Can Mutual Fund Managers Outguess Sectors?

May 2007

Jeffrey Junhua Lu

Jeffrey Junhua Lu

Global Wealth Management Citi UK

e-mail: [email protected]

1

Can mutual fund managers outguess sectors?

Abstract: Previous research demonstrates that actively managed mutual funds exhibit negative market timing abilities. However, few studies have investigated the sector timing abilities of these fund managers. By implementing a multi-factor model based on Sharpe (1992), I derive a time series of mutual fund sector risk exposure coefficients which are demonstrated to be good proxies for actual fund exposures to the eight sectors examined in this study. Based on these sector risk exposures, sector timing measures are derived and used to explore whether fund managers in general, or certain sub-groups of fund managers in particular, exhibit successful sector timing abilities. My sample covers 485 randomly selected mutual funds listed on Datastream for the period from April, 1997 to July, 2002. I conclude that mutual funds as a whole exhibit some evidence of negative sector timing abilities. However, particular groups of funds, such as aggressive growth funds, appear to possess better sector timing abilities than other types of funds, and this pattern is more manifest after controlling for market downturn conditions.

Key words: sector risk exposures, sector timing, style analysis, asset allocation

2

Can mutual fund managers outguess the sectors? 1. Introduction Recent studies 1 show that industry effects have grown markedly in importance in explaining cross-sectional stock returns variances. Clearly, the strengthening of sector effects is related to the market euphoria engendered by the New Economy. The dramatic and persistent outperformance of the Technology and Telecommunication sectors until March 2000 and the subsequent reversal of fortunes have obviously played a major role in the increase of sector effects. The recent years have also been characterized by a resurgence of exceptionally strong style effects compared to historical standards. The Growth style outperformed the Value style in the late 90s until March 2000 and then strongly underperformed it until mid-2001. The shapes of the relative performance graphs of Growth versus Value and the Technology and Telecommunication sectors versus the other sectors are strikingly similar, with a reversal occurring simultaneously in March 2000. Of course, the Technology and Telecommunication sectors are typically “Growth” sectors, showing superior earnings growth expectations and expensive valuations. Hence, the recent period has been characterized by an increasing correspondence between style and sector effects, with growth stocks being more and more concentrated in some sectors, or in other words, each sector becoming gradually more homogenous in terms of style. The increasing importance of the sector effect has led many fund managers to reconsider their investment processes. There is growing evidence from both academics and practitioners that some fund managers are allocating funds in more general investment portfolios based on decisions about sector and sub-sectors (Schwob, 2000 and Targett, 2000). Perhaps more radically, such a movement toward a sector based investment philosophy has resulted in the launch of several sectorspecific tracker funds whereby investors have the opportunities to invest in products which specialise in specific areas of the stock market – for example, consumer goods,

1

See example, Baca, Garbe, and Weiss (2000), Cavaglia, Brightman, and Aked (2000), and Cavaglia, Brightman and Akel (2000).

3

financial and technology stocks.2 In practice, more and more actively managed mutual funds are relying on sector bets.3

2. Rationale behind sector investing 2.1. Sector as an important dimension in portfolio management The sector investing procedure can be described as the following two steps: first identify the sectors that are going to outperform the market, then identify the stocks in those sectors that are going to outperform the sector. The rationale behind the sector based investment philosophy is its greater potential to add value and control risk. Generally speaking, fund managers desire to produce the most efficient portfolio given the stated aims of the fund in the sense that all possible benefits from diversification are achieved. Cavaglia, Brightman and Akel (2000) suggest that industry factors have become an increasingly important component of security returns and diversification across industries provides greater risk reduction than diversification across countries. This was supported by fund managers at UBS Asset Management. They reported that, over the past five years, it has been less risky to diversify across global industries than across countries. Unintended industry exposures that result from equity benchmarks that are biased toward the home market may result in increasing inefficiency. For example, the U.K. market has a small exposure to the information technology industry (about 1.5%) in comparison with the world market (about 11.3%). A home-biased U.K. portfolio would thus tilt the portfolio away from the global allocation to the information technology industry. Such a tilt would materially affect the portfolio's return relative to risk (Cavaglia, Brightman and Akel, 2000). On the other hand, some fund managers are active with regard to sector rotation and try to anticipate and tilt their portfolios toward the sector or sectors they expect to lead the market in future periods. According to a recent study from Ibbotson Associates, a balanced portfolio of industry sectors could beat the overall market by an average of

2

For example, during the year 2000, several investment companies such as Fidelity, Aberdeen Asset Management and Legal & General have offered a variety of global sector funds, including healthcare, financial, industrial, and consumer goods. 3

The recent law suit between the Unilever Pension Fund and Merrill Lynch Asset Management serves as a good example. One issue cited in the court case was that the weighting of the Unilever U.K. equity fund was “inconsistent with prudent risk management", with a heavy bet on property shares.

4

0.55% a year4. If fund managers exhibit some sector betting abilities, then their funds will consistently be weighted more heavily in those sectors with positive returns, while being more lightly weighted in other sectors with negative returns. Thus, it should be possible to observe that the fund risk exposures to those sectors with positive returns are higher than to those sectors with negative returns. 2.2. Sector investing in a style paradigm Sectors and industries have different characteristics in a style paradigm. In other words, sectors and industries are not randomly distributed around the value-growth, large-mid-small grid. Each style box is tilted toward certain industries. This means an investor should be aware of the sector tilting occurring when investing in a particular style box and adding it to a portfolio. Whether a particular style box is heavily weighted toward the health care or basic materials sector may greatly influence the performance, correlation and risk characteristics of the overall portfolio. The sector composition in a style box can affect a manager's effectiveness. Managers have styles that either continually favour stocks in certain industries or are better at discriminating among stocks in certain industries than in others. A manager that is good with large-cap value stocks might be that way because of the industries concentrated there. That same manager might not be proficient at analysing the stocks in the industries concentrated in another style box. Industries are not constant through time in their location on the style grid. As industries become more favoured, they move up and to the right as they get larger and higher in price-to-book value. Industries that fall out of favour move down and to the left -- relatively smaller and cheaper. As a result, a manager that develops an impressive track record over a period in one style box might have done so because of the industries which were in that universe. Over subsequent years, there might be different industries in that style box, perhaps those not suited to the manager's investment approach. 2.3. Contributions of this study Hence, it is worthwhile investigating the sector risk exposures of mutual funds and the sector investing behaviour of fund managers. This enables plan sponsors and individual investors to verify that the investment decisions taken by the manager are consistent with the desirable sector exposures, to ensure that the sector bets taken by fund managers do not exceed certain safe levels. 4

Ibbotson's study finds that the average correlation among sectors is 31%, which is quite low and represents a source of diversification benefit.

5

However, few studies have investigated the sector investing behaviours of these fund managers. One exception is Dellva, DeMaskey and Smith (2001) who use the Dow Jones Subgroup Total Return Indexes as benchmarks to test the timing performance of the Fidelity sector mutual funds during the 1989-1998 time period. Based on the traditional market timing models5, the authors find that the sector fund managers have difficulty timing the upward and downward movements of their respective industries. By implementing a multi-factor model based on Sharpe (1992), I derive a time series of mutual fund sector risk exposure coefficients which are demonstrated to be good proxies for fund actual exposures to the eight sectors examined in this study. Based on these sector risk exposures, sector timing measures are derived and used to explore whether fund managers in general, or certain sub-groups of fund managers in particular, exhibit successful sector timing abilities. The remainder of the paper is organized as follows. Section 3 reviews the style analysis model in Sharpe (1992) and lays out the theoretical foundation for this paper. Section 4 investigates fund sector exposures based on Sharpe’s (1992) model. Section 5 examines the sector timing abilities of fund managers and sub-group of fund managers. Section 6 discusses the potential issues for future research. Section 7 makes the conclusion.

3. Return-based style analysis Fund style analysis can be done directly by analysing a manager’s equity characteristics and investment philosophy. Morningstar Inc., for example, assign funds to general style groupings on the basis of the equity characteristics the manager considers, the manager’s view of where value can be found in the market, and other aspects of the investment process. 6 Such a detailed analysis of fund style requires information on portfolio composition, which may be difficult to obtain. Furthermore, the classification of individual securities into slots based on characteristics can involve a substantial amount of judgement. For example, a conglomerate firm would typically have operations in several different sectors of the economy and it may be

5

See Treynor and Mazuy (1966) , Henriksson and Merton (1981)

6

Morningstar’s style boxes depend on price-earnings and price-book-value ratios (P/Es and P/Bs). Morningstar classifies all funds with average P/Es and P/Bs that are at least 12.5% below those of the S&P 500 Index as “value” and those at least 12.5% above as “growth”; all other categories fall between the two benchmarks.

6

difficult to identify how much of the firm goes into each sector. To avoid the above problems, one can analyse fund style by using return-based style analysis. Return-based style analysis, pioneered by William Sharpe (1992), has gained widespread acceptance among investment practitioners. First described by Sharpe (1988) as “effective asset mix analysis”, the method is a constrained form of regression analysis used to form inferences about the influences on and composition of investment portfolios, based solely on the historical returns to those investment portfolios. 3.1. The factor asset pricing model Sharpe (1992) demonstrates that a multi-factor model can be used to explain asset return differences. ~ ~ ~ ~ Ri = [ β i1 F1 + β i 2 F2 + ... + β in Fn ] + ~ ei

(1)

~ ~ where Ri represents the return on asset i, F1 represents the value of factor 1, F2 the ~ value of factor 2, Fn the value of the n'th (last) factor and e~i the "non-factor"

component of the return on asset i. All these values are (potentially) unknown before~ ~ the-fact, as indicated by the tildes. The coefficient values ( β i1 through β in ) represent ~ ~ the sensitivities of Ri to factors F1 through Fn . 3.2. Fund risk exposures to asset classes

Traditional multiple regression calculations would provide a set of coefficients that would estimate how sensitive the return on the subject portfolio is to the returns of the set of market indices selected as factors. Coefficients arising out of the traditional regression process may take on either positive or negative values. Investment practitioners often find it intuitively unappealing that a market index exerts a negative influence on a portfolio return.7 In style analysis, two constraints are placed on the range of values that coefficients ~ β ij may take. The first is that the coefficients have only values between 0 and 1. The second constraint is that the sum of the coefficients must equal 1; that is, 7

In addition to a frequent lack of intuitive appeal, Lobosco and DiBartolomeo (1997) argue that traditional regression may exhibit multicollinearity. If the independent variables are highly correlated, as two indices representing different approaches to investing with the same asset class are likely to be, the reliability of the estimated coefficients in meaningfully describing the underlying true relationship is very much in doubt.

7

~ 0< β ij <1,

(2.1)

~ ∑ β ij =1

(2.2)

and

In this way, the coefficients have an intuitive interpretation as the weights of the assets within a portfolio and can be conveniently displayed as “slices” in a pie chart. Hence, style analysis can be thought of as the process of forming a portfolio of indices that mimics, as closely as possible, the historical performance of a given portfolio. From the weights ascribed to the various market indices (known also as Sharpe style weights), I form inferences regarding the behavioural influences on and likely composition of the given portfolio.8

4. Fund sector exposures analysis 4.1. Methodology

Based on Sharpe’s (1992) model, I analyse the sector risk exposures for a typical fund. First, sector indexes need to be identified. The first intuitive idea is to use the relevant industry index and derive the return data from the monthly changes in this industry return index. After choosing 29 DataStream America Industry Indices, I categorize these into 10 sectors based on Black, Buckland and Fraser (2001)9. However, the results of the sector analyses based on the above sector return indices are not satisfactory. Not only do the results misrepresent a typical fund’s sector risk exposure, but also the R-square is very small which indicates poor model fitting. Two potential reasons account for the poor results. First, the return-based style analysis relies crucially on the correct specification of the style benchmark asset classes. Inappropriate or inadequate choice of style benchmarks may lead to wrong inferences about performance and the level of risk exposure to related asset classes (Dor and Jagannathan, 2002). The Datastream America Industry Indices seem to do a poor job 8

By placing these constraints on the values of weights, the multicolinearity problem of traditional regression can be somewhat mitigated (Lobosco and DiBartolomeo, 1997). Placing bounds on the weights reduces the likelihood that high correlations between the independent variables will cause the coefficients to “blow up” to unrealistic values. However, weights that are calculated to be 0 or 1 are not unbiased estimators. A coefficient (weight) on a market index that is calculated to be 0 as derived by traditional regression might have come out negative but is prevented from doing so. Similarly, a value of 1 could have been greater than 1 in a traditional regression. 9

One can always derive such a categorization by implementing some factor analysis. Farrell (1975) gives a detailed description of how to group stocks into sectors. In his paper, he identifies four categories: growth, stable, cyclical and oil.

8

of representing asset classes related to sectors. Second, the composition of the industry return indices might be quite different from the investment universe faced by the fund manager. This could potentially damage the model’s abilities to identify fund sector exposures, since the constituent stocks of fund manager portfolios differ to the constituent stocks of the benchmark portfolios. Fortunately, since Sharpe’s (1992) research, new benchmarks for mutual funds have been developed that help to provide a solution to the problem. In 1997, Moody’s launched new benchmarks for mutual fund performance. Moody’s sector fund indices are constructed in a way that they represent the typical performance of a fund specialized in the specific sectors. It has long been demonstrated that mutual funds as a whole cannot beat the market, or at best, they can only match overall market performance. So the sector fund indices could represent the average performance that could be achieved in a specific sector and/or be closely related to sector average performance. Moreover, the sector fund indices mimic the general investment universe faced by fund managers. Thus sector funds are available investment portfolios for a specific sector in the market. One can always construct a portfolio of sector funds and achieve the same sector risk exposures as an actively managed fund. Table 1 summarizes the standard deviations and correlations for these sector fund indices.10 The relatively low correlations among sectors can provide greater benefits through sector focus investing. For sector timing to be successful, the correlation of sectors with each other has to be low enough for the correct investment decision to matter. The model for deriving fund sector risk exposures can be expressed as follows: RPit = β1 RPsec tor1,t + β 2 RPsec tor 2,t + ... + β 8 RPsec tor 8,t + ei

(3)

RPit , the excess return for fund i during month t, excess returns are calculated by using the monthly return on a three-month U.S. Treasury bill as the risk-free asset.

RPsec tor ,t , the excess return for sector i during month t

β , the risk exposures to sector i for fund i ei , the portion of the return on the fund not related to the sectors 10

Sharpe (1992) suggests two criteria for selecting the appropriate style benchmarks: asset class returns should either have low correlations with one another or, in cases in which correlations are high, different standard deviations. Actually, the average correlation among these sector indices is 43% and the indices’ standard deviations are quite different from each other. This shows that Moody’s sector fund indices satisfy Sharpe’s criteria for selecting appropriate benchmarks.

9

My model preserves the analytical tractability and ease of interpretation of Sharpe’s model. Rather than referring to traditional sector classes modelled by passive indexes, I simply use alternative sector styles represented by indexes of active sector funds. Therefore, for a given mutual fund, the beta coefficients can be seen as exposures to the different Moody’s sector styles. 4.2. Fund indices’ sector exposures

The results obtained from this sector analysis are quite satisfactory. Table 2 summarizes the sectors exposures for the nine Moody’s Managed Fund Indexes. In panel A estimated sector weights are given. Each row deals with one particular fund index, where the elements in columns 2 to 9 report the estimated sector weights. Figure 1 presents these results in a stacked bar diagram. It clearly shows value funds have the highest sector risk exposure to the financials sector, while all the growth funds (including growth and aggressive growth funds) have the highest sector risk exposure to the technology sector and the second highest sector risk exposure to the health sector. An interesting pattern identified in the above figure is that small capitalization value funds exhibit the highest sector exposures to the real estate sector. All the above analysis seems to be consistent with the traditional wisdom that growth funds favour technology and health sector stocks, while value and income funds take large stakes in financials sector. Figure 2 shows the sector risk exposures for a typical aggressive growth fund. That is, aggressive growth fund managers derive most of their fund performance from taking sector bets. The same analysis can be done on value fund managers, as shown in Figure 3. One can also examine the evolution of fund sector risk exposures over time. 11 Interestingly, aggressive growth fund managers have consistent portfolio exposures to technology and health sectors, while value fund managers have tended to decrease their portfolio exposures to the utilities sector during the sample period. Thus, nonzero style weights on all the eight sector indices provide a fuzzy picture of the true risk exposures of different active fund strategies. One potential problem with conventional style analysis is that it fails to distinguish significant style weights from insignificant ones. I therefore use the two-step procedure of Lobosco and DiBartolomeo (1997) to determine the statistical significance of sector weights. First, I calculate the standard deviation of the returns unexplained by the eight sector indices for each of the nine Managed Fund Indices. 11

See Figure 4 and Figure 5

10

Second, I conduct style analysis of each of the eight indices using the remaining seven indices as explanatory variables12. The objective is to obtain the standard deviation of the residuals from style analysis on each sector index relative to the remaining seven sector indices. Lobosco and DiBartolomeo (1997) show that the standard error of the style weight on index i is given by

σd σ B n − k −1

(4)

i

where i = index corresponding to the sector weight being estimated

σ d = standard error of the style analysis σ B = “unexplained Sharpe style index volatility” for index i i

n = number of returns used in the style analysis (e.g., 60 months for the full sample period) k = number of market indexes with nonzero style weights Using these standard errors, I determine the statistical significance of the style weights and report these in panel A of Table 2. In panel B 95% confidence intervals are given for all factor loadings, which show that the point estimates are relatively precise reflections of the portfolio sector weights. 4.3. Single fund sector exposures

The above analysis can be extended to a single mutual fund rather than a fund index. From Datastream, I identify a number of funds with stated investment objectives such as aggressive growth, growth, income, etc. I present below some results on the sector risk exposures analysis of these funds for illustrative purposes. First, Fidelity Select Biotechnology Portfolio is considered. This is a sector fund, which claims to invest “at least 80% of assets in common stocks of companies principally engaged in the research, development, manufacture, and distribution of various biotechnological products, services, and processes and companies that benefit significantly from scientific and technological advances in biotechnology.” Figure 6 12

The style is estimated without positive constraints on individual weights, though the weights must sum to 100%, thus giving an unexplained volatility for each index.

11

confirms its self-declaration. Results are similarly consistent for the Fidelity Select Utility Portfolio as Figure 7 illustrates in terms of its stated investment objectives of utility sector fund. Besides looking at sector funds, I also examine some general funds. The example I use to demonstrate here is the Fidelity Magellan fund. Table 3 compares sector sensitivities obtained in this study and the actual sector weightings provided by the Morningstar Fund Quicktake Report. Obviously, the fund sector exposures successfully identify the largest three sector bets. The major differences can be probably attributed to the different sector definitions used by Moody’s and Morningstar. Also the sample periods are different. By looking at the detailed portfolio holding information, one can be more confident about the sector risk exposures. Based on the above analysis, the sector sensitivity analyses methodology demonstrates it is possible to truly represent a fund’s sector risk exposures. So one can investigate how fund managers make sector bets based on the above analysis results.

5. Sector timing 5.1. Sector timing abilities

If the sector sensitivity coefficients truly represent the fund’s sector risk exposures, one can investigate the fund’s sector timing abilities based on the above results. The major assumption under the sector timing measures is that the portfolio holdings of an uninformed investor cannot be correlated with future sector returns. However, since an informed investor can predict when certain sectors will have either higher or lower than average returns, he can therefore profit from these changing expected returns by tilting his portfolio weights over time in favour of sectors with expected returns that have increased and away from assets with expected returns that have decreased. To figure out whether sector timing plays an important role in determining fund investment performance, Hypothesis 1 is developed and tested:

Hypothesis 1: Fund portfolio sector exposures differ significantly between winners and losers, with winners having larger exposures in winning sectors and smaller exposures in losing sectors when compared to losers. As mentioned earlier, style investing and sector investing are somewhat interrelated, one should expect to see that a fund manager who has implemented a successful style 12

investment strategy (basically value or growth) would have larger portfolio exposures to sectors that outperform the market during the same period.

Hypothesis 2: If growth fund managers outperform (underperform) value fund managers, growth fund managers will have larger portfolio exposures in winning (losing) sectors when compared to their value counterparts. Fund managers can generate additional performance if a sector betting strategy has time-varying expected returns that the manager can exploit by changing portfolio weights to exploit those sectors when they are the most profitable. To measure a fund manager’s success at timing the different sectors, I introduce a sector timing measure, which is similar to the characteristic timing measure used by Daniel, Grinblatt, Titman and Wermers (1997). The month t component of this measure is N

ST j ,t = ∑ (w j ,t R sj ,t − w j ,t − k R sj ,t − k )

(5)

j =1

ST j ,t , the sector timing measure for sector j at month t w j ,t , the sector exposures for sector j at month t w j ,t −1 , the sector exposures for sector j at month t-k R sj ,t , the sector return for sector j at month t R sj ,t −1 , the sector return for sector j at month t-k

Thus, if the Fidelity Select Biotechnology Portfolio, for example, increases its weighting in sector j in a month when the sector performs strongly, then the fund will have a positive ST measure for that month. The average ST measure over all sample months that the Fidelity fund is in existence is the overall ST measure for the fund. Grinblatt and Titman (1993) implement a similar portfolio change measure as follows:

∑∑ [ R

jt

( w jt − w j ,t − k )] / T

(6)

They argue that if fund managers do not possess superior information and cannot predict future asset returns, the above performance measure will converge to zero in large samples. I also examine the GT measure in this study as a comparison to the sector timing measure.

13

Since I have monthly sector risk exposure data, the interval k can be selected as 1,3,6 to represent monthly, quarterly or semi-annual portfolio sector weighting adjustments for actively managed mutual funds. It should be noted here that the purpose of the investigation of fund sector timings activities is first to figure out whether there exist such sector timing abilities among mutual fund managers. Further, if mutual funds possess sector timing abilities, do there exist significant differences between different types of mutual funds? Thus, two hypotheses are tested as follows:

Hypothesis 3: If actively managed funds possess sector timing abilities, there should be significant differences between the sector timing measures of these funds and those of the passive market index portfolios.

Hypothesis 4: If different types of mutual funds possess different sector timing abilities, there should be significant differences between the sector timing measures of different types of mutual funds. Hence, I am interested in the relative level of the sector timing measures (the differences between measures), rather than the absolute level of the measures. Also, the sign of the differences (either positive or negative) may well represent the superior or inferior sector timing abilities of different fund managers. 5.2. Data and methodology

Monthly total return index data for virtually all equity mutual funds (unit trusts) that existed during any given quarter between April, 1997 and July, 2002 13 (inclusive) were extracted from Datastream. The definition of return index is expressed as: RI t = RI t − k ×

Pt + Dt Pt − k

(7)

RI t , return index on day t RI t − k , return index on day t-k Pt , the closing bid price on ex-date t Pt − k , the closing bid price on date t-k 13

Although the sample period starts from April, 1997, in order to derive the sector risk exposures, each fund needs an observed time series of its previous 36 month excess returns. Therefore, the sector risk exposures for the earliest month should start from May, 2000, since there are exactly 36 months between April 1997 and May 2000. To obtain fund sector exposures for earlier months, either the sample period must be extended backward or the observed period needs to be shortened.

14

Dt , the dividend payment associated with ex-date t Hence, the monthly returns for each fund can be expressed as:

Rt = RI t / RI t − k − 1

(8)

The closing bid price of the fund includes the load charge, brokerage fees, and stamp duty. When fund managers calculate the bid price of the fund, the values of the underlying securities are valued at their offer prices. Because the impact of brokerage commissions and bid-ask spreads are not removed from returns, the fund returns will be gross of trading costs. Hence, the returns on the funds are approximate, as being gross of the load charge and trading costs but net of the management charge. Over 2,400 funds are identified in this database over the sample period. The fund category is identified by examining the fund name and more than 20 such categories are identified. The main U.S. mutual fund categories include bond fund, equity fund, international funds, global funds and sector funds. In this study, I focus on actively managed equity funds and investigate mainly 5 types of equity funds, including aggressive growth, growth, growth & income, value, and balanced funds. Funds within the aggressive growth and growth sector invest in stocks that provide capital growth. Funds within the value sector invest in stocks that have a good earnings track record and provide income. Funds within the growth & income sector invest in stocks that provide moderate income with decent growth prospects. Balanced funds invest in multiple assets in the market in an attempt to time the market. For aggressive growth and growth funds, I allocate small-cap growth funds to the aggressive growth sector while including large-cap growth funds in the growth sector, as small-cap companies on average exhibit a higher growth rate than large-cap companies. Among these five types of mutual funds, I randomly select an arbitrary number of funds in each category.14 Table 7 summarizes the number of funds in each category. As mentioned in the previous section, 8 sectors are used in this study. The performance characteristics for the 8 sectors during the sample period are summarized in Table 4. Obviously, the technology and metals sectors are very volatile during the sample period, while the real estate and utilities sectors exhibit the lowest volatility. The sector risk exposures of mutual funds are then derived using model (3) above.

14

Note that in this study, I just use self-reported investment objectives, because Datastream only includes the fund name in their database.

15

5.3. Results

To test Hypothesis 1, the winners and losers in each year during the 2000-2002 period are identified and a sector exposure analysis is carried out on these funds. Table 5 provides the results. During the 3-year period, winners tend to favour sectors with good performance (such as finance, metal, real estate) while losers tend to favour sectors with poor performance (such as technology) in each year. It seems sector investing has significant impacts on investment performance. An overview of the performance of the sample mutual funds over the 6-year period from 1997.4 to 2002.7 is presented in Table 6. The aggressive growth and growth funds performed very well during the sub-period of 1997 to 1999, outperforming their value counterparts. However, during the sub-period of 2000-2002, they performed very badly and underperformed value funds. Recalling that in table 2, I demonstrate that growth funds consistently exhibit larger exposures to technology and health sectors, while value funds consistently exhibit larger exposures to the finance sector. As reviewed in Panel B, Table 4, technology and health sectors performed very well during 1997-1999, while they did poorly during 2000-2002. This shows that style investing and sector investing are indeed closely related, which provides evidence for Hypothesis 2. Over the sample period, aggressive growth funds provide the lowest gross excess returns while growth funds are associated the highest gross excess returns. When we look at risk in terms of standard deviation of fund excess returns, balanced funds have the lowest risk, while aggressive growth funds have the highest. Table 7 provides the average sector risk exposures for all sample funds with different sectors of funds derived based on the sector risk exposures for each fund. The results are presented in Table 8. Consistent with the results of traditional market-timing performance studies on mutual funds, the sample funds as a whole exhibit negative sector timing. For the ST measure, the portfolio sector exposure adjustment is statistically significant at -0.16% (on a monthly basis), -0.13% (on a quarterly basis), and -0.25% (on a semi-annual basis). Also, the longer the adjustment interval, the larger the magnitude of the negative sector timing measure, which shows that fund managers tend to time sector selection more badly the longer the investment horizon. With the GT measure, the magnitude of the negative sector timing measure is much smaller than that according to the ST measure, but still significant at α=0.01. 16

When considering different mutual fund sectors, the results differ somewhat according to ST and GT measures. Under the ST measure, aggressive growth funds exhibit positive timing measures (0.03% for monthly, 0.31% for quarterly, and 0.62% for semi-annually), which indicates that their fund managers are better at sector timing compared to other fund managers over all adjustment intervals. Contrary to aggressive growth fund managers, value fund managers exhibit the worst sector timing abilities. Turning to the GT measure, aggressive growth fund managers also exhibit consistently better sector timing abilities than other fund managers. Looking at Panel B of Table 8, we can see that aggressive growth funds have the smallest (if not the smallest, then the second smallest) magnitude of negative sector timing measures (-0.03%, -0.13%, and -0.02% for monthly, quarterly, and semi-annual basis adjustments respectively), while growth funds have the largest magnitude of negative sector timing measures (-0.12% for monthly, -0.15% for quarterly basis adjustments respectively). 5.4. Adjusted for market movements

It should be noted that the sample period covers relatively volatile market conditions. As can be observed from Table 4, several sectors exhibit very high volatilities during the sample period. Figure 8 shows that different sectors of funds perform quite differently during sub-sample periods. Therefore, some sector exposure changes are the result of market value changes in the sectors. Without controlling for extreme market conditions, the sector timing measures could be biased downwardly during bear markets. This could partly explain why the sector timing measures are negative in this study, since all sectors fall during bear markets. The only contribution fund managers can make during such market conditions is that by reducing the exposures to those sectors with sharp falls, the sector timing measures could exhibit a smaller negative value. This has been demonstrated previously, as the sector timing measures are quite different across different sectors of funds. Hence, for robustness tests, I need to control for the different market conditions pertaining during the sample period.15

15

One way to achieve this task is to divide the sample period into two sub-sample periods, covering 1997.4 – 2000.3 and 2000.4 – 2002.7. The first half of the sample period exhibits some characteristics of bull markets, while the second half of the sample period exhibits some characteristics of bear markets. To obtain sector risk exposures for month before May, 2000, I rerun the quadratic programming based on 12 months and 24 months observation intervals. However, the sector risk

17

Actively managed mutual funds are supposed to provide extra services to fund investors; otherwise they cannot justify the higher management fees charged to customers when compared to their passive investment counterparts, such as index funds and ETFs (Exchange Traded Funds). Sector timing can possibly be one of such extra services, that is, if active fund managers can successfully time the sectors, they indeed provide superior performance to shareholders. Hence, a possible alternative is to compare the sector timing measures of active funds to those of passive investment portfolios (for example, index funds or ETFs), since these portfolios should not reflect any sector timing abilities. The only source of change in sector exposures of these portfolios is change in sector market value. For convenience, I simply use market index portfolios (such as the S&P 500) as benchmark portfolios for comparison. Nine market index portfolios are obtained from BARRA, which are summarized in Table 9. Based on the same techniques used in the previous sections, I obtain the sector timing measures for these index portfolios. The results in Table 10 show that the market indexes exhibit negative sector timing measures to some extent (Under the GT measure, some market indexes exhibit positive sector timing measures on a quarterly and semi-annually adjustment basis, however, these measures are not statistically different from zero16). Generally speaking, since market indexes represent a passive buy-and-hold investment strategy that follows market movements, the sector timing measures for these index portfolios should theoretically not be different from zero. However, Carlos and Angeles (2000) demonstrate that the existence of a passive timing effect 17 in portfolio management may create bias in the application of measurements used to evaluate market timing ability. During a rising (falling) market, the existence of passive timing creates a positive (negative) bias in the estimation of market timing. 18 This argument applies to the situation in this study: the market exposure coefficients obtained from the 12-month and 24-month quadratic programming are relatively volatile and may not be suitable as proxies for fund sector exposures. 16

One exception is the S&P Small Cap. 600 Growth Index.

17

Underlying the passive timing effects is the fact that variability in portfolio systematic risk exposures results from financial market trends and the accumulated evolution of the weightings of assets that make up the respective portfolio. Therefore, at an initial moment in time, the systematic risk exposure of the portfolio is the weighted sum of the weightings of its individual assets. As the prices of these assets vary in a non-homogenous manner, their weightings in the portfolio will change. Therefore, at a subsequent moment in time, and in the absence of new sales and purchases, the systematic risk exposure will have changed.

18

Faced with a rising (falling) market, the levels of systematic risk for the portfolio will increase (decrease). Consequently, passive timing will occur in a passively managed portfolio adopting a buy and hold strategy and, more importantly, this may artificially create evidence of market timing.

18

exhibited a downward trend during the sample period. Hence, we should expect to see negative timing measures for the market indexes (passively managed portfolios). What we are interested in here is whether the sector timing measures for the sample funds are statistically different from those for the market indexes. Table 11 compares the differences in sector timing measures between the sample funds and market indices. It seems that all mutual funds provide roughly the same magnitudes of sector timing measures as market index portfolios. Under the ST measure, all funds have the same sector timing measures as the S&P 500 market index over the one month adjustment interval, which indicates that mutual funds as a whole do not possess sector timing abilities. However, over 3-month and 6-month intervals, all funds exhibit increasing negative sector timing measures, suggesting that mutual funds as a whole exhibit poor sector timing abilities over these intervals. Applying the GT measure, however, mutual funds tend to have smaller magnitudes of negative sector timing measures than the market index, indicating that using this measure they appear to possess some sector timing abilities. Aggressive growth funds seem to possess better sector timing abilities over shorter adjustment intervals. For example, under the ST measure, although aggressive growth funds have consistently higher sector timing measures compared with their market benchmark (the small cap growth index), the significance level of these sector timing measures is decreasing with measurement interval length and becomes insignificant for the semi-annual interval. Under the GT measure, the sector timing measures are higher for monthly intervals but lower for quarterly and semi-annual intervals when compared to their market index counterpart. Value funds and balanced funds generally have poor sector timing abilities. According to “the fundamental law of active management” (Grinold, 1989), the value added by a specific investment strategy depends on the manager's skill and breadth. The investment manager's skill measures the correlation between the prediction and the outcome, which captures the degree of a manager's forecast accuracy. Breadth refers to the number of independent decisions a manager makes.19 Hence, if a manager possesses some sector forecasting skills, the more frequent he makes sector timing investments, the greater the

19

The number of active decisions is a function of both the number of securities in a manager’s investment universe and the frequency of decision-making. For instance, if a manager makes buy/sell decisions on 150 stocks a year, assuming each decision is independent of the others, we can say this manager has a breadth of 150. Also, an investment strategy that is rebalanced on a monthly basis has breadth three times that as rebalancing on a quarterly basis.

19

contribution he makes to the active management of his funds. Observations on aggressive growth fund managers are consistent with this argument, as these fund managers tend to achieve better sector timing measures over shorter adjustment intervals. 5.5. The value of sector timing skill/activity

The value of sector timing skill/activity is measured by employing a variation of the Black-Scholes option pricing model to value market timing investment proposed by Merton (1981). Merton shows that the pattern of returns from successful market timing has an isomorphic correspondence to the pattern of returns from following certain option investment strategies. This can be easily applied to sector timing in a particular sector. The value of sector timing per forecast interval is given by Equation 9:

1 mt = 2Φ[ σ γ ] − 1 2

(9)

where mt is the value associated with an individual fund manager’s sector timing decision, σ is the annualised standard deviation of the sector, γ is the length of forecast interval, and Φ ( ) is the cumulative normal density function. Similarly, the value of sector timing per year is given by Equation 10: 1 mt n = 2 n Φ n [ σ γ ] − 1 2

(10)

where mtn is the value associated with sector timing decisions by a fund manager over a year and n=1/ γ , the number of forecast periods (portfolio adjustment periods) contained in one year, indicating how often the fund manager makes sector timing decisions. Both option values in equation 9 and 10 are expressed as percentage of gross investment in the fund. Table 12 shows the option value associated with sector timing skill in a particular sector, based on the assumed timing interval. In panel A, the higher the volatility of the sector, the larger the value associated with the sector timing decisions in that sector. Also the option value increases as the timing interval increases. Panel B shows that the shorter the timing interval, the more often the manager invests in the market, the larger the option value associated with sector timing decisions in a year. To calculate the value associated with the sector timing activities of a fund, I assume the fund manager holds a bundle of options on eight sectors, with weights equal to the

20

fund’s average sector exposures (as showed in Table 7). The value of sector timing per forecast interval for a particular fund is given by Equation 11. mt F = β1 mt1 + β 2 mt 2 + ... + β 8 mt 8

(11)

Where mt i is the value of sector timing per forecast interval for sector i, β i is the fund’s sector exposure to sector i. Also the value of sector timing per year for a particular fund is given by Equation 12: mt nF = β 1 mt n1 + β 2 mt n 2 + ... + β 8 mt n8

(12)

Where mt ni is the value of sector timing per year for sector i. Table 13 shows that the value associated with the fund’s sector timing activities is larger for aggressive growth and growth funds, when compared to value funds. This is due to the fact that growth funds have consistent higher sector exposures to sectors with higher volatility (such as technology and communication). Although the above estimations are based on perfect sector timing and it is unrealistic for fund managers to successfully time the sectors continuously, it is still arguable that the incentives are higher for growth funds to play with sector timing, since the value associated with these activities is larger. An uninformed mutual fund manager will pick a portfolio with higher risk, gambling on a lucky outcome.20

6. Discussion 6.1. Trading behaviour

It could be argued that different fund managers exhibit different trading behaviour, since they are subject to different investment objectives and constraints placed on them. 21 Some fund managers may trade frequently in an attempt to capitalize on market inefficiencies or investment opportunities identified by them. Other managers may believe the market is efficient and do not try to time the market or sub-section of the market (such as sectors). These managers trade less frequently than the others. To 20

Carhart (1997) shows that the funds in the top decile differ substantially each year, with more than 80 percent annual turnover in the composition. In addition, last year’s winners frequently become next year’s losers and vice versa, which is consistent with gambling behaviour by mutual funds.

21

Grinold (1989) suggests that the value added by active portfolio management is attributed to two aspects: the breadth of the strategy and the skill in forecasting exceptional returns. The breadth of the strategy represents how often you play, that is, the number of bets you make based on superior information. The skill in forecasting exceptional returns relates to the information coefficient, that is, the correlation between forecast and realized exceptional returns, which actually demonstrates how well you play.

21

identify the sector timing abilities of these managers, one needs to apply different portfolio adjustment intervals to them. The adjustment interval can be quarterly, monthly, semi-monthly, weekly or even daily, depending on the trading frequency of the fund managers. The portfolio turnover ratio might serve as a potential measure for the trading frequency.22 My experience has been that the sector coefficients based on longer term adjustment interval are more stable than those based on shorter term; moreover, shorter term adjustment interval would result in high-turnover strategies in portfolio construction, which is not realistic due to the high transaction cost incurred. 6.2. Trading costs

Similar to other investment strategies, the profits obtained from sector timing does not come without expenses. Indeed, it may incur significant trading costs, depending on the trading volume, trading frequency and prevalent market conditions. Moreover, the trading costs are quite different across industries. For example, the total estimated transaction costs for purchasing a €100m MSCI Europe sector portfolio in May 2002 ranges from 20 bp for Financials to 117 bp for Information Technology. 23 Since different subgroups of funds exhibit different sector exposures, different trading costs associated with different sectors may place different constraints on different subgroups of fund managers’ sector timing activities. 6.3. Fund flow

Edelen (1999) suggests that fund managers provide a great deal of liquidity to investors and thus engage in a material volume of uninformed, liquidity-motivated trading. This could have an adverse effect on fund returns. During extreme market conditions, such as those volatile market behaviours observed in bear markets, investor flows will be associated with negative market timing in fund returns. That is, investors pulling out their investments from the funds could force the managers to sell the stocks at the wrong time; even though they anticipate that some sectors are going to recover. Thus, assessing fund managers’ sector-timing ability without considering investor fund flow can result in negatively biased inferences. This could partially

22

Value funds, which tend to buy stocks the manager feels are out-of-favour and under-priced, generally have lower turnover than aggressive-growth funds. For example, the Van Wagoner Emerging Growth fund, a well-known high-growth fund, has annual turnover of 353%, compared with 13% for the value-loving Oakmark fund. 23

The results are obtained from the ABN-AMRO report entitled ‘European Equity Liquidity Monitor – May 2002’. One might expect that the trading costs are also different across industry in the US market.

22

explain the fact that fund managers exhibit negative sector timing abilities during the sample period. This needs to be addressed in the future research.

7. Conclusion This study explicitly investigates the sector timing abilities of actively managed mutual funds. By implementing a multi-factor model extension to Sharpe (1992), the time series of sector risk exposures are derived for a large set of funds. Based on these fund sector risk exposures, sector timing measures are calculated to investigate whether actively managed funds or sub-groups of funds poccess significant sector timing abilities. Generally, mutual funds as a whole exhibit negative sector timing abilities during the sample period of 1997.4 – 2002.7. Different fund sectors possess different magnitudes of negative timing measures. Aggressive growth funds tend to have the smallest magnitude of negative timing measures (or the largest magnitude of positive timing measures), which suggests that these funds do best in terms of sector timing compared with other funds, especially when the portfolio adjustment interval is on a monthly adjustment basis (i.e., actively traded funds). After controlling for market downturn conditions, all funds are shown to demonstrate sector timing abilities to some extent. Aggressive growth funds again are the best among all funds, especially with monthly and quarterly adjustment intervals. However, value and balanced funds seem to exhibit poor sector timing abilities. These findings naturally lead us to ask questions about the reasons why different fund groups exhibit different sector timing abilities. Some possible explanations are presented here. First, sector investing, by its very nature, represents an effort to shift portfolio holdings into a sector that is going to have a run-up, which can be viewed as a kind of momentum investment strategy. 24 Wermers (1997) finds evidence that actively managed funds actively use momentum strategies to generate performance persistence. Moskowitz and Grinblatt (1999) argue that momentum is significantly influenced by industry effects. According to their findings, the investor’s willingness to be exposed to industry risks is a key determinant for the decision whether to invest in a momentum strategy. Most aggressive growth fund managers are more likely to 24

It should be noted that, in a strict sense, there are still differences between sector investing and momentum investing. Sector investors try to catch potential rising trends in the market, while momentum investors follow existing trends in the market. However, as long as such a trend has started, they are indistinguishable from each other.

23

implement momentum strategies, due to their better abilities and higher confidence in dealing with a specific subset of stocks in certain industries rather than others. Therefore, these managers exhibit better sector timing abilities than others. Second, as I demonstrate earlier, different groups of funds exhibit different weights on the sector indices. Aggressive growth funds show heavy weights on healthcare and technology sectors, while value funds show heavy weights on finance and utilities. In order to successfully time a specific sector, the sector should exhibit some sort of market inefficiency that could be exploited. The level of market inefficiency directly relates to the magnitude of information asymmetry in the sector. Generally, the magnitude of information asymmetry in healthcare and technology sectors is higher than that in more traditional sectors such as finance and utilities. Moreover, as sector timing can run up trading costs, the added value of a sector timing strategy needs to be able to offset the increasing trading costs associated with this strategy. Market timing strategy exhibits some sorts of option-like payoff;25 hence, the value of the strategy depends on the volatility of the underlying market return. The higher the volatility of return, the higher the value of the strategy. Similarly, the value of a sector timing strategy depends on the volatility of return in the target sectors. As shown in Table 4 and Figure 8, the return volatility in technology and healthcare sectors is higher than that in the traditional sectors. Therefore, a sector timing strategy is much more valuable for aggressive growth fund managers who are primarily investing in hightech sectors. Finally, the portfolio holdings of aggressive growth funds tend to concentrate on a relatively small set of stocks, since over-diversification is probably the greatest enemy of portfolio outperformance and the impact of a good idea is negligible in a well diversified portfolio. As a result, these fund managers are in a better position to demonstrate their sector timing abilities (or bets) than other managers who have more diversified portfolios.

25

See Henriksson and Merton (1981)

24

Table 1 Correlation and Standard Deviations for Moody’s Sector Equity Mutual Fund Indices SECTOR

STDEV

Real

Utilities

Estate

Natural

Health

Technology

Metals

Resources

ations

Real Estate

3.80%

Utilities

3.76%

0.47**

6.37%

0.44**

0.56**

Health

5.41%

0.29*

0.53**

0.33**

Technology

10.02%

0.16

0.53**

0.39**

0.73**

Metals

10.13%

0.18

0.12

0.50**

0.10

0.15

5.60%

0.31*

0.67**

0.49**

0.67**

0.87**

0.21

4.59%

0.55**

0.72**

0.57**

0.43**

0.42**

0.11

Natural Resources

Communicat ions Finance

Communic

0.61**

Note: Moody's Equity Mutual Fund Indices track the daily average total return performance of equity oriented mutual funds that are classified into 8 sectors, or peer groups. Funds are objectively classified on the basis of past weekly returns over a period of three years, using statistical factor analysis and cluster analysis. ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).

25

Table 2 Moody’s Managed Fund Index Sector Exposures for 60 Months (From May, 1997 to Jul, 2002) Panel A: Estimated sector weights Index

Real Estate

Utilities

Natural Resource s

Health

Technolog y

Metals

Communi cations

Finance

R2

Value Small Cap. Value Small Cap. Growth & Income Moderate Growth Growth Aggressiv e Growth Balanced Medium Cap.

4.16% 38.56%***

19.51%** 0.00%

12.06%*** 22.84%***

0.00% 0.00%

2.27% 15.86%***

0.00% 0.00%

6.60% 0.90%

55.40%*** 21.83%**

90.8% 81.0%

21.66%***

0.00%

16.44%**

7.81%

41.55%***

0.00%

12.54%

0.00%

89.7%

0.00%

18.82%**

0.00%

0.00%

21.65%***

0.00%

12.10%*

47.43%***

91.1%

0.00%

11.42%

0.00%

0.00%

42.51%***

0.00%

11.38%*

34.70%***

95.3%

6.71%* 1.26%

3.10% 0.00%

2.18% 3.53%

5.13%* 16.00%***

42.04%*** 72.05%***

0.00% 1.63%

22.65%*** 5.52%

18.20%*** 0.00%

97.2% 95.1%

14.70%** 14.71%***

34.53%*** 2.68%

0.00% 13.94%***

5.20% 0.00%

9.16%** 13.67%***

2.80% 0.00%

10.38% 8.97%

23.23%*** 46.04%***

73.8% 98.3%

Panel B: 95% Confidence intervals for sector weights Index

Real Estate

Utilities

Natural Resource s

Health

Technolog y

Metals

Communi cations

Finance

Value Small Cap. Value Small Cap. Growth & Income Moderate Growth Growth Aggressiv e Growth Balanced Medium Cap.

[0.00 – 0.13]

[0.06 -0.33]

[0.05 – 0.19]

[0.00 – 0.00]

[0.00 – 0.08]

[0.00 – 0.00]

[0.00 – 0.19]

[0.46 -0.65]

[0.24 – 0.53]

[0.00 – 0.00]

[0.11 – 0.35]

[0.00 – 0.00]

[0.06 – 0.26]

[0.00 – 0.00]

[0.00 – 0.22]

[0.06 – 0.38]

[0.07 – 0.36]

[0.00 – 0.00]

[0.04 – 0.29]

[0.00 – 0.19]

[0.31 – 0.52]

[0.00 – 0.00]

[0.00 – 0.34]

[0.00 – 0.00]

[0.00 – 0.00]

[0.00 – 0.35]

[0.00 – 0.00]

[0.00 – 0.00]

[0.15 – 0.29]

[0.00 – 0.00]

[0.00 – 0.27]

[0.36 – 0.59]

[0.00 – 0.00]

[0.00 – 0.26]

[0.00 – 0.00]

[0.00 – 0.00]

[0.36 – 0.49]

[0.00 – 0.00]

[0.00 – 0.25]

[0.24 – 0.45]

[0.00 – 0.14]

[0.00 – 0.15]

[0.00 – 0.09]

[0.00 – 0.11]

[0.37 – 0.47]

[0.00 – 0.00]

[0.12 – 0.34]

[0.10 – 0.27]

[0.00 – 0.15]

[0.00 – 0.00]

[0.00 – 0.16]

[0.05 – 0.27]

[0.62 – 0.82]

[0.00 – 0.07]

[0.00 – 0.26]

[0.00 – 0.00]

[0.04 – 0.25]

[0.17 – 0.52]

[0.00 – 0.00]

[0.00 – 0.14]

[0.02 – 0.17]

[0.00 – 0.07]

[0.00 – 0.26]

[0.11 – 0.35]

[0.05 – 0.24]

[0.00 – 0.18]

[0.06 – 0.22]

[0.00 – 0.00]

[0.07 – 0.20]

[0.00 – 0.00]

[0.00 – 0.23]

[0.36 – 0.57]

Notes: This table presents the parameter estimates of the Sharpe return-based model for 9 Moody’s Managed Fund Indexes. In panel A estimated sector weights are given. Each row deals with one particular index, where the elements in columns 2 to 8 report the estimated sector weights. Panel B reports the 95% confidence intervals for all estimated sector weights. *** Significantly different from zero at the 1% level ** Significantly different from zero at the 5% level * Significantly different from zero at the 10% level

26

Table 3 Comparison between Sector Exposures and Actual Sector Weightings

Sectors

Sector Risk Exposures (over past 36 months before July, 2002)

Sector Weightings (November, 2002, disclosed from Morningstar)

Financials

48.4%

19.53%

Technology

20.4%

21.76%

Communication

14.9%

17.36%

Natural Resources

0.0%

11.28%

Utilities

10.0%

8.78%

Health

0.0%

13.22%

Real Estate

0.0%

0.0%

Metals

6.4%

0.0%

Table 4 Sectors Performance during 1997.4 to 2002.7

Year 1997 1998 1999 2000 2001 2002 All Period

Real Estate 17.84% -16.61% -1.78% 28.89% 13.70% 6.40% 0.32%

Utilities 25.80% 16.38% 8.72% 15.99% -20.81% -22.78% -0.18%

Natural Resources 8.94% -23.05% 26.39% 25.26% -8.47% -3.89% 0.00%

Health

Technology

Metals

Communications

Finance

25.15% 24.53% 16.85% 55.58% -8.04% -25.87% 0.80%

20.84% 36.98% 103.99% -1.77% -25.40% -33.28% 0.82%

-35.29% -9.31% 18.37% -18.81% 19.82% 40.34% -0.08%

32.81% 29.52% 55.01% -10.26% -11.10% -10.43% 0.75%

33.63% 9.67% 1.76% 23.97% -3.58% -6.39% 0.42%

Note: For calculation purpose, the first month (1997.4) is excluded.

27

Table 5

Fund Sector Exposures Differences between Winners and Losers for 2000-2002

SECTORS Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

SECTORS Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

SECTORS Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

Sector Returns 28.9% 16.0% 25.3% 55.6% -1.8% -18.8% -10.3% 24.0%

Sector Returns 13.7% -20.8% -8.5% -8.0% -25.4% 19.8% -11.1% -3.6%

Sector Returns 6.4% -22.8% -3.9% -25.9% -33.3% 40.3% -10.4% -6.4%

Winners Losers

2000 Sector Exposures Winners Losers Differences 5% 1% 4%** 6% 1% 5%** 25% 3% 4% 3% 0% 53%

8% 19% 43% 8% 6% 14%

17%** -16%** -38%** -5%** -6%** 40%**

2001 Sector Exposures Winners Losers Differences 10% 0% 10%** 12% 2% 10%** 14% 3% 8% 4% 6% 43%

4% 17% 57% 4% 7% 9%

10%** -14%** -49%** 1%** -2%** 34%**

2002 Sector Exposures Winners Losers Differences 12% 0% 12%** 18% 4% 15%** 5% 5% 8% 7% 15% 30%

1% 15% 59% 4% 2% 16%

4%** -9%** -51%** 3%** 13%** 14%**

Top-10% Bottom-10%

** Significantly different from zero at the 5% level

28

Table 6

All funds Aggressive Growth Growth Growth & Income Value Balanced

Mutual Fund Monthly Excess Returns, 1997.4 – 2002.7

1997 1.20%

1998 0.45%

1999 0.91%

2000 -1.52%

2001 -1.50%

2002 -3.17%

6-Year -0.52%

Std Dev 6.70%

2.24%

0.24%

3.12%

-1.94%

-1.90%

-4.09%

-0.26%

9.62%

1.33%

0.98%

1.31%

-2.29%

-2.03%

-3.72%

-0.63%

7.47%

0.91%

0.41%

0.08%

-0.89%

-1.14%

-2.85%

-0.50%

5.16%

0.80% 0.46%

-0.26% 0.24%

-0.63% 0.14%

-0.30% -1.07%

-0.79% -0.84%

-2.55% -1.76%

-0.56% -0.43%

5.71% 3.96%

Table 7 Mutual Fund Average Sector Risk Exposures, 1997.4-2002.7

Number of Funds Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

All Funds

Aggressive Growth

Growth

Value

Balanced

193 0.77% 2.12% 5.15% 5.11% 7.37%

Growth & Income 50 3.23% 3.71% 13.89% 5.75% 12.22%

Mean Stdev Mean Stdev Mean

484 2.23% 2.67% 9.30% 5.50% 9.07%

73 2.11% 2.72% 1.15% 3.22% 3.71%

95 2.90% 2.54% 7.71% 4.47% 17.37%

69 4.71% 2.50% 28.23% 4.22% 5.75%

Stdev Mean Stdev Mean Stdev Mean Stdev Mean Stdev Mean Stdev

3.85% 9.97% 2.76% 21.84% 4.70% 5.36% 1.19% 10.57% 6.82% 31.68% 4.77%

3.40% 21.16% 2.07% 56.84% 3.02% 7.47% 1.28% 4.60% 5.77% 2.96% 3.46%

3.98% 11.77% 2.83% 28.70% 4.90% 5.42% 1.20% 15.15% 7.01% 25.66% 4.97%

2.45% 2.50% 1.84% 5.57% 2.32% 4.15% 0.77% 9.98% 3.87% 48.47% 3.97%

4.33% 1.41% 1.79% 2.67% 2.11% 3.68% 1.00% 2.77% 4.23% 61.50% 3.77%

2.48% 10.32% 2.00% 3.95% 2.27% 6.13% 0.77% 15.32% 5.17% 25.59% 3.61%

29

Table 8 Mutual Fund Sector Timing Measures, 1997.4-2002.7

Panel A

484

k=1 -0.16%***

ST measure Interval k=3 -0.13%***

k=6 -0.25%***

Number

All Funds Aggressive Growth Growth Growth & Income Value Balanced

73

0.03%**

0.31%***

0.62%***

193 50 95 69

-0.13%*** -0.25%*** -0.28%*** -0.22%***

-0.05%* -0.38%*** -0.47%*** -0.16%***

-0.09%* -0.76%*** -0.94%*** -0.32%***

Panel B

Number 484

k=1 -0.09%***

GT measure Interval k=3 -0.13%***

k=6 -0.07%***

73

-0.03%***

-0.13%***

-0.02%

193 50 95 69

-0.12%*** -0.11%*** -0.06%*** -0.10%***

-0.15%*** -0.13%*** -0.10%*** -0.10%***

-0.06%*** -0.13%*** -0.04% -0.15%***

All Funds Aggressive Growth Growth Growth & Income Value Balanced

*** Significantly different from zero at the 1% level ** Significantly different from zero at the 5% level * Significantly different from zero at the 10% level

Table 9 S&P/Barra Market Indexes Index Name S&P/Barra 500 S&P/Barra 500 Value S&P/Barra 500 Growth S&P/Barra MidCap 400 S&P/Barra MidCap 400 Value S&P/Barra MidCap 400 Growth S&P/Barra SmallCap 600 S&P/Barra SmallCap 600 Value S&P/Barra SmallCap 600 Growth

30

Table 10 Barra/S&P500 Market Index Timing Measures, 1997.4-2002.7

Panel A

S&P 500 S&P 500 Growth S&P 500 Value S&P Median 400 S&P Median 400 Growth S&P Median 400 Value S&P Small Cap. 600 S&P Small Cap. 600 Growth S&P Small Cap. 600 Value

k=1 -0.16% -0.08% -0.28% -0.20% -0.15% -0.17% -0.16% -0.15% -0.19%

Panel B

S&P 500 S&P 500 Growth S&P 500 Value S&P Median 400 S&P Median 400 Growth S&P Median 400 Value S&P Small Cap. 600 S&P Small Cap. 600 Growth S&P Small Cap. 600 Value

k=1 -0.11%* -0.09% -0.03% 0.04% -0.04% 0.13% -0.13% -0.11% -0.06%

ST measure Interval k=3 -0.09% 0.31% -0.53% -0.14% 0.08% -0.52% -0.14% 0.08% -0.19%

GT measure Interval k=3 -0.15% -0.24%** -0.02% -0.11% -0.08% 0.02% 0.09% 0.08% -0.13%

k=6 -0.17% 0.63% -1.07% -0.27% 0.15% -1.05% -0.28% 0.16% -0.38%

k=6 0.09% -0.07% 0.14% 0.55% 0.55% 0.30% 0.58% 0.32% 0.68%

*** Significantly different from zero at the 1% level ** Significantly different from zero at the 5% level * Significantly different from zero at the 10% level

31

Table 11 Comparing Sector Timing Measure between Mutual Funds and Market Indices

All Funds S&P 500 Differences Aggressive Growth Small Cap. Growth Differences Growth S&-P 500 Growth Differences

k=1 -0.16% -0.16% -0.01%

ST measure Interval k=3 -0.13% -0.09% -0.04%***

k=1 -0.09% -0.11% 0.02%***

GT measure Interval k=3 -0.13% -0.15% 0.02%***

k=6 -0.25% -0.17% -0.08%***

k=6 -0.07% 0.09% -0.16%***

0.03%

0.31%

0.62%

-0.03%

-0.13%

-0.02%

-0.15%

0.08%

0.16%

-0.11%

0.08%

0.32%

0.18%***

0.23%*

0.46%

0.08%***

-0.21%***

-0.34%***

-0.13% -0.08%

-0.05% 0.31%

-0.09% 0.63%

-0.12% -0.09%

-0.15% -0.24%

-0.06 -0.07%

-0.05%***

-0.36%***

-0.72%***

-0.03%***

0.09%***

0.01%***

-0.25%

-0.38%

-0.76%

-0.11%

-0.13%

-0.13%

Growth & Income Median Cap. 400 Differences

-0.20%

-0.14%

-0.27%

0.04%

-0.11%

0.55%

-0.05%***

-0.24%***

-0.49%***

-0.15%***

-0.02%***

-0.68%***

Value S&P 500 Value Differences

-0.28% -0.28% -0.01%

-0.47% -0.53% 0.06%

-0.94% -1.07% 0.13%

-0.28% -0.03% -0.25%***

-0.10% -0.02% -0.08%***

-0.04% 0.14% -0.18%***

Balanced Median Cap. 400 Growth Differences

-0.22% -0.15%

-0.16% 0.08%

-0.32% 0.15%

-0.10% -0.04%

-0.10% -0.08%

-0.15% 0.55%

-0.07%***

-0.24%***

-0.47%***

-0.06%***

-0.18%***

-0.70%***

-0.22% -0.17%

-0.16% -0.52%

-0.32% -1.05%

-0.10% 0.13%

-0.10% 0.02%

-0.15% 0.30%

-0.05%**

0.36%***

0.73%***

-0.23%***

-0.12%***

-0.45%***

Balanced Median Cap. 400 Value Differences

*** Significantly different from zero at the 1% level ** Significantly different from zero at the 5% level * Significantly different from zero at the 10% level

32

Table 12 Option Value Associated with Sector Timing Skill for Sectors Panel A: Option value per forecast interval for eight sectors Natural Health Technology Metals Resources Annual Standard Deviation of the Sector Return

Real Estate

Utilities

Length of Timing Interval

13.16%

13.03%

22.07%

18.74%

34.71%

1 Month 3 Months 6 Months

1.52% 2.63% 3.71%

1.50% 2.60% 3.67%

2.54% 4.40% 6.22%

2.16% 3.74% 5.28%

4.00% 6.92% 9.77%

Communications

Finance

35.09%

19.40%

15.90%

4.04% 6.99% 9.87%

2.23% 3.87% 5.47%

1.83% 3.17% 4.48%

Communications

Finance

Panel B: Annual option value for eight sectors

Length of Timing Interval 1 Month 3 Months 6 Months

Natural Health Technology Metals Resources Annual Standard Deviation of the Sector Return

Real Estate

Utilities

13.16%

13.03%

22.07%

18.74%

34.71%

35.09%

19.40%

15.90%

19.79% 10.92% 7.56%

19.56% 10.80% 7.48%

35.13% 18.79% 12.82%

29.20% 15.81% 10.84%

60.02% 30.66% 20.49%

60.84% 31.03% 20.72%

30.36% 16.39% 11.24%

24.33% 13.30% 9.17%

Note: The option value is expressed as a percentage of gross investment

Table 13 Option Value Associated with Sector Timing Skill for Sample Funds Panel A: Option value per forecast interval for sample funds Length of Timing Interval

All Funds

Aggressive Growth

Growth

Growth & Income

Value

Balanced

1 Month 3 Months 6 Months

2.52% 4.37% 6.17%

3.33% 5.76% 8.14%

2.70% 4.68% 6.62%

2.12% 3.67% 5.19%

2.07% 3.59% 5.08%

2.08% 3.60% 5.09%

Panel B: Annual option value for sample funds

Length of Timing Interval

All Funds

Aggressive Growth

Growth

Growth & Income

Value

Balanced

1 Month 3 Months 6 Months

35.64% 18.83% 12.78%

48.90% 25.29% 17.00%

38.56% 20.27% 13.72%

29.06% 15.62% 10.69%

28.23% 15.23% 10.43%

28.46% 15.30% 10.47%

Note: The option value is expressed as a percentage of gross investment

33

Figure 1 Moody’s Managed Fund Index Sector Exposures for 60 Months (From 1997 to 2002)

100% 90% 80% 70%

Finance Communications

60%

Metals Technology

50%

Health Natural Resources

40%

Utilities Real Estate

30% 20% 10% 0% Value

Small Cap. Small Cap. Growth & Moderate Value Value Income Growth

Growth

Aggressive Balanced Growth

Median Cap.

34

Figure 2: Sector Risk Exposures for Aggressive Growth Fund Index for 60 Months (from 1997 to 2002)

80.00%

70.00%

60.00%

50.00%

40.00%

30.00%

Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

20.00%

10.00%

0.00%

Figure 3: Sector Risk Exposures for Value Fund Index for 60 Months (from 1997 to 2002) 60.00%

50.00%

40.00%

30.00%

20.00%

Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

10.00%

0.00%

35

Figure 4: Fund Sector Exposures for Aggressive Growth Fund Index (from 2000 to 2002) 100%

90%

80%

70% Finance Communications Metals Technology Health Natural Resources Utilities Real Estate

60%

50%

40%

30%

20%

10%

0% 1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

Figure 5: Fund Sector Exposures for Value Fund Index (from 2000 to 2002) 100%

90%

80%

70% Finance Communications Metals Technology Health Natural Resources Utilities Real Estate

60%

50%

40%

30%

20%

10%

0% 1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

36

Figure 6: Risk Exposures for Fidelity Select Biotechnology Portfolio for 60 Months (from 1997 to 2002) 80.0%

70.0%

60.0%

Real Estate

50.0%

Utilities Natural Resources Health

40.0%

Technology Metals Communications

30.0%

Financials

20.0%

10.0%

0.0%

Figure 7: Risk Exposures for Fidelity Select Utility Portfolio for 60 Months (from 1997 to 2002) 45.0%

40.0%

35.0%

30.0% Real Estate Utilities 25.0%

Natural Resources Health Technology

20.0%

Metals Communications Financials

15.0%

10.0%

5.0%

0.0%

37

Figure 8 Sector Fund Return Index, 1997 – 2002 500.00

450.00

400.00

350.00 Real Estate Utilities Natural Resources Health Technology Metals Communications Finance

300.00

250.00

200.00

150.00

100.00

50.00

Sep-02

Jun-02

Mar-02

Sep-01

Dec-01

Jun-01

Mar-01

Sep-00

Dec-00

Jun-00

Mar-00

Sep-99

Dec-99

Jun-99

Mar-99

Dec-98

Jun-98

Sep-98

Dec-97

Mar-98

Jun-97

Sep-97

Mar-97

0.00

38

Reference Baca, S.P. and Brian L. Garbe (Sep/Oct2000) The Rise of Sector Effects in Major Equity Markets, Financial Analysts Journal 56, 34-41. Black, A.J., Roger Buckland and Patricia Fraser (2001)

Efficient portfolio

diversification: Changing UK stock market sector and sub-sector volatilities,19672000, Aberdeen Papers in Accountancy, Finance & Management, Working Paper. Cavaglia, S., Christopher Brightman and Michael Akel (2000)

The Increasing

Importance of Industry Factors, Financial Analysts Journal 56, 41-54.(Abstract) Daniel, K., Mark Grinblatt, Sheridam Titman and Russ Wermers (1997) Measuring mutual fund performance with characteristic based benchmarks,

The Journal of

Finance, 52( 3), 1035-1058. Dellva, Wilfred L., Andrea L. DeMaskey, Collen A. Smith (2001) Selectivity and market timing performance of Fidelity sector mutual funds, The Financial Review, 36, 39-54. Dor, A.B. and Ravi Jagannathan (2002) Understanding mutual fund and hedge fund styles using return based style analysis, Working Paper 9111 Edelen, R.M. (1999)

Investor flows and the assessed performance of open-end

mutual funds, Journal of Financial Economics , 53(3), 439-466. Farrell, J.L. (1975) Homogeneous stock groupings, Financial Analysts Journal 31, 50-57. Grinblatt, M. and Sheridan Titman (1993)

Performance measurement without

benchmarks: An examination of mutual fund returns, Journal of Business, 66(1 ), 4768. Grinold, R.C. (1989) The Fundamental Law of Active Management, Journal of Portfolio Management 15, 30-36. 39

Henriksson, R.D. and Robert C. Merton (1981) On market timing and investment performance II: Statistical procedures for evaluating forecasting skills, Journal of Business, 54(4), 513-534. Lobosco, A. and Dan DiBartolomeo (1997) Approximating the confidence intervals for sharpe style weights, Financial Analysts Journal 53, 80-85. Matallin-Saez, J.C. and Fernandez-Lzquierdo, A. (2001) Passive timing effect in portfolio management, Working Paper Moskowitz, T. J. and Grinblatt, M. (1999) Do industries explain momentum? Journal of Finance, 54(4), 1249-1290 Schwob, R. (2000) Style and style analysis from a practitioner's perspective: What is it and what does it mean for European Equity Investors?

Journal of Asset

Management, 1(1), 35-59. Sharpe, W.F. (1992)

Asset allocation: Management style and performance

measurement, Journal of Portfolio Management 18, 7-19. Targett, S. (2000) Time to redraw the investor's map, Financial Times Treynor, J.L. and Kay K. Mazuy (1966) Can mutual funds outguess the market? Harvard Business Review, 44(4), 131-136. Wermers, R. (1997) Momentum investment strategies of mutual funds, performance persistence, and survivorship bias, Working paper, Graduate School of Business and Administration, University of Colorado at Boulder, Boulder, Col.,

40