Mutual fund flows : Where does the money go ?

Received: 29-08-2014 Accepted: 05-10-2014 Available online: 27-10-2014


Introduction
As of the end of 2012, over $13 trillion had been invested in mutual funds with an additional $265 billion invested in closed-end funds, $1.3 trillion in ETFs, and $72 billion in unit investment trusts (UITs), (see 2003 Investment Company Fact Book).These dollars represent approximately 20.8% of total household wealth. 1 Mutual funds are an important financial intermediary for savings and investment with their holdings comprising more than 24% of all US corporate equities, 14% of US and international corporate bonds, and 12% of US Treasury and government agencies securities.
The industry has more than 776 fund sponsors (or investment companies) some with just a single fund and others with hundreds of funds.The plethora of mutual fund categories, styles, and objectives creates an investment mosaic of surprising depth which includes more than 8,750 mutual funds, 600 closed-end funds, 1,239 ETFs, and 5,787 UITs, (see 2003 Investment Company Fact Book).From amongst these many offerings, investors choose where to place their investable funds.But first they decide whether they want to invest in corporate equities, fixed income securities, or money market funds.These data are summarized below in Table 1.A plethora of forces influence an individual's decision about which sector to invest in.Amongst these are their age and life expectancy, holdings of other assets, tolerance for risk, liquidity preference and personal expectations of future returns from each investment sector.Our paper is concerned with the choice between equities, bonds, 2 and money market funds -probably the broadest categorization of investment vehicles.
By the end of 2012 as seen in Figure 1, investor's allocation choices placed approximately 50% of their mutual fund investments into the equities category, 28% into fixed income funds, and the remaining 22% in money market accounts.The mix between these three aggregate categories is not fixed.For example, in 2002 fewer than 44% of investment dollars were deployed to equities, slightly more than 37% went to fixed income investments, and about 19% was in money market accounts (2003 Investment Company Fact Book).It is the forces influencing this consumer choice between sectors that are studied in this paper as well as factors affecting investment returns.
For good reason, considerable attention has focused on the substantial and growing mutual fund industry.One topic which has attracted academic researchers is the determination of the causal relationship between fund flows and security returns.There is some evidence that returns are highest in the broad sectors where money has recently flowed into and additional evidence that to some extent money flows where returns are highest.Untangling these simultaneous processes is the key focus of this paper.
Unlike most of the previous work on this topic, this paper views the investor's sectoral financial-allocation decision to be as important as the question of whether higher returns attract new funds or whether the investment of new funds drives returns.We study sectoral allocation decisions, aggregate returns, and fund flows across three aggregate mutual fund sectors: equities, fixed income, and money market funds. 34Our work considers the investor's allocation decision within a simultaneous system of equations thereby recognizing that investment in any one sector is constrained by their total investable funds and that overinvestment of money flows into one sector must be balanced by underinvestment or outflows from another sector.The use of a simultaneous framework enables us to untangle the interrelationship between funds flows and sector returns.

Previous literature
Relatively few researchers have investigated how investors choose between equity, fixed income, and money market funds.The principal topic studied has been how investment flows influence security returns.These researchers posit the idea that additional money flows into an investment sector leads to incremental demand for 2 Most U.S. traded investment grade bonds are represented in our sample.Municipal bonds and Treasury Inflation-Protected Securities are excluded, due to tax treatment issues.The AGG index employed in this study includes Treasury securities, government agency bonds, mortgagebacked bonds, corporate bonds, and a small amount of foreign bonds traded in U.S. 3 Hybrid funds which generally combine elements of fixed income and equity investments are excluded from the analysis for two reasons: their mixed strategic nature and their relative small size. 4Mutual fund returns are captured by market returns as an alternative to creating a weighted average value across various mutual funds.Differences between the two series are presumed to be minor.securities in that sector which results in increased absolute and relative returns.Other researchers have evaluated the opposite relationship: how security returns contribute to the flow of funds.A smaller subset of studies has looked simultaneously at the two relationships, investment flows effect on returns and the impact of returns on flows.
In a key paper, Warther (1995) proposed a multistage-regression process in which to decompose aggregate mutual fund flows into expected and unexpected components.He tested this methodology using monthly data on stock funds, bond funds and gold funds.After decomposing fund flows into two parts, Warther (1995) security returns to be highly correlated with unexpected cash flows but to be unrelated to expected flows.Receipt of unexpected investor deposits result in funds reporting higher returns.Warther (1995) estimated model coefficients to determine an elasticity of market returns with respect to unexpected inflows that equaled 5.7.That is, additional (unexpected) funds given to mutual funds result in substantial boosts in aggregate market returns.He also found fund flows to be correlated with direct (i.e., same sector) security returns but not correlated with the returns of other types of securities.His finding conforms to a reasonable set of expectations about the forces influencing investor decisions.Oddly though, Warther (1995) found a negative relationship between fund flows and lagged security returns which suggests that investors withdraw money from fund categories in the month following their achieving positive returns.
Remolona, Kleiman, and Gruenstein (1997) also considered correlations between mutual fund flows and market returns.Like Warther (1995) they used an early stage regression to generate expected and unexpected mutual fund flows.Three types of stock funds and five types of bond funds were examined with a relatively small database of 118 observations each.Using an instrumental variables regression technique to account for the simultaneity between fund flows and returns, Remolona et.al, (1997) did not generally find a relationship between market returns and fund flows, except for funds with conservative investment objectives where they did find some level of significance for this relationship.For other types of funds, short term market returns had little effect on mutual fund flows.In other words, Remolona et.al, (1997) were unable to confirm Warther's (1995) results.They did not consider the reverse effect, the impact of mutual fund flows on market returns.
In an earlier study, Ippolito's (1992) proposed a different methodology than the two previously discussed studies.His work used annual data on 143 open-ended mutual funds representing 80% of the assets of all mutual funds for the period 1965-1984.He found that investors pay close attention to information about quality (i.e., returns).Ippolito's (1992) shows that investors move their money towards recent good performing mutual funds and away from recent poor performers.Evidence of serial correlation in mutual fund returns led Ippolito's (1992) to argue that better performing mutual funds maintain their return advantage which endorses the idea of investors chasing after good returns.Cashman, Nardari, Deli, and Villupuram (2012) chose a different methodological approach.They identified how persistence in funds flow is at least as important as fund performance to predict future fund flows.Though not identified as such by the authors, persistence may be functionally equivalent to expected funds flow as identified by both Warther (1995) andRemolona et.al, (1997).In addition, Cashman et.al, identify how some investors evaluate mutual fund performance in a short-term trading like context rather than, as is more traditionally assumed, in a longer run context.Moreover, Bensen, Faff, and Smith (2010) worked with data from 7,000 individual equity mutual funds.They looked for contemporaneous and lagged simultaneously relationships between fund flows and returns.Unlike findings attributed to other recent authors, Bensen et.al, (2010) did not find current and lagged funds flows to affect current returns, across all types of funds.However, they do find that current mutual fund returns significantly affect current investment flows.Ippolito's (1992) argued that higher returns attract investment funds while higher investment flows do not have any effect on returns.
Our research is motivated in part by the Bensen et.al, (2010) paper in that like them we assume endogeneity between fund flows and returns but unlike them we also recognize that investors have a choice between equity, fixed-income, and money market funds.Moreover, while their work analyzes individual mutual funds we propose to study aggregate sectoral mutual funds.Our contribution to the literature then is to evaluate the process of funds flow and security returns with a regression method design for simultaneous models that yields consistent and unbiased coefficient estimates.

Data
Fund flows Monthly mutual fund data were obtained for our study from the Investment Company Institute (ICI).The data spans the period 1990-2012.Mutual funds are classified into four categories based on their portfolio components; equity, bond, money market, and hybrid fund.Data for hybrid funds (which combine investment across several fund types) were not included as our research thesis is concerned with investor's discrete choices between specific types of funds.
For the three types of mutual funds the following data series are collected: 1. New Investment Sales -Dollar value of new purchases of mutual fund shares.Does not include shares purchased through the reinvestment of dividends on existing accounts.2. Reinvested Dividends -Dollar value of distributed income dividends used to purchase more shares of mutual funds.3. Redemptions -Dollar value of money returned to an investor who has sold shares of a fund (i.e.investor cashes in shares).
Monthly data series measuring investment flows by asset type are generated with the data series as follows: Flow of Investments i,t = Total Sales i,t -Redemptions i,t = New Sales i,t + Reinvested Dividends i,t -Redemptions i,t Where, i represent equity funds, fixed income funds, and money market funds and t represents months.Flow of Investments i,t represents the net dollars invested each month in each fund category.

Mutual fund returns
Aggregate equity fund returns are measured by the return on the S&P 500 index.Use of the S&P 500 index is an expedient alternative to the massive task of creating a weighted average of returns for all individual funds.The S&P 500 index is selected because of its substantial correlation with other broad indexes and because it is available for the period of our study.
Aggregate fixed income returns are measured by Barclay's aggregate bond index known as the AGG.The index tracks the broad bond market and is both well-known and highly regarded.The AGG is obtained from Bloomberg.

Model formulation
Similar to models proposed by Remolona et. al, (1997) and Bensen et. al, (2010), we study two related investment concepts in a simultaneous framework.Our study considers how the flow of dollars into three separate types of mutual fund investments (equities, fixed income, and money market funds) are affected by each sector's investment returns and how sectoral investment returns are affected by investment flows.By using a simultaneous estimation method, the model assumes that investors select between the three investment-vehicles using a conjoint analysis that compares opportunities from each type of investment.
The research paradigm tests whether investors make investment choices after reviewing current and lagged returns reported by equity and fixed income securities.In addition we ask whether investment returns earned in the equity and fixed income markets are affected by current or lagged flows into the three types of funds.The portion of the model representing funds flows is described in equations 1-3 where EQ is the annual net flow of money into equity funds, Bond is the annual net flow of money into fixed income funds, and MM is the annual net flow of money into money market funds.Returns earned by equity funds are measured with the current and lagged returns on the S&P 500, R_SP and R_SPlagged.Current and lagged returns in the fixed income sector are measured by the return on the aggregate bond index, R_AGG and R_AGGlagged.The final factors tested for their effect on investment flows are the flow of investment dollars into the other two markets in both the current and lagged period.That is, for example, we examine how the flow of investment dollars into equities affects the flow of dollars into fixed income and money market funds.For clarity reasons, the subscripts i and t are not included in equations 1 -3.The second portion of our work which explains the returns earned by equity and fixed income securities is described in equations 4-5.The basic hypothesis asks whether the flow of funds into a sector drives sectoral returns; our objective is not to explain total return which of course depends on a multitude of factors including economic conditions, interest rates, innovation, and other forces.Rather we focus on the narrower question of how the flow of funds into a sector affects that sectors returns.Returns earned by the S&P 500 is thought to depend upon money flows in both the current and lagged periods into money market accounts, equity funds, and fixed income funds and on the current and lagged returns of fixed income sector.Similarly, the return earned by the aggregate bond fund is thought to depend on the flow of money into the three sectors as well as on the current and lagged return earned by the equity sector.Again, in equations 4-5 the subscripts i and t are not included for clarity reasons.

Model estimation
The model's five equations are estimated using ordinary least squares (OLS) and then a second time to reflect the simultaneity in the decision process with seemingly unrelated regression techniques (SUR).The need for SUR estimation methods for equations 1-3 arises because the three error terms, ϵ1, ϵ2, and ϵ3, are thought not to be independent of the explanatory variables in each equation.When the error process is related to the covariates in a regression, OLS does not produce consistent or unbiased estimates.The same lack of independence affects regression results for equations 4-5 where the error processes, ϵ4 and ϵ5, are related to the independent variables in those equations.
A further complication affecting estimation of the first three equations is the presence of serially correlated error terms.Durbin Watson statistics equaled 0.94 for the equity flow model, 0.42 for the bond flow model, and 1.39 for the money market flow model.The presence of serial correlation is not surprising since there is a strong similarity between successive time periods in many of the factors that influence investor's choice between the three types of investments. 5 To alleviate this problem, lagged dependent variables is added to each equation (1-3).The models depicted above in equations (1-3) include the lagged dependent variables.Estimation results presented below do not include the uncorrected OLS results without the lagged dependent variables.Estimated coefficients on the lagged dependent variables represent the impact on the current dependent variable of past values of independent variables not already included in the equation.
Regression results are presented below in Tables 2a, 2b, and 2c and 3a and 3b.The discussion that follows addresses first the funds flow models (equations 1-3) and then looks at asset returns (equations 4-5).

Estimation results: funds flow models
The three models explaining the flow of funds into mutual fund sectors are first estimated with OLS.Overall model structures are confirmed with these regressions based on the high significance of the reported F statistics.Each of the three models appears to suffer from serial correlation based on the values of the Durbin Watson statistics.To remedy this situation, lagged dependent variables are added to the three funds flow equations.Re-estimation of the models yielded Durbin Watson statistics in the range that indicates the lack of serial correlation.The remaining discussion of these models is based on the SUR regressions.
The model explaining the flow of money into equity funds estimated with SUR regression has a high adjusted R 2 with a value of 0.52 as seen in Table 2a.The same set of significant independent variables observed in the OLS regression remains significant with SUR regression.A significant positive relationship exists between equity fund flows and the return on the S&P 500, the flow of money into bond funds, and the lagged flow of money into money market funds.Contrasting with these results, there are significant negative relationships found between the flow of money into equity funds and the returns earned on fixed income sector, the lagged flow of money into bond funds, and the flow of money into money market accounts.Two independent variables in the equity funds flow model, lagged returns of both equities and bond sectors are insignificant in the equity flow model.
Unlike equity fund model results, which had the same set of significant independent variables in both OLS and SUR regressions, SUR regression estimation of funds flows into fixed income funds had two additional significant variables beyond those which are significant in the OLS regression as seen in Table 2b.This finding is not unexpected given that OLS estimation is consistent but not as efficient as is SUR regression.A relatively efficient estimator has a smaller variance which may reduce the standard errors of coefficient estimates.Like the equity funds flow model above, the return earned in the fixed income sector is a positive determinant of the flow of money into fixed income funds; however, unlike equity fund flows which were unrelated to lagged returns, the lagged return earned in the fixed income sector is significantly related to the current funds flow into fixed income.The comparison between the two sector funds flow models also diverges when examining the impact of the opposite sector's returns on fund flows: equity fund flows are significantly and negatively impacted by bond returns while fixed income fund flows are not affected by equity returns.
Like the equity fund flows model which is significantly related to a) current and lagged fixed income flows and b) current and lagged money market flows, the fixed income flow is significantly related positively (negatively) to current equity (money market) fund flows and negatively (positively) related to lagged equity flows (money market flows).
As occurred in the fixed income flow model, the money market flow model also acquires two significant additional explanatory variables in the SUR regressions when compared with the OLS regression as seen in Table 2c.Fewer significant variables explain money market flows as compared with both equity and fixed income flows.equity funds.These results are consistent with the view that money market fund investments are a holding action for many people while they await opportunities in either equity or fixed income funds.That is, when money moves into equity or fixed income funds, balances are drawn down in money market funds to pay for the new equity or bond investments.
As occurred in the fixed income flow model, the money market flow model also acquires two significant additional explanatory variables in the SUR regressions when compared with the OLS regression as seen in Table 2c.Fewer significant variables explain money market flows as compared with both equity and fixed income flows.Flows of money into money market funds are negatively related to current flows into equity and fixed income funds and are positively related to lagged bond returns and lagged flows into equity funds.These results are consistent with the view that money market fund investments are a holding action for many people while they await opportunities in either equity or fixed income funds.That is, when money moves into equity or fixed income funds, balances are drawn down in money market funds to pay for the new equity or bond investments.
Overall our results have some similarity with previous work but it is the differences between the new and older studies that are of most interest.Our findings contrast with Warther (1995) in several ways: a) he did not find, as we did, equity fund flows to be significantly affected by the returns earned by fixed income securities and b) he found a significant negative relationship between flows and lagged returns which we did not replicate.Our findings also differ from Remolona et.al, (1997) who did not find a relationship between current or lagged returns and fund flows.The smallness of their database may influence the differences between our study and theirs.Our findings are consistent with Ippolito's (1992) results that returns influence fund flows though his work was directed only at equity funds and was not estimated simultaneously.Finally, our results support the findings of a positive relationship between current sector returns and fund flows observed by Remolona et.al, (2010) though our finding of an additional significant relationship between lagged returns and current flows extends beyond their results.

Estimation results: return models
Models explaining returns earned in the equity and fixed income sectors do not evidence the presence of serial correlation in their OLS regressions.Nor is serial correlation a problem in the simultaneous estimation versions of the return's models when using SUR regression.Insignificant independent variables are not dropped from the estimation equation since this paper is exploratory rather than predictive.The two SUR return's models have comparatively high adjusted R 2 s given the number of other factors such as economic conditions and corporate earnings that are known to influence investment returns that have purposefully been excluded as explanatory factors.
Returns earned in the equity sector are significantly affected by four independent variables as seen in Table 3a.Equity sector returns are higher the higher are the flow of funds into equities and the contemporaneous returns earned by bond sector; they are also higher the smaller are last period's money flows out of equities and money market funds.These results are not consistent with other researchers such as Remolona et.al, (1997) who did not find that money flows directly affect returns.
Similarly, the returns earned by fixed income funds are also influenced by four significant explanatory factors as seen in Table 3b.Fixed income returns are higher the higher are the flow of funds into fixed income accounts and the contemporaneous returns earned in the equity sector.Higher fixed income fund returns are significantly associated with smaller declines of a) money flows into bond funds last period and b) money flows into equity funds this period.Again, most previous studies do not find this association.
We believe that our findings are more intuitive than those reported by previous researchers; that is, we find that when money flows into either equity or fixed income funds that those sectors earn higher returns.The simple explanation of this phenomenon relies on nothing more than basic supply and demand: a higher demand for funds of either type leads to higher prices and therefore higher returns in corresponding sectors.
Our empirical work spans five separate models, three of which study money flows into equity, fixed-income, and money market funds while the other two models examine the returns earned in the equity and fixed-income sectors.The statistical significance and the direction of the relationship between independent variables and the five dependent variables are discussed above.In this section, we summarize the previous discussion to highlight the critical questions of do funds flow affect returns or do returns affect funds flow.Table 4 below provides an easy reference looking across the five different models with which to answer those questions.
Journal of Economic and Financial Studies.Page 69 Looking down the column labeled 'Funds Flow' it is clear that in our work current returns earned in the equity and fixed income markets significantly explain the flow of funds into both the equity and fixed income funds while lagged returns earned in the fixed income market also influences fixed income flows.Similarly, looking down the column labeled 'Returns', both current and lagged flows into the equity and fixed income markets are significant influences on the returns earned by both equity and fixed income sectors.Moreover, as our basic hypothesis of simultaneity in the decision process of individuals would suggest, our results show that all three fund types are significantly influenced by both the current and lagged flow of funds into other sectors.Returns earned in the fixed income sector are seen to be affected by the current flow of funds to other sectors while the returns earned in the equity sector is affected by lagged funds flow to other sectors.These results are substantially different from those found in previous studies.In some respects, our results are more intuitively obvious.For example, we find that investment returns are affected by funds flows.That is what one would expect based on supply and demand analysis.When money flows into sectors those sectors earn higher returns.Similarly, we find that investment decisions are influenced by current investment returns in equity and fixed income funds and lagged returns for fixed-income funds.This too is not surprising given our understanding of how momentum investing and general financial innumeracy are common amongst many investors.

Conclusions
The results presented in this paper provide a realistic view of forces effecting how investors allocate their investments between equities, fixed income, and money markets.In specific, we find that returns influence the decision to invest in equities and fixed income and lagged returns also influence fixed income investments.Moreover, money flows into the three investment sectors are influenced by the amount of money flowing into other sectors.These results, while unusual in comparison to other researcher, are both logical and intuitive.
Likewise we show that an influx of money from investors both now and in the previous period affects the returns earned in both the equity and fixed income markets.In addition we find that equity and fixed income returns are influenced by the flow of funds into other sectors and by the lagged return in other sectors.These results too are not surprising.

Table 01 :
Dollars investment in mutual funds, closed-end funds, ETFs, and UITs and the number of funds by category as ofDecember 31, 2012

Table 2a :
Money Flow into Equity Funds Note: EQ is the annual net flow of money into equity funds, Bond is the annual net flow of money into fixed income funds, and MM is the annual net flow of money into money market funds.Current and lagged (lag) returns are denoted with R. * Indicates statistical significance at the 0.01 level; ** indicates statistical significance at the 0.05 level.Note: * Indicates statistical significance at the 0.01 level; ** indicates statistical significance at the 0.05 level.Journal of Economic and Financial Studies.Page 66

Table 2c :
Money Flow into Money Market Funds Note: * Indicates statistical significance at the 0.01 level; ** indicates statistical significance at the 0.05 level.Journal of Economic and Financial Studies.Page 67

Table 3b :
Returns of Fixed Income Funds

Table 04 :
Identification of Causative Factors Which are Statistically Significant Influences on Models of Money Flows into Various Sectors and on the Returns of Those Sectors Bond, or MM indicate which sector's funds flow or returns are influenced by a specific causative factor on the left hand side of the table.