Feeling Rich: Disposable Income and Investor Rationality in the Market for Mutual Funds Swasti Gupta-Mukherjee * June, 2017 ABSTRACT This study shows that the representative investor s rationality and sophistication in the market for mutual funds has a significant relationship with changes in disposable income in the economy. I show that aggregate flows to actively managed funds relative to passively managed funds increase with the growth in disposable income, where active funds are more expensive and typically underperform passive funds on average. The representative investor s sensitivity to the price of retail S&P 500 index funds significantly decreases with the recent growth in disposable income. The representative investor shows inferior fund selection ability among S&P 500 index funds by driving higher flows to worse future performers in periods with high growth in disposable income, but superior ability in other periods. Investor choices among comparable actively managed funds reveal similar results, where the representative investor s price sensitivity and fund selection ability decreases with the recent growth in disposable income. There is evidence that fund sponsors engage in strategic fee-setting, where new funds offered in periods with high growth in disposable income charge higher fees and underperform post-inception. Taken together, the evidence is consistent with the market for mutual funds consisting of informed as well as uninformed investors. Further, the influence of uninformed investors is predictable based on disposable income levels, where they appear more influential in the market in periods when they feel rich, leading to distortions in the rational relationships between price, performance, and asset flows into mutual funds. Keywords: investor rationality, mutual fund flows, index funds, actively managed funds, disposable income JEL Classification: G11, G23, D14 * Author contact information: Quinlan School of Business, Loyola University Chicago, 16 E Pearson Street, Chicago IL, 60611; Tel. +1-312-915-6071; e-mail: sguptamukherjee@luc.edu. I thank Sris Chatterjee, Hae Mi Choi, Iftekhar Hasan, Tom Nohel, Clemens Sialm, Bhaskaran Swaminathan, Steven Todd, seminar participants at Fordham University and Loyola University Chicago, and conference participants at the Midwest Financial Association Annual Meeting 2015, Multinational Finance Society Annual Meeting 2016, and IFABS Barcelona Conference 2016 for many helpful comments and suggestions on a related paper titled When is Money Smart? Mutual Fund Flows, Energy Prices, and Household Disposable Income. I thank Alina Lamy of Morningstar Inc. for helping with the Morningstar Direct data. All remaining errors are my own.
I. Introduction Mutual funds have represented a sizeable proportion of household investments in the last two decades and have a material impact on life-cycle savings. Not surprisingly, a first order question in the literature on mutual funds is whether investors make informed decisions in the market for mutual funds. Moreover, U.S. equity funds, the largest asset class in the industry, managed over $6.5 trillion in assets with $2.98 trillion in passively managed funds and $3.54 trillion in actively managed funds in 2016. Despite the larger market share of actively managed funds, industry observers have noted a conspicuous upward trend in assets under passively managed funds in recent years, which doubled between 2007 and 2016. Given the substantial market share of active funds and remarkable growth of passive funds, the rationality of investor choices among both passive and active funds is of considerable economic importance. The study of investor rationality and sophistication in the existing literature has focused primarily on two aspects of investor choices. First, the existing literature on mutual funds has viewed the price sensitivity of investors as representing informed choices, since fund expenses have a negative relationship with performance. The rapid proliferation and survival of actively managed funds, which charge higher fees and generally underperform on a cost-adjusted basis compared to their passively managed counterparts, has raised questions about investor rationality. 2 Moreover, the significantly negative association between fees and performance exists within passively managed funds (Elton, Gruber, and Busse (2004)), as well as actively managed funds (e.g., Carhart (1997); Gil-Bazo and Ruiz-Verdu (2009)). Second, researchers have examined investors fund selection ability, i.e. the ability to pick future winners, as a reflection of informed choices. The premise in these studies is that if mutual fund investors reflect smart money, flows to funds should positively predict future risk-adjusted performance (e.g. Gruber (1996); Zheng (1999); Sapp and Tiwari (2004); Frazzini and Lamont (2008); Keswani and Stolin (2008)). In sum, the evidence on investor rationality and sophistication in the market for mutual funds has been inconclusive. Broadly, the existing 2 See, for example, Jensen (1968), Malkiel (1995), and Fama and French (2010) who find significant after-cost underperformance of actively managed U.S. equity mutual funds compared to passive funds. 1
literature posits that the market consists of a set of informed rational investors and a set of uninformed investors, where the existence of uninformed investors can explain the survival of inferior products (e.g. Gruber (1996); Elton, Gruber, and Busse (2004)). This study revisits investor rationality and sophistication in the market for mutual funds from a fresh angle the possibility that the observed attribute of the representative investor is time-varying, and related to disposable income in the economy. The median mutual fund-owning household earns a moderate income ($94,300 in 2016), suggesting that the average investor in the market for mutual funds faces financial constraints at the household level. 3 The constraints on disposable income at the aggregate level can influence the rationality and sophistication of the representative investor active in the market in two ways. First, previous studies have shown that wealthier households make better investment decisions and exhibit more financial sophistication, whereas less wealthy households are more uninformed and make investment mistakes (e.g. Campbell (2006); Calvet, Campbell, and Sodini (2009)). Less wealthy and uninformed investors face more constraints on disposable income, and may increase participation in the mutual fund market when the constraints decrease, especially in driving fund purchases that are a better indicator of investor ability than redemptions. 4 Thus, the fraction of uninformed investors driving fund inflows in the market in periods following higher growth in disposable income is likely to be higher than in periods when there are more constraints on disposable income. Second, an alternative but not mutually exclusive possibility is that investors could be putting more effort into the selection of mutual funds when they face more constraints on disposable income, since investment returns could reduce these constraints. The above notions form the basis for the central hypothesis in this study that the relative influence of uninformed investors compared to informed investors predictably increases with the growth in disposable income at the aggregate level. It follows that the representative investor s rationality and sophistication at the industry level should vary predictably over time, where the rationality of investor choices should 3 Only 11% of U.S. households which own mutual funds have a before-tax income of more than $200k (Figure 1). 4 Prior studies have noted that fund redemptions (outflows) are more likely to be influenced by factors other than future performance, such as liquidity concerns and taxes (see Keswani and Stolin (2008)). 2
decrease with the growth in disposable income. To test this central hypothesis, I mainly rely on three sets of empirical tests. First, I use time-variation in aggregate asset flows to actively managed funds relative to passively managed funds as an inverse proxy for investor rationality and sophistication, and study its relationship with growth in disposable income. Second, I use cross-sectional variation in fund-level asset flows to retail S&P 500 index funds to study price sensitivity and fund selection ability as a function of the growth in disposable income. As Elton, Gruber, and Busse (2004) note, S&P 500 index funds represent one of the simplest ways to examine investor rationality under the assumption of frictionless markets where the law of one price should hold, since all these funds hold virtually identical portfolios. Lastly, I repeat the analyses of price sensitivity and investor ability as a function of the growth in disposable income using cross-sectional variation in fund-level asset flows to actively managed funds. For the empirical investigation, I use the relatively new Morningstar Direct database on U.S. openend funds. I focus on retail U.S. equity funds with direct-market distribution during 1993 to 2017Q1, excluding funds sold via intermediaries (e.g. brokers or employer-sponsored retirement plans). Using this data, I first establish that aggregate asset flows to U.S. equity funds, including actively and passively managed funds, have the expected positive relationship with the growth in disposable income in the economy over the past six months. Annualized aggregate flows are around 4.3% higher in months following periods of highest monthly growth in disposable income compared to periods of lowest growth in disposable income in the sample period. The large economic magnitude of the inflows into the mutual fund industry when disposable income grows is consistent with more investors participating in the market when disposable income grows, as opposed to a constant group of investors increasing their investments. Further, in multivariate time-series regressions controlling for a variety of time-related factors, aggregate flows to actively managed relative to passively managed funds increase with the growth in disposable income. Annualized flows to active funds relative to passive funds are 3.7% higher in periods with the highest growth compared to periods with the lowest growth in disposable income. To the extent that active funds represent the less rational choice compared to passive funds on aggregate, these results 3
are consistent with the hypothesis that investor rationality at the industry level varies predictably over time, and has a significantly negative association with growth in disposable income. In multivariate cross-sectional regressions controlling for a variety of fund attributes to explain fund flows, the representative investor s price sensitivity in selecting among retail S&P 500 index funds significantly decreases with the growth in disposable income. For example, the negative effect of net expense ratio on fund flows is nearly two times higher following periods with the lowest growth in disposable income compared to other periods. I also find that the representative investor s ability to select the better future performers among S&P 500 index funds decreases with the growth in disposable income. Flows to S&P 500 index funds display a significantly negative relationship with future fund risk-adjusted returns in periods with the highest growth in disposable income, and a positive relationship otherwise. In other words, in periods following higher growth in disposable income, investors divert more flows towards index funds which subsequently underperform relative to their peers. These results on investor rationality are even more striking if, as Elton, Gruber, and Busse (2004) suggest, investors who buy index funds are among the most knowledgeable, making this asset class less likely to be affected by uninformed investors. Additional analyses using actively managed retail U.S. equity funds reveal similar results, with the representative investor s price sensitivity and fund selection ability significantly decreasing with the growth in disposable income. For example, the results show that the negative relation between fund flows and expense ratio and load of actively managed funds is significantly more pronounced in periods with low growth in disposable income. The representative investor s aversion towards funds with high beta risk and total risk also decreases with growth in disposable income. To the extent that high beta risk and total risk are associated with risk-taking by fund managers in response to agency problems, this finding suggests that the representative investor is less sensitive to funds agency problems in periods when disposable income experience higher growth. 5 Also, the main findings relating the growth in disposable income to price sensitivity and fund selection among passive and active funds cannot be explained by a secular time trend 5 For studies linking risk-taking by fund managers to agency problems see, for example, Brown, Harlow, and Starks (1996), Chevalier and Ellison (1997), and Christoffersen and Simutin (2015). 4
representing known and unknown time-related factors, such as trends in financial literacy and investor learning. Although there is some evidence of investor learning in the sensitivity to the pricing of mutual funds over time, there is no evidence to suggest that the representative investor has learnt to select funds which outperform on a risk-adjusted basis. In fact, fund selection ability among S&P 500 index funds has seemingly diminished over time as the number of passive funds in the market has grown. Finally, I explore some broader implications of the time variation in the attributes of the representative investor pertaining to the quality of products offered by fund sponsors. Specifically, I examine fee-setting at inception and post-inception performance of funds which enter the market in different regimes of growth in disposable income. I find evidence of a negative relationship between the quality of new mutual fund products and the growth in disposable income. Controlling for a variety of family and objective attributes and fund sponsor fixed effects, I find that new S&P 500 index funds and actively managed funds offered in periods with high growth in disposable income charge higher fees and underperform post-inception relative to new funds entering the market in other regimes. For example, new S&P 500 index funds entering the market in periods of high growth in disposable income are 8.1% to 26.1% more expensive than the median fund, ceteris paribus. It appears that fund sponsors engage in strategic feesetting and offer dominated products when the presence of uninformed investors is more pronounced. This study makes several contributions to the literature. First, it sheds new light on investor rationality and sophistication in the market for mutual funds. A small strand of the literature focuses on investor rationality in choices among S&P 500 index funds. Hortacsu and Syverson (2004) note the dispersion in fees of the seemingly homogenous set of retail S&P 500 index funds, and attribute the finding that even the more expensive funds among these attract substantial investments to product differentiation based on non-portfolio factors (e.g. fund age) and search costs. Elton, Gruber, and Busse (2004) also focus on S&P 500 index funds and find that inferior funds continue to survive contrary to rational expectations. They attribute this finding to various possible factors including potential investor irrationality and inferior financial advice from intermediaries. Another much larger strand of literature focuses on investor rationality and sophistication in the selection of actively managed funds, yielding mixed evidence (e.g. 5
Gruber (1996); Zheng (1999); Sapp and Tiwari (2004) and Frazzini and Lamont (2008); Keswani and Stolin (2008)). This study majorly deviates from the existing literature by showing that the attributes of the investor pool that drives flows in the mutual fund industry varies substantially over time, with investor rationality and sophistication varying predictably over time in the clientele for both passively managed and actively managed funds. Another key difference between the setting in this study and most previous studies on active as well as passive funds is that I use relatively new data on distribution channels to focus on funds sold directly to investors. Focusing on direct-market funds mostly alleviates the confounding influence of financial intermediaries and advisors in studying investor rationality. Additionally, economists have long been interested in understanding how disposable income and household wealth are related to household consumption and asset allocation. For instance, numerous economists have addressed how consumption decisions vary with the changes in disposable income (e.g. Bernanke (1983, 2004); Hamilton (2009)). Others have explored how income effects relate to portfolio allocation and savings (e.g. Guiso, Jappelli, and Terlizzese (1996); Angerer and Lam (2009)). However, as Campbell (2006) notes, our understanding of household finance and how it influences the broader capital markets is limited. Relative to the evidence on consumer behavior and price sensitivity in consumer products, far less is known about how changes disposable income are associated with investment choices and price sensitivity in financial products. I use the mutual fund industry setting to show that the rationality and sophistication of the average investor active in the market is associated with disposable income, thus addressing a potential channel by which household finance can have a broader impact on crucial aspects of financial markets, such as market efficiency. The rest of the paper proceeds as follows. Section II describes the data used for the empirical analyses. Section III examines the relationship between aggregate asset flows and time-varying disposable income. Section IV presents analyses of investor choices among S&P 500 index funds. Section V reports the results from analyses of investor choices among actively managed funds. Section VI presents additional results and explores some implications of the time variation in the rationality and sophistication of the representative investor for the quality of new funds entering the market. Section VII concludes. 6
II. Data and Summary Statistics II.A. Data on Disposable Income The data on disposable personal income is obtained from databases made available by the U.S. Department of Commerce s Bureau of Economic Analysis (BEA). I use BEA s DSPI time series data which is the seasonally adjusted disposable personal income in chained 2009 dollars, expressed in billions. Disposable income is the amount of net income a household or individual has available to invest, save, or spend after income taxes. 6 Since economic indicators such as DSPI often have monotonous time trends and are endogenous with other macro factors, I follow many studies on consumer behavior in using the short-run changes, as opposed to levels, in disposable income as the indicator that is comparable across time periods and could relate to investor behavior in the market for mutual funds. Figure 2 shows the distribution of changes in disposable income, measured as the rolling six-month cumulative growth rate of the seasonally-adjusted monthly disposable personal income (ΔDSPI (in %)) provided by the BEA over 1993 to 2017Q1. Over the sample period, the mean ΔDSPI is 2.3% and standard deviation is 1.4%, ranging from a minimum of -4.8% to a maximum of 7.41%, suggesting that there is significant dispersion in ΔDSPI. Table I reports summary statistics on DSPI as an economic indicator by calendar year. Summary statistics on the quarterly return on the S&P 500 stock market index is also reported. The highest mean ΔDSPI of 3.84% is recorded in 2000, with the only negative values of ΔDSPI recorded in 2009 and 2013 during the global financial crisis and its aftermath. The ΔDSPI does not coincide very closely with the S&P 500 return. The correlation between ΔDSPI in the three months in a quarter t and the return on the S&P 6 Disposable income indicators have a high correlation with indicators reflecting personal savings. Personal savings is the personal income that is left over after personal current taxes and outlays for personal consumption expenditures, interest payments, and net current transfers to government and to the rest of the world. Another economic indicator closely related to disposable income is consumer spending which closely tracks disposable income. 7
500 in quarter t, t 1, and t 2 is 0.05, 0.18, and 0.16, respectively, indicating relatively low positive correlations with changes in disposable income. II.B. Data on U.S. Open-end Funds and Sample Selection In this section, I describe the data on U.S. open end equity funds I use to test the hypotheses related to the rationality of investors in the market for mutual funds. The data on U.S. open end equity funds used in this study is provided by Morningstar, Inc. The data on aggregate asset flows and passively managed funds is from the relatively new survivorship-bias free Morningstar Direct (MD) database. MD provides a comprehensive coverage of open-ended funds in the United States, starting in 1993. The monthly aggregate flows data used in this paper covers the period 1993 to 2017Q1. It includes asset flows to actively-managed and passively-managed U.S. open-end equity funds, including exchange-traded funds (ETFs). Asset flows are the organic growth rate of funds TNA, defined as flows over the period divided by beginning net assets. To obtain aggregate flows, I exclude money market funds and fund of funds from the sample. To eliminate survivorship bias, I include obsolete funds in the sample. I include non-institutional funds distributed via direct market to focus on funds which have a retail clientele that is likely to redeem and purchase mutual fund shares as a response to householdlevel changes in disposable income, without the potentially confounding effects of financial intermediaries associated with broker-sold funds or funds sold via employer-sponsored retirement plans such as 401(k). U.S. equity funds sold via direct market channel are a significant portion of the sales of mutual funds in the sample period. For example, in April 2016, 50.6% of the total net assets (TNA) of retail U.S. equity funds came from direct-market funds, out of which about 60% is from actively managed and 40% is from passively managed funds. Figure 3 plots monthly total net assets (in $ billions) and relative flows over 1993 to 2017Q1 for retail direct-market passively managed (including ETFs) and actively managed U.S. equity funds provided by MD. The relative net cash flows to actively managed versus passively managed funds, 8
Flow_ActivePassive, is computed as the difference in monthly estimated dollar net flows between actively managed and passively managed funds, divided by the total net assets in actively managed and passively managed funds in the prior month. As noted in Morningstar s Asset Flow Commentary (January, 2017) and by other industry observers, the assets managed by passive funds shows an increasing trend over time relative to their active counterparts. 7 The report notes for 2016, Among U.S. equity funds, passively managed funds led by Vanguard s offerings took in $50.8 billion during December, which topped the previous monthly record of $41.9 billion set in November. Meanwhile, investors pulled $23 billion out of actively managed U.S. equity funds, extending the group s streak of outflows to 33 consecutive months. Figure 4 illustrates a similar trend on plotting asset flows measured as new money growth to passive and active U.S. equity funds over the sample period. The growth rate in ETFs is plotted separately to assess what proportion of the growth in assets under passive funds can be attributed to ETFs, with the data on ETFs available in MD starting in 1998. As has been noted widely by industry observers in recent years, assets under passive funds show positive growth in nearly all years, and is higher than the growth in assets under active funds in nearly all periods during 1993 to 2017Q1. On the other hand, active funds witness net outflows in all years from 2003 to 2016. For the first set of fund-level analyses, I obtain a sample of passively managed retail S&P 500 index funds from MD, which reports data at the fund share-class level. I use this sample of retail S&P 500 index funds to focus on funds which track the same underlying benchmark assets and are close substitutes in terms of portfolio risk, with any product differentiation arising from attributes unrelated to the underlying portfolio the funds track (see Hortacsu and Syverson (2004); Elton, Gruber, and Busse (2004)). A fund s net expense ratio, TNA, inception date (to compute fund age), manager tenure (of the longest-serving manager), number of funds in the family, number of share classes offered for the fund, tracking error, and alpha of returns is obtained from MD monthly if available, or on an annual basis. Tracking error is calculated in MD as the standard deviation of the error term in the linear regression of a fund s daily returns 7 https://corporate.morningstar.com/us/documents/assetflows/assetflowsjan2017.pdf 9
on the S&P 500 index s daily total returns for the past six months. I use MD s alpha computed monthly from the least-squares regression of the fund's excess return over Treasury bills and the excess returns of the fund's benchmark index, which is the S&P 500 index for this sample, based on a 12-month rolling window. MD calculates beta using the same regression equation as the one used for alpha, where beta equals the slope coefficient on the excess returns for the index. I obtain the standard deviation of returns each month from MD, calculated as the standard deviation of raw monthly returns based on a 12-month rolling window. The final sample of retail S&P index funds includes 222 different share classes represented 182 unique funds over the sample period. For the second set of fund-level analyses, I use a sample of retail U.S. equity actively managed funds obtained from MD over 1993 to 2017Q1. To form an initial sample of actively managed funds which are the comparable active counterparts of the S&P 500 index funds, I select direct-market retail active funds which have the S&P 500 index as the primary benchmark assigned to investments based on their Morningstar category. This sample includes funds with the following investment objectives listed in Morningstar: aggressive growth, asset allocation, balanced, equity income, growth, growth and income, income, small company. 8 Thus, Index, sector, bond, international, and money market funds are excluded. I obtain other fund attributes from MD monthly if available, or on an annual basis, and exclude funds with missing data. The final sample of actively managed funds includes 2,436 different share classes for 1,316 unique funds over the period 1993-2017Q1. Table II reports summary statistics on the sample of passively managed S&P 500 index funds and actively managed funds with S&P 500 as their benchmark. In Table II Panel A, there are 15 different share classes for 13 unique S&P 500 index funds at the start of the sample period in 1993. The number of funds in the sample represents 101 different share classes for 43 unique funds in 2016, representing a nearly seven-fold increase in the share classes offered between 1993 to 2016. The sample is comparable to 8 The results reported in the paper are robust to an alternative sample including the following objectives: aggressive growth, equity income, growth, and growth and income. 10
previous studies which focus on S&P 500 index funds. For example, the sample in 2000 includes available data on 53 unique index funds, whereas Hortacsu and Syverson (2004) record 50 funds in 2000. In Table II, the sample of active funds in 1993 includes 67 different share classes for 61 unique actively managed funds with S&P as benchmark, and in 2016 includes 728 different share classes for 375 unique actively managed funds. The summary statistics on fund characteristics in Table II Panel B shows that the sample of passive funds tend to include larger funds than the comparable sample of active funds, with the mean TNA of the passive funds being $1.6 bill versus $0.61 bill for active funds. The mean expense ratio of the active funds is nearly two times higher than those of their passive counterparts. As expected, the average tracking error of 0.38% for the passive funds relative to the S&P 500 benchmark index is much lower on average than the 4.79% for the active funds. III. Do Changes in Disposable Income Affect Asset Flows? III.A. Aggregate Flows To explore how household disposable income at the aggregate level may affect the mutual fund industry, I first investigate whether mutual fund flows are related to growth in disposable income in the economy. Theory predicts that investments in risky assets increase with disposable income, when precautionary motives for saving are lower. A substantial proportion of U.S. households own mutual funds, so a part of their investments in risky assets is funneled through mutual funds. Ceteris paribus, mutual fund flows should have a positive relation with growth in disposable income. Additionally, a significant proportion of mutual fund investments are through periodic payroll deductions contributed towards retirement plans, which are less likely to be affected by short term growth in disposable income. In this case, there could be no significant link between new money flows to mutual funds and changes in disposable income at the aggregate level. To explore the link between fund flows and short term changes in disposable 11
income, the empirical strategy is to estimate the sensitivity of asset flows to the recent growth in disposable income. In Table III, I examine the relation between monthly aggregate asset flows and recent growth in disposable income in a multivariate time-series regression setting that allows including other time-related controls, such as recent stock market performance. The dependent variable is the aggregate net cash flows to funds in month t, expressed in USD billions in columns (1)-(6) and new money growth rate in columns (7)-(12). The independent variable of interest is the growth in disposable income ( DSPI), measured on a rolling basis as the cumulative growth rate of disposable income over the months to t 5 to t. Alternatively, some specifications use the dummy variables Low_DSPI and High_DSPI to measure the growth in disposable income. Low_DSPI (High_DSPI) is a dummy variable assigned a value of one if the cumulative monthly growth in disposable income in the six months t 5 to t ( DSPI) is in the bottom (top) 20% of the sample period, and zero otherwise. In the remainder of the paper, I use the dummy variables Low_DSPI and High_DSPI as measures of changes in disposable income for ease of interpretation. However, the results in the remainder of the paper are robust to using the continuous variable DSPI. As controls, I consider the following time-related variables: Lag_AggregateTNA, Lag_SP500Ret, Lag2_SP500Ret, Lag_Recession, and Lag2_Recession. Lag_AggregateTNA is the aggregate total net assets in the prior month and accounts for industry size which could affect future net cash flows. Lag_SP500Ret and Lag2_SP500Ret measure the quarterly return on the S&P500 index in the two prior quarters preceding the measurement of flows, since aggregate stock market performance could affect flows to equity mutual funds. Controlling for stock market performance ensures that any observed effect of growth in disposable income on asset flows to U.S. equity funds is not explained by the possibility that the growth in disposable income repackages the effect of stock market returns. Also, I control for recessions using the indicators Lag_Recession and Lag2_Recession since some previous studies suggest that active funds outperform in recessions, perhaps making the preference for active funds in these periods more rational than in other periods (e.g. Moskowitz (2000); Kosowski (2006); and Staal (2006)). The p-values in the regression results reported in Table III are based on robust standard errors clustered by quarter. 12
The regression results reported in Table III show that aggregate asset flows to U.S. open-end funds have a positive and significant relationship with the recent growth in disposable income in the economy. For example, in column (12), annualized aggregate flows are around 4.3% higher in periods where High_DSPI equals one compared to periods where Low_DSPI equals one. In unreported regressions, these and other results reported in the remainder of the paper are robust to using the quarterly change in U.S. gross domestic product (GDP) as a proxy for changes disposable income. III.B. Asset Flows to Active versus Passive Funds In this section, I examine the aggregate asset flows to actively managed funds relative to passively managed funds (including exchange-traded funds). To the extent that actively managed funds are the less rational choice due to their higher costs and inferior performance compared to their passively managed counterparts on aggregate, periods in which asset flows to active funds relative to passive funds increase (decrease) can be viewed as reflecting less (more) rational choices of investors. In Table IV, I examine the relation between monthly asset flows to active and passive funds, and recent growth in disposable income in a multivariate time-series regression setting. The dependent variable is the difference in total net cash flows between active and passive funds in month t expressed in USD billions (Diff_ActivePassive) in columns (1)-(6), and new money growth rate in columns (7)-(12) (Flow_ActivePassive). Diff_ActivePassive is the difference in monthly estimated net flows between actively managed and passively managed funds in month t, expressed in USD billions. Flow_ActivePassive is the difference in monthly estimated dollar net flows between actively-managed and passively-managed funds in month t, divided by the total net assets in actively managed and passively managed funds in month t 1. I note that there is considerable variation in Flow_ActivePassive over the sample period in this study. The mean of Flow_ActivePassive is -0.014, which is statistically indistinguishable from zero. However, there is significant dispersion in Flow_ActivePassive, which ranges from an annualized -24.24% to 25.8% and has a standard deviation of 7.74%. 13
In Table IV, using both the level of growth in disposable income DSPI as well as the dummy variables Low_DSPI and High_DSPI measured over the months to t 5 to t, the regression results suggest that asset flows to actively managed funds relative to passively managed funds increase with the recent growth in disposable income. The coefficient on DSPI in all specifications is positive and significant at least at the 5% level. Moreover, the relation between growth in disposable income and fund flows is economically significant. In column (1), a one standard deviation increase in DSPI increases the net cash flows to active U.S. equity funds relative to passive U.S. equity funds by an annualized $40 billion. In column (12), annualized flows to active funds relative to passive funds are 3.67% higher in periods with highest growth compared to periods with lowest growth in disposable income, ceteris paribus. In sum, the results support the hypothesis that the representative investor in the market for mutual funds makes less (more) rational choice when recent growth in disposable income is high (low). It is worth noting that the time-series regression approach is vulnerable to secular time trends that raise questions about whether the results on disposable income could be driven by spurious correlations with time-related variables omitted from the analyses. To eliminate this concern, the remainder of the empirical tests focus on cross-sectional analyses which eliminate the effect of time trends by absorbing time-specific heterogeneities in the market for mutual funds. IV. Investor Choices among S&P 500 Index Funds In this section, I study investor rationality and sophistication based on the asset flows to retail S&P 500 index funds. IV.A. Price Sensitivity The existing literature on mutual funds has viewed the price sensitivity of investors as a proxy for rational behavior, since expensive funds tend to underperform after-cost. This view exists in studies on 14
passively managed funds (Elton, Gruber, and Busse (2004)), as well as actively managed funds (e.g., Carhart (1997); Gil-Bazo and Ruiz-Verdu (2009)). Researchers have almost unanimously suggested that rational investors should avoid funds with high expenses and fees. Based on this premise, in this section I examine price sensitivity of investors in choices among S&P 500 index funds to study time variation in investor rationality as a function of recent changes in disposable income. To study price sensitivity, I follow previous studies in examining the impact of a fund s expenses and loads on the fund s propensity to attract flows. Based on rational expectations and evidence in the existing literature, price sensitivity should manifest in a negative relationship between measures of fund expenses and fund flows. Moreover, if the representative investor is less rational when the growth in disposable income is higher, we should expect the negative fee-flow relationship to be weaker in periods with higher growth in disposable income. In other words, the representative investor should be less (more) averse to more expensive funds when growth in disposable income is high (low). Table V presents regression results on the relation between asset flows and pricing of retail S&P index funds using cross-sectional regressions, where all specifications include month fixed effects. The independent variables representing fund prices are the net expense ratio (Expense Ratio), dummy variables representing the more expensive funds (HighExpense) signifying expense ratios above the median levels in the sample in the month, and a dummy variable representing funds that charge loads (Load). The coefficients on measures of fund pricing in the cross-sectional regressions reflect investor preference for funds which vary in pricing. I control for standard fund attributes that previous studies suggest could be related to fund flows, namely, fund size represented by the natural logarithm of the fund s TNA (log (TNA)), the natural logarithm of fund age plus one (log (Age)), the natural logarithm of the number of share classes offered by the fund (log (#Share Classes)), the natural logarithm of the family size (measured as the number of funds in the complex) (log (Family Size)), the natural logarithm of the longest-serving manager s tenure (log (Manager Tenure)), the tracking error relative to the S&P 500 benchmark index (Past Tracking Error), and risk-adjusted returns measured as alpha (Past Alpha). The statistical significance tests are based on p- values from robust standard errors clustered by fund and month. 15
In Table V, various specifications are reported where the measure of expenses is interacted with the indicator variables Low_DSPI and High_DSPI. In columns (1) and (3), consistent with rational expectations, flows have a significantly negative relationship with HighExpense and Expense Ratio. However, the relation between Load and fund flows is statistically insignificant in column (5). Supporting the hypothesis outlined earlier, the interaction terms of higher expenses with Low_DSPI are negative and significant at standard levels in all specifications, and with High_DSPI are insignificant. In column (2), in periods when Low_DSPI equals one, the annualized flows to the more expensive funds (HighExpense= 1) are 3.11% lower than the less expensive funds (HighExpense= 0). The negative relation between HighExpense and flows arises from the significantly negative relationship when Low_DSPI equals one, with a statistically insignificant relation in other periods when Low_DSPI equals zero. That is, the representative investor appears particularly sensitive to expenses in periods with low growth in disposable income. Similar results are obtained in column (3)-(4) where Expense Ratio is used as the measure of fund expenses, with the negative effect of Expense Ratio on fund flows being 1.8 times higher when Low_DSPI equals one than in periods when Low_DSPI equals zero. In columns (5)-(6) when the dummy variable Load is used to examine price sensitivity, the results remain consistent with the findings based on net expense ratios. For example, in column (6), the representative investor s sensitivity to a fund s load is only significant when Low_DSPI equals one. In sum, investors are substantially less averse to selecting the more expensive S&P 500 index funds when choosing across the cross-section of these index funds in periods following high growth in disposable income, compared to periods following low growth in disposable income. So, investor rationality in the choice of S&P 500 index funds decreases with the growth in disposable income in the economy. IV.B. Future Performance Elton, Gruber, and Busse (2004) show that S&P 500 index funds have predictable risk-adjusted returns, as measured by alpha. Moreover, past alphas predict future alphas for these funds. Fund selection 16
ability is the ability of investors, at a point in time, to select funds which outperform in the future from the cross-section of available funds. In this section, I consider the fund selection ability of investors as revealed by the relationship between fund flows and future risk-adjusted performance of the funds, where a sophisticated investor should drive more flows into funds which outperform in the future. If investor rationality and sophistication decreases with the growth in disposable income, then flows should be a weaker (stronger) positive predictor of future performance in periods with high (low) growth in disposable income. In Table VI, the coefficients and statistics are estimated from cross-sectional regressions of monthly fund performance for the S&P 500 index funds. Fund performance is measured in month t on a rolling basis over a 12-month window as the alpha based on the monthly excess returns over the months t+1 to t+12 (obtained from Morningstar). 9 The independent variable representing fund flows in column (1)-(2) is the dummy variable representing funds drawing the higher flows, HighFlow, which equals one if the fund s monthly new money growth on month t is above the median of all retail S&P 500 index funds in the same month, and zero otherwise. In columns (3)-(4), flows are represented by the continuous variable Flow in month t. The regressions in Table VI reveal fund selection ability in the representative investor s choice of S&P 500 index funds, as reflected in the significantly positive coefficients on HighFlow and Flow in columns (1) and (4). However, the relationship between fund flows and future risk-adjusted performance is strikingly different in periods with the highest growth in disposable income (High_DSPI= 1). For example, in column (2), HighFlow is a negative predictor of future performance when High_DSPI equals one, but is a positive predictor in other periods. A similar conclusion can be drawn from column (5). In columns (3) and (6) when a variety of fund characteristics are included as control variables, the coefficient on HighFlow and Flow is only significant when High_DSPI= 1. Therefore, the representative investor 9 These and other results reported in this paper are robust to measuring alpha based on the four-factor model of Carhart (1997). 17
displays poor fund selection, or less rational and sophisticated choices, in periods when there is high growth in disposable income. V. Investor Choices among Actively Managed Funds funds. In this section, I repeat the tests of investor rationality based on the asset flows to actively managed V.A. Price Sensitivity First, I repeat the analyses of price sensitivity using the sample of actively managed funds with the S&P 500 as their passive benchmark. In studying the representative investor s sensitivity to the price of actively managed funds, I consider the two main channels by which owning mutual funds is costly to investors: expense ratios and loads. Table VII reports the results on price sensitivity of retail actively managed funds and time-varying disposable income. The dependent variable in the cross-sectional regression is monthly fund flows, with all specifications including month and objective fixed effects. Various specifications are reported where different measures of expenses are interacted with the indicator variables Low_DSPI and High_DSPI, computed from the changes in disposable income over the prior six months. The variables representing expenses are HighExpense and LowExpense in columns (1)-(2), Expense Ratio in columns (3)-(4), and Load in columns (5)-(6). The results based on price sensitivity of actively managed funds in Table VII support the results reported earlier for S&P 500 index funds. The interaction terms of higher expenses with Low_DSPI are negative and significant at the 1% level in all specifications, and with High_DSPI are insignificant or positive at the 10% level (column (6). In column (2), in periods when Low_DSPI equals one, the annualized flows to the more expensive funds (HighExpense= 1) are 3.49% lower than to the less expensive funds. In periods when Low_DSPI equals zero, the annualized flows to the more expensive funds 18
(HighExpense= 1) are 1.78% lower than to the less expensive funds. Thus, the outflows from the more expensive funds are nearly two times more in periods following the lowest growth in disposable income compared to other periods. Similar results hold in columns (3)-(6). In sum, the representative investor appears significantly more sensitive to the prices of actively managed funds in periods with low growth in disposable income compared to periods with high growth in disposable income. V.B. Future Performance In this section, I consider the rationality and sophistication of the representative fund investor revealed by fund selections based on the sample of actively managed funds. Prior research on actively managed funds has asked if investors exhibit the ex ante ability to purchase mutual fund shares of future winners and divest shares of future losers, i.e., are they smart money? However, the evidence is inconclusive. For example, Gruber (1996), Zheng (1999), and Keswani and Stolin (2008) find evidence of money being smart, where funds experiencing inflows outperform those experiencing outflows in the short term. In contrast, Sapp and Tiwari (2004), and Frazzini and Lamont (2008) do not find evidence of ability in mutual fund selections by investors. Notably, the task of selecting future winners among active funds is a substantially more difficult task than selecting future winners among passive funds, since the dispersion in future risk-adjusted performance is much larger for active funds which differ more their portfolio strategies, risk, and returns. In Table VIII, the coefficients and statistics are estimated from cross-sectional regressions of monthly fund performance. Fund performance is measured in month t on a rolling basis over a 12-month window as the alpha (in %) based on the monthly excess returns over the months t+1 to t+12. HighFlow (columns (1)-(2)) and Flow (columns (3)-(4)) in month t proxy for the representative investor s fund selection, and the presence of fund selection ability should drive a positive relationship between fund flows and future alpha. Supporting the studies which find evidence of fund selection ability among investors in actively managed funds, I fine some evidence that flows positively predict future risk-adjusted returns. In 19
column (1), the coefficient of HighFlow is positive and significant at the 5% level, suggesting that funds which attract higher flows exhibit better future performance. However, in column (2), when the interaction terms HighFlow x Low_DSPI and HighFlow x High_DSPI are also included, the coefficient of HighFlow in periods when High_DSPI = 1 is significantly negative, and in periods when High_DSPI = 0 is positive. Consistent with earlier results based on passively managed funds, the representative investor drives more flows into funds which subsequently underperform when High_DSPI equals one relative to periods when Low_DSPI equals one. This result suggests that the representative investor makes poor choices among active funds following periods with high growth in disposable income, but makes substantially better choices in other periods. The results are qualitatively similar in columns (3)-(4) when Flows is used as an explanatory variables instead of the dummy variable HighFlow. In sum, the results are consistent with the representative investor s effectiveness in allocating capital to actively managed funds decreasing with the growth in disposable income. The results based on active funds reiterate the earlier findings based on passive funds which suggest that there is considerable time variation in investor rationality and sophistication. V.C. Fund Risk In this section, I conduct an additional test which could reveal the time-varying attributes of the representative investor based on the preference for funds with different risk-taking behavior. Active funds differ substantially in their risk-taking behavior. I rely on existing studies to identify risk-taking behavior by fund managers that is likely to arise from perverse incentives. For example, Brown, Harlow, and Starks (1996) and Chevalier and Ellison (1997) associate career concerns with fund managers increasing risktaking. Additionally, Christoffersen and Simutin (2015) and others argue that a manager who faces benchmarking pressures has an incentive to hold high-beta (often negative-alpha) stocks since these stocks have a higher likelihood of beating the benchmark. To the extent that more rational and sophisticated investors avoid funds with more agency problems which engage in risk-taking arising from perverse 20
incentives for the manager, an (indirect) test of whether the representative investor is more rational and sophisticated when disposable income is lower would be to observe if there are lower allocations to funds which exhibit these types of risk-taking behavior in these periods. This section provides evidence of the relation between risk-taking attributes of the funds and the flows they receive, conditional on the recent rate of growth in household disposable income. Table IX presents cross-sectional regression results on the relation between a fund s monthly flows in month t and its past risk-taking attributes. All specifications include month and objective fixed effects. Various specifications are reported where two measures reflecting a fund s risk-taking, High Beta and High Total Risk, are interacted with the indicator variables Low_DSPI and High_DSPI. High Beta is a dummy variable assigned a value of one if the fund s beta in month t is above the median beta across all funds with the same objective in month t, and zero otherwise, where beta is the slope coefficient from regressions of the fund s monthly excess returns on the excess return of the S&P 500 Composite index measured on a 12- month rolling basis over months t 12 to t 1. High Total Risk is a dummy variable assigned a value of one if the fund s total risk in month t is above the median total risk across all funds with the same objective in month t, and zero otherwise, where total risk is computed in month t on a 12-month rolling basis as the standard deviation of raw monthly returns over the months t 12 to t 1. The results in Table IX lend empirical support to the hypothesis that the rationality of the representative investor decreases with the growth in disposable income. The representative investor is more discerning of funds which tend to be linked to agency problems in periods with low growth in disposable income, and less discerning in periods with high growth in disposable income. For example, in column (2), the coefficient on High Beta x Low_DSPI is negative and significant at the 1% level, whereas the coefficient on High Beta x High_DSPI is positive and significant at the 1% level. In column (2), the flows to funds with higher than median beta in periods when Low_DSPI equals one is 3.06% lower than to funds with lower than median beta on an annualized basis, ceteris paribus. In striking contrast, the flows to funds with higher than median beta in periods when High_DSPI equals one is 6.53% higher than to funds with lower than median beta on an annualized basis, ceteris paribus. Thus, the representative investor s preference for 21
funds with high betas in periods with high growth in disposable income contrasts the preference in periods with low growth in disposable income. The results are similar in columns (3)-(4). Funds associated with higher total risk (i.e. volatility) attract significantly lower flows in periods with low growth in disposable income compared to periods with high growth in disposable income, ceteris paribus. In sum, the representative investor s aversion towards funds which engage in higher risk-taking behavior decreases with the recent growth in disposable income. VI. Other Results and Implications VI.A. Investor Learning To conclude the empirical analyses, I consider the possibility that the findings on disposable income and investor choices can be explained by secular time trends in factors that could affect investor rationality and have a spurious correlation with trends in disposable income. For instance, if there is a downward time trend in the growth in disposable income which coincides with an upward time trend in financial literacy and investor learning in the market for mutual funds, it could explain the results uncovered so far. As an initial test of this possibility, I observe that the correlation between a time trend variable and the six-month rolling growth in disposable income is only -0.20. This does not suggest that the measure of growth in disposable income coincides with a time trend. Nevertheless, to formally explore the role of time trends, I examine the representative investor s price sensitivity and selection ability for passive and active funds over time using multivariate regressions. If there is an upward trend in investor learning, the negative priceflow relationship reflecting price sensitivity, and the positive flow-performance relationship representing selection ability should become stronger over time. Table X reports the results of cross-sectional regressions exploring time trends in price sensitivity and selection ability reflected in investor choices among passive and active funds. The main variables of interest are the interactions of log (Time) and expenses in explaining flows, and the interaction of log (Time) 22
and flows in explaining future risk-adjusted performance. Here, log (Time) is a time trend variable, measured as the natural logarithm of a time unit that advances by one every year starting with a value of one in 1993. The coefficients on HighExpense x log (Time), Expense Ratio x log (Time), and Load x log (Time) are negative and statistically significant for both passive funds (columns (1)-(3)) and active funds (columns (6)-(8)). Thus, the price sensitivity of investors in choosing among passive funds as well as active funds increases over time. This is consistent with the notion that investors have learnt to make more rational and sophisticated choices with regards to the costs of owning mutual funds, exhibiting a stronger preference for cheaper funds in later periods, ceteris paribus. This result confirms the observation by industry analysts that investors have strongly preferred cheaper funds in recent years, explaining the exodus of capital from active to passive funds in recent years and leading to increased price competition which is forcing funds to lower fees. 10 However, the results on fund selection ability are not consistent with an increasing trend in investor rationality and sophistication over time. In fact, for passive funds, the representative investor s ability to select better future performers diminishes over time, as indicated by the significantly negative coefficients on the interactions terms HighFlow x log (Time) and Flow x log (Time) in columns (4)-(5). For active funds, there is no significant time trend in the representative investor s ability to select future outperformers (columns (9)-(10)). In sum, learning over time does not explain the earlier results relating investor rationality and sophistication to the growth in disposable income in the economy. Although there is some evidence of investor learning in the sensitivity to the pricing of mutual funds, there is no evidence to suggest that investors have learnt to select funds which outperform on a risk-adjusted basis in the future. I speculate that this could be due to increased awareness about pricing of mutual funds which is a salient attribute that is easy to observe and compare between funds even for uninformed investors, whereas selection ability 10 "Investors have been selling expensive funds and buying lower-cost ones, exerting pressure on the industry, in general, and on active managers, in particular, to lower fees in order to stay competitive," said Alina Lamy, Morningstar's senior analyst for quantitative research. 23
requires an understanding of less salient attributes that could predict fund performance. 11 I leave a more rigorous investigation of learning in the market for mutual funds as a promising avenue for future research. VI.B. Implications: Quality of New Funds Earlier sections of this paper documented that the representative investor s rationality and sophistication varies predictably over time. The evidence lends empirical support to the notion that the market for mutual funds includes both informed and uninformed investors, and their relative influence in the market is predictable over time. In this section, I conduct a preliminary investigation of some broader implications of the time-varying attributes of the representative investor. Specifically, I examine whether there are systematic differences in the quality of mutual fund products offered by fund sponsors in different market regimes which vary in the representative investor s rationality and sophistication. Conventional wisdom and academic research have considered low fees and high risk-adjusted returns to be primary features of high quality mutual fund products which produce better investor returns. Moreover, the time-varying attributes of the representative investor (i.e. the clientele) is more pertinent to the quality of the new products offered by fund sponsors in different market regimes, since the fund sponsor could offer a product that caters to the nature of the contemporaneous clientele. Thus, I explore fee-setting at inception and post-inception performance of new funds as a function of the recent growth in disposable income in the economy. VI.B.1. Fee-setting in New Funds Table XI reports the results of OLS regressions explaining fee-setting. The dependent variable is the net expense ratio in year t. Columns (1)-(5) report regression results for the sample of passive S&P 500 index funds. Columns (6)-(10) report results for the sample of comparable active funds. The main 11 See Bordalo, Gennaioli, and Shleifer (2013) who model a consumer whose attention is drawn to salient attributes of goods, such as quality or price. 24
hypothesis tested in Table XI is that new funds offered in periods where the growth in disposable income is high are more expensive, suggesting potential strategic fee-setting by fund sponsors targeting uninformed investors. Inception is an indicator variable assigned a value of one if the fund s inception year is year t, and differentiates new funds from incumbent funds. Since the exact date of decision-making around the inception of a fund is not clear, I assume that the inception is at year-end. 12 High_DSPI (Low_DSPI) is a dummy variable assigned a value of one if the cumulative monthly growth in disposable income in the 12 months in year t is in the top (bottom) 20% in the sample period, and zero otherwise. The main variables of interest are the interaction terms Inception x Low_DSPI and Inception x High_DSPI, which differentiate between new funds which enter the market following periods of low and high growth in disposable income, respectively. All specifications include family fixed effects or year fixed effects, with the regressions based on the sample of active funds also including objective fixed effects. Following previous studies, I include family fixed effects to eliminate sponsor-level heterogeneities in fee-setting (e.g. Tufano and Sevick (1999)). Additionally, a variety of family-level and objective-level control variables that could explain mutual fund fees are included. The results in Table XI show that fund sponsors offer new funds with higher net expense ratios in periods of high growth in disposable income. In columns (1)-(5) based on retail S&P 500 index funds, the coefficient on Inception as an independent variable is mostly insignificant, with the coefficient on Inception x High_DSPI is positive and statistically significant at standard levels. Although new index funds do not have significantly different expenses than incumbent funds in general, the ones that enter the market following periods of high growth in disposable income in the economy tend to charge higher fees. The results are also economically significant. New S&P 500 index funds entering the market in periods of high growth in disposable income (High_DSPI =1) are 8.1% (column (3)) to 26.1% (column (5)) more expensive relative to the median net expense ratio, ceteris paribus. 12 In unreported results, I assume the inception is that the year-beginning, with the explanatory variables measured in the previous year. The results remain qualitatively similar, and are available upon request. 25
The regression results in columns (5)-(10) of Table XI based on fee-setting in active funds corroborate the findings based on passive funds. The coefficients on Inception x High_DSPI are positive and significant at the 1% level in almost all specifications for the sample of active funds. Active funds entering the market in periods of high growth in disposable income (High_DSPI =1) are 2.3% (column (3)) to 9.8% (column (5)) more expensive relative to the median net expense ratio, ceteris paribus. Thus, the results in Table XI based on passive and active funds are consistent with strategic feesetting by fund sponsors in periods when uninformed investors are likely to form a larger fraction of the retail clientele, where uninformed investors are less sensitive to fund pricing. VI.B.2. Post-inception Performance of New Funds Table XII reports the results of OLS regressions explaining funds alpha. The dependent variable is the alpha for a fund in month t, computed on a rolling basis over the next 12 months t+1 to t+12. Columns (1)-(5) report regression results for the sample of passive S&P 500 index funds. Columns (6)-(10) reports results for the sample of comparable active funds. The regressions in Table XII test the hypothesis that new funds offered in periods where the growth in disposable income is high exhibit post-inception underperformance, suggesting potentially lower quality of mutual funds products targeting uninformed investors in these periods. The results reported in Table XII reveal systematic differences in short-term post-inception riskadjusted performance of new funds conditional on the recent growth in disposable income in the economy. The regression specifications in columns (1)-(5) based on the sample of S&P 500 index funds show that new funds typically underperform over the short term post-inception, except the ones which enter the market in periods of low growth in disposable income. The coefficient on Inception is negative and statistically significant in columns (1)-(5). Additionally, the coefficient on Inception x Low_DSPI is significantly positive significant at least at the 10% level in all specifications. Therefore, the new S&P 500 index funds entering the market in period of low growth in disposable income perform significantly better 26
post-inception than those entering in period of high growth in disposable income. The results are materially unchanged in columns (6)-(10) based on the sample of active funds. VII. Conclusion Investor rationality has long been of interest to researchers studying the functioning of capital markets. One strand of the literature has focused on investor rationality and sophistication in the market for mutual funds. This paper seeks to shed new light on investor rationality and sophistication by examining whether time-varying disposable income of U.S. households is related to investor rationality observed in the market for mutual funds. In analyses based on aggregate asset flows, I find that flows to actively managed funds relative to passively managed funds increase with the growth in disposable income, ceteris paribus. Previous literature generally views active funds as the less rational choice compared to passive funds on average, since active funds are more expensive and typically underperform their passive counterparts on an after-cost basis. These results support the central hypothesis in this study that investor rationality and sophistication at the industry level varies predictably over time, and has a significantly negative association with growth in disposable income. Further, in multivariate cross-sectional regressions controlling for a variety of fund attributes and time-related variables to explain fund flows, the representative investor s price sensitivity in choosing retail S&P 500 index funds significantly decreases with the growth in disposable income. The mutual fund literature posits that informed investors should generate a negative relation between prices and the demand for the fund captured by flows. The evidence in this study suggests that this predicted negative relationship between the cost of owning a mutual fund and the demand for the fund is less pronounced (i.e. less rational) in periods with higher growth in disposable income. For example, loads do not significantly affect fund flows, except in periods with the lowest growth in disposable income. I also find that the representative 27
investor s ability to select the better future performers among S&P 500 index funds decreases with the growth in disposable income. One set of tests reveal that flows positively predict future fund performance in an unconditional setting, but flows have a significantly negative link with future performance in periods when there is high growth in disposable income, alluding to inferior selection ability in these periods. The results from another set of tests based on actively managed retail U.S. equity funds are also consistent with the hypothesis that investor rationality at the industry level decreases with disposable income. The representative investor s price sensitivity and ability to drive flows into future winners among active funds significantly decreases with the growth in disposable income. For example, the results show that the rational negative relation between fund flows and expenses of actively managed funds is significantly less pronounced in periods with higher growth in disposable income. Moreover, the representative investor also reveals less aversion towards funds associated with more agency problems in periods with higher growth in disposable income. Learning over time does not explain the results relating informed decisions to the growth in disposable income in the economy. Although there is some evidence of investor learning in the sensitivity to the pricing of mutual funds increasing over time, there is no evidence to suggest that investors have learnt to select funds which outperform on a risk-adjusted basis. Lastly, there is evidence that fund sponsors offer new mutual fund products with higher expenses and postinception underperformance in periods with higher growth in disposable income when the representative investor exhibits less rationality and sophistication. In sum, this study presents novel evidence that the representative investor s rationality and sophistication varies predictably over time. The evidence in this study suggests avenues for future research. For example, it would be interesting to explore whether investor rationality shows significant variation over time and is a function of household disposable income in other segments of financial markets, such as in the trading of individual stocks, which could have broad implications for market efficiency. 28
References Angerer, X., and P. Lam, 2009. Income Risk and Portfolio Choice: An Empirical Study. Journal of Finance 64 (2), 1037 1055. Angrist, J., V. Chernozhukov, I. Fernández-Val, 2006. Quantile regression under misspecification, with an application to the U.S. wage structure. Econometrica 74, 539 563. Barber, B., T. Odean, and L. Zheng, 2005. Out of sight, out of mind: The effects of expenses on mutual fund flows. Journal of Business 78 (6), 2095 2119. Berk, J., and R.C. Green, 2004, Mutual fund flows and performance in rational markets, Journal of Political Economy 112, 1269 1295. Bernanke, B.S., 1983. Irreversibility, uncertainty, and cyclical investment. Quarterly Journal of Economics 98, 85 106. Bernanke, B.S., 2004. Oil and the economy. Remarks at the Distinguished Lecture Series, Darton College, Albany, Georgia, October 21. Bernanke, B.S., 2006. Energy and the economy. Speech to the Economic Club of Chicago, June15. Bordalo, P., N. Gennaioli, and A. Shleifer, 2013. Salience and Consumer Choice, Journal of Political Economy 121, 803 843. Brown, K.C., W.V. Harlow, and L.T. Starks, 1996. Of tournaments and temptations: An analysis of managerial incentives in mutual fund industry, Journal of Finance 51, 85 110. Campbell J.Y., 2006. Household Finance, Journal of Finance 61,1553 1604. Calvet, L., J. Campbell, and P. Sodini, 2009. Measuring the Financial Sophistication of Households, American Economic Review: Papers and Proceedings 99, 393 398. Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52, 57 82. Chevalier, J., and G. Ellison, 1997, Risk taking by mutual funds as a response to incentives, Journal of Political Economy 105, 1167 1200. Christoffersen, S., and M. Simutin, 2015, On the demand for high-beta stocks: Evidence from mutual funds, Working Paper, University of Toronto. Elton, E., M. Gruber, and J. Busse, 2004. Are Investors Rational? Choices among Index Funds, Journal of Finance 59, 261 288. Fama, E.F., and K.R. French., 1993. Common risk factors in returns on stocks and bonds. Journal of Financial Economics 33, 3 56. Fama, E., French, K., 2010. Luck versus skill in the cross section of mutual fund returns. Journal of Finance 65, 1915 1947. Frazzini, A., and O. Lamont, 2008. Dumb money: Mutual fund flows and the cross-section of stock returns, Journal of Financial Economics 88, 299 322. Gil-Bazo, J., and P. Ruiz-Verdu, 2009, The relation between price and performance in the mutual fund industry, Journal of Finance 64, 2153 2183. Gruber, M., 1996, Another puzzle: The growth in actively managed mutual funds, Journal of Finance 51, 783 810. Guiso, L., T. Jappelli, and D. Terlizzese, 1996. Income Risk, Borrowing Constraints, and Portfolio Choice. American Economic Review 86 (1), 158 172. Hamilton, J.D., 1983, Oil and the macroeconomy since World War II, Journal of Political Economy 91, 228 248. Hamilton, J.D., 1985, Historical causes of postwar oil shocks and recessions, Energy Journal 6, 97 116. 29
Hamilton, J.D., 2009. Oil and the macroeconomy. In: Durlauf, S. and L. Blume (Ed.), New Palgrave Dictionary of Economics, Second Edition. Hortaçsu, A., and C. Syverson, 2004. Product differentiation, search costs, and competition in the mutual fund industry: A case study of S&P 500 index funds, Quarterly Journal of Economics 119, 403 456. Jensen, M., 1968. The performance of mutual funds in the period 1945 1964, Journal of Finance 23, 389 416. Kacperczyk, M., and A. Seru, 2007. Fund manager use of public information: New evidence on managerial skills, Journal of Finance 62, 485-528. Kacperczyk, M., C. Sialm, and L. Zheng, 2005. On the industry concentration of actively managed equity mutual funds, Journal of Finance 60, 1983 2011. Keswani, A., and D. Stolin, 2008. Which money is smart? Mutual fund buys and sells of individual and institutional investors, Journal of Finance 63, 85 118. Malkiel, B., 1995. Returns from investing in equity mutual funds 1971 to 1991, Journal of Finance 50, 549 572. Moskowitz, T., 2000. Mutual fund performance: an empirical decomposition into stock-picking talent, style, transaction costs, and expenses: discussion, Journal of Finance 55, 1695 1703. Pástor, L., and R.F. Stambaugh, 2003. Liquidity risk and expected stock returns, Journal of Political Economy 111, 642 685. Petersen, M.A., 2009. Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches, Review of Financial Studies 22, 435 480. Sapp, T., and A. Tiwari, 2004. Does stock return momentum explain the smart money effect? Journal of Finance 59 (6), 2605 2622. Sialm, C., L. Starks, and H. Zhang, 2014. Defined contribution pension plans: Sticky or discerning money?, Journal of Finance, forthcoming. Sirri, E., and P. Tufano, 1998. Costly search and mutual fund flows, Journal of Finance 53, 1589 1622. Stoughton, N.M., Y. Wu, and J. Zechner, 2011. Intermediated investment management, Journal of Finance 66, 947 980. Thompson, S.B., 2011, Simple formulas for standard errors that cluster by both firm and time, Journal of Financial Economics 99, 1 10. Tufano, P., and M. Sevick, 1999, Board structure and fee-setting in the U.S. mutual fund industry, Journal of Financial Economics 99, 321 355. Wermers, R., 2000, Mutual fund performance: An empirical decomposition into stock-picking talent, style, transactions costs, and expenses, Journal of Finance 55, 1655 1695. Zheng, L., 1999. Is money smart? A study of mutual fund investors fund selection ability, Journal of Finance 54, 901 933. 30
Figure 1 Distribution of U.S. households and households owning mutual funds by income The chart reports the percentage distribution of mutual fund-owning U.S. households by before-tax household income ranges in 2016 (Source: Figure 7.1, Investment Company Fact Book, February 2017). ΔDSPI Figure 2 Histogram of Changes in Disposable Income The graph plots the histogram of the rolling six-month cumulative growth rate (ΔDSPI) of the seasonallyadjusted monthly disposable personal income provided by the U.S. Bureau of Economic Analysis over 1993 to 2017Q1. 31
Figure 3 Total Net Assets and Flows of Passive and Active U.S. Equity Funds (in $ billions) The graph plots the annual average of (1) the sum of monthly total net assets in USD billions over 1993 to 2017Q1 for retail direct-market passively managed (Passive TNA) and actively managed (Active TNA) U.S. equity funds provided by Morningstar, Inc. (2) the relative net cash flows to actively managed versus passively managed funds (Flow_ActivePassive), computed as the difference in monthly estimated dollar net flows between actively managed and passively managed funds, divided by the total net assets in actively managed and passively managed funds in the prior month, expressed in %. Figure 4 Flows to Passive and Active U.S. Equity Funds (in %) The graph plots the flows measured as the annual average of the monthly new money growth (in %) over 1993 to 2017Q1 for retail direct-market passively managed (Passive Flow), actively managed (Active Flow), and exchange-traded ( ETF Flow ) U.S. equity funds provided by Morningstar, Inc. The flows to passively managed funds include flows to passive mutual funds and ETFs. 32