Short‐term investment indicator:
How short‐term oriented is your fund manager?
Mark T. Hömmen (1483919) Instructor: dr. Auke Plantinga University of Groningen Faculty of Economics February 26, 2009 Abstract To supplement return‐based mutual fund analysis, an alternative holdings‐based measure called the STI indicator is developed in this paper. It describes the amount of the mutual fund’s net asset value allocated to short‐term investments (STI) in relation to trading activity (turnover) for a given period of time. With the STI indicator, investors can match their preference for short‐term orientation and activeness with mutual funds. Behavioral finance concepts are linked with this new measure as well as Jensen’s alpha that appears to have a relation with the STI indicator. Using the STI indicator in conjunction with other ‘rating’ systems enables the investor to better understand mutual funds.
JEL classification: C51, G11
2
1
Introduction
Mutual fund performance measurement systems are a popular tool for private and institutional investors to screen investment opportunities among mutual funds and manage their investment portfolios. According to several methodologies developed by Morningstar, LipperLeaders, and ValueLine, certain rates are assigned to different kinds of mutual funds. All these methodologies are based on historical returns of mutual funds in the calculus of their rating. Additionally a benchmark is often used to measure ‘relative’ performance of the mutual fund. Historical returns have only very limited predictive power, as James Montier states in his book Behavioural Investing published in 2007: page 575‐576 “… we couldn’t find any evidence of predictive power from a wide variety of leading indicators.” The same is the case for equity returns, according to Montier (2007). Therefore, it is in my view incorrect just to rate mutual funds on the basis of historical return data. A further observation is that mutual fund portfolio holdings and turnover are not examined whereas these elements certainly have a considerable bearing on performance. Using historical mutual fund performance data involves the implicit assumption that the derived information has at least some predictive content for the future (Sharpe, 1998). According to Markowitz (1952) the way of constructing an investment portfolio takes into account the best possible estimates of relevant future risk and returns. In an attempt to use portfolio holdings instead of historical returns to ‘rate’ mutual funds, a two‐ dimensional approach called the Short‐Term Investment (STI) indicator is developed in this paper. The STI indicator describes the amount of the mutual fund’s net asset value allocated to short‐term investments in relation to the trading activity (turnover) for a given period of time. In addition, the STI indicator will give valuable insights in the behavioral component of asset management.
Nowadays, traditional mutual fund performance measurement systems are used extensively for different reasons. One reason is the perceived convenience in their use to select investment opportunities and their goodwill among many investors. As Damato (1996) showed, 90% of the money that has to be invested will be allocated to mutual funds with a four or five star rating of Morningstar. Financial advisors are influenced by their clientele to invest preferably in high rated mutual funds and as a mark of ‘good housekeeping seal of approval’; institutional investors of 401 (k) pension funds also invest in those high rated mutual funds (Franecki, 2000). The goodwill created by rating agencies is widely used in marketing campaigns of individual mutual funds; this enhances the reputation of, for example, the star system from Morningstar.
However, the popularity of the rating services has another aspect. Changes in ratings of mutual funds can have a severe impact on the fund’s performance. Morey (2005) discovered side‐effects, for example like subsequent to receiving the first five star rating the mutual fund management decides to take excess risk and use strategies that cause their performance to suffer. According to Del Guercio and Tkac (2002) significant cash outflow might be the result of a decrease in the rating provided by rating services.
performance of mutual funds (Reichenstein, 2004). In 2002 Morningstar Inc adjusted this system after some criticism and extended the categories of mutual funds, resulting in more objective ratings. Investors wish to use these ratings, because the amount of mutual funds available to them is too large to analyze quickly. However, after the first selection using mutual fund ratings, the investor should check the prospectuses of the funds to come to a sound decision instead of making a decision based on only a fund rating. Still, investors consider the ratings for the Holy Grail and that is not the right attitude towards those rating services. Warshawsky (2000) mentions that investors tend to use these systems as vision for the future, because investors are in search of systems that are able to predict future performance.
The focus on historical performance data may not help investors, as historical performance is no guarantee for future performance. Alternatively, it is probably better to study the allocation and turnover of mutual funds. With the STI indicator the investor uses portfolio holdings based information and trading activity of mutual funds. The STI indicator dimensions give an indication of how short‐term oriented and active a mutual fund management invests relative to a population of mutual funds. Then, the investor can select the fund that suits his preferences toward short‐term orientation and activeness according to the STI and turnover dimension. For example, a risk‐averse investor who wants a secure income in, say, five years would choose a mutual fund that is less short term oriented (relative small STI ratio) and with less active management (relative low turnover ratio).
After all, this STI indicator attempts to take into account the problems of using historical performance data as opposed by different authors about the Morningstar star, LipperLeaders, and ValueLine rating mechanisms. The STI indicator provides the investor holdings‐based information and matches the investor’s preference for active management with mutual funds, compared to methods relying on historical returns only. To validate the STI indicator, it will be linked with mutual fund portfolio alphas and betas and it is examined if the STI and turnover dimension are linearly related with each other.
4
2
Literature
The introduction provided a short overview on the current use and limitations of existing mutual fund performance measurement systems. This section gives an impression of literature on mutual fund performance measurement systems as well as literature about behavioral issues in managing a mutual fund. At the end of each subsection, the literature is summarized and linked to both dimensions of the STI indicator. 2.1 Financial elements related to the STI indicator Financial literature that is not focused on explaining behavioral elements in asset management offers a variety of explanations that affect both dimensions of the STI indicator. This subsection provides a point‐by‐point outline of financial elements affecting the STI indicator.
Wermers (2000) examined comprehensively the mutual fund industry, by examining mutual fund performance through skills in picking stocks, characteristics of stock holdings, costs of implementing the style of a manager, fund expenses and fees charged, and differences between gross stock portfolio returns and net fund returns. The costs of implementing the style of the manager can be significant, a study by Morey (2005) found that after receiving an increase in performance rating, managers adapt their strategies in an attempt to sustain those higher ratings. According to Morey 2005: “…investors should be very wary about using the 5‐star rating as a signal of future 3‐year performance.” Whereas, subsequent mutual fund performance seems to suffer from those changes in strategy. Wermers (2000) states that high turnover mutual funds, although they incur more transaction costs and charge higher fees, hold stock with much higher average returns compared to low turnover mutual fund. He concludes: “Our evidence supports the value of active mutual fund management.” High turnover means that the portfolio holdings do change often and mutual fund management should indeed hold stocks with high returns, otherwise the transaction cost will not be covered at all. The risk of such stocks can be significant, because their volatility is high.
Marcin Kacperczyk (2007) argues that professional managers do have access to superior information and that this is related to the skills of the manager. He developed a measure of reliance on public information (RPI) that “…clarifies whether traditional performance measures indeed reflect skill, in that managers who produce high values of these measures should also have low sensitivities of their portfolio holdings to changes in public information.” Of course, the manager can adopt a strategy that actively trades on information, which can be related to the overconfidence concept, on which will be elaborated further in the next subsection. According to Kacperczyk et al. (2005), mutual fund managers tend to concentrate their funds in industries where they assume they have informational advantages (familiarity). This study found that there is a relation between industry concentration and the performance of a mutual fund. He found that industry concentrated funds tend to overweigh growth and small stocks; whereas managers of more diversified funds hold portfolios that more closely resemble the total market portfolio.
Perceived informational advantage can be an argument, as proposed by Kacperczyk et al. (2005). They state that asymmetric information between local and non‐local investors drives the preference for geographically proximate investments. They divide the preference into two categories; one based on the distance from the home country and the other which relies on national or governmental frictions.
Chevalier and Ellison (1997) tested the relationship between the cash inflows and the performance of mutual funds, they “…examine portfolio holdings of mutual funds in September and December and show that mutual funds do alter the riskiness of their portfolios at the end of the year in a manner consistent with these incentives.” Investors want a secure return and the fund management would like as many inflows of cash as possible. This relation generates incentives for mutual funds to increase or decrease the risk of their portfolio.
Cohen et al. (2005) developed a performance evaluation approach in which the skill of a fund manager is judged by the extent to which his investment decisions resemble the decisions of managers with distinguished performance records. This proposed approach uses historical returns and the holdings of many funds. They state that their developed measure is useful in ranking mutual fund managers from poor to good ones and “…skilled managers tend to make similar investment decisions, because they interpret information similarly.”
Grinblatt, Titman, and Wermers (1995) analyzed the extent to which mutual funds purchase stocks based on their past return. They find that 77 percent of the mutual funds were "momentum investors," buying stocks that were past winners; however, most did not systematically sell past losers. On average, funds that invested on momentum realized significantly better performance than other funds.
The two dimensions, STI and turnover, of the STI indicator are related to the previous financial literature. A source that influences the STI ratio is the adjustment of the investment strategy by fund managers. This can be a result of, for example, a rise in performance rating or managers that have informational advantages in specific markets and tend to invest like momentum investors. Managers that trade due to change in strategies and potential superior information may affect both dimensions. Furthermore, persistence in mutual fund performance is awkwardly interpreted from historical performance data. Markov (1952) already suggested that only the present state of a variable is relevant for predicting the future and the way the present has emerged from the past is irrelevant.
2.2 Behavioral elements related to the STI indicator
Behavioral finance literature is focused on explaining financial decisions from behavioral elements in asset management. Many circumstances can influence the investment behavior of asset managers; therefore managers can deviate from their initial investment strategy. This subsection provides a point‐by‐point outline of behavioral finance elements affecting the STI indicator.
6 differential performance may simply reflect differences in transaction costs that are incurred.”
Overconfidence can be related to turnover, because if investors think they are above average they are likely to trade more actively (Glaser and Weber (2003). De Bondt and Thaler (1995) argue that overconfidence is the best behavioral indicator to reason about the trading puzzle.
The illusion of knowledge is a source of overconfident trading behavior, where investors believe that the more information they possess the more accurate their decisions become. Theoretically, the investor that uses his superior information should cover its transaction costs, so its active trading behavior is justified. However, according to Barber and Odean (2000): “Overconfident investors will overestimate the value of their private information, causing them to trade too actively and, consequently, to earn below‐ average results.” More obvious is the result of a study by Odean (1999): “those who trade the most lose most.” Gervais and Odean (2001) do summarize the threat within overconfidence in the following quote: “…overconfidence does not lead to greater profits, greater profits lead to overconfidence.”
The position of overconfidence has to be interpreted in a broader perspective; the market condition should also be considered. As can be concluded from the research by Gervais and Odean (2001); investors become overconfident about the precision of their information after favorable market conditions. In a market downturn investors tend to be less overconfident and downsize their trading activity, however this distinction between good and bad market times is not symmetrical. This difference explains some psychological elements in the behavior of people. If an investor in a good market discovers an investment opportunity, it is surrounded with other investments that increase in value. If this investor sees the same investment opportunity in a bad market he is more anxious about the return. Therefore, the uncertainty surrounding the precision of their knowledge about an investment opportunity experienced by a risk averse investor seems to be higher in bad markets. Before interpreting the active trading behavior (turnover) of an asset portfolio, it is necessary to be aware of the market conditions.
Statman et al. (2006) found that market turnover tends to increase after months with positive market returns and they establish the link with the disposition effect. This effect can be described as the desire by investors to realize gains by selling stocks that appreciated and delaying realizations of losses. Trading activity is dependent on past returns and explains the attitude of investors in realizing losses and gains according to the disposition effect.
Glaser and Weber (2004) consider the overconfidence in relation with trading activity, but in addition they consider another strand of literature: the differences‐of‐opinion concept. “Differences of opinion can arise due to differences in prior beliefs or due to differences in the way investors interpret public information.” This study provided a psychological foundation for the differences of opinion concept applied in examining trading activity and tries to explain observed trading activity.
market clearing condition requires that he and no other trader sees the profitable investment opportunity. In other words, if one investor has information that induces him to trade, other rational investors are unwilling to trade with him, as he might have superior information. Therefore, some participants in the financial markets clearly behave irrational. This has an empirical implication: the introduction of irrational investors will alter the behavior or rational market participants and therefore change the nature of equilibrium (Black, 1986). In line with this reasoning it is important for an investor to recognize the fraction of opinion and fraction of information from another investor’s belief to determine the reliability of the other investor’s trading behavior.
Differences in opinion are caused by differences in beliefs, risk attitudes, existing portfolio holdings, hedging activities, etc. To link the difference of opinion concept with turnover, the illusion of knowledge seems to be a critical part. Trading is generated first of all by differences in opinion about prior beliefs. Prior belief induces trading, because investors think they possess superior information and use this information to trade. Varian (1989) describes the relation between beliefs and asset values like: “An increase in the "spread" of the probability beliefs of investors may increase or decrease equilibrium asset values depending on the value of a parameter of the utility function.” The more diverse the differences in opinion, the more trade will be possible. Think of an equilibrium, without difference of opinion, no investor would have an incentive to trade, as the equilibrium is assumed to be a product of efficient market theory. Therefore, trade can only be introduced if investors have different prior beliefs based on their illusion of knowledge. This can explain the trading activity that is measured with the turnover ratio used in the STI indicator. For example, a mutual fund management with a particular strategy can see many profitable investment opportunities, based on this the management seems to differ in opinion with other established mutual fund providers and market participants.
8
3
Methodology and data description
This section provides a complete manual of how the STI indicator has been developed and can be used by individual investors or mutual fund companies. The essence of the STI indicator is to categorize mutual funds according to trading activity and allocation to short‐term investments. By using the STI indicator, an investor should be able to link his preference toward active investment management with the two dimensions of the STI indicator. Both STI dimensions are tested for the existence of a linear relationship and a statistical approach is suggested to test if the alphas and betas do have some explanatory power if they are used in conjunction with the STI indicator. Furthermore, summary statistics explain characteristics of the mutual fund population and tables are included that describe the distribution of the mutual funds over certain intervals of the STI and turnover.
To clarify the process of how the STI indicator is determined, an ongoing example throughout this section is developed: virtual mutual fund X. This section first describes the derivation of both STI indicator dimensions. Second, the necessary data is described and analyzed. Third, both dimensions are combined to form the STI indicator and finally, the STI indicator is applied in a general format on a virtual mutual fund.
3.1 STI indicator fundamentals
The next four subsections define active and passive management, the STI and turnover dimension, and the in‐sample alpha and beta, which serve as the building blocks of the STI indicator.
3.1.1 Active and passive management
According to Elton et al. (2007), active portfolio management is defined as: “taking a position (stock selection) different from that which would be held in a passively managed portfolio, based on a forecast about the future.”
Passive portfolio management is based on trading through mechanical rules by using past data, like replicating an index or form a portfolio of a specified number of stocks based on mathematical programming (Elton et al., 2007).
3.1.2 Shortterm investment (STI)
The STI ratio of a mutual fund is the fraction of the fund’s net asset value that is held less than one year in a portfolio with an investment horizon of at least three years. Typically, the STI is characterized by disposed investments that do not satisfy the goal of the investment manager’s strategy. It is possible to use more frequent portfolio holdings data, like quarterly holdings data, however, quarterly data has to be compared with quarterly data and not with yearly data and so on. The STI is combined with turnover, because the STI does not explain total activeness of mutual fund management, due to allocation in long term‐oriented positions. Section 3.3 will elaborate on the combination of both dimensions of the STI indicator.
proposed long‐term investment horizon. Likewise, from a STI ratio of 40% for a long term investing fund can be concluded that the fund manager does not follow its proposed strategy. Singleton (2005) describes that short‐term investments (also called satellites) include the riskier part of a portfolio. However, Singleton (2005) argues that professional investors should all use the core‐satellite portfolio management model. Hence, it is not necessary to exhibit a core‐satellite portfolio management structure for mutual funds to derive a STI ratio that just measures investments that last for less than one year in a portfolio.
The process of calculating the STI ratio of a mutual fund portfolio is straightforward. First, the mutual funds are selected that have portfolio data over three recent successive years. Second, in deriving the STI ratio, we want to filter out investments that are exclusively apparent in the mutual fund portfolio of year t‐1 (the middle year). Therefore, we need to compare the portfolio of year t‐1 with investments in portfolios of year t and t‐2. For example, this study uses the years 2005, 2006, and 2007; and the investments held exclusively in 2006 are called the STI component of a portfolio. The virtual example of mutual fund X and a mathematical explanation will clarify this.
Suppose fund X is a defensive long term oriented mutual fund, focused on investing worldwide. Table one represents the latest three years of portfolio holdings. Yearly portfolio holdings are represented by t‐2, t‐1, and t, respectively. The allocated funds to single securities in the mutual fund portfolio for each year are represented as percentage of the fund’s net asset value (NAV) by ! Ii,t, where i stands for an individual stock and t stands for a particular year (t, t‐1, or t‐2). For example, ! I2,t"1 represents stocks of ING relative to the fund’s NAV in year t‐1. For every year, subscript i stand for the same stock. Table 1: Portfolio holdings fund X for year t, t1, and t2, holdings of mutual fund X relative to NAV, represented by
!
Ii,t is value of stock i at time t relative to fund NAV, Portfolio t (portfolio holdings of most recent year), Portfolio t‐1 (one year old portfolio holdings), and Portfolio t‐2 (two year old portfolio holdings), used to derive the STI component of fund X in t‐1. The STI component is I10,t1 in the shaded cell.
Portfolio t Portfolio t1 Portfolio t2
! I1,t ! I1,t"1 ! I1,t"2 ! I2,t ! I2,t"1 ! I2,t"2 ! I3,t ! I3,t"1 ! I3,t"2 ! I4,t ! I4,t"1 ! I4,t"2 ! I5,t ! I10,t"1 ! I6,t"2
To derive the STI component from this portfolio, a comparison of the three portfolios must be made, respectively the portfolios in year t, t‐1, and t‐2. It is clearly visible from table one that investments I1, I2, I3, and I4 are investments
are at least three years in the portfolio. These investments are called core investments in the STI indicator methodology. The position of I6,t2 is initiated in
t‐2 or in previous years, however it is certainly disposed in t‐2, so I6,t2 is not
included in the STI ratio. I5,t is initiated in the current year t, so it is not possible
to judge the amount of time it will be included in the portfolio until it will be disposed. Therefore, I5,t is not included in the STI ratio. I10,t1 is an investment
10 the criteria of a STI investment within fund X. If the fund’s investments are not stated relative to the fund’s net asset value, the STI can be calculated according to the following equation, where Oi,t represents the amount of money allocated to
stock i at time t. ! STIx,t"1= O10,t"1 NAV (O1,t"1,O2,t"1,O3,t"1,O4,t"1,O10,t"1)= O10,t"1 O1,t"1+ O2,t"1+ O3,t"1+ O4,t"1+ O10,t"1. According to table one, a frank mathematical approach can be derived to calculate the STI ratio by using a dummy variable and sets of investment portfolios. Ii,t is, as stated earlier, the relative value with respect to the mutual
fund’s net asset value of stock i at time t. A convenient way of performing this filtering process is to compare sets of investments (the portfolios). The set of investments in t‐1 we call St‐1 and is compared with the comparison set
consisting of investments in t and t‐2. The comparison set is called Sc, where
subscript c refers to the comparison set, see equations under one. The dummy variable equals one for an individual investment if it appears only in St‐1 and not in Sc, so that investment satisfies the STI definition. The dummy variable is zero if a particular investment appears in St‐1 as well as in Sc. The dummy conditions are provided in equations two and three. From this, the last step in calculating the STI ratio is simply adding the investments that are held exclusively in year t‐ 1, like in equation four.
!
St"1= {I1,t"1,...,Ii,t"1} and
!
Sc = {I1,t,...,Ii,t},{I1,t"2,...,Ii,t"2} (1)
! di,t = 1 if ! St"1# Sc,$i = 1,2,...,n (2) ! di,t = 0 if ! St"1# Sc,$i= 1,2,...,n (3) !
STIt"1= Ii,t"1# di,t i=1
n
$
(4)
The core ratio is calculated according to the same principle as in equation four, however by adding positions that last for at least three years. Again, the portfolios of year t, t‐1, and t‐2 have to be compared. Yet, the dummy conditions of equation two and three have to be adjusted, instead of using a comparison set, we need a particular set for t and t‐2. The set for t and t‐2 will become St and St2, respectively. The dummy becomes one if an investment in t‐1 is also present in t and t‐2, like in equation five. With a dummy of zero, the particular investment is not present in both years t and t‐2, as in equation six. Though it is possible that an investment is present in one of the years t or t‐2, but than it does not satisfy the definition of the core component. With changed dummy conditions, the formula of equation five will transform into one used to calculate the core ratio in t‐1 by replacing STI with CORE as in equation seven. ! di,t = 1 if ! St"1# St$ St"2,%i= 1,2,...,n (5) ! di,t = 0 if ! St"1# St$ St"2,%i= 1,2,...,n (6) !
COREt"1= Ii,t"1# di,t
n=1 n
3.1.3 Turnover
The turnover ratio is defined according to the generally accepted description offered by the Center for Research in Security Prices (CRSP) as the minimum
aggregate purchases or sales of securities divided by the average total net assets of a mutual fund over a calendar year. According to this definition the turnover
ratios used in the STI indicator are based on the annual reports of mutual funds. Turnover does not make any difference in what allocation of investments generates it, therefore it is combined with the STI that concentrates on short‐ term investments.
Turnover ratios are calculated in order to give investors insight in trading activity and transaction costs involved and is formulated as in equation eight:
!
turnover =value sales and purchases of investments " subscriptions " redemptions
net asset value (8)
In the literature section the turnover ratio was introduced with additional background information in combination with behavioral finance literature. 3.1.4 Alpha and beta According to Elton et al. 2007 the Jensen’s alpha is described as the excess return of a stock or portfolio of stocks relative to the return of the benchmark index adjusted for risk with the risk‐free rate (treasury rate). The beta of a stock or portfolio of stocks can be described as a measure of volatility relative to the market. The in‐sample Jensen’s alpha and in‐sample beta are calculated using a single index model (Elton et al., 2007) as of equation nine. Ri,t is the return of
stock i at time t and Rf,t is the risk‐free treasury rate at time t. The alpha is (αi,t)
and the beta (βi,t) for stock i at time t, respectively. Rm,t is the return of the
benchmark at time t.
!
Ri,t" Rf ,t =#i,t+$i,t(Rm,t" Rf ,t) (9)
The beta for stock i at time t is calculated by dividing the covariance of the return of the stock i at time t (Ri,t) and return of the market at time t (Rm,t) with the
variance (σ2m,t) of Rm,t (equation 10). The second step to calculate the beta of
the portfolio p at time t is adding all betas for the stocks included in the portfolio according to portfolio weights Xi,t (equation 11). The annualized three‐year beta
12 equation 13; and dividing the outcome by three to calculate the annualized three‐year Jensen’s alpha (called alpha in the rest of the paper).
!
"i,t = Ri,t# Rf ,t #$i,t(Rm,t# Rf ,t) (12)
! "p,t = Xi,t i=1 n
#
"i,t (13)In‐sample three‐year annualized alphas and betas are used in this paper and section 3.5 describes how to test for the significance of alpha and beta in the STI indicator.
3.2 Data description
This subsection provides data requirements and assumptions to perform the mutual fund analyses according to the STI indicator. Four topics are covered, namely population selection, data on holdings, data on turnover, and data on alpha and beta.
3.2.1 Population selection
The mutual funds that are used in this paper are selected according to the following conditions:
• The mutual fund has portfolio holdings and allocation data over the past three years;
• Yearly turnover ratio is available in annual report of mutual fund and is calculated according to equation eight;
• Only all‐equity open‐end mid‐ to long‐term investing mutual funds are included.
sample. Comparing the geographical versus the sector sample shows that the median and average values for the turnover and STI are less and that the core ratio and number of positions held are larger. The median and average turnover for the geographical and sector oriented sample are 66% and 96%, and 83% and 102%, respectively. The sector‐oriented sample does have a maximum turnover of 378%. The population’s standard deviation with respect to turnover is approximately 70%.
14 Table 2: Summary statistics of data, panel a, b, and c of this table represent summary statistics for the population, geographical sample, and sector sample, respectively. The turnover, STI, core ratio, three‐year annualized alpha and beta are calculated using the definitions presented in section three. The net asset value (NAV) in € times 106 is the value of the mutual fund’s assets and positions represent the number of holdings within a mutual fund (excluding cash and/or equivalents). The Morningstar rating shows the number of stars. The summary statistics include a median, a minimum and maximum value, a mean, a standard deviation (SD) and the number of observations (#obs) of individual mutual funds. All the figures are calculated using data from table one in appendix A. Panel a: Summary statistics for the population
Median Minimum Maximum Average SD #obs
Table 3: Population (2006) of allequity mutual funds, sorted by the STI dimensions. The mutual funds are classified in intervals for STI and turnover ratio. The amounts of mutual funds that lie within a certain interval for the STI or turnover ratio are summed up in the last column (for STI) and last row (for turnover). The STI and turnover ratio satisfy the definitions as mentioned in section three. Turnover (%) STI (%) 0‐30 30‐ 60 60‐ 90 90‐ 120 120‐ 150 150‐ 180 180‐ 210 210‐ 240 240‐ 270 >270 All 55‐60 1 1 50‐55 0 45‐50 1 1 40‐45 0 35‐40 1 1 2 30‐35 2 1 1 4 25‐30 1 1 20‐25 1 2 3 2 1 1 10 15‐20 1 3 1 3 2 1 11 10‐15 1 6 7 3 1 18 5‐10 1 6 8 4 1 2 22 0‐5 11 4 2 3 2 1 23 All 15 19 19 16 11 4 5 1 0 3 93
Table 4: Median net asset value and expense ratio’s, calculated from the population consisting of all equity open‐end mutual funds in 2006. The median values for certain intervals of STI and turnover are based on at least three observations. Panel a displays the median net asset value in € x 106 derived
16
3.2.2 Data on positions
Portfolio positions are acquired through Morningstar and meet the criteria as stated above, so only all equity open‐end mutual funds are selected. The most important part of the holdings data is the fraction of each investment to the total net asset value (on which also the turnover ratio is based on). Table two provides summary statistics for the data of the number of positions held within mutual funds. The median number of positions held within a mutual fund of the population is 76. The median and average number of positions held within a mutual funds in the geographical and sector oriented sample are 96 and 51, and 147 and 66, respectively.
3.2.3 Data on alpha, beta, and Morningstar rating
The three‐year annualized in‐sample alphas and betas of the population show a large gap between the minimum and maximum values, according to table two. For example, the population has a minimum and maximum value for the three‐ year annualized alpha of approximately ‐29% and 19%, respectively. The three‐ year beta of the population varies from 0.69 to 1.92. The sector sample includes these maximum and minimum values for the three‐year annualized alpha. The average three‐year annualized alpha and beta for the sector sample are large, compared to the geographical one. The median values of the three‐year annualized alpha and beta for the population are ‐0.41% and 1.09, respectively.
The Morningstar rating ranges from zero to five stars, as is also applicable to this population with an average rating of three stars, according to table two. The geographical sample does have a minimum amount of stars equal to zero, whereas the sector sample does have one star as minimum amount.
3.3 Value of combining of both dimensions
The combination of the STI and turnover (the STI indicator) gives the investor insight in the short‐term orientation and trading activity of his or her investment manager. The drawback of using the dimensions separately is that the turnover dimension concentrates on the activeness of a mutual fund and does not make any difference in what generates the turnover (portfolio allocation changes within STI or long term investments). Separate use of the STI dimension will not explain the total activeness of the mutual fund management, due to allocation in core and mid‐term investments. Single use of the dimensions would only be of value if they where compared with a benchmark of mean turnover or STI component of a sample of other mutual funds. The STI dimension is different from the Active‐Share, as is proposed by Cremers and Petajisto (2007), in that the STI dimension does not need a benchmark to compare for dissimilarities with that benchmark. For example, the mutual fund ‘g1’ from table one in appendix A has a turnover of 130% with a STI component of 2.4%. It is unlikely that this STI component is responsible for the trades having a value of 1.3 times the net asset value of the mutual fund ‘g1’. Therefore, a possible explanation is that in 2006 the fund’s management decided to change the strategy and bought many new investments or sold investments that where in portfolio more than one year, but less than three.
geographical and sector sample are characterized with an r‐squared of 47% and 7%, respectively for the linear relationship between STI and turnover. Results of the F‐test and coefficients of determination are summarized in table five and include the same analysis for CORE investments.
18 The graph of figure two is divided into four quadrants based on median values of the STI and turnover from a population. The first quadrant includes mutual funds with the relative lowest STI and turnover and the second quadrant contains mutual funds with the relative lowest STI, but relative highest turnover. Quadrant three and four represent mutual funds with, respectively, the relative highest STI and lowest turnover and relative highest STI and highest turnover. All the quadrants represent values for STI and turnover based on a population of mutual funds. For example, geographical oriented investing mutual funds, as in the samples above, are characterized by relative low turnover and STI ratios (quadrant one). The fund’s management invests long term oriented and initial investments are likely to become core investments with the passage of time. Short‐term investing and quick response to market circumstances is not a common habit for these mutual funds, this can be concluded from the fact that both STI dimensions have relative low values.
3
4
S T I1
2
0 TurnoverFigure 2: STI indicator graph with four quadrants separated by median values of STI and turnover, based on the underlying population.
Especially quadrant four is important in examining behavioral finance elements within mutual fund management, because it includes funds with relative large STI and turnover. If a particular mid‐ or long‐term investing mutual fund from the population stays more than one year in quadrant four, one may argue that this fund’s management exhibits overconfident behavior. Unless, of course, if the fund’s alpha is positive. Overconfident behavior can be caused by the illusion of knowledge and, consequently, results in more trading. According to many studies, low turnover funds eventually outperform high turnover funds in the long term. The investment focus of these fund managers, suffering from the illusion of knowledge, is concentrated on new and/or small companies that are risky short‐term investments (Statman et al., 2006). These short‐term oriented investments are captured by the STI ratio and enhance the fund’s overall turnover.
investments, modern portfolio theory statistics and the STI and turnover. Using the contents of table six, we are able to apply the STI indicator. Make the assumption that the median values for the STI and turnover are 15% and 80%, respectively.
Table 6: Data for mutual fund X to apply STI indicator, summary of STI, core, and turnover ratio and the strategy of the investment managers in t‐1, supplemented with NAV, number of positions (#pos), alpha and beta, Sharpe ratio, and strategy.
STI % Core % Turnover % Strategy
10% 60% 30% Defensive Worldwide
LT
NAV #pos Alpha/Sharpe Beta
6 billion 150 7%/‐0,5 0,95
The result of the STI indicator of mutual fund X is presented in figure three; the spot marked with ‘X’ in quadrant one indicates the position associated with the STI (10%) and turnover (30%). The median values of the STI and turnover are represented by the horizontal and vertical line and have values of 15% and 80% based on the population, respectively. Y
3
4
Z S T I 1 5% X1
2
0 80% Turnover Figure 3: STI indicator for mutual fund X20 can speculate that the mutual fund management of fund X did some major new investments, or sold those to become cash or equivalents.
To provide an investor with a pallet of information, the mutual fund provider can design a fact sheet with the STI indicator as a supplementary tool. The principle of applying the STI indicator will not change, however the investor is able to weigh all alternatives according to its risk preferences. For example, the modern portfolio statistics as the alpha and beta can be used to accompany the quadrants of the STI indicator. In combination with return‐based or style‐ based mutual fund rating systems the STI indicator provides the investor maximum possibility to select mutual funds according to a snapshot of important information. The STI indicator used in practice is elucidated in section four. 3.5 How to test for significance of alphas and betas in quadrants This subsection describes a statistical test that can be used to assess if the alphas and/or betas from the STI quadrants deliver additional insight that is statistically significant using a t‐test. This t‐test compares the mean from a sample (one of the four quadrants) with the known value of the population mean.
An unpaired one‐sample Student’s t‐test that assumes a normal distribution is used to test if the alpha and/or beta for a particular quadrant has any explanatory power. The test is unpaired, because there is no one‐to‐one correspondence between the samples, likewise the samples are independent of each other. Although the sample size of two quadrants is less than 30 observations, it is not suitable to use a non‐parametric test, like a Wilcoxon signed rank test. The samples consist of data based on direct measurements instead of ordinal data that can be arranged in a particular order. We need to clarify if the sample mean alpha for each quadrant differs significantly (at 5% and 10% significance level) from the population mean alpha. The main and alternative hypotheses stated below are used to test for significance of alpha or beta in each quadrant’s sample (Q1, Q2, Q3, and Q4). H0: the quadrant’s mean alpha/beta does not differ from the population mean alpha. H1: the quadrant’s mean alpha/beta does differ from population mean alpha. Calculating the t‐statistic as is defined in equation 14 will test these hypotheses. ! t = X " µ0 #x/ nx , (14) where ! Xis the sample mean and ! µ0is the mean of the population, ! "x and ! nxare
4
Results: STI indicator in practice
In practice the STI indicator can be used as a stand‐alone and as a supplementary tool in combination with other return or style based mutual fund rating mechanisms. This section will test if the alphas and betas for each quadrant do have explanatory power and a selection of two mutual funds from the population will be used to apply the STI indicator. 4.1 Determining quadrants and validation of STI indicator To determine the relative position of an individual mutual fund, the STI indicator graph is divided into four quadrants based on median STI and turnover values from the population. In this case, the population of 93 mutual funds does have a rounded median value for STI and turnover of 11% and 80%, respectively. The STI indicator graph for this paper’s population is presented in figure four. In addition of the scatter plot from figure one of section three, the four quadrants are drawn into figure four based on the population’s median values of STI and turnover and the 93 mutual funds are plotted. Quadrant one is characterized by a three‐year annualized mean alpha and beta of ‐1.72 and 1.09 based on 35 observations, respectively. Quadrant two has a three‐year annualized mean alpha and beta based on 14 observations of 1.18 and 1.14, respectively. The third quadrant has 14 observations and a three‐year annualized mean alpha and beta of, respectively 4.14 and 1.25. Quadrant four has a three‐year annualized mean alpha of 0.96 and a mean beta of 1.17 based on 30 observations. From this small population one can observe that the funds are not equally distributed among the quadrants, however there is an indication for the alpha and beta in each quadrant. From this data, it is possible to conclude that high turnover and a high STI ratio (quadrant four) has a small positive alpha. The mutual funds with quadrant one characteristics (low STI and turnover ratio) do have a negative alpha that may be explained that very passive investing fund do not produce excess return on their benchmark.
22
Table 7: Tstatistics for testing significance of threeyear annualized mean alpha and beta for each quadrant: the mean, standard deviation (SD) and sample size (#obs) for alpha and beta of the quadrants are used to calculate the t‐value. Panel ‘a’ shows t‐ and p‐values for one‐year average in‐sample alpha and panel ‘b’ gives the t‐ and p‐values for the in‐sample beta. The figures in panel ‘a’ are based on average one‐year in‐sample alpha. The row represented by ‘Pop’ contains the mean, standard deviation, and size of population for alpha and beta.
Panel a: information for calculus on significance of (one‐year average) alpha for quadrants
Mean SD #obs t‐value p‐value
Q1 ‐0,5735 2,4187 35 ‐1,7652* 0,0865 Q2 0,3655 1,6425 14 0,4951 0,6288 Q3 1,3793 2,1551 14 2,1375* 0,0521 Q4 0,3199 2,0801 30 0,4522 0,6545 Pop 0.1482 2.2397 93 ‐ - Panel b: information for calculus on significance of beta for quadrants Q1 1.0877 0.2125 35 ‐1.6365 0.111 Q2 1.1346 0.1868 14 ‐0.2377 0.8159 Q3 1.2500 0.2700 14 1.4343 0.1751 Q4 0.9597 6.2404 30 0.4473 0.6578 Pop 1.1465 0.2244 93 ‐ ‐ t‐values significant at the 10% significance level are marked with an asterisk (*)
Figure 4: STI indicator for the population used in this paper, the horizontal and vertical lines represent median values of STI and turnover, respectively.
To deliver the investor quick and understandable information about short‐term allocation and activity of the mutual fund, figure three may be too complex. Therefore, in the application of the STI indicator the format of figure one in section three will be used.
mutual funds in quadrant four have relative large net asset values contrasted by relative small net asset values for funds in quadrant one. From panel ‘b’ of table four can be argued that funds in quadrant four that have relative large net asset values do have relative low expense ratios. Tables like table three and four can be useful to show investors the characteristics of the quadrants from the STI indicator at a glance.
4.2 STI indicator applied
This subsection proposes an application format for the STI indicator on two mutual funds selected from this paper’s population. The selection is characterized by geographical and sector investment focus. Each selected mutual fund is accompanied with its investment policy and mutual fund related ratio’s and modern portfolio statistics as well as with return and style based rating systems.
First the STI indicator will be applied as a stand‐alone measure; thereafter the STI indicator is used as a supplementary tool with return‐based performance measures. The selected mutual funds are Fortis Obam (flagship of Fortis Investments) and the ING Daily Consumer Goods (sector oriented). 4.2.1 STI indicator for Fortis Obam Fortis Obam is a mutual fund of the population as is described in section three and is denoted by ‘g4’ in table one of appendix A. The chairman of the board of supervisory directors of Fortis Obam (prof R.A.H. van der Meer) describes the strategy of the fund in the annual report of 2006/2007 as in the next paragraph.
“Fortis OBAM aims at a balanced international portfolio of listed shares with interests in Europe, the United States of America, South East Asia and Japan and strives for increase of the net assets in the long term. Traditional, a considerable part of the assets are invested in Dutch companies. Within the portfolio of Fortis OBAM there is special attention for large companies that are strongly international orientated and capable of adjusting well to the global investment climate, but also for smaller companies.“
The quadrants of the STI indicator are based on median values form the population of the STI (11%) and turnover (80%) in 2006. The STI and turnover ratio of Fortis Obam are 1,0625% and 20,66% in 2006, respectively. Quadrant one in figure five displays the relative position of Fortis Obam within the population in the STI indicator. 3 4 ST I 1 1% 1 ● 2 0 80% Turnover Figure 5: STI indicator for Fortis Obam in 2006, the spot in quadrant one represents the place of Fortis Obam in the STI Indicator with an STI and turnover ratio of 1,0625% and 20,66%, respectively.
24 population. The STI component of the portfolio is considerably low in a way that it cannot produce a turnover of 20%. Turnover can be generated by new investments or by selling of core or mid‐term investments. To add more perspective to the STI indicator, table seven delivers a wide variety of return based mutual fund ratios. Nearly seventy percent of Fortis Obam its investments last for at least three years and 20% of NAV is allocated to the top 10 investments. With the portfolio composition and strategy of Fortis Obam in 2006 it gained a five star rating from Morningstar to reflect good historical return‐based performance. The alpha of nearly 10% reflects that the long‐term investment approach with little room for short‐term investments pays off.
Table 8: Summary statistics for Fortis Obam 2006, threeyear annualized return is based on fund values in 2005, 2006, and 2007; the core ratio are positions within Fortis Obam that last for at least three years; Morningstar indicates past performance; alpha is the excess abnormal return of a portfolio compared to prediction from the CAPM, NAV is the net asset value; #inv is the number of positions held; %NAV in top10 explains the percentage of NAV allocated to 10 major positions; and beta is defined as a measure of systematic risk compared to the benchmark.
3Y ann. return% Core % Morningstar Alpha
18,80 % 70% 9,8%
NAV #inv %NAV in top10 Beta
€4,21 billion 245 20,50% 1,92
4.2.2 STI indicator for ING Basic Materials Fund
The sector fund is identified by ‘s33’ in table one of appendix A and ING Fund Management B.V. describes in the prospectus the investment strategy as in the next paragraph.
“The ING Basic Materials Fund uses active management to outperform the
result in investments that are risky, short term oriented, and producing high turnover for this long term oriented ING Basic Materials Fund.
Additional return‐based information is summarized in table nine, explains that that the historical performance is valued with four stars from Morningstar relative to its peer group at Morningstar. However, the fund does not outperform its benchmark as is intended by the mutual fund investment objective with an alpha of ‐1,92%. The active management is resulting in that the portfolio is 20% more volatile as the benchmark (beta equals 1,2).
Table 9: Summary statistics for ING Basic Materials Fund 2006, threeyear annualized return is based on fund values in 2005, 2006, and 2007; the core ratio are positions within ING Basic Materials Fund that last for at least three years; Morningstar indicates past performance; alpha is the excess abnormal return of a portfolio compared to prediction from the CAPM, NAV is the net asset value; #inv is the number of positions held; %NAV in top10 is the fraction of NAV allocated to the 10 major positions; and beta is a measure of systematic risk compared to the benchmark.
3Y ann. return% Core % Morningstar Alpha
23,04% 45% ‐1,92%
NAV #inv %NAV in top10 Beta