Substitution and Competition Among Different Types of
Open-end Mutual Funds: Evidence from China
RUOCHEN HU
∗August 11, 2019
ABSTRACT
Previous literature shows that either mutual funds with similar characteristics or their managers compete intensively with each other for drawing investments. However, little literature stresses competition among different types of funds. By investigating the change in the fund demand after the rise of money market funds in China, I find that the rise of money market funds substitutes the demand for other types of open-end funds. Moreover, the substitution effect is significant between money market funds and funds with less risky assets because they share similar return and risk patterns. Accordingly, actively managed funds are aware of such substitutions and increase their risk-taking and trading expenses to defend. These defensive strategies exaggerate their comparative advantages over money market funds, which is in line with industrial organization models about the entry deterrence. My finding consistently demonstrates that the competition exists among different types of open-end mutual funds in China.
∗This is my master thesis for Research Master in Economic and Business, the University of Groningen, the Netherlands. I
1.
Introduction
The mutual fund has been among the most successful financial products for individual investors over the past decades (Gruber, 1996; Khorana et al., 2005). The worldwide total net assets managed by open-ended mutual funds doubled from 29 trillion USD in 2010 to 49 trillion USD at the end of 2017 (Investment Company Fact Book, 2018). However, the development of the fund market of China conflicts with this worldwide trend. On the one hand, the total scale of mutual funds increased slowly in China despite its fast-growing capital market (Investment Company Fact Book, 2018; SSE Annual Report, 2018). On the other hand, the total net asset of money market funds boomed from 368 billion CNY at the end of 2012 to 4,968 billion CNY at the end of 2018. This fact leads to the research question of whether the recent rise of money market funds in China inhibits the growth of other mutual funds.
I propose that there is competition between money market funds and other mutual funds because of the potential substitution effect between these funds. First, in China, money market funds and other funds are distributed with the same type of fund retailers. During the rise of money market funds, their fund families intensively advertised “how fantastic” money market funds are, which in turn influenced the behaviour of fund investors by changing their risk preferences (Chuang and Schechter, 2015) or through anchoring effects (Jain and Wu, 2000; Dhami, 2016). Second, previous research shows that fund investors are informative and tend to allocate their money among different types of funds (Chalmers and Phillips, 2013). Hence, as money market funds become more and more attractive, fund investors may shift their money from other funds to money market funds.
If the substitution effect exists, the rise of money market funds will lead to two observations in the fund market of China. First, the rise of money market funds competes for investments with other funds. Funds which face high competitive pressures from money market funds will receive lower subsequent invest-ments (Wahal and Wang, 2011). Second, funds under competition will apply defensive strategies such as cut down fee rates or change their risk-taking to mitigate the substitution effect (e.g. Christoffersen and Musto, 2002).
the money market funds (Idzorek and Bertsch, 2004). Intuitively, a fund faces higher competitive pressures when money market funds attract investors and when its return pattern is similar to money market funds.
Second, I use the widely applied measure “fund flow” as a proxy for the demand for funds (e.g. Sirri and Tufano, 1998; Chevalier and Ellison, 1997). I then regress one quarter ahead fund flows on current competitive pressures to check whether the demand for open-end funds decreases when they face higher competitive pressures from money market funds. Third, I check whether other types of funds react to competitive pressures by changing their risk-taking. Because the fund fee rates are fixed in China, changing risk-taking is the most feasible way for funds to react. The research sample in the latter phase is the actively managed hybrid funds and equity funds, and I use the percentage of stock holdings to proxy for their risk-taking.
The finding confirms the existence of the substitution effect between money market funds and other funds in China, by showing that money market funds induce other funds to change their risk-taking. First, I find that the competitive pressure from money market funds is negatively correlated with one quarter ahead flows of other funds, and such correlations are significant only after the rise of money market funds. This finding is in line with the competition of investors’ money across different types of funds, which demonstrates that the rise of money market funds substitutes the demand for other funds. Moreover, the result shows that the substitution effect is more pronounced between money market funds and funds with low risk-taking. Second, I find that actively managed hybrid funds and equity funds react to the rise of money market funds by increasing their risk-taking to distinct themselves from money market funds. Such findings suggest equity funds and hybrid funds are aware of the competitive pressure, and apply defensive strategies which mitigate the substitution effect.
This thesis contributes to previous literature on fund competitions by demonstrating the existence of competition between different types of funds. Hoberg et al. (2017) suggest that the literature about competi-tion among funds are less developed, and a considerable stream of it only focuses on the competicompeti-tion between similar funds (Christoffersen and Musto, 2002; Khorana and Servaes, 2004; Wahal and Wang, 2011; Chua et al., 2018). All these studies identify peers based on the assumption that the fund market is segmented, which comes from the heterogeneous taste of fund investors. Even though such assumptions may be consis-tent with the evidence from developed markets with sophisticated fund investors, fund investors in emerging markets are largely confined with behavioural biases (Huang et al., 2007; Ferreira et al., 2012). My findings reveal the fact that in China, the taste of fund investors is affected by advertising, which induces competition even between different types of funds. Similar patterns might be found in other emerging markets, which provide avenues for future research.
China and the rise of money market funds. Section 3 reviews relevant literature and develops the hypothesis. Section 4 explains the research methods, and Section 5 explains and presents the data. Section 6 and 7 presents and discusses my empirical results. Section 8 discusses the limitations and section 9 concludes the thesis.
2.
Institutional Background: The Mutual Fund Market of China
2.1.
Mutual Funds and Their History of Development in China
Mutual funds are financial products that pool money from many fund investors and manage the pooled money (i.e., the portfolios) for them. As a reward, mutual funds charge management fees that are propor-tional to the size of the portfolios (i.e., the total net asset or the TNA) from fund investors. Moreover, mutual funds also charge front-end loads (back-end loads) that is proportional to the dollar amount of the purchase (redemption) from fund investors. Similar incentive schemes are applied to the fund manager’s salaries as well (Ma et al., 2019). According to such incentive schemes, one of the main goals of mutual funds and their managers is to draw ample attention from investors and to get as much of their investments as possible (Ibert et al., 2017).
In China, the formal name of mutual funds is “security investment fund”, which includes equity funds, hybrid funds, bond funds, and money market funds (MMFs). According to the regulation of China Securities Regulation Commission (CSRS), the equity funds have to invest at least 80% of their assets into stock markets, and the hybrid funds can diversify between stocks and fixed incomes (bonds) and can invest at most 95% of their assets into stock markets. These two types of funds tend to take higher risk than bond funds and MMFs, and their performances are more dependent on the stock market. In this respect, they are more similar to mutual funds in the US. In contrast, bond funds and MMFs are designed for investors who prefer low risks and have stable incomes. The bond funds (or MMFs) can only be invested in the bond market (or the money market).
In China, all these four types of funds are distributed by fund retailers and are always together referred to as “funds” by most Chinese investors, commercial websites, databases, and academic papers. Moreover, they are all specially designed for individual investors who want to diversify among securities, which is the same as the mutual funds in the US. Therefore in this thesis, I do not specifically distinguish between the term “security investment funds” and “mutual funds”.
fund named “Huaan Creativity”. For the first time, this offers the vast majority with limited budgets a chance to participate in the stock market and diversify among many stocks. Because of the huge demand of small investors, “stock market oriented” funds grow rapidly afterwards until the 2008 stock market crisis, as is shown in Figure 1 panel C. After the stock market crisis, though the number of hybrid funds keeps increase (Figure 1 panel A), the aggregate TNA of hybrid funds and equity funds is more invariant than it was before the 2008 crisis.
Figure 1. The number and total net asset of different types of funds from 2005 to 2017. Note: The above figure reports how the number and the total net asset (TNA) value of different types of funds evolve from 2005 to 2017 using quarterly data. Different colours from darkest blue to lightest blue indicate money market funds (MMFs), bond funds, hybrid funds and equity funds respectively. Panel A reports the numbers of existing funds for different types from 2005 Q1 to 2018 Q4, and panel B reports the relative numbers of existing funds for the same time period. Panel C reports the aggregate TNAs of existing funds for different types from 2005 Q1 to 2018 Q4, and panel D reports the relative values of aggregate TNAs for different types of funds for the same time period.
2.2.
The Rise of Money Market Funds in China
total net asset (TNA) of MMFs is no more than 20% in the total volume of the mutual fund market between 2007 and 2013 when other types of funds grow rapidly. However, there are still three small rises of MMFs before 2013, and all of them are caused by fund investors shifting their money from stocks, equity funds, and hybrid funds to MMFs because of the bad performance of the stock market (see in Table IV for 2005, 2008 and 2011).
The true rise of MMFs started on June 2013 when Alibaba initialized its MMF named as “Tianhong Yu’ebao” (code: 000198) which could be directly purchased via the most popular Chinese digital payment software Alipay. Due to such a good promotion and the stable return of Yu’ebao itself, Tianhong Yu’ebao grew very fast and has now become the largest MMF in the world (ICI Factbook, 2019). Such a huge success led to the subsequent rise of MMFs in China, after which almost all other fund companies began to promote their own MMFs to their extent fund investors. As a result, the proportion of MMFs’ TNA in all investment funds grew from around 10% before 2013 to more than 60% at the end of 2018 (Figure 1 panel D), which has now become the largest “player” of the Chinese mutual fund market only within only six years.
The rise of MMFs is largely attributed to individual investors. According to the annual report of Yu’ebao, at the end of 2018, there are 559 million people holding Yu’ebao, which are more than half of the concurrent labour force in China (902 million, China National Bureau of Statistics). Individual investors are willing to invest a large amount of money into MMFs. The 2018 annual report of Yu’ebao shows that nine out of ten largest holders are individual investors, and their holding shares add up to 450 million CNY (65 million USD). These statistics show the rise of MMFs successfully attracts individual investors.
3.
Literature Review and Hypothesis Development
3.1.
Mutual Funds and Demand Substitution
Mutual funds tend to compete fiercely on drawing new investments (i.e., fund flows), which comes from the substitution effect between funds. The substitution effect means that the increase in demand for certain funds decreases the demand for its rival funds. The substitution effect is likely to influence the demand for funds because compared to products in other markets, mutual funds are more similar to each other, and there are fewer entry barriers in the mutual fund market (Baumol et al., 1986).
Previous literature has documented that the substitution effect exists between index funds which hold almost identical portfolios (e.g. Horta¸csu and Syverson, 2004), and between funds which have similar styles (Wahal and Wang, 2011; Hoberg et al., 2017). Wahal and Wang (2011) find that incumbent funds which have higher overlaps with entrant funds receive lower future investments from fund investors. More-over, the substitution effect is more pronounced when the fund market is highly competitive, and when there are many incumbent funds. Hoberg et al. (2017) show that the competition from other funds with similar investment styles may drive down the performance of funds. As a result, the decrease in performances leads to the decreases in the future flows because fund investors are sensitive to the performance of funds and tend to chase the “past winner” (Ippolito, 1992; Chevalier and Ellison, 1997; Ferreira et al., 2012).
However, previous literature does not stress whether the substitution effect exists across funds which hold different types of assets. This ignorance comes from their measures of the “competitive pressure”. For example, Wahal and Wang (2011) measure the competitive pressure of funds with the overlap in fund assets between competitors, and Hoberg et al. (2017) measure the competitive pressure with the number of funds which hold similar assets. Such holding-based measures fail to capture the flow competition between funds which hold different types of assets. According to their logic, in China, funds which hold bonds and stocks should feel little competitive pressure from the rise of money market funds (MMFs) because the asset overlap between these funds and MMFs is small.
preference of (individual) investors is unstable (Chuang and Schechter, 2015). Hence, fund investors are likely to be affected by MMF advertisements and become more risk-averse. Accordingly, they will choose to shift part of their money from other types of mutual funds to MMFs. Third, the TNA of MMFs becomes very large soon after the rise of MMFs (60% of the TNA of the investment fund market), which helps to draw the attention of fund investors as well (Sirri and Tufano, 1998). Therefore, the rise of MMFs may be accompanied by the future decrease in demand for other mutual funds. Such mechanisms make the MMFs become substitutes of other types of mutual funds.
Hypothesis 1 : The money market funds and other types of mutual funds are substitutes for fund investors.
I propose the substitute effect between MMFs and other mutual funds varies for funds hold different types of assets. This variation comes from the heterogeneous risk preference of fund investors (Ferson and Lin, 2014), which shares the same theoretical basis with previous literature about flow competitions (e.g. Hoberg et al., 2017). Fund investors who invest in funds with different risk-taking have different risk preferences. Fund investors who invest in risky funds ask for a lower risk premium and are less affected by the rise of MMFs than fund investors who invest in safe funds (Dhami, 2016). As a result, safe funds (e.g., funds which mainly hold bonds) are easier to be substituted by the MMFs than risky funds (e.g., funds which mainly hold stocks).
Hypothesis 2 : The substitution effect varies across mutual funds which hold different types of assets. Last but not least, the substitution effect is different from the “hedge behaviour” of fund investors. Chalmers and Phillips (2013) find that fund investors allocate their money across different types of funds according to economic conditions, which leads to correlations between concurrent aggregate flows of different types of funds. Such observations are driven by fund investors’ hedging market risks, while not because of the substitution effect. The substitution effect comes from the entry of new rivals which draws the attention of consumers (i.e., fund investors) and affects their decision making. Therefore, the substitution effect should be captured by how the current competitive pressure from rival funds affect the future demand of funds under competition. This is the theoretical basis of the regression approaches in this thesis.
3.2.
Fund Reaction
draw new investments (Christoffersen and Musto, 2002; Wahal and Wang, 2011). Such decreases in fee rates make funds “cheaper” and help to draw fund investors that are sensitive to holding costs. This strategy is more likely to be applied by passively managed funds, for example index funds (Horta¸csu and Syverson, 2004). Second, funds may also change their investment styles to enter a different “pool of competitors” where the competitions are less fierce (Sensoy, 2009; Chua et al., 2018). This strategy is often applied by actively managed funds which hold stocks (i.e., equity funds and hybrid funds) and is achieved by changing risk-taking of funds’ stock portfolios (Khorana, 1996; Brown et al., 1996; Ma et al., 2019).
Similar to the research about fund demand substitution, literature about fund reactions to the competitive pressure does not stress the competition between funds with different types of assets (e.g. Brown et al., 1996). However, once the substitution effect between MMFs and other funds exists, these funds should feel the competitive pressure from MMFs and apply defensive strategies accordingly. That is, the competitive pressure should be captured by the actual demand substitution between funds, while not by the overlap of held assets between funds. Moreover, because the rise of MMFs draws ample attention of fund investors, other funds may view MMFs as “latent rivals” and apply defensive strategies (Milgrom and Roberts, 1982), even though there is no flow competition between them.
Therefore, if there is substitution effect between MMFs and other funds in China, the actively managed funds in China will react to the rise of MMFs with defensive strategies. Such strategies are the most likely to be achieved by changing their risk-taking because fee rates are fixed in China, and are difficult to be changed by fund companies. When funds are initialized, their fee rates are determined based on their categories. In this case, changing the risk-taking becomes the most feasible defensive strategy of actively managed funds, which may be achieved by changing the percentage of stocks in their assets.
Hypothesis 3 : In China, actively managed equity funds and hybrid funds react to the demand substi-tution from money market funds by changing risk-taking.
4.
Methods
4.1.
Variables
In order to investigate the substitution effect between money market funds (MMFs) and other funds, I run regressions that capture how current competitive pressure from MMFs is related to the future demand for other types of mutual funds.
et al., 2007). Hereby, f lowi,q+(1)presents the relative change in a fund’s total net asset (TNA) from quarter q to quarter q + 1 that is contributed by the purchase and the redemption of fund investors, where T N Ai,q (T N Ai,q+1) is the fund’s TNA at the end of quarter q (q + 1) and Ri,q+(1) is the fund’s return in net asset value (NAV) per share from quarter q to quarter q + 1.
f lowi,q+(1)=
T N Ai,q+(1)− T N Ai,q· 1 + Ri,q+(1) T N Ai,q· 1 + Ri,q+(1)
. (1)
The chosen measure of fund flows assumes fund flows are stable over the whole quarter, which is the most robust measure among different flow measures. Previous research (Chevalier and Ellison, 1997; Cashman and Villupuram, 2008) also proposes different flow measures, which assume flows happen at the beginning of each quarter q and replace the denominator of equation (1) with T N Ai,q. I present the regression results with these two flow measures in the Appendix with Table A1, and its result is consistent with the result with the chosen measure.
The main explanatory variable CPi,q is the competitive pressure of MMFs for fund i during quarter q. The competitive pressure is determined as the product of two factors: the similarity between fund i and MMFs (i.e. Overlapi,q) and the attractiveness of MMFs (i.e. F lowM M Fq ). By combining these two factors, CPi,q is able to capture both he cross-sectional and the time-serial variation of the competitive pressure from MMFs. The competitive pressure from MMFs tends to be high for quarters when MMFs are more attractive to fund investors, and for funds which share similar return patterns with MMFs. The calculation is shown with equation (2).
CPi,q = Overlapi,q· F lowM M Fq . (2)
In equation (2), Overlapi,q measures the “return overlap” between fund i and the value-weighted MMF index during quarter q, which is calculated with equation (3). Hoberg et al. (2017) propose that both the “return overlap” and the “holding overlap” present the similarity between funds. Because there is little holding overlap between MMFs and other funds, I choose the return overlap to measure the similarity between given funds and MMFs.
Overlapi,q = −log(Distancei,q−(1))
In equation (3), the variable Distancei,q−(1) is a proxy for the average difference between the daily return of fund i (i.e. Ri,t) and the daily return of the MMF index (i.e. RtM M F) for each day t in quarter q. Intuitively, Distancei,q−(1) measures the average “tracking error” on MMF index for funds (see e.g. Idzorek and Bertsch, 2004). Hence, if the daily return of a fund is very similar with (different to) the daily return of MMF index, this fund tends to have a low (high) value with the measure Distancei,q−(1) and accordingly, a high (low) value with the overlap measure. For the overlap measure, I apply logarithm transition to Distancei,q−(1) to mitigate the effect of extreme values.
The attractiveness of MMFs during quarter q is measured by F lowM M F
q , which is the “dollar value” of aggregate flow for all the MMFs during quarter q. The aggregation is calculated as equation (4). This variable equals to the amount of money that flows in or out of all MMFs during quarter q, which reveals the “attractiveness” of MMFs perceived by fund investors during quarter q (Warther, 1995; Christoffersen et al., 2014). A higher F lowM M F
q means that people are more attracted by the MMFs and decided to invest more money into it.
F lowM M Fq = X m∈M M F s
(T N Am,q− T N Am,q−1· (1 + Rm,q)) (4)
There are two latent concerns about this measure. First, the aggregate flow of MMFs correctly reveals the attractiveness of MMFs only if there is no structural change in the population of fund investors, which means that the proportion of rich investors or poor investors do not vary too much over time. The Gini coefficient measures the income inequality of a country, which reveals the proportion of poor and rich people (De Haan and Sturm, 2017). The Gini coefficient of China is stable from 2005 to 2018 and only varies between 0.46 and 0.49 (CEIC), which rules out this concern. Second, the aggregate flow of MMFs may be affected by the performance of the stock market (e.g. the three small rise of MMFs before 2013), which could blur the estimation. I will check this “hedge behaviour” of fund investors by adding interactions with stock market returns.
present variable F lowqM M F is the best proxy for the attractiveness of MMFs at quarter q under my current availability of data.
4.2.
Demand Substitution: Baseline Regression
I construct my baseline regression model with equation (5), which is similar to Wahal and Wang (2011) and Franzoni and Schmalz (2017). In addition to my main explanatory variable CPi,q, I also include control variables to make my model in line with other literature (Wahal and Wang, 2011; Ferreira et al., 2012; Spiegel and Zhang, 2013). Xi,q, wqand ziare time-varying fund level controls, time-varying market status controls and invariable fund level controls respectively.
f lowi,q+(1)= α + β1· CPi,q+ γ · Xi,q+ θ · wq+ δ · zi+ i,t (5)
Xi,q includes fund past performances, which largely affect future flows of funds because fund investors chase high returns (Chevalier and Ellison, 1997; Berk and Green, 2004). I use the peer rank of fund i based on his accumulated return during the past quarter to measure fund performances. To control for the well-documented convex flow-performance relationship (Chevalier and Ellison, 1997; Sirri and Tufano, 1998), I convert funds’ peer rank of past quarter into three variables: Highi,q−(4), M idi,q−(4) and Lowi,q−(4), which are constructed in the same way as previous research (Sirri and Tufano, 1998; Ferreira et al., 2012):
Lowi,q−(1)= min(0.2, Ranki,q−(1))
M idi,q−(1)= min(0.6, Ranki,q−(1)− Lowi,q−(1)) Highi,q−(1)= Ranki,q−(1)− (Lowi,q−(1)+ M idi,q−(1))
(6)
Ranki,q−(1)is fund i’s percentile rank of its accumulated performance during the past quarter for quarter q, where 1 means it performs better than all its peers (i.e. all the existing funds that hold the same type of assets) and 0 means it performs worse than all its peers. As a result, the parameter for Highi,q−(1), M idi,q−(1) and Lowi,q−(1) presents the marginal effect of past performances in attracting future flows for funds with top 20%, middle 60% and bottom 20% past performances respectively.
difference between young and old funds (Chevalier and Ellison, 1997; Berk and Green, 2004). log(F amilyi,q) is the logarithmic value of aggregate TNA of the fund family of fund i at the end of quarter q (Nanda et al., 2004; Del Guercio et al., 2010). log(P eeri,q) is the logarithmic number of peer funds for fund i at the end of quarter q, which controls for the substitution effect between funds with similar assets (Wahal and Wang, 2011; Hoberg et al., 2017). Std(Ri,t−(12)) is the standard deviation of fund i’s monthly return for the past 12 months at the end of quarter q (t ∈ q), which controls for his risk-taking in the near past (Wahal and Wang, 2011).
wq includes time-varying market level variables Rmq and Rmq+1 (i.e., F.Rm in tables), which is the ac-cumulated stock market return during quarter q and quarter q + 1 respectively. Warther (1995) finds that the market return affects the concurrent and future demand for funds (i.e. fund flows) and higher market returns are accompanied by unconditional higher flows. zi includes invariable fund level control F eei and asset type fixed effect indicators. F eei is the fee rates of fund i, which controls for the “price effect” of mutual funds (Huang et al., 2007). F eei is invariable over time because in China fee rates are fixed when funds are initialized, and only differ on the cross-sectional dimension. Therefore, I use the current fee rate of funds to construct F eei to control for the price effect. I also include asset type fixed effect indicators equity and hybrid that equals to 1 for equity funds and hybrid funds respectively.
4.3.
Demand Substitution: Interactions
The substitution effect between MMFs and other funds is affected by other market level variables. As I explain in section 3.1, such competitions are conducted through drawing attention of fund investors. To capture such interactions, I add interactions between my main explanatory variable (i.e. CPi,q) and other relevant market level variables (i.e. Intq), and the regression model becomes:
f lowi,q+(1)= α + β1· CPi,q+ β2· CPi,q· Intq+ β3· Intq+ γ · Xi,q+ θ · wq+ δ · zi+ i,t (7)
The substitution effect may be affected by market returns (Chalmers and Phillips, 2013), and I use the stock market return (Rm
In the same vein, the substitution effect is affected by the returns of MMFs (Investment Company Fact Book, 2018). I use the difference between the value-weighted return of MMFs and the return of stock market as Intq to control for it. I denote the return difference as Rdif fq , where Rdif fq = RM M Fq − Rmq . When returns of MMFs are high, the aggregate flows of MMFs are largely contributed by fund investors that are sensitive to returns, and they are likely to leave MMFs and go back to other funds when the return of MMFs drops. The substitution effect is also affected by the public attention to MMFs per se. Before the rise of MMFs, the total size of MMFs is relatively small and MMFs draw little attention of fund investors. Hence, the current aggregate flow of MMFs may not affect the future demand of other funds. I use the aggregate size of MMFs as Intqto control for this effect, because larger funds tend to capture more attention of investors (Sirri and Tufano, 1998). I measure the aggregate size of MMFs with absolute value log(T N AM M Fq ) which equals the logarithm of aggregate TNA of MMFs at the end of quarter q, and with relative value RatioM M Fq which equals to the aggregate TNA of MMFs divided by the aggregate TNA of all the sample funds at the end of quarter q. Moreover, after the rise of MMFs, the MMFs draw ample public attention due to intensive advertisements of Tianhong Yu’ebao and other MMFs (see section 2). Hence, I also use a dummy variable Af terq as Intq to control for the interaction from the rise of MMFs. Af terq equals to 1 for quarters q after June 2013 and equals to 0 for quarters before June 2013.
However, the aggregate flow of MMFs (CPi,q) is seriously correlated with its interactions with the aggregate size of MMFs (i.e. CPi,q· log(T N AM M F) and CPi,q · RatioM M F), which cannot be removed even with centering or Z-score transformation. Such correlations will cause multicollinearity if I include CPi,q · log(T N AM M F) or CPi,q · RatioM M F into regressions. To avoid this problem, I replace interaction CPi,q· log(T N AM M F) (CPi,q· RatioM M F) with CPi,q· DHigh (CPi,q· DLarge), where DHigh (DLarge) is a dummy which indicates whether log(T N AM M F) (RatioM M F) is larger than average or not. Under this ad-justment, the parameter of interactions (β2) presents how the substitution effect changes when the aggregate size of MMFs is large enough.
4.4.
Fund Reaction: Change in Stock Holdings
during a quarter presents the change of risk-taking. I construct my dependent variable ∆Stock as:
∆Stocki,q+1= Stocki,q+1− Stocki,q (8)
In equation (8), Stocki,q (Stocki,q+1) is the percentage of stocks fund i holds at the end of quarter q (q + 1). By replacing the dependent variable of equation (7) with ∆Stocki,q+1, the empirical model now becomes:
∆Stocki,q+1= α + β1· CPi,q+ β2· CPi,q· Intq+ β3· Intq+ γ · Xi,q+ θ · wq+ δ · zi+ i,t (9)
For equation (9) I include the current percentages of stocks (i.e. Stocki,q) into control variables Xi,q. This is because that the change of percentages of stocks (∆Stocki,q+1) is affected by the current percentages of stocks of funds (Taylor, 2003). It is more costly for funds with high percentages of stocks to increase their percentages of stocks.
4.5.
Model Estimation
I estimate my models with OLS with robust standard errors double-clustered at fund and year level, which corrects for heteroscedasticity overtime and between funds (e.g. Ferreira et al., 2012; Franzoni and Schmalz, 2017). Some previous literature corrects heteroscedasticity only between funds and reports estimations with robust standard errors clustered at fund level (e.g. Sirri and Tufano, 1998). My results are unchanged with the fund level clusters. My results are not confined with the serial correlation problem of fund flows, and the results remain unchanged with and without adding the lag flow into explanatory variables (Spiegel and Zhang, 2013).
5.
Data
5.1.
Sample Selection
The main sample is all the open-end funds (excluding money market funds) traded in the Chinese market from the beginning of 2005 to the end of 2018. I exclude funds with other operation types for several reasons. Based on operation types, funds can be categorized into open-end funds, close-ended funds, exchange-traded funds (ETFs), and qualified domestic institutional investors (QDIIs) in the mutual fund market of China. I exclude close-ended funds because they cannot be purchased or redeemed in the most period of the year. I exclude ETFs because they are traded in the stock exchanges, while MMFs are distributed by the fund retailers; MMFs compete for little flow with ETFs because they are traded in different channels. I exclude QDIIs because they invest in foreign markets and their development is less related to the domestic market. In short, I use open-end funds because they are in the same “playground” with MMFs, and are directly affected by the rise of MMFs. This sample is representative of funds in the Chinese market as more than 80% of funds are open-end funds at the end of 2018.
All the sample funds can be categorized into equity funds, hybrid funds, and bond funds based on their holding assets. Equity funds and hybrid funds tend to hold a large share of stocks in their assets and have higher risk-taking. Contrarily, the bond funds mainly hold bonds and tend to take lower risk. The difference between equity funds and hybrid funds is that equity funds have to hold at least 80% of their assets as stocks, while hybrid funds can hold 30% to 95% of their assets as stocks. This difference offers hybrid funds larger freedom to alter their risk-taking by changing the percentage of stocks.
5.2.
Data Adjustment
I collect data from two databases: JoinQuant database and EastMoney website. The data provider of EastMoney website is the Choice database. Both databases offer audited data and are adopted by many financial institutions and universities in China. The data is survival-bias free and includes all the open-end funds that ever exist in China. The fund numbers are consistent with previous literature about mutual funds in China which uses different data providers (Yang et al., 2013; Chua et al., 2018). In line with previous liter-ature, I exclude fund record with TNA being smaller than 1400 million (i.e. 200 million USD) because small funds are sensitive to large flows, which yields volatile fund flows and blurs the estimations (Kostovetsky, 2015). My results are consistent both with and without this adjustment.
the quarterly data of portfolio compositions for all funds (i.e. the percentages of stocks, bonds and cash they hold); (4) the semi-annually data of total expenses for all funds (manager’s salary, transaction fees, etc.); (5) compositions of fund investors (i.e. percentages of individual investors and of institutional investors) for all funds. I collect the daily data of market returns (HS300 index returns) from JoinQuant database.
I apply two adjustments to my collected data due to data availability. First, both databases do not offer data about returns of funds, which is necessary for calculating the quarterly flows with equation (1). Therefore, I calculate the monthly return of funds using the accumulated NAV per share from JoinQuant for all the funds, as is shown in equation (10):
Ri,t=
N AVacc
i,t − N AVi,t−1acc N AVacc
i,t−1
(10)
In equation (10), Ri,tis the monthly return for fund i in month t, and N AVi,tacc(N AVi,t−1acc ) is the accumulated NAV per share of fund i at the end (beginning) of month t. I use accumulated NAV per share rather than the NAV per share to calculate returns because it adjusts for dividends or fund splits. The quarterly return is then calculated by compounding monthly returns within the quarter. Moreover, I have checked the quarterly returns obtained in this way with the quarterly returns presented on EastMoney website, and I find they are consistent with each other.
The second adjustment is to fund families. In China, the mutual funds are “contract-based” and operated by fund companies, which means funds are more similar to products of fund companies (Chua et al., 2018). Hence, the Chinese counterpart of fund family is the fund companies, and in the following chapters, I use “fund families” to refer to fund companies in China. My data of fund family size is also clustered with fund companies.
5.3.
Descriptive statistics
Table I lists the descriptive statistics of dependent and explanatory variables, where column 1 to 3, column 4 to 6, and column 7 to 9 presents the statistics of the full time period, the time period before the rise of MMFs, and time period after the rise of MMFs respectively. The statistics are similar to previous research about mutual funds in China (e.g. Yang et al., 2013; Chua et al., 2018), which confirms the validity of my data.
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Full Before June 2013 After June 2013
VARIABLES Obs. Mean SD. Obs. Mean SD. Obs. Mean SD.
DEPENDENT VARIABLE: flow 54,940 0.37 65.14 12,782 0.18 55.62 42,158 0.42 67.77 ∆Stock 54,743 0.85 11.49 12,538 0.86 10.07 42,205 0.85 11.88 EXPLANATORY VARIABLE: CP 56,598 0.19 0.34 12,874 -0.01 0.10 43,724 0.25 0.36 Rm 56,583 0.99 13.22 12,782 1.39 17.48 43,801 0.83 11.75 Rdiff 51,929 -0.20 9.20 11,538 -0.84 9.70 40,391 0.01 9.05 log(TNAMMF) 58,417 28.00 1.26 12,905 25.90 0.51 43,801 28.59 0.61 RatioMMF 56,681 0.41 0.18 12,880 0.14 0.11 43,801 0.49 0.10 log(TNA) 56,681 20.16 1.57 12,880 21.01 1.50 43,801 19.92 1.50 log(Age) 56,681 6.49 0.96 12,880 5.38 0.87 43,801 6.83 0.70 log(Family) 56,681 24.32 1.46 12,880 23.77 1.21 43,801 24.49 1.49 log(Peer) 56,681 6.50 0.96 12,880 5.38 0.87 43,801 6.83 0.70 Fee 56,681 1.34 0.49 12,880 1.53 0.37 43,801 1.28 0.51 Std(Rt−(12)) 43,349 0.04 0.03 9,876 0.04 0.03 33,473 0.04 0.04
Table I: The descriptive statistics of variables.
other hand, fund investors become more informative efficient. The standard deviation of fund flows is larger after the rise of MMFs than before the rise, which means investors are more sensitive to new information and trade more often during the latter period (Huang et al., 2007).
Specially, I present the statistics of the variable Overlapi,q in Table II. The statistics show there are variations both within and between different types of funds. This offers cross-sectional variation to the explanatory variable CPi,q. On average, the return overlap is larger (smaller) between bond (equity) funds and MMFs. The statistics show that the average return difference between MMF index and other funds is 0.4%, while this measure is 1% (or 0.1%) for equity (or bond) funds. If fund investors pick funds based on their returns, bond funds would be the most similar “products” to MMFs.
(1) (2) (3) (4) (5) (6) (7) VARIABLES Mean SD. 10% 25% 50% 75% 90% Full 11.00 2.59 8.22 8.94 10.29 13.01 14.81 Bond funds 13.63 1.95 11.06 12.32 13.78 15.04 15.94 Hybrid funds 10.29 2.18 8.11 8.76 9.71 11.46 13.62 Equity funds 9.12 1.49 7.63 8.34 8.93 9.68 10.56
Table II: The descriptive statistics of the overlap measure.
Note: This table presents the descriptive statistics of the overlap measure (Overlapi,q) that is used to construct the main explanatory variable “competitive pressure” (CPi,q). Column 1 and column 2 presents the average and the standard deviation. Column 3 to 7 presents the 10%, 25%, 50%, 75% and 90% percentile respectively.
Year All funds Equity funds Hybrid funds Bond funds MMFs Entr./Incu. (1) (2) (3) (4) (5) (6) 2005 149 (0.36) 9 (0.01) 100 (0.15) 14 (0.01) 26 (0.19) 0.295 2006 229 (0.63) 13 (0.03) 157 (0.51) 19 (0.01) 40 (0.07) 0.349 2007 273 (2.53) 15 (0.18) 197 (2.18) 21 (0.06) 40 (0.11) 0.168 2008 360 (1.62) 17 (0.08) 249 (0.99) 54 (0.17) 40 (0.39) 0.244 2009 458 (2.07) 28 (0.17) 311 (1.56) 76 (0.08) 43 (0.26) 0.214 2010 541 (1.93) 42 (0.17) 363 (1.49) 90 (0.12) 46 (0.15) 0.152 2011 663 (1.68) 52 (0.13) 436 (1.14) 124 (0.11) 51 (0.29) 0.184 2012 842 (1.78) 68 (0.15) 504 (1.12) 178 (0.15) 92 (0.37) 0.214 2013 1,090 (1.97) 77 (0.12) 598 (1.15) 279 (0.17) 136 (0.54) 0.230 2014 1,380 (3.04) 119 (0.18) 716 (1.14) 332 (0.20) 213 (1.51) 0.218 2015 2,002 (4.66) 236 (0.30) 1,135 (1.89) 388 (0.45) 243 (2.02) 0.324 2016 2,965 (5.24) 289 (0.26) 1,643 (1.79) 725 (1.23) 308 (1.96) 0.327 2017 3,678 (7.54) 370 (0.32) 2,006 (1.76) 930 (1.26) 372 (4.20) 0.213 2018 3,567 (7.81) 383 (0.26) 1882 (1.05) 926 (1.53) 376 (4.97) 0.052
Table III: The overview of open-ended funds from 2005 to 2018.
Note: the column 1 presents the number (aggregate total net assets in brackets, and in trillion CNY) of existing open-ended funds in the Chinese market at the end of each year, and the column 2 to 5 present the number (aggregate total net assets in brackets, and in trillion CNY) of existing open-ended funds for different asset types. The column 6 presents the entrant ratio, which is the number of new entrant funds divided by the number of incumbent fund for each year.
well. On the other hand, equity funds and hybrid funds experience outflows for these two years. This shows that after 2013, the intensive advertisements of MMFs have significant impacts in influencing the behaviour of fund investors.
Year Equity funds Hybrid funds Bond funds MMFs Total flow (billion) Market return
(1) (2) (3) (4) (5) (6) 2005 -24% -510% -26% 660% 4.40 0.95% 2006 5% 37% 4% 54% -257.58 121.02% 2007 16% 74% 5% 5% 665.63 161.55% 2008 22% 199% -7% -114% -239.58 -65.95% 2009 -17% -2% 56% 64% -212.83 96.71% 2010 5% 61% 1% 32% -347.53 12.51% 2011 15% 139% 35% -89% -149.60 25.01% 2012 2% 51% 24% 27% -186.24 7.55% 2013 23% 100% 53% -76% -145.06 -7.65% 2014 -3% -55% -4% 161% 483.57 51.66% 2015 23% 177% -26% -75% -556.97 5.58% 2016 4% 123% -93% 66% -253.76 -11.28% 2017 -1% -25% -24% 150% 1,368.61 21.78% 2018 3% -393% 44% 446% 133.86 -25.31%
Table IV: The overview of aggregate flows of open-ended funds from 2005 to 2018.
Note: this table shows the “dollar value” of aggregate open-ended fund flows for each year from 2005 to 2018 (column 5), and how they are contributed by different types of funds (column 1 to 4). The column (6) shows the annual return of HS300 index (i.e. the stock market return) for each year from 2005 to 2018.
Figure 2. The average fee rates of entrant funds.
Note: this figure shows the average annual fee rates for funds that are initialized in each year from 2009 to 2018. The average rate is calculated for each year and different lines present the average rate for different types of funds. The annual fees are charged by fund companies annually and do not include the front- and back-loads that are charged to fund investors when they purchase and redeem funds.
6.
Results
6.1.
Hypothesis 1: Demand Substitution
This section presents the result about hypothesis 1, in which I hypothesized that the competitive pressure from money market funds (MMFs) decreases the future demand of other types of open-end mutual funds. The results in Table V show that the current level of competitive pressure from MMFs is negatively correlated with one quarter ahead flows of other types of funds, which indicate that the substitution effect exists between MMFs and other funds. Column 1 to 4 shows such correlations are robust after controlling for fund level effects (e.g. past performances) and market returns. Moreover, the results of regressions with interactive terms reveal such negative correlations are more pronounced when the stock market return is high and when MMFs draw enough attention of fund investors.
VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) CP -0.11*** -0.06*** -0.08*** -0.09*** -0.08*** -0.08*** -0.03 -0.11*** 0.07* 0.07 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.03) (0.03) (0.04) (0.04) CP · Rm -1.10*** (0.15) CP · Rdiff 0.65*** (0.25) Rdiff -0.18*** (0.06) CP · DHigh -0.07** (0.03) log(TNAMMF) -0.01 (0.01) CP · DLarge 0.03 (0.02) RatioMMF -0.03 (0.03) CP · After -0.17*** -0.16*** (0.05) (0.05) After 0.01 (0.01) Rm 0.11*** 0.21*** 0.12*** 0.11*** 0.11*** 0.11*** (0.03) (0.04) (0.03) (0.03) (0.03) (0.03) F.Rm 0.22*** 0.22*** 0.24*** 0.22*** 0.23*** 0.22*** 0.22*** (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) Low 0.12* 0.08 0.15** 0.15** 0.13** 0.15** 0.15** 0.15** 0.15** (0.06) (0.06) (0.06) (0.06) (0.07) (0.06) (0.06) (0.06) (0.06) Mid 0.10*** 0.10*** 0.10*** 0.09*** 0.09*** 0.10*** 0.10*** 0.10*** 0.10*** (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) High 0.67*** 0.79*** 0.70*** 0.71*** 0.68*** 0.70*** 0.70*** 0.70*** 0.70*** (0.10) (0.09) (0.10) (0.10) (0.10) (0.10) (0.10) (0.10) (0.10) log(TNA) -0.04*** -0.04*** -0.04*** -0.04*** -0.04*** -0.04*** -0.04*** -0.04*** -0.04*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) log(Age) 0.03*** 0.03*** 0.02*** 0.02*** 0.02*** 0.02*** 0.02*** 0.02*** 0.02*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) log(Family) 0.01*** 0.01*** 0.01*** 0.01*** 0.01*** 0.01*** 0.01*** 0.01*** 0.01*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) log(Peer) -0.06*** -0.04*** -0.04*** -0.04*** -0.05*** -0.03** -0.04*** -0.04*** -0.05*** (0.01) (0.00) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Std(Rt−12) 0.58*** 0.58*** 0.57*** 0.51*** 0.56*** 0.57*** 0.56*** 0.54*** (0.10) (0.11) (0.11) (0.11) (0.11) (0.11) (0.11) (0.12) Hybrid -0.12*** -0.10*** -0.11*** -0.10*** -0.11*** -0.09*** -0.10*** -0.11*** -0.11*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) Equity -0.04*** -0.04*** -0.04*** -0.04*** -0.04*** -0.05*** -0.05*** -0.04*** -0.04*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) Fee 2.63*** (0.96) Constant 0.02*** 0.75*** 0.60*** 0.66*** 0.63*** 0.73*** 0.77*** 0.64*** 0.66*** 0.69*** (0.00) (0.08) (0.07) (0.08) (0.08) (0.08) (0.16) (0.08) (0.08) (0.09) Observations 54,888 42,668 52,122 40,097 40,097 39,893 40,097 40,097 40,097 40,097 R-squared 0.00 0.02 0.02 0.03 0.03 0.03 0.03 0.03 0.03 0.03 AIC 108,418 79,671 99,891 74,917 74,857 74,603 74,916 74,919 74,912 74,914
Table V. The substitution effect between MMFs and other open-end funds.
low. The marginal effect of these results indicate that when stock market performs badly, a high aggregate flow of MMFs is contributed by investors shifting their assets from risky funds to MMFs. The high flow of MMFs driven by such hedge behaviour of fund investors is less likely to decrease the future flows of other funds because these “hedged” money will flow back to risky funds when the stock market performs better (Chalmers and Phillips, 2013). As a result, fund managers may also interpret the same amount of aggregate MMF flow as different levels of the substitution effect.
When looking at Column 7 to 10 of Table V, we can see that the substitution effect is more pronounced when MMFs draw more public attention. Column 7 shows the current competitive pressures from MMFs lead to lower future flows of other funds only when the aggregate size of MMFs is large enough. This means that MMFs start to substitute the demand for other funds only when MMFs draw enough attention from public or fund investors (Sirri and Tufano, 1998). Column 9 and 10 shows the substitution effect only exists after the rise of MMFs, which is in line with my argument that the substitution effect between MMFs and other funds is conducted through the attention of fund investors. To check the robustness, I also estimate the substitution effect with subsamples before and after the rise of MMFs (see Appendix with Table A3). The result also shows that the substitution effect only exists with the time period after the rise of MMFs, which confirms the conclusion as well. Notably, the parameter of the time dummy “Af ter” is insignificant, which means that the low flow of other funds after June 2013 is fully explained by the substitution effect between peer funds (log(P eer)) and by the rise of MMFs. This shows that more new entrant funds drive down the expected flows for each existing funds (Wahal and Wang, 2011).
The substitution effect may inhibit the growth of mutual funds other than MMFs by decreasing their expected future flows. According to the result in column 11, for a fund with average level return overlap (i.e., Overlapi,q equals to 11.00), an average quarterly inflow of MMFs (which equals to 0.19 trillion CNY or 23 billion Euros) decreases its flow of next quarter by 3.34% after the rise of MMFs. Moreover, the inflow of MMFs increases the aggregate size of MMFs (log(T N AM M F)) and the relative share of MMFs (RatioM M F
q ). According to column 7, large aggregate sizes of MMFs help MMFs to draw more attention of fund investors and increase the substitution effect between MMFs and other funds further in the future. The substitution effect could explain why the mutual fund market in China does not grow as fast as the mutual fund markets of other developing countries (ICI Factbook, 2018), despite the fast-growing financial market of China (Khorana et al., 2005).
correlated with R-squared being 0.00, which means the aggregate flow of MMFs is not driven by the stock market performance. Previous research also shows such hedge activities tend to be done quickly, and can only be captured with contemporary flow competitions or with daily flows (Chalmers and Phillips, 2013; Cashman and Villupuram, 2008). Second, the result is not caused by changes in regulations. Because the fund market in China is an emerging market, its regulations are changed more often than developed markets such as the US or the European fund markets. Changes in regulations are likely to systematically affect fund flows for specific years (Nanda et al., 2004). To check this concern, I estimate the substitution effect with year fixed effects (see Appendix with Table A2) and the substitution effect result is still robust.
My result also confirms conclusions of other literature as well. The parameters of the past performance variables low, mid and high show that the marginal effect of past peer ranks to the future flows is positive and most pronounced for funds with top 20% peer ranks, which confirms the well-documented convex flow-performance relationship (Sirri and Tufano, 1998; Choi et al., 2016). This fact means that the Chinese fund investors also chase funds with best past performances, which may incentive (small and young) mutual funds to take high risks and gamble for outstanding performances (see empirical evidence e.g. Yang et al., 2013). The negative parameters of log(T N A) show that the larger funds tend to have lower expected future flows, and this may be because it is more difficult for larger funds to outperform the benchmarks or the indexes (P´astor et al., 2015; Harvey and Liu, 2017). The robustly positive parameter of log(age) shows that older funds attract more future flows, which may come from their better reputation or investors believing they have better management skills according to Berk and Green (2004)’s model. My results also offer evidence for “fund family effect” that funds with larger fund families tend to attract larger flows (Nanda et al., 2004). This may be because larger fund families tend to pay more on advertising themselves or their funds. In conclusion, regarding the control variables, my result shows the behaviour of Chinese fund investors is similar to those in the US, which makes my result be comparable with previous research.
6.2.
Hypothesis 2: The Variation of Substitution
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
VARIABLES BOND FUNDS HYBRID FUNDS EQUITY FUNDS
CP -0.14*** -0.15*** -0.14*** 0.15 -0.12*** -0.06*** -0.05*** 0.02 -0.02 0.00 0.01 0.09 (0.02) (0.03) (0.03) (0.14) (0.01) (0.01) (0.02) (0.03) (0.02) (0.03) (0.03) (0.09) CP · Rm -1.77*** -0.64*** -0.42 (0.32) (0.16) (0.51) CP · After -0.29** -0.08** -0.07 (0.15) (0.04) (0.09) After 0.20*** -0.01 -0.14*** (0.04) (0.02) (0.03) Rm 0.02 0.27*** -0.06 0.12*** 0.17*** 0.12*** 0.14 0.18 0.20** (0.08) (0.09) (0.08) (0.04) (0.04) (0.04) (0.09) (0.13) (0.10) F.Rm -0.33*** -0.33*** -0.40*** 0.36*** 0.35*** 0.36*** 0.42*** 0.42*** 0.46*** (0.09) (0.09) (0.09) (0.03) (0.03) (0.03) (0.11) (0.11) (0.11) Low 0.82*** 0.80*** 0.78*** -0.03 -0.03 -0.03 0.14 0.14 0.13 (0.19) (0.19) (0.19) (0.07) (0.07) (0.07) (0.12) (0.12) (0.12) Mid 0.21*** 0.20*** 0.21*** 0.06*** 0.06*** 0.06*** 0.14*** 0.14*** 0.14*** (0.05) (0.05) (0.05) (0.02) (0.02) (0.02) (0.04) (0.04) (0.04) High 0.44 0.46* 0.48* 0.81*** 0.82*** 0.81*** 0.44* 0.44* 0.43* (0.27) (0.27) (0.27) (0.11) (0.11) (0.11) (0.24) (0.24) (0.24) log(TNA) -0.08*** -0.08*** -0.08*** -0.04*** -0.04*** -0.04*** -0.02*** -0.02*** -0.02*** (0.01) (0.01) (0.01) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) log(Age) 0.00 0.01 0.00 0.03*** 0.03*** 0.03*** 0.02** 0.02** 0.03*** (0.01) (0.01) (0.01) (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) log(Family) 0.03*** 0.03*** 0.03*** 0.01*** 0.01*** 0.01*** 0.01 0.01 0.01 (0.01) (0.01) (0.01) (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) log(Peer) -0.03** -0.02* -0.11*** -0.06*** -0.06*** -0.05*** -0.00 -0.00 0.05*** (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Std(Rt−12) 0.15 0.15 -0.37 0.70*** 0.69*** 0.70*** 0.58** 0.58** 0.86*** (1.04) (1.04) (1.08) (0.11) (0.11) (0.12) (0.23) (0.23) (0.27) Constant 0.09*** 0.77*** 0.72*** 1.13*** 0.00 0.75*** 0.73*** 0.73*** 0.01 0.06 0.06 -0.10 (0.01) (0.20) (0.20) (0.22) (0.00) (0.10) (0.10) (0.12) (0.01) (0.17) (0.17) (0.17) Observations 13,017 8,946 8,946 8,946 35,976 26,967 26,967 26,967 5,895 4,184 4,184 4,184 R-squared 0.01 0.03 0.03 0.03 0.00 0.03 0.03 0.03 0.00 0.03 0.03 0.04 AIC 32,567 22,162 22,136 22,138 64,206 43,814 43,800 43,816 7,581 5,176 5,176 5,165
Table VI. The variation of the substitution effect among different types of funds.
Figure 3. The percentage of assets held by individuals for different types of funds.
Note: this figure shows the percentage of assets held by individuals for each type of funds at the end of each year from 2005 to 2018, and the data of percentage held by individuals comes from the annual reports of funds. The percentage of individual holdings is calculated by dividing the total value of fund (outstanding) assets held by individual investors with the total value of fund (outstanding) assets at the end of each year.
with all the interactions are presented in the appendix with Table A4, A5 and A6 respectively. Moreover, the negative parameter of log(P eer) indicates that the decreases of hybrid funds’ flows are more likely caused by new entrant equity funds or hybrid funds (Wahal and Wang, 2011), rather than by the rise of MMFs.
investors tend to have larger bargaining power and can align their liquidity demands better with the term structures of bond funds beforehand. Figure 3 shows the percentage of assets held by individuals for each type of funds from 2005 to 2018. There is a large and persistent downward trend only for bond funds after the rise of MMFs. Hereby, the evidence above supports my hypothesis 2 and demonstrates the rise of MMFs hurts the bond funds the most.
As for the control variables, the results in Table VI are similar to the results in Table V with the pooled sample except for two notable issues. First, the age of funds (log(Age)) only affects the flow of equity funds and hybrid funds. This is because the management skills are more important for equity funds and hybrid funds than for bond funds, and fund investors tend to infer those skills from the age of funds Berk and Green (2004). Second, previous risk-taking of funds (Std(Rt−(12)) significantly affects flows for equity funds and hybrid funds, which means fund investors are sensitive to the investment styles of them. This is because these funds have larger freedom than bond funds to alter their investment styles than bond funds. These two facts show that equity funds and hybrid funds are more likely to alter their risk-taking to send signals to fund investors; on the other hand, fund investors also actively search for these signals and interpret them (Berk and Green, 2004).
Though the above result buttresses hypothesis 2 that the substitution effect differs among different types of funds, it does not necessarily demonstrate that such differences come from different risk-taking of their assets. Bond funds and hybrid funds (equity funds) may differ not only in their risk-taking, but also in other ways such as distribution channels. To test whether the substitution effect comes from the difference in assets, I estimate the substitution effect with hybrid funds with different assets. I divided all the open-end hybrid funds into “safe funds” and “risky funds” based on whether their percentages of stocks are below the median level (i.e. 75%) or not. This classification (“safe fund” and “risky fund”) is not used in any fund websites or databases, so the only difference between these two types of funds is the risk-taking of their assets.
(1) (2) (3) (4) (5) (6)
VARIABLES SAFE FUNDS RISKY FUNDS
CP -0.14*** -0.13*** 0.05 -0.07*** 0.05** -0.01 (0.01) (0.02) (0.05) (0.01) (0.02) (0.04) CP · After -0.18*** 0.06 (0.06) (0.04) After 0.06** -0.05** (0.03) (0.02) Rm 0.13** 0.12** 0.09** 0.12** (0.06) (0.06) (0.05) (0.05) F.Rm 0.43*** 0.40*** 0.30*** 0.32*** (0.05) (0.04) (0.05) (0.05) Low 0.01 0.03 -0.03 -0.01 (0.15) (0.15) (0.08) (0.08) Mid 0.08** 0.08** 0.04 0.04 (0.03) (0.03) (0.02) (0.02) High 0.28 0.28 1.07*** 1.05*** (0.18) (0.18) (0.14) (0.14) log(TNA) -0.05*** -0.05*** -0.04*** -0.04*** (0.01) (0.01) (0.00) (0.00) log(Age) 0.03*** 0.03*** 0.02** 0.02*** (0.01) (0.01) (0.01) (0.01) log(Family) 0.01*** 0.01*** 0.01* 0.01* (0.00) (0.00) (0.00) (0.00) log(Peer) -0.04*** -0.08*** -0.08*** -0.05*** (0.01) (0.02) (0.01) (0.01) Std(Rt−12) 0.66*** 0.55*** 0.69*** 0.83*** (0.18) (0.19) (0.14) (0.17) Constant -0.01** 0.64*** 0.81*** 0.02*** 0.91*** 0.76*** (0.01) (0.14) (0.18) (0.01) (0.14) (0.15) Observations 17,473 12,082 12,082 18,495 14,878 14,878 R-squared 0.01 0.03 0.03 0.00 0.04 0.04 AIC 33,273 21,460 21,454 30,593 22,009 22,007
Table VII. The substitution effect for hybrid funds with different risk-taking.
6.3.
Hypothesis 3: Fund Reaction
The result presented in Table VIII shows that actively managed hybrid funds change their percentages of stock holdings in response to the substitution effect from MMFs, which is in line with hypothesis 3. After the rise of MMFs, when facing higher competitive pressure from MMFs, actively managed hybrid funds tend to increase their risk-taking to “distinguish” themselves from MMFs. However, the result shows that equity funds react little to aggregate flows of MMFs in changing percentages of stocks, which may be because equity funds are required to maintain high percentages of stocks by CSRC and with less freedom to change them. Similar to previous estimations, the result is estimated with robust standard errors double-clustered at fund and year level to correct for heteroscedasticity (Franzoni and Schmalz, 2017).
Moreover, regressions with interactions reveal that such reactions are more pronounced when the sub-stitution effect of MMFs is more pronounced. The results of the subsub-stitution effect (e.g. Table V) shows that when the market returns are high, a same amount of competitive pressure leads to larger substitution effect and competes for more future flows of other funds than when market returns are low. Therefore, a same amount of competitive pressure from MMFs signals higher concerns of the substitution effect to other funds when market returns are higher. Table VIII column 3 and 7 shows that funds recognize this signal and react accordingly. For a given amount of competitive pressure, the reaction of funds almost doubles when the market return increases from 0 to 10%. In the same vein, compared to the time period before the rise of MMFs, actively managed hybrid funds tend to increase their percentages of stocks rather than decrease them. This is because after the rise of MMFs, flows of MMFs are less contributed by the “hedge behaviour” of fund investors. A high aggregate flow of MMFs is interpreted as high competitive pressures, and is not caused by bad performance of the stock market. The results are robust with Logit regressions (see Appendix with Table B1), in which the dependent variable is a dummy for increase.
Such reactions of hybrid funds can be explained by the law of comparative advantage. Under the compe-tition of MMFs, the hybrid funds increase their percentages of stocks and stake higher risk-taking to distinct themselves from MMFs, because they have a comparative advantage over the MMFs in taking high risks. Such behaviour is also in line with previous literature about entry deterrence (e.g. Milgrom and Roberts, 1982; Bagwell and Ramey, 1988) that incumbents (i.e. hybrid funds) apply strategies when they face com-petitive pressures from new entrants (i.e. MMFs), and such strategies are relatively more costly for entrants than for incumbents. Intuitively, to deter new entrances, incumbents tend to play strategies that entrants are not good at.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
VARIABLES POOLED HYBRID FUNDS EQUITY FUNDS
CP -0.79*** 1.21*** 1.02*** -2.88** -0.90*** 1.21*** 1.01*** -2.93** 0.53 1.79** 1.59* -2.06 (0.24) (0.33) (0.33) (1.15) (0.25) (0.35) (0.35) (1.17) (0.67) (0.80) (0.83) (3.85) CP · Rm 15.39*** 16.65*** 8.86 (3.33) (3.50) (8.16) CP · After 4.55*** 4.59*** 4.00 (1.21) (1.23) (3.95) After -3.05*** -3.09*** -1.34 (0.23) (0.23) (1.12) Rm -0.04*** -0.05*** -0.02*** -0.04*** -0.05*** -0.02*** -0.03* -0.04** -0.02 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) F.Rm 0.17*** 0.18*** 0.18*** 0.18*** 0.18*** 0.19*** 0.07*** 0.07*** 0.07*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) Low -2.17 -2.15 -1.63 -2.09 -2.06 -1.48 -3.94 -3.97 -3.85 (1.74) (1.74) (1.74) (1.89) (1.89) (1.88) (2.93) (2.94) (2.93) Mid -1.00** -0.97** -1.00** -1.15*** -1.12*** -1.15*** 1.36* 1.38* 1.32* (0.40) (0.40) (0.40) (0.43) (0.43) (0.43) (0.72) (0.72) (0.72) High 5.47*** 5.41*** 5.10*** 5.96*** 5.88*** 5.52*** -1.07 -0.93 -1.19 (1.64) (1.64) (1.63) (1.77) (1.77) (1.77) (2.84) (2.84) (2.83) log(TNA) 0.29*** 0.28*** 0.26*** 0.35*** 0.34*** 0.32*** -0.21** -0.22** -0.23** (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.10) (0.10) (0.09) log(Age) 1.03*** 1.05*** 1.21*** 1.05*** 1.06*** 1.22*** 0.28 0.31 0.33 (0.10) (0.10) (0.10) (0.11) (0.11) (0.11) (0.20) (0.20) (0.22) log(Family) -0.31*** -0.30*** -0.30*** -0.35*** -0.34*** -0.34*** -0.11 -0.11 -0.11 (0.06) (0.06) (0.06) (0.06) (0.06) (0.06) (0.08) (0.08) (0.08) log(Peer) -0.25** -0.31*** 1.26*** -0.21* -0.28** 1.35*** -0.15 -0.16 0.32 (0.11) (0.11) (0.16) (0.12) (0.12) (0.17) (0.25) (0.25) (0.55) Std(Rt−12) 0.33*** 0.33*** 0.41*** 0.36*** 0.37*** 0.44*** 0.11*** 0.11*** 0.14*** (0.02) (0.02) (0.02) (0.02) (0.02) (0.03) (0.04) (0.04) (0.04) Stock% -14.24*** -14.30*** -14.72*** -14.42*** -14.48*** -14.88*** -19.77*** -19.78*** -20.02*** (0.46) (0.46) (0.46) (0.47) (0.47) (0.47) (5.19) (5.19) (5.12) Equity -2.50*** -2.41*** -5.12*** (0.21) (0.21) (0.30) Constant 1.28*** 6.84*** 7.22*** 0.39 1.26*** 3.84** 4.39*** -5.63*** 1.45*** 22.49*** 22.51*** 21.02*** (0.06) (1.44) (1.43) (1.55) (0.07) (1.53) (1.52) (1.75) (0.17) (6.80) (6.80) (7.55) Observations 37,603 28,382 28,382 28,382 34,413 26,274 26,274 26,274 3,190 2,108 2,108 2,108 R-squared 0.00 0.12 0.12 0.12 0.00 0.12 0.12 0.12 0.00 0.15 0.15 0.15
This is because hybrid funds with lower percentages of stocks tend to possess worse management skills of high risk-taking than their peers who hold higher percentages of stocks. As a result, it is more costly for them to increase percentages of stocks (Chua et al., 2018). To test this proposition, I estimate equation (9) with interactions to show how funds with different percentages of stocks react to the substitution effect differently. I interact the current competitive pressure (CP ) with concurrent percentages of stocks (Stock) for each funds, which reveals how funds’ reaction varies in their risk-taking. To avoid multicollinearity, I center both CP and Stock before generating the interaction. I also create an interaction between current competitive pressure and the risky dummy (DRisky), which equals to 1 (0) when the percentage of stocks of a fund is above (below) the cross-sectional median.
Table IX shows that after the rise of MMFs, high competitive pressure from MMFs is more likely to drive actively managed funds with high percentages of stocks to increase their percentages of stocks in the future, ceteris paribus. Contrarily, such heterogeneous reactions between “risky funds” and “safe funds” do not exist before the rise, which is because the aggregate flow of MMFs does not signal competitive pressures at that period of time. The results are robust with Logit regressions as well (see in the appendix with Table B2).
Such observations are in line with the “style shift” model of Chinese mutual funds which suggests that funds change their investment styles based on (1) their management skills and (2) the number of rivals with similar investment styles (Chua et al., 2018). Hence, after the rise of MMFs, hybrid funds with high percentages of stocks increase their risk-taking by holding more stocks to avoid competitions with MMFs. However, hybrid funds with lower percentages of stocks are faced with the dilemma that they possess worse management skills of high risk-taking than their risky peers and worse management skills of low risk-taking than bond funds or MMFs. Therefore, they tend to defence with strategies other than changing risk-taking. For example, they may apply waives or discounts to fund fees to draw more investors (Wahal and Wang, 2011).
(1) (2) (3) (4) (5) (6)
VARIABLES FULL BEFORE AFTER
CP 2.31*** -0.58 -2.64* -3.48** 3.92*** -0.12 (0.33) (0.47) (1.54) (1.75) (0.39) (0.52) CP · Stock% 8.00*** 1.26 10.05*** (0.98) (10.20) (1.08) CP · Risky 4.47*** 1.47 5.91*** (0.58) (2.26) (0.67) Risky -2.50*** -1.76*** -2.98*** (0.23) (0.53) (0.31) Rm -0.03*** -0.04*** -0.03*** -0.04*** 0.02** 0.00 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) F.Rm 0.18*** 0.18*** 0.19*** 0.20*** 0.19*** 0.20*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Low -1.04 -1.85 -3.62 -4.00 1.76 0.45 (1.89) (1.88) (2.56) (2.56) (2.41) (2.39) Mid -1.28*** -1.20*** 0.79 0.82 -2.26*** -2.13*** (0.43) (0.43) (0.56) (0.56) (0.55) (0.55) High 5.04*** 5.31*** -2.04 -1.86 5.88*** 6.62*** (1.76) (1.76) (2.68) (2.65) (2.19) (2.19) log(TNA) 0.37*** 0.46*** 0.71*** 0.80*** 0.22*** 0.30*** (0.05) (0.05) (0.08) (0.08) (0.07) (0.07) log(Age) 1.01*** 0.75*** -0.20 -0.48** 1.32*** 1.15*** (0.11) (0.11) (0.24) (0.23) (0.13) (0.14) log(Family) -0.33*** -0.35*** -0.17* -0.21** -0.35*** -0.37*** (0.06) (0.06) (0.09) (0.09) (0.07) (0.07) log(Peer) -0.42*** 0.20 4.91*** 5.00*** 0.43* 1.11*** (0.12) (0.14) (0.31) (0.31) (0.22) (0.24) Std(Rt−12) 0.41*** 0.42*** 0.55*** 0.55*** 0.55*** 0.54*** (0.02) (0.02) (0.05) (0.05) (0.03) (0.03) Stock% -15.48*** -11.50*** -21.67*** -17.45*** -16.56*** -12.09*** (0.50) (0.66) (2.07) (1.75) (0.59) (0.76) Constant -4.66*** -7.28*** -38.73*** -37.89*** -11.10*** -13.87*** (1.53) (1.55) (2.59) (2.65) (2.32) (2.39) Observations 26,274 26,274 7,393 7,393 18,881 18,881 R-squared 0.12 0.12 0.22 0.23 0.11 0.12
(1) (2) (3) (4) (5) (6)
VARIABLES TOTAL EXPENSE ADM. EXPENSE TRADING EXPENSE
CP -0.15* -0.25** -0.19** -0.28*** -0.02 -0.27 (0.09) (0.11) (0.08) (0.10) (0.14) (0.17) CP · Risky 0.34*** 0.31*** 0.78*** (0.11) (0.09) (0.17) Risky -0.02 -0.04 0.23*** (0.04) (0.04) (0.08) Rm -0.05 -0.07 -0.23*** -0.24*** 0.55*** 0.46*** (0.09) (0.09) (0.08) (0.08) (0.17) (0.17) F.Rm -0.11 -0.14 -0.06 -0.08 0.04 -0.09 (0.11) (0.11) (0.11) (0.11) (0.16) (0.15) Low -0.01 -0.00 -0.01 -0.00 -0.02*** -0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Mid 0.00 0.00 0.00* 0.00* 0.00 0.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) High -0.01* -0.01** -0.01 -0.01** -0.01 -0.01** (0.00) (0.00) (0.00) (0.00) (0.01) (0.01) log(TNA) -0.19*** -0.19*** -0.07*** -0.07*** -0.31*** -0.32*** (0.02) (0.02) (0.02) (0.02) (0.03) (0.03) log(Age) 0.13*** 0.09*** 0.11*** 0.07*** 0.53*** 0.39*** (0.03) (0.03) (0.02) (0.02) (0.04) (0.04) log(Family) 0.02 0.02 0.02 0.02 0.01 0.03 (0.02) (0.02) (0.02) (0.02) (0.03) (0.03) log(Peer) -0.10* -0.16*** -0.02 -0.07 -0.09 -0.30*** (0.06) (0.06) (0.05) (0.05) (0.08) (0.08) Std(Rt−12) 0.02*** 0.02*** 0.00 0.00 0.12*** 0.10*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Constant 4.28*** 4.85*** 0.88* 1.36** 2.01* 4.08*** (0.60) (0.64) (0.53) (0.57) (1.14) (1.10) Observations 9,759 9,759 9,726 9,726 9,693 9,693 R-squared 0.12 0.13 0.04 0.06 0.23 0.29 AIC 25461 25300 24041 23905 33450 32745
Table X. Change in expenses for actively managed hybrid funds.
Table X presents the results, which shows that actively managed hybrid funds react to the substitution effect by changing their expense ratios, and such reactions differ between safe funds and risky funds. Column 2 shows that when facing high competitive pressure, risky funds tend to increase their total expenses, while safe funds do not. Moreover, column 4 reveals that safe funds react to the competitive pressure by decreasing their administration expenses. Such decreases may be achieved by hiring fewer managers and indexing a part of their portfolios (Berk and Green, 2004; Cremers et al., 2016). On the other hand, column 6 shows that risky funds trade more and become more “active” when facing the competitive pressure, which exaggerates their comparative advantages over safe funds or MMFs. These results are consistent with my previous results on the fund risk-taking, both of which are in line with the prediction of entry deterrence models.
7.
Robustness Test: Capital Asset Pricing Model
The result in the previous section demonstrates actively managed hybrid funds react to the substitution effect by increasing their percentages of stocks, which indicates they tend to stake higher risks and exaggerate their comparative advantages. However, the percentage of stocks may not correctly reveal the risk-taking of funds for two reasons. First, it may not correctly reveal risk-taking because of the variety of stocks. Second, changing the percentage of stocks is confined to exogenous regulations.
As a robustness check, I measure the risk-taking of funds with the Beta of Capital Asset Pricing Model (CAPM) and investigate how the Beta changes before and after the rise of MMFs. The Beta measures how the return of a portfolio varies relative to the (stock) market return and presents the risk-taking of funds more precisely than the percentage of stocks does (Cochrane, 2009). However, the estimation of Betas for funds involves rolling regressions, which generates the autocorrelation problem. To avoid this problem, I compare the change of Betas for actively managed equity funds and hybrid funds before and after the rise of MMFs, rather than regress them on explanatory variables.
In this part of the thesis, I treat the rise of MMFs as an exogenous shock to all the equity funds and hybrid funds because it substitutes the demand for these funds. If funds react to this shock by changing risk-taking, their Betas with estimation periods before and after the rise of MMFs should differ. However, the change in Betas may be affected by various factors. To exclude the effect of other factors, I compare the first 24-month Beta of entrant funds that are initialized just before and after the rise of MMFs.
