mutual funds time the market? Evidence from portfolio holdings. J. Financ. Econ. 86 (3), 724–758. Jun, S.-G., Marathe, A., Shawky, H.A., 2003. Liquidity and stock returns in emerging equity markets. Emerg. Mark. Rev. 4 (1), 1–24. Kang, W., Zhang, H., 2014. Measuring liquidity in emerging markets. Pacific-basin Financ. J. 27, 49–71. Karstanje, D., Sojli, E., Tham, W.W., van der Wel, M., 2013. Economic valuation of liquidity timing. J. Bank. Financ. 37 (12), 5073–5087. Kearney, C., 2012. Emerging markets research: trends, issues and future directions. Emer. Market Rev. 13 (2), 159–183. Kon, S.J., 1983. The market-timing performance of mutual fund managers. J. Bus. 323–347. Kon, S.J., 1984. Models of stock returns–A comparison. J. Finance 39 (1), 147–165. Kon, S.J., Jen, F.C., 1979. The investment performance of mutual funds: an empirical investigation of timing, selectivity, and market efficiency. J. Bus. 263–289. Kostakis, A., Muhammad, K., Siganos, A., 2012. Higher co-moments and asset pricing on London stock exchange. J. Bank. Financ. 36 (3), 913–922. Kraus, A., Litzenberger, R.H., 1976. Skewness preference and the valuation of risk assets. J. Finance 31 (4), 1085–1100. Lam, K.S., Tam, L.H., 2011. Liquidity and asset pricing: evidence from the Hong Kong stock market. J. Bank. Financ. 35 (9), 2217–2230. Lambert, M., Hübner, G., 2014. Size Matters, Book Value Does Not! The Fama-french Empirical CAPM Revisited. The Fama-french Empirical CAPM Revisited. Available at. http://2015.eurofidai.org/Lambert_2014_1725.pdf. Lesmond, D.A., 2005. Liquidity of emerging markets. J. financ. econ. 77 (2), 411–452. Lesmond, D.A., Ogden, J.P., Trzcinka, C.A., 1999. A new estimate of transaction costs. Rev. Financ. Stud. 12 (5), 1113–1141. Levy, H., Sarnat, M., 1972. Investment and Portfolio Analysis. Wiley, New York. Li, B., Stork, T., Chai, D., Ee, M.S., Ang, H.N., 2014. Momentum effect in Australian equities: revisit, armed with short-selling ban and risk factors. Pacific-basin Financ. J. 27, 19–31. Massa, M., Rehman, Z., 2008. Information flows within financial conglomerates: evidence from the banks–mutual funds relation. J. Financ. Econ. 89 (2), 288–306. Mehran, H., Stulz, R.M., 2007. The economics of conflicts of interest in financial institutions. J. Financ. Econ. 85 (2), 267–296. Mills, T.C., 1995. Modelling skewness and kurtosis in the London stock exchange FT-SE index return distributions. J. R. Stat. Soc. Ser. D 44 (3), 323–332. Moreno, D., Rodríguez, R., 2009. The value of coskewness in mutual fund performance evaluation. J. Bank. Financ. 33 (9), 1664–1676. Nathaphan, S., Chunhachinda, P., 2012. Determinants of growth for Thai mutual fund industry. Int. Res. J. Financ. Econ. 86, 120–131. Newey, W.K., West, K.D., 1987. Hypothesis testing with efficient method of moments estimation. Int. Econ. Rev. (Philadelphia) 777–787. Pástor, Ľ., Stambaugh, R.F., 2003. Liquidity risk and expected stock returns. J. Polit. Econ. 111 (3), 642–685. Peiró, A., 1999. Skewness in financial returns. J. Bank. Financ. 23 (6), 847–862. Prommin, P., Jumreornvong, S., Jiraporn, P., 2014. The effect of corporate governance on stock liquidity: the case of Thailand. Int. Rev. Econ. Financ. 32, 132–142. Ritter, J.R., Zhang, D., 2007. Affiliated mutual funds and the allocation of initial public offerings. J. Financ. Econ. 86 (2), 337–368. Roll, R., 1984. A simple implicit measure of the effective bid-ask spread in an efficient market. J. Financ. 39 (4), 1127–1139. Romacho, J.C., Cortez, M.C., 2006. Timing and selectivity in Portuguese mutual fund performance. Res. Int. Bus. Financ. 20 (3), 348–368. Samuelson, P.A., 1970. The fundamental approximation theorem of portfolio analysis in terms of means, variances and higher moments. Rev. Econ. Stud. 37 (4), 537–542. Sirri, E.R., Tufano, P., 1998. Costly search and mutual fund flows. J. Financ. 53 (5), 1589–1622. Smith, D.R., 2007. Conditional coskewness and asset pricing. J. Empir. Finance 14 (1), 91–119. Tchamyou, V.S., Asongu, S.A., 2017. Conditional market timing in the mutual fund industry. Res. Int. Bus. Financ. 42, 1355–1366. Treynor, J., Mazuy, K., 1966. Can mutual funds outguess the market? Harv. Bus. Rev. 44 (4), 131–136.

13