The timing ability of the U.S.
mutual fund investors
Master Thesis Finance
Panagiotis Michas (s3299600)
Supervisor: Dr Egle Karmaziene
June 2018
Abstract
1. Introduction
The importance of accessing the timing ability of investors arise from the size of the U.S. mutual fund industry which is the largest in the world with $ 18,7 trillion in total net assets (and an estimate of 100 million individual investors), as of the end of 2017. In particular, 85% of the industry funds’ assets were held in long-term mutual funds, which are the equity, the bond and the hybrid mutual funds, while the rest 15% were held in short-term mutual funds which are the money market funds. Another critical reason to justify this study is the main participants of mutual funds, which are the retail investors (e.g. households). In particular, 95% of the investors that held long-term funds during 2017 were retail investors, who use the mutual funds for retirement goals, education and home ownership. On the other hand, institutional investors held 5% of the long-term funds and 38% of the money market funds, which they use to manage their cash balances1.
In the literature, there are two documented ways that the mutual fund investors can enhance their returns, by selecting superior funds (Gruber, 1996; Zheng, 1999) or by advantageously timing their cash flows into the fund. In this paper, I focus on the second method, and I ask whether mutual fund investors benefit from varying their capital exposure over the investment period. As Dichev (2007) points out, the fact that investors vary their capital exposure to the equity market during the investment period, influences their realised returns. This effect can lead to superior investor returns if they experience inflows to the market prior to high equity returns, or to inferior investor returns if they experience inflows to the market prior to low equity returns. The same relationship holds for the investors of mutual funds, meaning that the net cash flows of investors affect their returns (Nesbit, 1995; Braverman et al., 2005; Friesen and Sapp, 2007; Hsu et al., 2016). However, the effect of varying the capital exposure during the investment period is not captured by the return that the mutual fund managers' report. Because, the past performance of mutual fund managers' is measured as the geometric average return, which ignores the month-to-month variation of assets under management, by applying equal weights to all past returns. Besides, this is the return that an investor earns if he invests in a fund and does not vary his capital exposure during the investment period, also known as a buy-and-hold strategy. In contrast, the investors of mutual funds vary their capital exposure into the fund during the investment's life; thus a better proxy to capture their return experience is that of dollar-weighted return, which is a well-accepted model in the literature. The dollar-weighted return is obtained as the internal rate of return (IRR) from an investment which takes into account the magnitude of investors net cash flows into or out of the fund, as well as their timing, by allowing the cash
flows to influence the weights of the past returns. Thus, in this paper, I use the geometric average to measure the return of a buy-and-hold strategy, the dollar-weighted return to measure the investors' actual realised return, and their difference labelled as the performance gap (Friesen and Sapp, 2007) to measure the timing skills of the investors.
As Dichev (pp. 389, 2007) points out "the advantage of aggregate specifications is that they offer a high level of generality and allow a comprehensive investigation of dollar-weighted returns". Hence, I use aggregate categories of mutual funds to investigate the timing skills of investors since they offer a good overview of the whole U.S. mutual fund industry. In particular, I use two long-term macro categories, which are the domestic equity and the bonds mutual funds, and one short-term category, which is the money market funds. Also, I analyse two subcategories of domestic equity mutual funds, that of growth and value-objective. By analysing these categories, I am able to test the timing ability of investors and also to see whether it differs among them.
and the current period. This suggests that when the return of the previous and current period is high, the underperformance will be minimal. On the other hand, when after high previous period return, a low current period return will follow, the magnitude of underperformance will be larger.
My study adds to the growing literature of the performance of mutual funds investors. I contribute to the existing literature by evaluating the documented timing ability of U.S. investors in more recent years and by finding that the underperformance of the investors is sensitive to the investment horizon considered, which differs for each one of them. Overall, the results suggest that the investors should not try to time the market but follow a buy-and-hold strategy.
The rest of the paper is organised as follows. Section 2 presents a literature review of the papers that I use as a guide for my study and the hypotheses. Section 3 presents the empirical methodology. Section 4 describes the data. Section 5 presents the empirical results and Section 6 concludes.
2. Literature review and hypotheses
There are numerous studies in the literature that have examined the timing ability of active mutual fund managers, but the same cannot be said about the timing ability of investors. Dichev’s study (2007) is the first to differentiate the equity market returns and the returns of the investors in these stocks. He argues that the reported returnsby the media are different from those which investors actually earn. The reason for this discrepancy is that the reported equity return, is measured by the geometric average return and it represents a strategy consisting of investing in a security and holding it throughout its life without varying the capital exposure2, also known as a buy-and-hold strategy. The author points out that the
geometric average is a poor proxy for the actual return obtained by the average investor because it does not take into account the effect of varying capital exposure. To tackle this problem, he introduces another measure, that of dollar-weighted return. Essentially, it is computed as the internal rate of return (IRR) from an investment, and it takes into account the magnitude of investors cash flows, as well as their timing. The reason that these two measures differ is that in the computation of the geometric average return all the returns are considered to have equal weights, while in the computation of dollar-weighted returns, the cash flows (variation of capital exposure) of investors during the investment period influence the weights of the returns. After the computation of these two measures, he subtracts the
2 An investor can vary his capital exposure by buying or selling shares. Buying shares is a cash inflow into the
dollar-weighted from the geometric average return. In this framework, if the difference is positive, it represents poor timing skills for the average equity investor and/or underperformance of a buy-and-hold strategy, while if it is negative, it represents superior timing skills and/or outperformance of a buy-and-hold strategy. The author hypothesises that this difference between the geometric average and the dollar-weighted return exists if there is a correlation between the time that cash flows occur and the past or future returns. To examine that, the author uses the market capitalisation and the past returns of NYSE/AMEX and NASDAQ indices for the period 1926 to 2002 and 1973 to 2002, respectively. The author finds a gap of 1,3% for NYSE/AMEX and a 5,3% for NASDAQ, which indicates an underperformance of the average investor. As Dichev (2007) points out, this gap exists because investors on average experience cash inflows into the equity market after superior past returns and before low future returns, whereas the opposite is true for cash outflows.
At the same time, another equally important contribution to the literature of investor’s timing ability comes from Friesen and Sapp (2007), which uses the same intuition as Dichev (2007), but in this case the authors focus on the experienced returns of mutual fund investors. In particular, they are the first to investigate the timing ability of mutual fund investors at the individual fund level. They use a sample of U.S. mutual funds for the period 1991 to 2004, and they pose the question: “Do equity fund investors put cash in and take cash out at the right time on average?” (Friesen and Sapp, 2007, p. 2797). To reach this conclusion, the authors compute the geometric average, the dollar-weighted return and also the difference between those two, labelled as the performance gap, for each fund in the sample. The interpretation of the performance gap is identical to Dichev’s (2007), where a positive value indicates poor investor's timing ability, while a negative value the opposite. They find that on average the equity fund, the bond fund and money market fund investors earn on average 1,56%, 0,02% and 0,004% less annually than the average fund, as a result of their timing decisions. Their results indicate that the underperformance due to poor timing is a characteristic of equity mutual funds and that it is higher in the funds with relatively large risk-adjusted returns. The explanation that the authors provide for their results is related to the existence of return-chasing behaviour on behalf of mutual fund investors. Furthermore, two additional papers, that of Dichev and Yu (2011) and Chieh-Tse Hou (2012), follow this methodological approach and conclude to similar results. In the first paper the authors investigate the timing ability of hedge fund investors, and in the second one, they examine the timing skills of equity investors in the Taiwanese market.
both value and growth-objective funds underperform the geometric average return of the funds and that growth investors experience a higher underperformance compared to value investors. These results indicate some categories of investors are more prone to have a larger performance gap, which implies that certain categories of investors time their cash flows more poorly. The authors conclude that investors who time poorly tend to be less sophisticated, as evidence of their investment choices in growth-objective, large-cap and high-fee mutual funds.
Although the dollar-weighted return is a well-accepted framework in the literature to calculate the actual return that an investor earned, it should be handled with caution. As Keswani and Stolin (2008) point out the dollar-weighted return, and hence the performance gap is very sensitive to the investment horizon considered. In particular, they follow the same methodology approach with the same sample as Dichev (2007), but they subdivide the sample period in three equal time-periods and recalculate the performance gap. Their results indicate that the performance gap is a fraction of what Dichev reports if sub-periods are considered. Thus, the magnitude of underperformance depends also on the investment horizon of each equity investor, which differs among them.
The above literature review will be used as a guide to access the timing ability of mutual fund investors and to test the following hypotheses:
H1: The mutual fund investors underperform a buy-and-hold strategy on average due to their erroneous timing decisions.
H2: Investors that invest in different categories of mutual funds, experience a different magnitude of underperformance.
H3: The timing ability of mutual fund investors depends on the investment horizon considered.
3. Methodology
The required dataset for each aggregate category of mutual funds that I will examine consists of the monthly total net assets, the monthly realised past returns and the monthly net cash flows. Bellow, I present the methodology.
The framework in equation (1) for estimating the net cash flows for a fund or a stock is a well-accepted model in the literature. Dichev (2007) and Friesen and Sapp (2007) among others use it. Furthermore, Ber and Ruenzi (2006) investigate the appropriateness of using net cash flows instead of actual inflows and outflows and find that the net cash flows are a suitable and unbiased measure.
Here, the NCFt stands for the monthly net cash flow of the aggregate mutual fund, TNAt
denotes the monthly total net assets of the aggregate mutual fund, and rt is the realised return
for that month. Since the dataset does not include the monthly realised returns for the aggregate mutual fund, but reports the monthly net cash flows, I obtain the returns by reverse engineering equation (1) into equation (2) below.
𝑟𝑡= 𝑇𝑁𝐴𝑡−𝑁𝐶𝐹𝑡
𝑇𝑁𝐴𝑡−1 − 1 (2)
After the collection of the monthly realised returns for the whole sample period, all the required information to test for the hypotheses are available. Hence, as the literature proposes it, I will compute the difference between the geometric average return (Equation 3) and the dollar-weighted return (Equation 4). The geometric average return is the appropriate return to measure the past performance of fund managers as it is a buy-and-hold strategy, and it measures the return on a dollar invested during the entire sample period while disregarding the impact of cash flows to the fund. Thus, the geometric return is the return that an investor would earn, if he would invest in a fund at the beginning of the sample period, without varying his capital exposure till the end of the sample period.
𝑟𝑔 = [ ∏𝑇𝑡=1(1 + 𝑟𝑡) ] 1𝑇− 1 (3)
Whereas, the dollar-weighted return is the appropriate measure to calculate the actual realised return by investors since it takes into account the magnitude of their monthly net cash flows as well as the timing of these flows. Therefore, to compute the actual realised return of investors, I will use equation (4) of Friesen and Sapp (2007). In this framework, the dollar-weighted return is defined as the return that will equalise the accumulated beginning value of total net assets plus the accumulated value of net cash flows with the ending value of total net assets.
𝑟𝑑𝑤 ∶ 𝑇𝑁𝐴0(1 + 𝑟𝑑𝑤)𝑇+ ∑𝑇𝑡=1𝑁𝐶𝐹𝑡(1 + 𝑟𝑑𝑤)(𝑇−𝑡)= 𝑇𝑁𝐴𝑇 (4)
Then, the measure for investor’s timing ability is the difference between the geometric average return in equation (1) and the dollar-weighted return in equation (2), labelled as the performance gap (Equation 5). Positive values of the performance gap indicate poor timing skills and/or underperformance of a buy-and-hold strategy incurred by the investors, while negative values superior timing skills and/or outperformance of a buy-and-hold strategy. Additionally, a more substantial absolute value of the performance gap indicates a higher magnitude of timing underperformance or outperformance.
4. Data description and statistics
In order to evaluate whether the investors of mutual funds underperform a buy-and-hold strategy as a result of their timing decisions, I gather time-series data from Thomson Reuters Datastream from January 2001 to December 2017. Specifically, the Investment Company Institute (ICI) conducts every month a comprehensive survey of the U.S. mutual fund industry to collect the actual dollar value of total net assets (TNA) and compile the dollar value of net new cash flow (NCF) of every mutual fund. Then it aggregates the values of total net assets and net cash flows of each mutual fund based on its investment style and reports these aggregates on the Datastream database. The components of net cash flows are the net new sales combined with the net exchanges. The net new sales are the dollar value of purchases of mutual fund shares by investors minus the dollar value returned to investors that sold shares, and it shows the money that were added to and removed from mutual funds. The net exchanges show the dollar amount of shareholder switches into or out of funds in the same fund family. A positive number indicates exchanges into funds are greater than exchanges out of funds, while a negative number the reverse. In short, a positive value of NCF is an inflow to the aggregate mutual fund or a contribution from investors, while a negative value is an outflow from the aggregate mutual fund or a distribution to investors.
The ICI reports these monthly data in various categories of aggregation, from level 1 which is the broadest category to level 5 which is the most detailed. The level 5 contains forty-two investment objective aggregates of mutual funds such as growth, value and blend investment style3. The time-series for every mutual fund category includes actively managed
and index funds while it excludes mutual funds that invest primarily in other mutual funds. The inclusion of index funds does not bias the results since the scope of this paper is to see whether investors of mutual funds underperform a buy-and-hold strategy on average, as a result of their timing decisions, independently from their investment vehicle choice. The mutual fund categories4 which I use in this paper are the following:
• Domestic Equity (level 2)5, which contains all the mutual funds that invest in
stocks of companies that are listed in the U.S. and two subcategories of it;
• Growth objective (level 5), which includes the funds that invest primarily in common stock of companies with above-average growth;
3 https://www.ici.org/research/stats/iob_update/iob_definitions
4 Source of definitions: https://www.ici.org/research/stats/iob_update/iob_definitions
5 The, reported by ICI, subcategories of domestic equity mutual funds are growth, sector, alternative strategy,
• Value objective (level 5), which consist of funds that invest primarily in common stock of value companies. These are the companies that usually are out of favour with investors, appear under-priced by the market, and/or have high dividend yields. • Bond funds (level 2), which tend to be less volatile than equity funds and to produce regular income;
• And lastly, e) Market funds (level 2), which contains the funds that invest in a pool of short-term, interest-bearing securities.
As Dichev (pp. 389, 2007) points out "the advantage of aggregate specifications is that they offer a high level of generality and allow a comprehensive investigation of dollar-weighted returns". Thus, I choose the above aggregate categories because they offer a good overview of the whole U.S. mutual fund industry. As of the end of 2017, the U.S. mutual fund industry had $18,746 billion in total net assets, of which $7,481 billion (40% of the whole industry) were in domestic equity mutual funds; $ 4,067 billion (22% of the whole industry) were in bond mutual funds; and $ 2,847 billion (15% of the whole industry) were in money market funds. In addition, I choose the growth and value-objective mutual funds as they are the biggest and the most well-known subcategories of domestic equity funds with $ 1,647 billion (22% of the domestic equity) and $ 1,424 billion (19% of the domestic equity) in total net assets, respectively6.
Table 1 below presents the descriptive statistics for the categories of mutual funds that I examine. As it is expected, the realised average monthly return of domestic equity mutual funds is almost two times the bonds' return and close to five times the return of money market funds. In addition, the standard deviation is much higher for domestic equity mutual funds than for bond and for money market funds, which is also expected; modern portfolio theory, pioneered by Harry Markowitz (1952) assumes that investors are risk-averse, implying that for a higher level of risk, they require a higher reward. Furthermore, I observe that in the selected sample period (2001 to 2017) the value-objective mutual funds provided higher returns while simultaneously had also a lower risk level than growth-objective mutual funds. This result is expected since value stocks perform better than growth stocks during recession periods, and during the sample period there were two recession periods; Early 2000s, March 2001 till November 2001 and the Great Recession, December 2007 till July 20097. The most significant
variation of monthly total net assets is present at the bond mutual funds, which from a minimum of $823,991 million grew at a maximum of $4,067,308 million, which is almost 5
6 https://www.ici.org/research/stats/factbook
Table 1. Descriptive Statistics Domestic Equity funds Growth-objective funds Value-objective funds Bond funds Money Market funds Average number of mutual funds 3,457 1,001 735 2,043 735 Arithmetic monthly return (%) 0.58 0.55 0.63 0.34 0.11 Median monthly return (%) 1.13 1.07 1.16 0.41 0.06 Standard deviation of returns (%) 4.29 4.66 4.25 1.00 0.16 Average total net assets ($
millions)
4,311,153 1,068,959 901,378 2,263,417 2,567,676 Max total net assets ($ millions) 7,481,781 1,652,547 1,425,536 4,067,308 3,894,863 Min total net assets ($ millions) 2,133,284 576,329 400,324 823,991 1,830,082 Average net cash flows/total net
assets (%)
0.28 0.41 0.47 0.65 1.41
Average net cash flows ($ millions) -5,080 -3,229 -648 8,900 2,151 Net cash outflow max ($ millions) 48,791 17,466 15,426 59,257 155,927 Net cash outflow min ($ millions) 163 340 77 154 842 Net cash inflow max ($ millions) 30,255 9,733 11,275 47,252 159,339 Net cash inflow min ($ millions) 262 112 56 139 198 Number of months with positive
cash flows
74 44 90 160 94
Number of months with negative cash flows
130 160 114 44 110
Number of observations 204 204 204 204 204
Notes: The table presents summary statistics, of five different datasets of aggregates of mutual funds obtained from Thomson Reuters Datastream and created by the Investment Company Institute (ICI). The sample period includes 204 monthly observations from January 2001 till December 2017. The aggregates of mutual funds include actively managed funds as well as index funds, while exclude mutual funds that invest primarily in other mutual funds. The average number of funds is the average number of funds that are included in the aggregate mutual fund. The arithmetic monthly return is obtained by 𝑟𝑡 = [(𝑇𝑁𝐴𝑡− 𝑁𝐶𝐹𝑡)/𝑇𝑁𝐴𝑡−1− 1], where TNAt is the monthly total net assets, and NCFt is the monthly net cash flow. The median and standard deviation are obtained from the return time-series of each fund category. The average, maximum and minimum total net assets are calculated from the total net assets time-series. The average net cash flows, maximum and minimum of net cash inflows/outflows, and the number of months with positive and negative cash flows are obtained from the net cash flow time-series. Positive net cash flows indicate higher inflows than outflows, thus money poured into the mutual fund by investors, while negative the reverse. The negative dollar values of net cash outflows are presented on absolute values. The average of net cash flows over total net assets is calculated on absolute value.
monthly total assets under management for every aggregate category. In addition, it is worth to notice that the average net cash flows of domestic equity, growth and value-objective mutual funds are negative, meaning that on average the outflows in these funds were higher than the inflows, while for bond and money market mutual funds the reverse is true. Additionally, the highest monthly cash inflows and/or outflows took place at the money market mutual funds, which match with their short-term investment style. Finally, all categories except the bond mutual funds experienced more months of negative net cash flows. Bonds had only 44 out of 204 months of negative net cash flows, which implies that on 78% of the selected period the contributions to the funds by investors were higher than the distributions from the funds to investors.
5. Empirical results
5.1 The performance gap
Table 2. The Performance Gap
Domestic Equity funds Growth-objective funds Value-objective funds Bond funds Money Market funds Arithmetic return (%) 6.92 6.63 7.56 4.12 1.37
Geometric average return (%) 5.80 5.30 6.46 4.06 1.37 Dollar-weighted return (%) 5.50 4.23 6.22 3.87 1.34
Performance Gap (%) 0.31 1.07 0.23 0.19 0.03
Notes: The returns and the performance gap are reported in annual percentages. For each aggregate category of mutual funds, the arithmetic return, the geometric average and the dollar-weighted return is calculated over the entire sample period. The performance gap is the difference between the geometric average and the dollar-weighted return. Positive values of the performance gap indicate poor timing skills and/or an underperformance of a buy-and-hold strategy incurred by the investors, while negative values the reverse.
hypothesis and the literature. In particular, growth-objective investors experience the most considerable gap of 1.07% annually, while the value investors underperform the average value fund by 0.23% annually. Hsu et al. (2016) also find that investors who choose growth-oriented mutual funds time their cash flows more poorly than the ones that invest in value-oriented mutual funds, and attribute it to a potentially lower level of sophistication of growth investors. This could also be explained by the fact that value stocks are usually bought for their high dividend yields, thus the investors who own shares of a value mutual fund vary their invested capital exposure less frequently.
In the bond mutual funds, the average investor earns 0.19% less annually, while for the money market funds the gap is only 0.03% annually. These results indicate that the underperformance of the investors caused by their timing decisions is mostly linked with the long-term investment funds, which are the domestic equity, including its subcategories growth and value-orientation, and the bond funds. As I mentioned in the introduction, the main players of the long-term investment funds are retail investors, while institutional investors own a big part of the short-term funds. Thus, the existence of underperformance for the long-term funds is connected to a lower level of sophistication of retail compared to institutional investors. However, this result is contrary to Friesen and Sapp (2007), who find that the performance gap is a characteristic of only equity mutual fund investors. Furthermore, I observe that there is a positive correlation between the variation of returns of the aggregate category and the underperformance of investors. In particular, the categories that include riskier securities exhibit a higher performance gap. The growth-objective funds that had the highest variation of returns (Table 1), also experienced the highest gap, while the money market funds that had a minimal variation (Table 1), experienced an almost flat gap.
5.1 The performance gap in sub-periods
It is interesting to note that investors of bond mutual funds experience the second highest gap if annual sub-periods are considered, instead of fourth at the whole sample period gap.
However, when I move from annual periods (Table 3, panel A) to quarter sub-periods (Table 3, panel B), the gap does not reduce as much as from whole sample period (Table 2) to annual sub-periods. Instead, it is stable for the domestic equity mutual funds at 0.03% annually, it is increasing for the value-oriented funds at 0.04% annually, and it keeps decreasing for the rest fund categories. This enhancement of investors’ timing skills, when sub-periods are considered, could be attributed to some sophisticated (institutional) participants that gain exposure to the funds for shorter terms in better periods, meaning that they can predict the market movements and invest before higher future returns.
Table 3. The performance gap by sub-periods Domestic Equity funds Growth-objective funds Value-objective funds Bond funds Money Market funds Panel A: annual sub-periods
Arithmetic return (%) 6.92 6.63 7.56 4.12 1.37
Geometric Average return (%) 5.96 5.50 6.60 4.07 1.37 Dollar-weighted return (%) 5.93 5.40 6.58 4.00 1.36
Performance Gap (%) 0.03 0.10 0.02 0.06 0.01
Panel B: quarter sub-periods
Arithmetic return (%) 6.92 6.63 7.56 4.12 1.37
Geometric Average return (%) 6.29 5.89 6.93 4.08 1.37 Dollar-weighted return (%) 6.26 5.85 6.89 4.05 1.37
Performance Gap (%) 0.03 0.04 0.04 0.03 0.00
Notes: The returns and the performance gap are reported in annual percentages. In panel A, I subdivide the sample period in calendar years and calculate the annual geometric average dollar-weighted return for each aggregate category of mutual funds, and present their average multiplied by twelve. In panel B, I subdivide the sample period in quarters and calculate the quarter geometric average and dollar-weighted return for each aggregate category of mutual funds, and present their average multiplied by twelve. The performance gap is the difference between the geometric average and the dollar-weighted return. Positive values of the performance gap indicate poor timing skills and/or an underperformance of a buy-and-hold strategy incurred by the investors, while negative values the reverse.
5.2 Determinants of the performance gap
average size of funds. The money market funds category is excluded since the timing underperformance of the investors in this category is zero. I estimate the parameters of the following time-series model:
𝑃. 𝐺.𝑡= 𝑎 + 𝛽1𝑅𝑡+ 𝛽2𝑅𝑡−1+ 𝛽3𝑁𝐶𝐹𝑡
𝑇𝑁𝐴𝑡+ 𝛽4 𝑇𝑁𝐴𝑡+ 𝜀𝑡 (6)
Where P.G.t is the performance gap of the quarter sub-period; Rt is the return of the current
quarter sub-period; Rt-1 is the return of the previous quarter sub-period; NCFt/TNAt is the
average net cash flows of the period as a percentage of total net assets; and TNAt is the natural
logarithm of the average total net assets of the period or else the size of the aggregate mutual fund.
Table 4. Determinants of the performance gap Domestic Equity funds Growth-objective funds Value-objective funds Bond funds Intercept 1.31** 2.04** 1.85** 0.47** (0.61) (0.81) (0.72) (0.22) Arithmetic Return -0.85* -1.19** -1.33** -1.95 (0.46) (0.50) (0.53) (1.57)
Arithmetic Return (previous
period) -0.68* -1.04** -0.26 -2.24
(0.41) (0.44) (0.54) (1.37)
Average net cash flows (% of
total net assets) 0.67 6.13 -1.13 -1.76
(4.57) (4.71) (3.52) (1.56)
Average total net assets
(natural logarithm) -0.08** -0.14** -0.13** -0.03**
(0.04) (0.06) (0.05) (0.02)
Adj. R2 0.18 0.23 0.20 0.16
Observations 68 68 68 68
Notes: For each aggregate category of mutual funds I calculate the quarter geometric average and the dollar-weighted return and label their difference as the performance gap. The performance gap is the dependent variable in a linear regression on the funds' characteristics listed in the first column of the table. All the coefficients and the standard errors of the independent variables are multiplied by x1,000. Standard errors are reported in parenthesis, and ***,** and * indicate statistical significance at the 1%,5% and 10% levels, respectively.
listed in the first column of the table. In addition, the first row of the table lists the aggregates categories that I am evaluating. By examining the domestic equity mutual funds and its subcategories, I observe that the return of the current period is a significant predictor of the performance gap. In addition, the return of the previous period is also significant in every category, except the value-oriented mutual funds. The correlation between the returns of the current and/or previous period, and the current period performance gap is negative, which indicates that high returns in the previous and current period will result to a low timing underperformance in the current period. While if there are high returns in the previous period, and low returns in the current period, the effect on the current period gap is opposite. Nevertheless, the second case leads to a higher timing underperformance, which is in terms with the literature. The non-significance of the current and previous return for the bond mutual funds can be explained by the fact that these securities are quite different than equities and thus the investors of these funds do not react to the reported return. In addition, bond securities face different types of risks such as interest rate, credit and prepayment risk. Furthermore, the average net cash flows (as a percent of total net assets) is not a significant predictor of the timing performance in any aggregate category of mutual funds. As I observe from Table 1, the average net cash flows are less than 0.65% of the total net assets for every aggregate category. This result suggests that when the magnitude of net cash flows is small related to the fund size (TNA), the overall rate of non-investment growth is irrelevant to investor timing performance. The last independent variable in Table 4 is the fund size, which I measure it as the natural logarithm of the average total net assets, to scale down its immense value and eliminate its high variation. I observe that it is a significant predictor of the performance gap for every long-term aggregate category of mutual funds and in particular that the timing underperformance of investors decreases with the fund size. This result indicates that the investors who choose bigger funds on average are more likely to exhibit a lower magnitude of underperformance.
However, it is worth to notice that all the coefficients in the table have a marginal effect on the performance gap since they are already multiplied by x1,000. In domestic equity funds, an increase of 1% in the return of the previous period leads to a 0.68% (divided by 1,000) decrease in the gap; an increase of 1% in the return of the current period leads to a 0.85% (divided by 1,000) decrease in the gap; and a 1% increase in the fund size leads to a 0.0008%8 (divided by 1,000) decrease in the gap. The rest categories of mutual funds exhibit
similar results. Hence, in real-world terms the effect of the previous and current period return on the performance gap is small but still significant. On the contrary, the effect of the fund size on the gap is rather insignificant.
6. Conclusion
My study examines the timing ability of mutual fund investors using cash flow data at the aggregate level. For every category of mutual funds, I calculate the geometric average and the dollar-weighted return over the period 2001-2017 and find that the investors of domestic equity, growth-objective, value-objective, bond and money market funds underperform a buy-and-hold strategy by 0.31%, 1.07%, 0.23%, 0.19%, 0.03% annually, respectively. The performance gap is more pronounced in the long-term mutual funds, which are the domestic equity, its subcategories and the bond mutual funds. This result is contrary to Friesen and Sapp (2007) who find that it is a characteristic of domestic equity investors only, but it is in line with the fact that the main players of long-term mutual funds are retail investors, while institutional investors hold a significant portion of short-term funds. Thus, the timing underperformance is a characteristic of retail investors and can be attributed to a lower level of sophistication of the retail compared to institutional investors. Furthermore, growth investors in comparison with value investors time their cash flows more poorly, which is in line with Hsu et al. (2016) regarding that some categories of investors time their cash flows more poorly, because of a lower level of sophistication and/or less frequent trades. Moreover, the timing underperformance is higher for the aggregates of funds that include securities with higher volatility, which is also in line with the literature.
By subdividing the sample period in annual and quarter sub-periods the performance gap becomes a fraction of the gap that I present above, but it is still present. This result is in line with Keswani and Stolin (2008), and it suggests that the performance gap is sensitive to the investment horizon considered, which is different for every investor. Therefore, the magnitude of underperformance I present above is related to an average investor for the period of 2001 to 2017. Even though investors underperform the funds due to their timing decisions in general, the magnitude of this underperformance depends on the investment horizon of each investor. Furthermore, by conducting a linear regression of the determinants of the performance gap, I find that there is a negative correlation between current and previous period returns and the performance gap, thus high previous and current returns suggest a small performance gap. On the other hand, if there are high returns on the previous period and low returns on the current period, the performance gap will rise. This is in line with the literature, which expects that the underperformance will exist if after high returns and inflows to the fund, low returns will follow.
strategy in 19 major international stock markets. Furthermore, the time-series dataset that I use does not allow me to test for the significance of the performance gap that I report in Table 2 and Table 3. One way to do so is by bootstrapping, meaning that the dollar-weighted return is calculated 1,000 times with randomised net cash flows in each case.
References
Ber, S., Ruenzi, S., 2006. On the usability of synthetic measures of mutual fund net-flows, No 06-05, CFR Working Papers, University of Cologne, Centre for Financial Research (CFR).
Braverman, O., Kandel, S., Wohl, A., 2005. The (bad?) timing of mutual fund investors. Working Paper.
Chieh-Tse Hou T., 2012. Return persistence and investment timing decisions in taiwanese domestic equity mutual funds. Managerial Finance 38(9):873–91.
Dichev, I., 2007. What are stock investors actual historical returns? evidence from dollar-weighted returns. American Economic Review (Evanston), Vol. 97, No. 01, P. 386-401.
Dichev, I., & Yu, G., 2011. Higher risk, lower returns: what hedge fund investors really earn. Journal of Financial Economics, 100(2), 248-263.
Friesen, G., Sapp, T., 2007. Mutual fund flows and investor returns: an empirical examination of fund investor timing ability. Journal of Banking and Finance, 31(9), 2796-2816.
Gruber, M., 1996. Another puzzle: the growth in actively managed mutual funds. Journal of Finance 51, 783–810.
Hsu, J., Myers, B., & Whitby, R., 2016. Timing poorly: a guide to generating poor returns while investing in successful strategies. Journal of Portfolio Management, 42(2), 90-98.
Keswani, A., Stolin, D., 2008. Dollar-weighted returns to stock investors: a new look at the evidence. Finance Research Letters, 5(4), 228-235.
Markowitz, H., 1952. Portfolio selection. The Journal of Finance, 7(1), 77-91.
Nesbitt, S., 1995. Buy high, sell low: timing errors in mutual fund allocations. Journal of Portfolio Management 22, 57–60.
