Hedge Fund, Mutual Fund, and Institutional Fund Conglomerates

Hedge Fund, Mutual Fund, and Institutional Fund Conglomerates: Risk and Return Choices for a Sophisticated Investor Janis Berzins Indiana University jaberzin@indiana.edu Crocker Liu Arizona State University Charles Trzcinka Indiana University December, 2006 Abstract We investigate investment choices that a sophisticated investor would face by analyzing a unique sample of investment conglomerates, which simultaneously offer investment products in mutual fund, hedge fund, and institutional investment industries. We control for backfill bias in our study and correct hedge fund returns for smoothing effects caused by stale prices. We compare performance of funds across industries by using Sharpe ratios and alphas from multifactor models. We find that for the whole sample institutional funds perform better than hedge funds, and hedge funds tie with mutual fund Sharpe ratios in regression settings with fixed conglomerate effects. Hedge funds have the highest gross returns, but the lowest seven factor alpha. Our study is the first to offer a comparative analysis across all three investment industries. JEL classification: G2, G1, L1, L2 Key words: Institutional funds, hedge funds, mutual funds, performance evaluation 1 I Introduction This study compares the investment performance of three types of delegated portfolios: mutual funds, hedge funds, and institutional funds. These portfolios differ fundamentally in the contract between the client and the decision maker. The systematic differences in these contracts are over cash flow rights, ownership claims, regulation and reporting practices. We investigate the impact of these differences on the risk and return of these portfolios. While the finance literature has studied mutual funds and hedge funds separately, there are few studies of institutional money management and no study has directly compared all three. Whether variation in organizational form matters is an important question for investors, regulators and financial economists. Investors must choose between selecting a contract and managing money themselves. Regulators are concerned with the sufficiency of contract regulations and whether unregulated contracts should be regulated. Financial economists are interested in how the form of a contract affects the risk-return relationship. Comparative studies suffer from the drawback that variation in the risk-return relationship may have more to do with differences in institutional arrangements that are unrelated to a given research hypothesis. For companies that offer these forms of portfolios, there are large differences in assets under management, centralization of information gathering, economies of scale, transactions costs and risk control. This gives rise to risk-return differences which can obscure a researcher’s focus on characteristics directly generated by the form of a portfolio. To mitigate this problem of unobserved control variables, we propose examining portfolios that are owned by the same organization. We call these organizations “Investment Conglomerates”. Our maintained (that is, untested) assumption is that a conglomerate has a degree of control over many of these variables. If investment conglomerates have the ability to centralize trading, centralize or partially centralize information gathering, spread legal costs over all portfolios being offered, monitor risk and issue consistent reports, then the differences that we observe in the risk- 2 return relationship across the form of a contract are more likely attributable to the contract’s form than to the organization. The literature has convincingly established that mutual funds, hedge funds and institutional funds differ substantially. (see Del Guercio and Tkac (2002), Lakonishok, Schleifer, and Vishny (1992), Cici, Scott, and Moussawia (2006), Ackermann, McEnally, and Ravenscraft (1999) among others). The mutual fund industry is the most regulated of these investment venues. A mutual fund has to report returns and assets on a regular basis; the use of derivatives, short selling, leverage, concentrated investments, and advertising is restricted, and fiduciary responsibility is policed and ordinarily restricted. Mutual funds must allow daily redemptions which means that they can only invest in liquid securities. The hedge fund industry is the least restricted of these contracts and is characterized by flexible investment strategies, a lack of government regulation, and large managerial investment. Reporting is not mandatory unless the hedge fund operates in the US and has grown to a certain size. Hedge fund information is very elusive and a few competing data sources provide voluntarily reported returns and fund characteristics (see data section for example). The clients of institutional money managers are largely pension funds and foundations. These clients have accounts that are larger than hedge fund accounts and over a thousand times larger than retail mutual funds. The securities are directly owned by clients instead of the fund. Government regulation covers the client but not the manager and the manager – client relationship is determined almost exclusively by private contracts. While institutional managers frequently have performance-based incentives, assetbased fees are typical and high watermark provisions and large incentive fees of (20 percent) are not common in the institutional fund industry. The large incentive fees and wide latitude in the choice of investment strategy make the earning potential for hedge fund managers the highest among the three industries. It is possible that the highest possible remuneration attracts the best talent. In one of the few comparative studies, Cici, Scott, and Moussawia (2006) argue that mutual fund managers highly value the 3 option to simultaneously manage mutual funds and hedge funds because of the high incentive compensation associated with hedge funds. Therefore, it may be possible that the best chance for an investor to find a talented manager who can “beat” the market is to invest in the hedge fund industry. On the other hand, the opaqueness created by looser regulation, and the consequent increases in monitoring costs, restrictions on asset movement (e.g., holdup periods, front and end loads) and unavailability of information creates an environment where superior performance may not be passed on to an investor but rather extracted in fees (e.g. Berk and Green (2002)). Our sample consists of 67 investment conglomerates between June 1993 and September 2005. These conglomerates control $2.28 billion in assets by the end of the sample period. Portfolios offered by the conglomerates span equity, fixed income, and derivative asset classes. To examine the risk-return differences we estimate Sharpe (1966) ratios and seven factor alphas. Sharpe ratios are influenced by two factors: return autocorrelations (Lo (2002) and Getmansky, Lo, and Makarov (2004)) and a manager’s ability to manage Sharpe ratios via certain derivative strategies (Goetzmann, Ingersol, Spiegel, and Welch (2002)). The first influence we mitigate by unsmoothing the hedge fund returns with the moving average process as in Getmansky et al. (2004). The second issue forming our empirical priors is that the hedge fund industry managers with the most latitude in derivative applications should produce the highest Sharpe ratios of the three industries. Surprisingly we document that hedge fund Sharpe ratios net of fees are similar to mutual fund Sharpe ratios, but smaller than institutional fund Sharpe ratios. The relationship between hedge fund Sharpe ratios and the mutual fund Sharpe ratios varies over time. To account for differences in portfolio exposure to various risk factors we use a multifactor model and estimate time-varying alphas. Similar models have been used by Hubner, Corhay, and Capocci (2005) and Agarwal and Naik (2000a). In the regression setting we find that institutional fund and mutual fund alphas dominate hedge fund alphas. 4 The paper proceeds as follows. Section 2 briefly reviews the relevant literature on crossmanaged fund industry comparisons. Section 3 presents the sample collection process and introduces descriptive statistics. Section 4 discusses biases in our sample. Section 5 presents hypotheses and methodology; Section 6 discusses results, and Section 7 concludes. II Relevant Literature While several recent studies suggest examining the interaction between fund complex and portfolio performance (see Zhao (2002), Gaspar, Massa, and Matos (2004), Guedj and Papastaikoudi (2003), Kempf and Ruenzi (2005), Gallaher, Kaniel, and Starks (2005)), few studies have conducted cross industry comparisons of managed portfolios. Del Guercio and Tkac (2002) examine the performance-flow relationship in a sample of mutual funds and fiduciary pension fund segments. They find differences in the relationship which they attribute to the type of clientele. Namely mutual fund clients are slow to withdraw money from poorly performing managers, but they chase winners. In contrast institutional clients withdraw money from poorly performing managers, but do not chase winners. Del Guercio et al. also document that institutional management clients use such risk adjustment methods as tracking error, while their retail mutual fund counterparts use less sophisticated methods in performance evaluation. Cici, Gibson, and Moussawi examine a sample of managers that manage side-byside mutual funds and institutional funds. They find evidence that side-by-side managers strategically transfer performance from mutual funds to hedge funds. Specifically Cici et al. demonstrate that mutual funds managed by these managers underperform their matched-sample mutual funds. The matched sample is selected from the rest of the mutual fund industry. Side-by-side mutual funds are allocated fewer underpriced IPOs than the rest of the mutual fund industry. 5 Ackermann, McEnally, and Ravenscraft (1999) examine a sample of hedge fund data from MAR and HRF databases and find that hedge funds as an industry consistently outperform mutual funds in terms of Sharpe ratios (1966), but do not beat standard market indices. Hedge funds are more volatile than both mutual funds and market indices. They attribute some of the higher performance to incentive fees, but find that incentive fees cannot account for increased risk. They also find that the net effect of the two opposing biases: termination and self selection are negligible. Ackermann and Ravenscraft (1998) suggest that regulatory differences between mutual and hedge fund industries lead to dramatic differences in how the two industries use lockup periods, illiquid securities, short selling, derivatives, leverage, and concentration. Mutual fund performance may suffer from these limitations. With the exception Cici et al., the aforementioned studies have not been able to control for fund complex. Ackermann et al. (1998) results may be influenced by the effect of return serial correlation on the risk and return measures. Our paper extends the current literature by offering a cross industry performance comparison within a multifactor risk adjustment setting while controlling for conglomerate effects on fund performance. It is the first study to examine portfolios across the three managed portfolio industries. 6 III Data Collection and Descriptive Statistics We identified investment conglomerates that offer investment products in mutual funds1, hedge funds, and in institutional money management industries where institutional funds are separately managed accounts. We first identified all hedge fund managers and selected those that offer both institutional funds and mutual funds between 1993 and 2005. This process resulted in a sample of 67 investment conglomerates offering 473 hedge funds, 813 institutional funds, and 1487 mutual funds. The subsequent sections describe the sample construction process and present sample descriptive statistics. Databases used in selection process The databases used in our sample collection include: for hedge funds- TASS/Lipper, CISDM, Barclays Hedge Fund Data, Global Fund Analysis (for returns and fund descriptions), Mobius Hedge Fund Panel Data2; for mutual funds- CRSP mutual fund database, Morningstar website, Morningstar CDs of various months between 1994 and 2005; for institutional data- a unique compilation of Mobius database quarterly survey panels; for addresses and legal company relationships: SEC Edgar Filings, Thompson Fund Database, websites of investment conglomerates, Lexis-Nexis, Dow Jones, and general internet searches. Matching Process We identified hedge fund families that are listed in TASS or Lipper CISDM databases. From this set we selected families that were also offered mutual funds and were listed on the Morningstar mutual fund website. Most of the companies were matched by name. We then checked the address match between hedge funds and mutual funds. Several companies did not have a precise name match as in their fund conglomerate mutual funds and hedge funds were offered by different subsidiaries. We included these companies in the sample. This step yielded 114 fund families. It is possible that some families have ceased to exist before 2005. Our initial 1 2 Institutional mutual funds are included in this industry. We cross-checked Mobius Hedge Fund returns against TASS and CISDM database returns and found them accurate. 7 procedure would not have captured these families making our sample subject to family-level survivorship bias. To mitigate this bias and capture families that have ceased to exist we searched for name matches between TASS, CISDM, Mobius hedge fund families and CRSP dead mutual fund families. First we collected a list of hedge fund names and mutual fund management company names for which no mutual funds have survived to the present in any of the families. For practical purposes we insisted that the CRSP family names should have been on record for at least two years with reported returns in one of those years. Then we proceeded with a mechanical match. Within the mechanical match we used the first 12 letters from CRSP family name to search the list of hedge fund names. We insisted that the match start no later than on the sixth letter in the hedge fund name. In the second step we scanned the matched result list for match candidates. In the third step we cross-checked addresses of Mutual fund and hedge fund families. This procedure added 14 families to our sample of families that have offered both mutual funds and hedge funds for a total of 132 families in the initial sample. The next step involved identifying families that offer institutional funds. In this step we used a unique survivorship bias free database of institutional managers. The institutional database is created from twelve years of Mobius institutional managers quarterly surveys (see BT 2005 for a more complete database description). 114 of 132 families had funds in all three industry databases3. Of the 114 families, 71 families listed returns for at least one fund in each industry. Of 71 families 67 families reported fees for at least one portfolio in each industry. Definition of a Family Name and Address Check Our initial list of hedge fund families was created from TASS and CISDM databases4. While the CISDM database has a family name 3 We manually checked all names of portfolios that investment conglomerates offer. This procedure eliminated the possibility that the same fund is entered in more than one industry. For example it appears that occasionally TASS lists some institutional portfolios, which should not be classified as hedge funds. Most often TASS does not provide returns for these funds. Also it is possible that the institutional database is listing some institutional mutual funds, which should be classified as mutual funds not institutional funds. 4 Mobius Hedge Fund database was used to supplement return series, but was not used in the initial sample creation. 8 identifier, the TASS database does not. Therefore, we most frequently used the following steps to create a family name. First we extracted the family name from the hedge fund itself. This step alone might lead to name errors so we also used the Global Fund Analysis database to search for the fund. This in turn yielded the name of the fund family, their address as well as the names of other hedge funds in the fund family. In general CISDM and TASS do not provide hedge fund legal addresses, while the Mobius hedge fund database provides addresses. Therefore we supplemented our hedge fund address searches with Lexis-Nexis, SEC filings, hedge fund web address, and general internet queries. The CRSP mutual fund database does not have a family identification number for the whole time period, but list the family names. Family names are subject to alternative spellings and typographic errors. Therefore, for each family name we created a list of alternative names. On the other hand, a mutual fund has a unique identifier in CRSP, the ICDI number. A particular fund may change its membership in a family either via acquisition, merger, or strategic alliance arrangements. Therefore a particular mutual fund may be affiliated with several management family names throughout its life. In the construction of our sample we used the following rule. If a change in name affiliation between ICDI and mutual fund family is just a result of a mutual fund family name change, we add this name to the mutual fund family name-ICDI affiliation list. If a fund family is merged into a larger complex, just the fund ICDI numbers from the initial family are included for the period after the merger. Finally, if a fund is taken over by an unaffiliated family, the period of this affiliation is not included in our sample. If a subsidiary of a large family offers funds in all three industries, we list it as a separate family (e.g. Mercury in Merrill Lynch). The CRSP database provides legal addresses for very few mutual funds. The institutional database has a unique family name identifier and follows name changes for families. It also lists family legal addresses. 9 Fund Level Filters and Joining of Assets Classes To recap, of 132 initial conglomerates 18 do not have institutional accounts5 and 4 do not have returns in the CRSP database. 34 conglomerates miss all HF returns. The link between industries for 5 conglomerates cannot be established with certainty. This leaves us with a starting sample of 71 total conglomerates, of which 67 had complete fee information. Next we applied the following filters to the initial dataset. We considered mutual and hedge funds that reported at least twelve monthly returns and institutional funds that reported at least eight quarterly returns. This eliminated 23 hedge funds, 220 institutional funds, and 77 mutual funds but no conglomerates from our sample. For Mobius we took funds that reported net assets at least once; this eliminated very few newly created funds. For hedge funds we eliminated the very few funds that never reported asset values if other hedge funds in the same complex reported asset values. We kept one small complex with one hedge fund and no hedge fund assets. Mutual funds with no assets were excluded because they most often were an additional share class to a fund that already reported one share class assets. We eliminated funds that did not report fees. We also did not include money market or municipal bond funds in our sample (CRSP codes MF, MG, MQ, MS, MT, MY). These funds are mostly tax exempt and passively managed, and we have no similar funds in the hedge fund or institutional fund sector. This eliminated 530 funds (we found that about one percent of institutional portfolios are municipal portfolios and excluded them from further analysis). We joined assets of mutual and hedge funds with multiple asset classes or return reports in multiple currencies. In combining the asset classes we used the following rules. We kept the longest return series. If two return series were of similar length we took class “A” or institutional shares to avoid load impact on returns. If a particular asset class had assets tens of times higher than all other return series, we kept the returns from the series with the highest assets. For mutual funds we joined regular and institutional share class assets. If an institutional fund was the only 5 3 of the 18 complexes reported institutional returns for less than 5 quarters. 10 asset class available and a fund with a similar name was reported by the institutional database as a “separate account“, we excluded this fund from the sample. Conversely, if an institutional database reported an institutional mutual fund, we excluded it from the institutional sample. This way the mutual fund industry contains both retail and institutional accounts. Sometimes the institutional database offers both equally and asset weighted returns. We chose equally weighted returns when available. If two asset classes reported time series of the same length, but just one had turnover indicated, we took the latter. We chose “common” over “investor”, and I over N asset classes. If a fund asset class with no load was offered we took that. Similarly we preferred to keep USD denominated to other currency denominated hedge fund returns. If assets of multiple currency funds were equal across funds, we did not add the assets. We viewed that in this case the total assets were already redundantly reported in all different share classes. We also preferred onshore to offshore funds. Again we retained the longest return time series. We maintain that the main difference between the onshore and offshore returns is the tax impact which is not the subject of our study. We kept funds from three industries that reported returns for at least a year. Equity Identifier Next we identify equity mutual hedge and institutional funds. For equity mutual funds we modify the selection method used in Cici, Gibson, and Moussawia (2006). We mark funds that have equity position between 50% and 105%, but do not belong to categories that are not identified as equity in the subsequent sentences6. For the funds that do not report equity position, we mark funds as equity if they belong to the following equity classes (ICDI_obj_cd code in CRSP): aggressive growth (AG), growth and income (GI), global equity funds (GE), international equity funds (IE), long term growth funds (LG), total return funds (TR), sector funds (SF)7. 6 7 In identifying funds’ equity percentage, or class membership, we take the last reported variable. Some sector funds include real estate funds. 11 Funds with missing entries were manually identified by examining the fund name. Among 1556 mutual funds 1093 are equity funds in our sample. We also marked and excluded index funds from our sample as they represent a passive investment strategy. Variable identifying index funds are available from 2003 only; we searched for “index” and “500” in names for the rest of funds. In the institutional database we marked either international equity or domestic equity portfolios as equity. Of 875 funds 529 are equity funds. For hedge funds we examined styles reported in CISDM, TASS, or Mobius Hedge Fund Databases. If styles suggested use of equity, we marked the fund as an equity fund. For example, CISDM equity funds have the following styles: sector, relative value multi strategy, merger arbitrage, equity market neutral, equity long/short, while in Mobius the strategies are long/short equity, small/micro cap, healthcare sector, long/short equity, market neutral equity. Of 485 hedge funds 237 are equity funds; this excludes funds of funds. Descriptive Statistics Figure 1 graphically presents return availability for each complex in the three industries. Our sample starts with 433 million in 1993 and reaches 2.28 billion in 2005 across all three industries (Table 1). In 2005 equity portfolios contribute 1.46 billion to assets, while the rest contribute 0.82 billion. Panel B and C also exhibit a shift away from equity during the internet burst time in 2000. The hedge fund industry is by far the smallest in our sample, while the institutional fund industry is the largest. The table two reports median returns for the sample. Surprisingly, returns are not statistically different between the three industries, if we calculate average for each portfolio and then report equal average cross sectional return (Table 2). The returns are different, if we compute an equally weighted cross sectional average first and then average the result over time. Hedge funds perform best relative to the other funds during the market downturn. This suggests a significant benefit of investing in the hedge fund industry. This result will need to be verified with regression analysis. 12 Conglomerates differ significantly in their commitment to the three industries. We compute the Herfindahl index to account for differences in commitment. We use the average fund assets for each portfolio to calculate the percentage of assets that a conglomerate has invested in each industry. The sum of the squared percentages forms the Herfindahl index. The index for a conglomerate with assets spread equally across all industries would be 3333, while the index for a conglomerate with all assets in one industry would be 10000. The Herfindahl index of the most concentrated quarter of conglomerates is 9741, while the index for the least concentrated quarter is 5026. The most (least) concentrated quartile conglomerates have 1% (34%) invested in hedge funds, 52% (18%) in institutional funds, and 35% (50%) in mutual funds (see Table 3 panel a). The total assets of the smallest quartile of conglomerates are 104.0 million, while the same number for the largest quartile is 13.4 billion, where the assets are accumulated by summing the average assets for each portfolio regardless of when it existed. 13 IV Dealing with Biases in Data Several studies have argued that hedge fund returns exhibit substantial serial correlation, where the main source of the correlation is the presence of illiquid assets in a portfolio (see Getmansky, Lo and Makarov (2004), Okunev and White (2003), and Jagannathan, Malakhov, and Novikov (2006)). Manager performance measurement could be biased if one does not account for serial correlation. Asness, Krail, and Liew (2001) suggest that style alphas adjusted for serial correlation are lower than unadjusted alphas. Jagannathan et al (2006) show that occurrence of positive alphas decreases in a sample after accounting for serial correlation. Lo (2002) shows that Sharpe ratios for auto-correlated returns could be upwards biased. Several methods have been suggested for dealing with serial correlation. Getmansky et al. use MA 2 process to correct for the smoothing bias. Jagannathan et al suggest that serial correlation is especially severe for hedge funds in specific sectors. We follow Getmansky et al. and unsmooth all hedge fund returns with MA2 process and consider the following model: X ti = θ 0iηti + θ1iηti−1 + θ 2iηti− 2 where ηti is unsmoothed, or true return i, at time t, X ti = Rti − μi is a demeaned return. We assume that ηti ~ N (o,σ i2 ) . The equation 1 is moving average process with lag 2. To (1) identify this system we impose the invariability constraint θ 0 + θ1 + θ 2 = 1 . This constraint has an economic meaning that the smoothing takes place over the most recent two periods. If Rti follows the MA(2) process, we can estimate ηti from (1) by maximum likelihood. Getmansky et al show that the concentrated likelihood function for equation 1 is: 14 ˆ L0 (θ ) = log[T −1 ∑ ( X t − X t ) 2 / rt −1 ] + T −1 ∑ log rt −1 t =1 t =1 T T (2) To estimate this process we use the SAS ARIMA procedure. To ensure invariability we divide the estimate MA factors by 1 + θ1 + θ 2 . This follows a well known statistical property of the MA process, where any non invertible process can be transformed into an invertible process by adjusting the parameters and variance. We estimated MA(2) process for all hedge fund returns that had at least 30 observations to mitigate the well known small sample problem of the estimated coefficients. Further, we proceeded with smoothing for those funds that exhibited at least a positive and significant first autocorrelation. Jagannathan et al document that just a few of 32 hedge fund styles in the Hedge Fund Research (HFR) database are susceptible to return smoothing; e.g. “convertible arbitrage”, “distressed securities”, and “emerging markets”. Getmansky et al. reported lowest first MA coefficients for European Equity, Fixed Income Directional, Convertible Fund (Long Only category), Event Driven, Nondirectional/Relative Value, Pure Emerging Market, and Fund of Funds in the TASS database. 514 hedge funds in our sample had sufficient return series for the estimation. Of these, 190 hedge funds had significant first lag autocorrelation, and 35 had both first and second lags significant. 180 funds yielded significant positive θ1 and 20 yielded significant positive θ1 and θ 2 . We unsmoothed returns for the 180 hedge funds by first unsmoothing the demeaned returns and then adding back the mean. The average MA2 parameters for the 180 hedge funds are as follows: θ 0 is 0.744, θ1 is 0.246, θ 2 is 0.01, while average σ 1 is 0.093 and σ 2 is 0.094. This means that just 74.4% for returns from a current month are directly reflected in the reported returns for these 180 funds. 15 Back-fill Bias Another bias in hedge fund data is back-fill bias. Funds frequently bring all their history with them when joining the sample. Funds can choose how much and what if any history to bring. It is possible that just the funds with a superior past history would choose to take their history with them. This would bias conclusions about a manager’s superior performance in previous years (see Ackerman, McEnally, and Ravenscraft (1999), Liang (2000), Fung and Hsieh (2000)). Several databases identify the date funds enter the sample. In this case the back-filled returns are deleted from the sample. Jagannathan et al. estimate the length of back-fill bias in the HFR database. They attempt to replicate the style indices reported by the HFR database. The style indices are constructed from hedge funds in the same investment style, but are not updated when a new fund enters the sample and brings in back-fill bias. By sequentially left truncating all database returns Jagannathan et al. examine how closely the style indices from truncated returns match the ones reported by the database. The two series converge at about 12 return truncation and max out at 25 returns. The difference in gain between 12 and 25 seems negligible. By truncating too long of a tail the bias of excluding early returns for managers with no back-filled returns is introduced. In our data we chose to truncate all return series by 12 returns. Outliers in data We tested for outliers in hedge fund returns by looking up returns that differ from the mean by more than 8 standard deviations. We found 4 suspects. For example, “GAM Money Markets Fund (BP)” reported a return of 904.48% in 1/31/1993, the NAV 16 increased ten fold in that month, while “GAM Money Markets Fund (DM)” reported a return of -94.11 12/31/1996, its assets decreased tenfold in that month. The other two were around negative 30% return and were related to the 1987 market crash. As a result the identified occurrences were not marked as data errors. 17 V Hypotheses and Methodology Principal agent relationships form between managers and clients. The portfolio manager is the agent who has more information about securities than the client who is the principal. There is a difference between the interests of the agent and the principal which gives rise to agency costs. Bhattacharya and Pfleiderer (1985) prove that if the preferences of managers can be observed then it is possible to design a contract that forces managers to reveal their information. If the information is known then a principal can write a contract that eliminates agency costs. In practice it is probably impossible for such a contract to be written. As a consequence, the investment business has evolved into three segments with common contract features that attempt to control agency costs for the clients in those segments. Ackermann et al. suggest (1999) that four mechanisms help align these interests: incentive contracts, ownership structure, market forces, and government. Our study examines the differences between the risk and return relationship that a wealthy investor can expect to earn by investing in one of the three portfolio types. Key differences between fund organizational form and performance implications. It is reasonable to expect that hedge funds would produce a superior risk-return relationship based on a gross return measure such as the Sharpe Ratio. Hedge fund managers may invest in a wider variety of assets than mutual and institutional fund managers and incentive schemes differ across industries. Stark (1987) analyzes the relationship between incentive contracts and portfolio manager investment decisions. Here the manager can choose portfolio risk level and level of commitment towards improving portfolio performance. Stark evaluates manager choice between symmetric and bonus reward systems. Symmetric plans penalize managers if they do not reach the goal, but align risk preferences of investors and managers best. Bonus plans encourage managers to select more risk and invest fewer resources in production of superior returns. Hedge funds and to a lesser extent institutional funds offer performance related, or bonus fees, while all three industries offer a symmetric reward system. Stark also 18 suggests that bonus plans enhance managerial effort with respect to no plans. Carpenter (1998) develops a model that examines the link between risk and bonus incentive fees and concludes that increase in incentive fees decreases managerial risk taking. It is conceivable that an institutional funds, and to a greater extent hedge funds attract better managers though this mechanism and produce superior returns with possible higher volatility. Ackermann et al. (1999) find empirical evidence linking performance fees to improved hedge fund performance, but not to increased risk. A manager’s personal investment in funds, fund size, and age may act to decrease the differences in performance between portfolio types. Ownership mechanisms discourages risk taking by managers; it may also align the interests of managers and investors. Hedge funds managers typically invest significant capital in their funds; naturally, this capital accumulates over time. We can expect that older management companies are more risk averse than newer ones and possibly produce inferior returns to these younger companies. Agarwal, Daniel, and Naik (2004), Goetzmann, Ingersoll and Ross (2003), and Getmansky (2004) find decreasing returns to scale among hedge funds. Boyson and Cooper (2004) examine hedge fund persistence and find that no performance persistence when funds are selected based on past performance alone. If funds are selected using both fund performance and manager tenure, hedge funds exhibit persistence on a quarterly horizon. Manager tenure is negatively related to future performance. Agarwal, Daniel, and Naik (2004) establish that management companies’ age negatively relates to their performance. Management ownership is the most common in the hedge fund industry. Some management ownership is also evident in institutional business. Market factors as argued by Ackermann are very important in aligning managers’ and investors’ interests. Vast literature on performance-flow relationship within mutual fund, institutional fund, and hedge fund industries has demonstrated that investors care about a fund’s past and expected performance and reallocate assets according to their beliefs. Chevalier and Ellison (1997), Gruber (1996), and Sirri and Tufano (1998) show the convex relationship between 19 performance and flows in the mutual fund industry. Del Guercio and Tkac (2002) find the same result for mutual fund industry, but find that institutional investors do not chase winners, but withdraw from poorly performing managers. Agarwal, Daniel, and Naik (2004) find a convex relationship in hedge fund flow-performance relationship. Baquero, Horst, and Verbeek (2005) show that hedge fund money inflows are sensitive to past long-run performance, outflows exhibit an immediate and sustained response to past performance in the short run. Also part of the market factor is the power that clients have in their relationship with the manager. Arguably the mutual fund clients have the least amount of power and client power is restricted in the hedge fund industry with a lockup-up period and minimum notification times to withdraw the assets as well as the opaque nature of reporting in the business. Institutional clients have the most power over their manager. This power increases with the client size and sophistication. For example, a wealthy individual with $20M to invest via Merrill Lynch private equity has much less power over a manager than the California Teacher Federation Pension Fund that decides to hire Merrill Lynch to manage their equity assets. This potentially aligns the interests of managers and investors in an institutional industry the best and could translate in superior risk and return relationships. Performance comparison across organizational forms is significantly influenced by the differences in fees. In basis points, institutional funds have the lowest management fees (74 basis points on average across all industry) and performance fees significantly lower than in the hedge fund industry. 10 to 20 percent performance fees are not uncommon in hedge fund industry. In summary, due to the fact that hedge fund managers invest in their own funds, receive high bonus fees, have the freedom to choose from wide range investment strategies, and frequently are general partners with large liability in the case of bankruptcy, they should have better aligned interests with investors than mutual funds. We expect that hedge funds outperform mutual funds in terms of risk adjusted returns, but the high fees may decrease the hedge fund 20 advantage over other organizational forms. However, institutional funds also offer a range of attractive features. As opposed to hedge funds, institutional funds have very large clients, therefore even smaller fees can generate substantial fee revenue and therefore attract top quality manager. We view that the different sample period and inability to control for serial autocorrelation in hedge fund returns in the previous study calls into question the mutual fund and hedge fund comparison1. It is especially so because the hedge fund industry has experienced a record growth in 1980s and 1990s, and now is approaching the size category of the other two industries- in other words it is maturing. Chan, Getmansky, Haas, and Lo (2005) suggest that hedge fund returns are decreasing over our sample period. Thus, it is an empirical question, which industry provides the best risk and return relationship. We formally state hypotheses in the next section. Hypotheses Our first hypothesis is the management alignment hypothesis where the greater the alignment of interests between management and clients, the higher the return. If interests are better aligned agency costs will be reduced in two ways. First, management will reveal the information (by actively trading securities) that adds value to the portfolio. Second, management will charge lower expenses for the value adding information and reduce hidden expenses such as soft dollars. Thus we expect that the risk and return relationship will differ between the organizational forms reflecting the degree of alignment. The alternative hypothesis is an argument from Berk and Green (2004), who argue that managers will eventually capture the wealth created by each portfolio in a competitive market. Under the alternative hypothesis, we should observe no difference in performance across organizational forms. Our second hypothesis is the flexibility hypothesis which is the fewer restrictions on the manager, the higher the gross return. Management with fewer restrictions is better able to achieve 1 Ackermann et al (1999) find that hedge funds outperform mutual funds in terms of Sharpe Ratios. 21 an optimal return for the risk in an efficient market and better able to use the information that they have if they get information that is not in the price of the security. Our third hypothesis is the Grossman- Stiglitz (1980) hypothesis where higher expenditures on manager skill and on information result in a better return. With the exception Cici et al., the previous studies comparing performance across portfolio types do not control for differences that have nothing to do with type of portfolio2. For example, many hedge funds are younger and more active than mutual funds and institutional funds, yet this difference has nothing to do with being a hedge fund. Younger and more active mutual funds may have exactly the same risk-return characteristics. Transactions costs per share are probably lower for the more active and larger portfolios. Institutional portfolios are often much larger than hedge funds and mutual funds and may be charged smaller transactions costs. While the aforementioned variables can be controlled for, fund complexes introduce omitted variables that are hard to operationalize, but may influence the risk and return outlook of a fund. We suggest that that when these investment segments are all owned by the same company the effect of omitted variables is mitigated. Investment complexes provide a variety of services that are centralized and tend to reduce differences in the segments not due to the contract. Investment complexes provide the following benefits to investors: partially or fully centralized information collection, have a common resource base to pay for information, engage in risk control over all managed portfolios, use common reporting and monitoring reducing the risk of fraud, “blowup” risk and allow diversification, have economies of scale in costs (legal cost, transactions costs). 2 See Ackermann, McEnally, and Ravenscraft (1999) and Del Guercio and Tkac (2002) for example. 22 The maintained hypothesis assumes that these benefits are at least partially associated with omitted variables that cause systematic differences between the investment segments but have nothing to do with controlling agency costs. Methodology To compare the risk and return relationship across industries we use Sharpe ratios (SR) and the seven factor model that includes Carhart (1997) four factors, an additional equity factor and two fixed income factors. Alternatively self selected benchmarks can be used to estimate factor model alphas (e.g. Getmansky et al. (2004)). While this approach gives an insight into funds performance persistence, it does not allow for a comparison across funds. Use of the Sharpe ratio is limited in this type of research due to managers’ ability to manipulate the Sharpe ratio with derivative strategies (Goetzmann, Ingersol, Spiegel, and Welch (2002)) and due to bias that is introduced in Sharpe ratios from the return serial correlation (Lo 2002, Getmansky, Lo, and Makarov (2004)). While we unsmooth hedge fund return to control for the serial correlation in returns, the return manipulation induced bias remains. Goetzmann et al. suggest ways how managers can boost Sharpe ratios in the absence of investment restrictions. Therefore it is reasonable to assume that Sharpe ratios are most boosted in the hedge fund industry, somewhat less in the institutional industry, and even less in the mutual fund industry. Therefore, simple Sharpe ratios for hedge funds should be viewed as optimistic estimates. Goetzmann et al. (2002) also argue that manager performance based income in institutional and hedge fund industries somewhat mitigate the inclination to boost the Sharpe ratio. Let’s recall that Sharpe ratio is defined as the ratio of excess expected return to the standard deviation of return: SR ≡ μ − Rf , σ 23 where µ and σ are mean and variance of simple return of portfolio. SR is calculated relative to the risk free rate. Both mean and variance are unobservable and are estimated from ˆ returns: μ = ˆ SR ≡ 1 T 1 T ˆ ∑t =1 Rt and σˆ 2 = T ∑t =1 ( Rt − μ )2 . Then the estimate SR follows: T ˆ ˆ μ − Rf . ˆ σ Lo (2002) demonstrates that the standard error of Sharpe ratio is as follows a ⎛ 1 ⎞ ˆ SE ( SR) = ⎜1 + SR 2 ⎟ / T , and 95% confidence intervals can be estimated by ⎝ 2 ⎠ a ⎛ 1 ˆ ⎞ ˆ SR ± 1.96 × = ⎜1 + SR 2 ⎟ / T ⎝ 2 ⎠ Note that institutional returns are reported on quarterly basis while all other returns are reported on monthly basis. Lo proves that the Sharpe ratio aggregates over time as a function of square root of time intervals. SE (q ) = E[ Rt (q )] − R f (q ) Var[ Rt (q )] = q( μ − R f ) qσ = q SR , Where q is time interval over which returns are aggregated. Different multi-factor models have evolved over time since CAPM and Jensen’s alpha (1969). Many of them have stemmed out of research on cross-section variation in average returns. Research on book-to-market (Rosenberg et al. 1985, Fama and French, 1992), dividend yield (Litzenberger and Ramaswamy, 1982), and momentum (Jagadeesh and Titman, 1993, Carhart 1997) have led to the development of several multi-factor models. Fama and French (1993) propose a three factor model, Carhart (1997) extends it to four factors, Grinblatt and Titman (1989) propose an 8 factor model, Ferson and Khang (2002) propose to examine a conditional factor model. Fama and French 1998 suggest an international book-to-market factor to account 24 for international value premium. More recently Edwards and Caglayan (2001) propose an eight factor model, while Bares, Gibson, and Gyger (2003) propose an eight factor APT model. Capocci, Corhay, and Hubner (2005) propose a ten factor model to study hedge fund returns. The factor model includes the Fama and French (1993) size, and value factors, Carhart’s (1997) momentum factor, Agarwal and Naik (2002) factors: a factor for non-US equities funds (MSCI World excluding US), two factors to account for the fact that hedge funds invest in US and foreign bond indices (Lehman High Yield Index, Salomon World Government Bond Index), Capocci and Hubner (2004) JP Morgan Emerging Market Bond Index and Commodity factor (GSCI Commodity Index) and commodity factor (GSCI Commodity Index), Hubner et al. (2005) factors: Lehman Mortgage-Backed Securities Index and Lehman High-Yield Credit Bond Index. Because our sample period is limited to 12 years with many funds in existence for part of the sample period, we chose the first seven factors from Hubner et al. model. The market factor is the value-weighted returns on all NYSE, AMEX, and NASDAQ stocks and the risk free rate is the one-month Treasury bill rate from Ibbotson Associates as in Fama and French (1993)3. The first four factors are Carhart’s (1997) four factors: market risk premium, size, value, and the momentum factor. The next three factors are “MSCI world ex US” (MSWXUS), “Merrill Lynch high yield bond C” (MLHYB), and “Merrill Lynch global governments” (MLGG). Rpt – Rft = αp + βp1(Rmt-Rft) + βp2SMBt + βp3HMLt + βp4PR1YRt + β5(MSWXUSt-Rft) + βp6(MLHYB-Rft) + βp7(MLGG-Rft) +εpt. We use Flexible Least Squares method in the estimation (see Berzins 2005 for details). In the next section we compare the fund performance across the three industries by using Sharpe (1966) ratios and time-varying alphas from the seven factor model. 3 Agarwal and Naik (2002) use Russell 3000 Index as a market proxy. Capocci and Hubner (2004) find that the two indices are almost perfectly correlated. 25 VI Discussion and Results A. Univariate Return and Sharpe Ratios Statistics In order to estimate Sharpe ratios we dropped the first 12 observations from all hedge fund returns to control for backfill bias. Our institutional and mutual fund samples are free from this bias. We allowed mutual fund and hedge fund returns to go back as far as April 1990 provided that they were reported before 1993 for estimation purposes only. Institutional returns are subject to backfill bias before June 1993 and no returns before this date are included for the Sharpe ratio calculation. The average one month T-Bill rate for our sample period is 0.3316 percent. Institutional returns are reported gross of fees. While we do not know precise fees, they can be estimated from fee schedules and average account sizes. We use the estimated fees to calculate net institutional returns. We excluded the index mutual funds from the average industry Sharpe ratio calculation. The Sharpe ratio is calculated for each portfolio for the complete return series that is not backfilled. We assign the portfolio Sharpe ratio to every date for which the portfolio exists in our sample. We include negative Sharpe ratios in our analysis. A priori one cannot expect to observe negative Sharpe ratios. However, negative Sharpe ratios occur quite often for several reasons. First, we may have too short of a time series to estimate the true Sharpe ratio for a fund. Second, a manager may lose his competitive advantage and earn a negative true Sharpe ratio. The first case introduces unknown bias into our study. We Windsorized the few outliers that reported Sharpe ratios larger than two in absolute terms. We start by examining the cross-sectional univariate properties of the Sharpe ratios and alphas. To test hypotheses we will use multivariate regressions. Table 4 reports the average Sharpe ratios for all portfolios. Sharpe ratios are calculated over all available dates for each portfolio and cross sectional averages are reported. The three industry averages are statistically significantly different. Institutional funds have the highest monthly Sharpe ratios (0.15), followed by hedge funds (0.13) and mutual funds (0.11). Hedge funds have the lowest kurtosis of all three industries (1.49). 26 Kurtosis for institutional funds is 2.01, while for mutual funds it is 6.42. Figure 2 presents cross-sectional average plots where the Sharpe ratios from Table 4 are assigned to the dates a portfolio reported returns. The time series average of cross sectional average Sharpe ratio is 0.1037 for the mutual funds, 0.1218 for the hedge funds, and 0.1368 for the institutional funds. Institutional funds dominate the mutual funds by producing statistically significantly higher Sharpe ratios in every period. Therefore institutional funds outperform hedge funds by 12% on this measure, and hedge funds outperform mutual funds by 17%. This is very close to the Ackerman et al. (1999) finding that hedge funds between 1987 and 1995 have 21% higher average Sharpe ratios than mutual funds. The smaller difference may suggest that the hedge fund industry is maturing and showing lower returns in general. While the ranking is statistically significant it is important to note that the difference is time varying. Therefore for formal tests we will resort to multivariate regression. The results are robust to splitting the sample up into equity and non equity portfolios (Panels B and C). The hedge funds seem to dominate the other two industries in the later part of sample. However, Panel C reveals that the hedge fund dominance is present among equity, but not present among non-equity portfolios. Figure 3 further examines Sharpe ratios by reporting a different time varying statistic. Here for each report date we calculate Sharpe ratios for a minimum of 12 previous months (or 8 quarters for institutional funds) and a maximum of 36 previous months. The panels A though C report similar results to Figure two. The time series average of the cross-sectional averages is 0.1714 for institutional portfolios, 0.1645 for hedge funds, and 0.1277 for mutual funds. Hedge funds overall perform statistically insignificantly worse than institutional funds but statistically significantly better than mutual funds. Equity hedge funds have the highest average for the three industries. Equity Sharpe ratios exhibit the best performance during the market downturn. This further strengthens the return findings that hedge funds have the potential to make an important contribution to an overall investor portfolio along the lines of Elton, Gruber, and Rentzler (1987). Elton et al develop a methodology for assessing the contribution of an alternative investment portfolio to an existing portfolio. Fixed income 27 hedge fund and mutual fund Sharpe ratios are indistinguishable. In general, the hedge fund Sharpe ratios exhibit less variation than the other two industries. B. Univariate Alpha Statistics We attempt to account for portfolio exposure to various sources of risk by calculating seven factor model alphas. To estimate the seven factor model, we allowed mutual fund and hedge fund returns to go back as far as April 1990 if they were reported before 1993. We use institutional returns from as far back as June 1990 for time varying coefficient estimation. For analysis we exclude the coefficient dates for which funds backfilled returns. We estimate time varying alphas with the flexible least square (FLS) method. The cross-section averages of alphas are reported in Figure 4. The net value weighted hedge fund alpha is negative for the most of the sample period with the exception of the very beginning of the period (-0.12% value weighted per month on average). This concurs with a common belief that more than half of hedge funds do not earn back their fees. The worst hedge fund alphas are reported around the Asian crisis in 1998 and the best are noted around the burst of the “internet bubble” in 2000. During the latter period equally weighted alphas are noticeably higher than the value weighted alphas. Thus perhaps small hedge funds offered the best “hedging’ during the internet bubble burst. Institutional funds produce alphas that are insignificantly negative until 1999, and significantly positive alphas from 2001. Mutual funds report negative alphas for almost the whole time period; this coincides with conclusions from many mutual fund studies starting with Jensen (1968), Elton et al. (1993), Carhart (1997). Similar to hedge funds, more than half of mutual funds do not earn back their fees (the value weighted alpha of -0.12% per month on average). The institutional fund alpha clearly dominates the other two industries since 1997 (0.03% on average for the whole period). Mutual fund alphas are surprisingly higher than the hedge fund alphas after 2001, but the difference is statistically insignificant. This finding indicates that while hedge funds dominate mutual funds with respect to the Sharpe ratio, the difference disappears for the whole sample when one accounts for hedge fund risk exposure. This supports the conclusion that hedge funds, 28 despite their seemingly better organizational form, do not recover fees as well as mutual funds and thus appear to “underperform” from the perspective of a sophisticated investor. This is an opposite conclusion to that of Ackermann et al. (1999). The next section tests the hypotheses within a multivariate regression framework. C. Regression Analysis and the Test of Hypotheses. To test the hypotheses we estimate a fixed effect model where the net-of-fees Sharpe ratio (or net-of-fees alpha) is the dependent variable. Table 5 reports regressions for time invariant Sharpe ratios. Here we have calculated one Sharpe ratio per portfolio over its lifespan. The first column includes just the industry dummies, while the rest of the columns include conglomerate fixed effects dummies. We insisted that the conglomerate dummies add to zero. Thus the intercept reflects the average conglomerate effect and the omitted industry dummy. We use the following independent variables. Age, which is estimated in months over all reported returns for a fund including backfill return dates (the distribution is Windsorized at 1980 for funds that report longer time series). Asset time series average per portfolio in millions of dollars. Average revenues - the product of monthly fees and assets in millions. Average fees are reported on a monthly basis in decimals. Mutual fund expense ratios are used when available. For institutional funds, we use fee estimates, while for hedge funds performance fees are included for periods of positive returns. Family assets are a sum of average portfolio assets in conglomerate in millions. Equity dummy is one for equity. We normalize fees around man of one within an industry group for each regression. This allows us to estimate the impact of fees on performance relative to the peer group. The normalization is necessary as the average fees differ significantly across industries: 6.20% per year for hedge funds, 1.37% per year for mutual funds, and 0.45% per year for institutional funds (see row 1 in Panel C of Table 2). Mutual fund and institutional fund dummies are significantly larger than the hedge fund dummies for the pooled regression. Mutual fund dummy is 0.0197 and significant at 10%, while institutional fund dummy is 0.0543 and significant at 1%. However, mutual fund dummy 29 significance vanished in regressions with the fixed conglomerate effects, while institutional fund dummy is 0.036 and significant at 5 percent. Opposite to Ackerman (1999) we conclude that mutual funds and hedge funds are indistinguishable in terms of Sharpe ratios. This finding does not reject the first or agency alignment hypothesis, but rejects the alternative to our first hypothesis (the Berk and Green hypothesis). It does appear that institutional funds have found the best way to align interests between managers and investors. For robustness, we estimate the same regression separately for equity and fixed income samples (columns 4 to 7). The institutional dummy is significantly positive for equity funds (0.0548 and 0.0357 at 1% in two different specifications), but positive and insignificant for fixed income funds. The mutual fund dummy has no significance. To test the flexibility hypothesis, we estimate the sample difference in gross performance. We do not observe gross mutual fund or hedge fund returns, but can estimate the gross returns from fees and net returns. We did not estimate other gross performance measures to avoid compounding of fee estimation errors for mutual funds and hedge funds. Overall hedge funds post statistically significantly higher gross returns than mutual and institutional funds (see row 2 of Panel B in Table 2). Cross sectional average of the portfolio time series average monthly returns are 1.00% for hedge funds, 0.83% for institutional funds, and 0.66% for mutual funds. The significance between mutual fund and hedge fund gross returns vanishes just in the early part of the sample. We cannot reject the flexibility hypothesis and conclude that the flexibility in hedge fund investing generates the higher gross returns among all three sectors. The normalized fees for overall regression are significantly negative (-0.0128 at 1 %). The significance comes from the fixed income portfolios. The fee variable is insignificant for the equity portfolios. This is consistent with the view that lower fees promote better alignment between manager and investor interests. The revenues are positively significantly related to the Sharpe ratios. For pooled fixed effects regression, the revenue coefficient is 0.0174 (significant at 1%). When split on equity and 30 fixed income dimension, only the equity coefficient is significant (0.0247 at 1%). This finding does not reject the third hypothesis- the Grossman Stiglitz argument. It appears that higher revenues attract managers that produce higher Sharpe ratios. Table 6 reports a robustness check for the Sharpe ratios. Here the Sharpe ratios are calculated for four non overlapping periods. The Sharpe ratio is calculated for a minimum of 12 months (or 8 quarters for institutional sample) and a maximum of 36 months. The industry dummy results hold with the exception of the third period where predictably the hedge funds did well against the other two industries. In the third period mutual fund dummies are significantly negative (-0.0814 and -0.0782) in two separate specifications, while the institutional fund dummies are insignificant. Mutual fund coefficient sign varies over time. The revenue and fee conclusion is robust in this specification. The control variables reveal that portfolio age is negatively related to the Sharpe ratio (with one exception). This confirms for the three industries the results from previous hedge fund studies about the negative impact of tenure on performance. The size is not related to the Sharpe ratio in a consistent way. It is negative in pooled regressions, but the sign varies in time varying regressions. In other words, we do not observe negative returns to scale across all portfolio types. Table 7 reports alpha regression results. The time varying seven factor alphas are calculated over all available dates, excluding backfill return dates. The hedge fund and institutional fund industry dummies are significantly positive, which implies that the hedge fund alphas are the smallest. For the pooled fixed effects regression the mutual fund dummy is 0.2253 and the institutional fund dummy is 0.7168; both are significant at 1%. Therefore, we conclusively do not reject the agency alignment hypothesis, but reject the Berk and Green hypothesis. The finding is robust across asset classes and time periods (Table 8) with an exception for the first time period. The noticeably smaller sample size during the first time period could have caused this exception. 31 The revenues across different specifications are positively related to the alpha, yet in several time varying specifications the relationship is insignificant. Thus, we cautiously do not reject the Grossman Stiglitz hypothesis. The fees are either negative and significant or insignificant, but never positive and significant. Again the significance is observed for fixed income, but not equity portfolios. Among the control variables, the age is negatively related to the alpha. The result is statistically significant and robust across all specifications. Thus we observe negative performance to tenure across performance measures and portfolio classes. However, the portfolio assets are negative and significant for the pooled fixed effects regression and equity regression. Thus, we cautiously conclude that equity portfolios are more subject to negative returns to scale than the fixed income portfolios. 32 VII Conclusion We investigate investment choices that a sophisticated investor would face by analyzing a unique sample of investment conglomerates, which simultaneously offer investment products in mutual fund, hedge fund, and institutional investment industries. We compare performance of funds across industries by using the Sharpe ratios and alphas from a multifactor model. We find that in a univariate setting hedge funds produce higher gross returns than the other two industries. Institutional funds have higher Sharpe ratios than hedge funds; while hedge fund and mutual fund Sharpe ratios are overall indistinguishable in a regression setting. Hedge funds report the lowest alphas in cross sectional fixed effects regressions. We test three hypotheses. We do not reject agency alignment hypothesis, but reject the alternative- Berk and Green hypothesis. We do not reject the investment flexibility hypothesis by finding that the most flexible contract form- the hedge fund, produces the highest gross return. We also do not reject the Grossman Stiglitz hypothesis. In addition we find that the Sharpe ratio and alpha are negatively related to tenure, but fixed income funds have a negative relationship between fund size and alpha. We cannot concur with the conclusion of Ackermann et al (1999) that hedge funds offer a superior riskreturn relationship to mutual funds. But we observe that small hedge funds offered superb performance during the “internet bubble burst”. We control for backfill bias in our study and correct hedge fund returns for smoothing effects caused by stale prices with a MA(2) process. We find that a considerably smaller percentage of our funds exhibit significant autocorrelation than reported by Getmansky, Lo, and Makarov (2004). To our knowledge, our study is the first to offer a comparative analysis across all three investment industries. 33 REFERENCES Ackerman, C., R. McEnally, and D. Ravenscraft, 1999, The performance of hedge funds: risk, return, and incentives, Journal of Finance, 54, 833—874. Ackermann, C., D. Ravenscraft, 1998, The impact of regulatory restrictions on fund performance: a comparative study of hedge funds and mutual funds, Working Paper, University of Notre-Dame. Agarwal, V., and N. Y. Naik, 2000a, Multi-period performance persistence analysis of hedge funds, Journal of Financial and Quantitative Analysis, 35, 327—342. Agarwal, V., and N. Y. Naik, 2000b, On taking the alternative route: risks, rewards, and performance persistence of hedge funds, Journal of Alternative Investments, 2, 6—23. Agarwal, V., and N. Y. Naik, 2000c, Performance of hedge funds with option-based buy-and-hold strategies, Working paper. Agarwal, Vikas, N. D. Daniel and N.Y. Naik, 2004, Flows, performance, and managerial incentives in hedge funds, EFA 2003 Annual Conference Paper No. 501. Asness, C., R. Krail, and J. Liew, 2001, Do hedge funds hedge?, Journal of Portfolio Management, Fall, 6—19. Baquero, G., and M. Verbeek, 2005, A portrait of hedge fund investors: flows, performance and smart money working paper, Erasmus University, 40, 493—517. Baquero, G., J. T. Horst, and M. Verbeek, 2005, Survival, look-ahead bias and the persistence in hedge fund performance, Journal of Financial and Quantitative Analysis 40, 493-518. Bares, P., R. Gibson, S. Gyger. 2002, Performance in the hedge fund industry: an analysis of short and long-term persistence, Working paper. Berk, J. B., and R. C. Green, 2004, Mutual fund flows and performance in. rational markets, Journal of Political Economy, 112 no. 6. Berzins, J. and C. A. Trzcinka, 2005, Survival and performance of pension fund money managers, Working paper, Indiana University. Bhattacharya, S., P.Pfleiderer, 1985, Delegated Portfolio Management, Journal of Economic Theory 36, 125. Boyson, N. M. and M. J. Cooper, November 2004, Do hedge funds exhibit performance persistence? a new approach, Working paper. Capocci, D., A. Corhay, and G. Hubner, 2005, Hedge fund performance and persistence in bull and bear markets, The European Journal of Finance, Vol. 11, No. 5, 361–392, October 2005. Carpenter, J. N., 1998, The optimal dynamic investment policy for a fund manager with an incentive fee, Working paper, New York University. Chan, N., M. Getmansky, S. M. Haas, A. Lo, 2005, Systemic risk and hedge funds, working paper MIT. Chevalier, J., and G. Ellison, 1997, Risk taking by mutual funds as a response to incentives, Journal of Political Economy 105, 1167–1200. Cici, G., S. Gibson, and R. Moussawi, 2006, For better or worse? mutual funds in side-by-side management relationships with hedge funds, Working paper. Del Guercio, D. and P. A. Tkac, 2002, The determinants of the flow of funds of managed portfolios: mutual funds versus pension funds, Journal of Financial and Quantitative Analysis, 523. Edwards, F.R. and M.O. Caglayan, 2001, Hedge Fund and Commodity Fund investments in Bull and Bear Markets, Journal of Portfolio Management, 27, nr.4. 34 Elton. E.J., M.J. Gruber, and J. Rentzler, 1987, Professionally managed, publicly traded commodity funds, Journal of Business 60, 175-199. Fama, E.F., French, K.R., 1993, Common risk factors in the returns on stocks and bonds, Journal of Financial Economics, 33, 3 – 56. Ferson, W., K. Khang, 2002, Conditional performance measurement using portfolio weights: evidence for pension funds, Journal of Financial Economics, 65, 249-282. Fung, W., Hsieh, D., 2000, Performance characteristics of hedge funds and commodity funds: natural versus spurious biases, Journal of Financial and Quantitative Analysis 35, 291–307. Gallaher, S., L. Starks, and R. Kaniel, 2005, Madison avenue meets wall street: mutual fund families, competition and advertising, Working paper. Gaspar, J.M., M. Massa. and P. Matos, 2004, Favoritism in mutual fund families? evidence on strategic cross-fund subsidization, Working paper. Getmansky, M., 2004, The life cycle of hedge funds: fund flows, size and performance, unpublished working paper, MIT Laboratory for Financial Engineering. Getmansky, M., A. W. Lo, and I. Makarov, 2004, An econometric model of serial correlation and illiquidity in hedge fund returns, Journal of Financial Economics, 74, 529-609. Getmansky, M., A. W. Lo, I. Makarov, 2004, An econometric model of serial correlation and illiquidity in hedge fund returns, Journal of Financial Economics, 74, 529-609. Goetzmann, W., Ingersoll, J. and S. Ross, 2003, High water marks and hedge fund management contracts, Journal of Finance 58, 1685-1718. Goetzmann, W., J. Ingersoll, M. Spiegel, I. Welch, 2002, Sharpening Sharpe ratios, National Bureau of Economic Research, working paper no. W9116. Grinblatt, M., Titman, S., 1989, Mutual fund performance: an analysis of quarterly portfolio holdings. Journal of Business, 62, 393– 416. Gruber M., 1996, Another puzzle: the growth in actively managed mutual funds, The Journal of Finance 51, 783–810. Guedj, I. and J. Papastaikoudi, 2003, Can mutual fund families affect the performance of their funds? Working paper. Jagadeesh, N., S. Titman, 1993, Returns to buying winners and selling losers: implications for stock market efficiency, Journal of Finance, 48, 93– 130. Jagannathan, R., A. Malakov, D. Novikov, 2006, Do hot hands persist among hedge fund managers? an empirical evaluation, Working paper. Jensen, M. C., 1969, Risk, The pricing of capital assets, and the evaluation of investment portfolios, The Journal of Business, Vol. 42, No.2, 167-247. Kempf, A, and S. Ruenzi,2003, Tournaments in mutual fund families, Working paper. Lakonishok, J., A. Shleifer, and R. Vishny, 1992, The structure and performance of the money management industry, Brookings Papers: Microeconomics, 339-391. Liang, B., 1999, On the performance of hedge funds, Financial Analysts Journal, 55, 4, 72-85. Liang, B., 2000, Hedge funds: the living and the dead, Journal of Financial and Quantitative Analysis, 35, 309—326. Litzenberger, R., K. Ramaswamy, 1982, The effects of dividends on common stocks returns: tax effects or information’s effects?, Journal of Finance 37, 429– 443. Lo, A., 2002, The statistics of Sharpe ratios, Financial Analysts Journal 58, 36–50. Okunev, J., and D. White, 2003 Hedge fund risk factors and value at risk of credit trading strategies, 35 Working paper. Rosenberg, B., K Reid, and R. Lanstein, 1985 Persuasive evidence of market inefficiency, Journal of Portfolio Management, 11, 9 –17. Ross, Stephen A., 1973, The economic theory of agency: the principal’s problem, American Economic Review, 63, 134 -139. Sharpe, William F., 1966, Mutual fund performance, Journal of Business, 39, 119-138. Sirri, E. and P. Tufano, 1998, Costly search and mutual fund flows, The Journal of Finance 53, 1589–1622. Starks, Laura T., 1987, Performance incentive fees: an agency theoretic approach, Journal of Financial and Quantitative Analysis 22,17-32. Zhao, X., 2005, Entry decisions by mutual fund families, Working paper. 36 Figure 1 Conglomerate Reporting Time Series This figure presents the return time series for the sample of investment conglomerates. Y axis indicates the time, while the 66 conglomerates from our sample are listed across on the X axis. This figure identifies the dates for which a conglomerate reported a return in a particular managed fund segment. Hedge Funds 2005 12 Institutional Funds Mutual Funds 2004 12 2003 12 2002 12 2001 12 2000 12 1999 12 1998 12 1997 12 1996 12 1995 12 1994 12 1993 12 1993 6 37 Table 1 Asset Time Series This table presents the asset time series for the sample of investment conglomerates. These conglomerates manage funds simultaneously in hedge fund, institutional fund, and mutual fund industries. Year end assets are aggregated across all portfolios and reported in millions of dollars. Panel A: All Assets Year Hedge Funds Institutional Mutual Fund Total 2005 73,492.7 1,654,141.7 555,271.0 2,282,905.4 2004 71,042.3 1,621,905.3 518,208.0 2,211,155.5 2003 51,507.4 1,513,309.3 460,433.0 2,025,249.6 2002 38,064.3 1,326,577.7 359,330.8 1,723,972.8 2001 31,526.1 1,487,121.1 420,749.8 1,939,397.0 2000 27,184.8 1,350,332.5 445,640.9 1,823,158.2 1999 27,047.7 1,455,422.9 464,344.1 1,946,814.7 1998 23,695.4 1,317,616.1 364,241.9 1,705,553.4 1997 21,987.1 919,483.6 313,886.7 1,255,357.5 1996 14,230.9 764,287.9 209,929.6 988,448.4 1995 11,546.9 591,613.8 167,362.0 770,522.6 1994 8,939.0 418,165.0 123,514.3 550,618.3 1993 8,261.3 315,773.0 109,471.9 433,506.1 Panel B excluding equity 2005 38,361.5 2004 42,042.2 2003 31,998.7 2002 22,530.7 2001 16,956.8 2000 13,098.8 1999 13,921.5 1998 13,337.2 1997 12,596.0 1996 8,222.5 1995 6,904.6 1994 5,352.8 1993 4,799.2 Panel C Equity 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 666,174.3 659,158.4 649,265.2 630,229.0 670,170.8 528,030.1 580,588.3 605,068.9 389,793.8 329,913.2 243,637.1 178,694.6 146,379.5 116,060.4 115,186.7 117,101.3 111,883.6 102,368.8 81,807.9 86,593.2 83,106.7 75,336.1 54,446.0 49,328.6 41,299.1 41,855.4 820,596.2 816,387.3 798,365.1 764,643.3 789,496.4 622,936.8 681,103.0 701,512.8 477,725.9 392,581.7 299,870.2 225,346.5 193,034.1 35,131.1 29,000.0 19,508.7 15,533.6 14,569.2 14,086.0 13,126.1 10,358.1 9,391.1 6,008.4 4,642.3 3,586.2 3,462.1 987,967.4 962,746.9 864,044.1 696,348.7 816,950.3 822,302.4 874,834.6 712,547.2 529,689.8 434,374.7 347,976.7 239,470.4 169,393.5 38 439,210.6 403,021.3 343,331.7 247,447.2 318,381.0 363,832.9 377,750.9 281,135.3 238,550.7 155,483.6 118,033.4 82,215.2 67,616.4 1,462,309.2 1,394,768.2 1,226,884.5 959,329.5 1,149,900.5 1,200,221.4 1,265,711.6 1,004,040.6 777,631.6 595,866.7 470,652.4 325,271.8 240,472.0 Table 2 Descriptive Statistics: Tests of Mean Differences This table report means of monthly returns and annual fees from portfolios offered by investment conglomerates. We include all portfolios that have at least 12 monthly or 8 quarterly returns. We drop 12 monthly returns from the hedge fund returns to control for the backfill bias. If hedge fund is started during our sample period and has less than 24 returns we do no drop any returns. No index mutual funds are included. We exclude institutional portfolio backfill returns. Mutual funds do not suffer from the backfill bias. Institutional fund returns are gross of fees and we have estimated fees from the fee schedules. The sample period is from April 1993 to December 2005 (mutual fund data ends in September 2005). For "cross-section", time series averages are calculated for each portfolio and cross sectional averages reported in the table. For "time-series", cross-sectional averages are calculated first and time series averages are reported in the table. For hedge funds, performance fee is included in every month with positive returns. Panel A: Net Returns HF INST MF Inst HF Inst MF MF HF p-Value p-Value p-Value Type Mean Median StDev N Mean Median StDev N Mean Median StDev N Means Means Means Time Series 0.70 0.60 1.68 150 0.78 1.05 1.84 150 0.73 1.03 3.18 150 0.70 0.86 0.93 Cross-section All Years 0.54 0.53 0.85 473 0.78 0.73 0.60 813 0.55 0.56 0.84 1487 0.00 0.00 0.82 Cross-section 1993 -1996 0.86 0.95 1.24 169 1.03 0.99 0.64 411 1.10 1.05 0.73 438 0.09 0.09 0.01 Cross-section 1997 -1999 1.22 0.83 2.09 222 1.45 1.11 1.91 550 1.41 1.14 1.65 772 0.17 0.75 0.21 Cross-section 2000 -2002 0.36 0.43 1.13 294 -0.36 -0.29 1.25 627 -0.71 -0.70 1.26 1226 0.00 0.00 0.00 Cross-section 2003 -2005 0.56 0.51 0.83 328 1.31 1.34 0.88 547 1.23 1.25 0.87 1263 0.00 0.06 0.00 Panel B: Gross Returns HF INST MF Inst HF Inst MF MF HF p-Value p-Value p-Value Type Mean Median StDev N Mean Median StDev N Mean Median StDev N Means Means Means Time Series 1.22 1.07 1.88 150 0.83 1.09 1.84 150 0.84 1.14 3.18 150 0.07 0.96 0.21 Cross-section All Years 1.00 0.93 1.06 473 0.83 0.78 0.61 813 0.66 0.66 0.84 1487 0.00 0.00 0.00 Cross-section 1993 -1996 1.38 1.26 1.56 169 1.07 1.04 0.65 411 1.22 1.17 0.73 438 0.01 0.00 0.19 Cross-section 1997 -1999 1.85 1.35 2.52 222 1.49 1.16 1.91 550 1.53 1.26 1.66 772 0.05 0.68 0.07 Cross-section 2000 -2002 0.84 0.74 1.28 294 -0.31 -0.24 1.25 627 -0.59 -0.60 1.25 1226 0.00 0.00 0.00 Cross-section 2003 -2005 0.98 0.79 1.04 328 1.36 1.38 0.89 547 1.34 1.35 0.88 1263 0.00 0.64 0.00 Panel C: Fees HF INST MF Inst HF Inst MF MF HF p-Value p-Value p-Value Type Mean Median StDev N Mean Median StDev N Mean Median StDev N Means Means Means Time Series 6.20% 5.73% 2.71% 150 0.56% 0.56% 0.02% 150 1.37% 1.36% 0.04% 150 0.00 0.00 0.00 Cross-section All Years 5.53% 4.73% 4.27% 473 0.55% 0.50% 0.29% 813 1.30% 1.26% 0.53% 1487 0.00 0.00 0.00 Cross-section 1993 -1996 6.35% 5.48% 5.50% 169 0.55% 0.49% 0.34% 411 1.39% 1.30% 0.56% 438 0.00 0.00 0.00 Cross-section 1997 -1999 7.54% 6.03% 6.85% 222 0.54% 0.50% 0.31% 550 1.42% 1.32% 0.57% 772 0.00 0.00 0.00 Cross-section 2000 -2002 5.80% 4.51% 4.25% 294 0.56% 0.53% 0.30% 627 1.34% 1.30% 0.52% 1226 0.00 0.00 0.00 Cross-section 2003 -2005 4.96% 4.13% 3.50% 328 0.58% 0.55% 0.31% 547 1.30% 1.27% 0.52% 1263 0.00 0.00 0.00 39 Table 3 Sample Averages by Assets and Herfindahl Index This table presents the conglomerate sample statistics based on the Herfindahl index in panel A and assets in panel B. The Herfindahl index is calculated by first calculating the average assets for each portfolio for the whole time series that portfolio lived. These numbers are aggregated to form conglomerate asset figure for each industry. We calculate the percentage assets investment in the three industries for each conglomerate. The sum of the percentage squares forms the Herfindahl index. We report the average index for each quartile (Herfindahl Avg Assets). We repeat the same exercise by using the portfolio count (Herfindahl Nr Portfolios). "Total Nr HF" is the number of hedge funds in the quartile. These funds may not have existed concurrently. "Avg HF Nr %" is the number of hedge fund portfolios as a percentage of all portfolios in a quartile. "Avg HF Ast %" is hedge fund assets as a percentage of the total quartile assets. The same statistics are reported for mutual fund and institutional fund industries. The table finishes with the total number of portfolios and the total assets in millions. Herfindahl Herfindahl Total Nr Total Avg Avg HF Avg HF Total Nr Total Avg Avg Inst Avg Inst Total Nr Total Avg Avg MF Avg MF Total Total Quartile Nr Avg Assets HF HF Ast Nr % Ast % Inst Inst Ast Ast % Nr % MF MF Ast Ast % Nr % Portfolios Assets Portfolios Panel A: Asset Herfindahl Rank 203 166 67 288 410 149 680 603 1 5,164 5,026 32% 34% 18% 42% 50% 24% 107 108 201 2,571 552 244 860 2,923 2 4,848 7,749 23% 16% 26% 72% 51% 12% 90 76 239 3,015 261 66 590 3,157 3 4,154 9,062 16% 2% 40% 95% 44% 3% 73 58 306 8,863 264 57 643 8,978 4 4,580 9,741 13% 1% 52% 99% 35% 1% Panel B: Total Asset Rank 1 4,625 2 4,735 3 5,120 4 4,348 Total 4,699 6,599 7,807 7,921 9,095 7,837 59 114 109 191 473 25 90 119 174 409 30% 22% 18% 15% 21% 27% 17% 7% 2% 13% 44 96 122 551 813 52 358 1,420 12,906 14,736 27% 34% 33% 40% 34% 54% 73% 84% 95% 76% 115 253 343 776 1487 27 53 143 293 516 43% 44% 49% 45% 45% 19% 9% 9% 3% 10% 218 463 574 1518 2773 104 501 1,682 13,373 15,661 40 Table 4 Sharpe Ratio Distribution This table and chart reports the distribution of Sharpe ratios for all portfolios. The Sharpe ratios are reported by industry. Normal distribution is plotted for reference. We include all portfolios that have at least 12 monthly or 8 quarterly returns. We drop 12 monthly returns from the hedge fund returns to control for the backfill bias. If hedge fund is started during our sample period and has less than 24 returns we do no drop any returns. No index mutual funds are included. We exclude institutional portfolio backfill returns. Mutual funds do not suffer from the backfill bias. Institutional fund returns are gross of fees and we have estimated fees from the fee schedules. The sample period is from April 1993 to December 2005 (mutual fund data ends in September 2005). HF Inst MF Mean 0.13 0.15 0.11 Median 0.13 0.13 0.10 St Dev 0.23 0.15 0.21 Kurtosis 1.49 2.01 6.42 Skewness -0.34 0.26 0.89 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% 0.1 -0.8 -0.7 -0.6 -0.5 -0.4 HF -0.3 -0.2 -0.1 Inst 0.0 0.2 MF 0.3 0.4 0.5 0.6 0.7 0.8 Norm 41 Figure 2 Sharpe Ratios-All Time Series per Portfolio This figure presents monthly Sharpe (1966) ratios for all portfolios in the conglomerate sample. We drop the first 12 return observations from hedge funds older than 24 months to control for backfill bias. We used mutual and hedge funds with at least 12 observations and institutional funds with at least 8 quarterly observations. Our institutional and mutual fund samples are free from this bias. We considered monthly returns from April 1993 to December 2005. Institutional quarterly returns start in June 1993 as they are subject to backfill bias before June 1993. The one month T-Bill rate is Ibbotson Associates. All returns are net of fees. Just the positive Sharpe ratios are included. We exclude the index mutual funds. We “Windsorized” a few outliers that were larger than 2 in absolute terms. The Sharpe ratio is calculated for each portfolio for the complete return series that are not backfilled. We assign the portfolio Sharpe ratio to every date for which the portfolio exists in our sample. We report the cross-sectional mean Sharp ratios and standard deviation of the mean Sharpe ratio in the table. Panel A: Sharpe Ratios Over All Time Series For All Portfolios- Time Invariant 0.25 0.20 0.15 0.10 0.05 0.00 Jun-93 Jun-94 Jun-95 Jun-96 Jun-97 Jun-98 Jun-99 Jun-00 Jun-01 Jun-02 Jun-03 Jun-04 Dec-93 Dec-94 Dec-95 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00 Dec-01 Dec-02 Dec-03 Dec-04 Jun-05 SR_HF SR_Inst SR_MF SR_HF 0.1218 SR_Inst 0.1368 SR_MF 0.1037 SE_HF 0.0095 SE_Inst 0.0097 SE_MF 0.0041 42 Figure 2 (contd.) Sharpe Ratios-All Time Series per Portfolio This figure presents monthly Sharpe (1966) ratios for equity portfolios in the conglomerate sample. We drop the first 12 return observations from hedge funds older than 24 months to control for backfill bias. We used mutual and hedge funds with at least 12 observations and institutional funds with at least 8 quarterly observations. Our institutional and mutual fund samples are free from this bias. We considered monthly returns from April 1993 to December 2005. Institutional quarterly returns start in June 1993 as they are subject to backfill bias before June 1993. The one month T-Bill rate is Ibbotson Associates. All returns are net of fees. Just the positive Sharpe ratios are included. We exclude the index mutual funds. We “Windsorized” a few outliers that were larger than 2 in absolute terms. The Sharpe ratio is calculated for each portfolio for the complete return series that are not backfilled. We assign the portfolio Sharpe ratio to every date for which the portfolio exists in our sample. We report the cross-sectional mean Sharp ratios and standard deviation of the mean Sharpe ratio in the table. Panel B: Sharpe Ratios Over All Time Series For Equity Portfolios- Time Invariant 0.25 0.20 0.15 0.10 0.05 0.00 Jun-93 Jun-94 Jun-95 Jun-96 Jun-97 Jun-98 Jun-99 Jun-00 Jun-01 Jun-02 Jun-03 Jun-04 Dec-93 Dec-94 Dec-95 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00 Dec-01 Dec-02 Dec-03 Dec-04 Jun-05 SR_HF SR_Inst SR_MF SR_HF 0.1198 SR_Inst 0.1168 SR_MF 0.0949 SE_HF 0.0150 SE_Inst 0.0124 SE_MF 0.0050 43 Figure 2 (contd.) Sharpe Ratios-All Time Series per Portfolio This figure presents monthly Sharpe (1966) ratios for NON-equity portfolios in the conglomerate sample. We drop the first 12 return observations from hedge funds older than 24 months to control for backfill bias. We used mutual and hedge funds with at least 12 observations and institutional funds with at least 8 quarterly observations. Our institutional and mutual fund samples are free from this bias. We considered monthly returns from April 1993 to December 2005. Institutional quarterly returns start in June 1993 as they are subject to backfill bias before June 1993. The one month T-Bill rate is Ibbotson Associates. All returns are net of fees. Just the positive Sharpe ratios are included. We exclude the index mutual funds. We “Windsorized” a few outliers that were larger than 2 in absolute terms. The Sharpe ratio is calculated for each portfolio for the complete return series that are not backfilled. We assign the portfolio Sharpe ratio to every date for which the portfolio exists in our sample. We report the cross-sectional mean Sharp ratios and standard deviation of the mean Sharpe ratio in the table. Panel C: Sharpe Ratios Over All Time Series For Non-Equity Portfolios- Time Invariant 0.25 0.20 0.15 0.10 0.05 0.00 Jun-93 Jun-94 Jun-95 Jun-96 Jun-97 Jun-98 Jun-99 Jun-00 Jun-01 Jun-02 Jun-03 Jun-04 Dec-93 Dec-94 Dec-95 Dec-96 Dec-97 Dec-98 Dec-99 Dec-00 Dec-01 Dec-02 Dec-03 Dec-04 Jun-05 SR_HF SR_Inst SR_MF SR_HF 0.1236 SR_Inst 0.1688 SR_MF 0.1278 SE_HF 0.0122 SE_Inst 0.0155 SE_MF 0.0076 44 Figure 3 Sharpe Ratios-Limited Time Series per Portfolio This figure presents monthly Sharpe (1966) ratios for all portfolios in the conglomerate sample. In contrast to Figure 1, The Sharpe ratios are time varying. For each report date we calculate Sharpe ratios over the previous n months. N is minimum 12 months or 8 quarters and maximum 36 months. The one month T-Bill rate is Ibbotson Associates. All returns are net of fees. Just the positive Sharpe ratios are included. We exclude the index mutual funds. We “Windsorized” a few outliers that were larger than 2 in absolute terms. The Sharpe ratio is calculated for return series that are not backfilled. We report the cross-sectional mean Sharp ratios and standard deviation of the mean Sharpe ratio in the table. Panel A: Time Varying Monthly Sharpe Ratios- All Portfolios 0.35 0.25 0.15 0.05 Sep-95 Sep-96 Sep-97 Sep-98 Sep-99 Sep-00 Sep-01 Sep-02 Sep-03 Sep-04 Mar-95 Mar-96 Mar-97 Mar-98 Mar-99 Mar-00 Mar-01 Mar-02 Mar-03 Mar-04 -0.05 -0.15 -0.25 SR_HF SR_Inst SR_MF SR_HF 0.1645 SR_Inst 0.1714 SR_MF 0.1277 SE_HF 0.0161 SE_Inst 0.0167 SE_MF 0.0074 45 Mar-05 Sep-05 Figure 3 (contd.) Sharpe Ratios-Limited Time Series per Portfolio This figure presents monthly Sharpe (1966) ratios for equity portfolios in the conglomerate sample. In contrast to Figure 1, The Sharpe ratios are time varying. For each report date we calculate Sharpe ratios over the previous n months. N is minimum 12 months or 8 quarters and maximum 36 months. The one month T-Bill rate is Ibbotson Associates. All returns are net of fees. Just the positive Sharpe ratios are included. We exclude the index mutual funds. We “Windsorized” a few outliers that were larger than 2 in absolute terms. The Sharpe ratio is calculated for return series that are not backfilled. We report the cross-sectional mean Sharp ratios and standard deviation of the mean Sharpe ratio in the table. Panel B: Time Varying Monthly Sharpe Ratios- Equity Portfolios 0.35 0.25 0.15 0.05 Sep-95 Sep-96 Sep-97 Sep-98 Sep-99 Sep-00 Sep-01 Sep-02 Sep-03 Sep-04 Mar-95 Mar-96 Mar-97 Mar-98 Mar-99 Mar-00 Mar-01 Mar-02 Mar-03 Mar-04 -0.05 -0.15 -0.25 SR_HF SR_Inst SR_MF SR_HF 0.1760 SR_Inst 0.1538 SR_MF 0.1212 SE_HF 0.0245 SE_Inst 0.0216 SE_MF 0.0090 46 Mar-05 Sep-05 Figure 3 (contd.) Sharpe Ratios-Limited Time Series per Portfolio This figure presents monthly Sharpe (1966) ratios for NON-equity portfolios in the conglomerate sample. In contrast to Figure 1, The Sharpe ratios are time varying. For each report date we calculate Sharpe ratios over the previous n months. N is minimum 12 months or 8 quarters and maximum 36 months. The one month T-Bill rate is Ibbotson Associates. All returns are net of fees. Just the positive Sharpe ratios are included. We exclude the index mutual funds. We “Windsorized” a few outliers that were larger than 2 in absolute terms. The Sharpe ratio is calculated for return series that are not backfilled. We report the cross-sectional mean Sharp ratios and standard deviation of the mean Sharpe ratio in the table. Panel B: Time Varying Monthly Sharpe Ratios- Non-Equity Portfolios 0.35 0.25 0.15 0.05 Sep-95 Sep-96 Sep-97 Sep-98 Sep-99 Sep-00 Sep-01 Sep-02 Sep-03 Sep-04 Mar-95 Mar-96 Mar-97 Mar-98 Mar-99 Mar-00 Mar-01 Mar-02 Mar-03 Mar-04 -0.05 -0.15 -0.25 SR_HF SR_Inst SR_MF SR_HF 0.1605 SR_Inst 0.2069 SR_MF 0.1557 SE_HF 0.0216 SE_Inst 0.0265 SE_MF 0.0134 47 Mar-05 Sep-05 Figure 4 This figure presents the cross sectional averages of time varying alphas for all three industries. Both equally and asset weighted alphas are reported. To estimate the seven factor model, we allowed mutual fund and hedge fund returns to go back as far as April 1990 if they were reported before 1993. We use institutional returns from as far back as June 1990 for time varying coefficient estimation. For analysis we exclude the coefficient dates for which funds backfilled returns. The seven factors are: four Carhart (1997) factors: market risk premium, size, value, and the momentum factor; MSCI world index excluding the U.S., Merrill Lynch high yield bond C index, Merrill Lynch global governments bond index. We use Flexible Least Squares in estimation. The time series averages are reported below. 0.6 Time Varying Alphas: All Portfolios 0.4 0.2 0 pr -9 3 O ct -9 3 A pr -9 4 O ct -9 4 A pr -9 5 O ct -9 5 A pr -9 6 O ct -9 6 A pr -9 7 O ct -9 7 A pr -9 8 O ct -9 8 A pr -9 9 O ct -9 9 A pr -0 0 O ct -0 0 A pr -0 1 O ct -0 1 A pr -0 2 O ct -0 2 A pr -0 3 O ct -0 3 A pr -0 4 O ct -0 4 A pr -0 5 a_ew_hf a_vw_hf a_ew_inst a_vw_inst a_ew_mf a_vw_mf a_vw HF A -0.2 -0.4 -0.6 a StD HF AvgAst HF Cnt HF a_vw Inst a StD Inst AvgAst Inst Cnt Inst a_vw MF a StD MF AvgAst MF Cnt MF -0.12 0.08 141 174 0.03 0.04 2,602 48 394 -0.10 0.03 461 709 Table 5 Time Invariant Sharpe Ratio Regressions The dependent variable is the portfolio Sharpe Ratio. The Sharpe ratio is calculated over all available dates. Dum mf and dum inst are mutual and institutional fund dummies respectively, hedge fund dummy is omitted. Age is estimated in months over all reported returns for a fund including backfill return dates, the distribution is Windsorized at 1980. Average assets are average assets in millions of dollars. Average revenues is an average of the product of monthly fees and assets. The revenues are reported in millions. Average fees are reported on monthly basis in decimals. For mutual funds expense ratios are used when available. For institutional funds estimates fees are used, while for hedge funds performance fees are added for periods of positive returns. Family assets are a sum of average portfolio assets in conglomerate in millions. Equity dummy is one for equity. Last four columns are reported separately for equity and fixed income portfolios. All regressions except for the first column are estimated with fixed effects- LSDV1 for industry dummies and LSDV3 for conglomerate dummies. Fees are normalized within portfolio type group around the All Variable Intercept Dum MF Dum Inst Age Since Start Avg Assets Avg Revenues Normalized Fees Family Assets Equity Dummy Adj R-Sq Simple OLS 0.1653 *** 0.0197 * 0.0543 *** -2.1E-04 *** -3.6E-06 * 0.0212 *** 0.0000 ** -0.0624 *** 0.0297 0.1735 0.0124 0.0360 -1.1E-04 -3.5E-06 0.0174 *** ** * * *** 0.1899 *** 0.0005 0.0262 * -8.2E-05 1.4E-06 -0.0128 *** -0.0679 *** 0.0991 -0.0557 *** 0.0974 Fixed Income Equity *** *** *** *** Fixed Income 0.2164 *** -0.0058 0.0218 -1.1E-04 1.1E-06 -0.0320 *** Equity 0.1159 *** 0.0050 0.0357 ** -6.3E-05 1.5E-06 -0.0038 Fixed Effects 0.1736 *** 0.0978 0.0071 0.0188 0.0290 0.0548 -1.3E-04 -1.0E-04 -1.6E-06 -9.9E-06 0.0140 0.0247 0.0371 0.1329 0.0525 0.1253 49 Table 6 Sharpe Ratios Over Four Non Overlapping Time Periods This table reports results of pooled cross sectional fixed effects regressions. The dependent variable is the portfolio Sharpe ratio. At least 8 quarterly or 12 monthly observations are used over non overlapping time periods. The first time period is for 1993 through 1996; the second -through 1999, the third- through 2002, the fourth - through 2005. All regressions are estimated with fixed effects- LSDV1 for industry dummies and LSDV3 for conglomerate dummies. Age TV is age of all available returns to the end of fund life or period end, including the backfilled returns. Fees are normalized within portfolio type group around the mean of 1. *** indicate significance at 1 %, while * at 10%. Variable Intercept Dum MF Dum Inst Age TV Avg Assets Avg Revenues Normalized Fees Equity Dummy Adj R-Sq t1 0.1716 *** -0.0029 0.1097 *** -5.3E-04 *** 9.0E-06 -0.0169 0.1170 *** 0.1759 0.1916 *** -0.0010 0.1220 *** -6.1E-04 *** 1.6E-06 -0.0205 ** 0.1232 *** 0.1796 -0.2673 0.2570 0.4625 -2.2E-03 1.1E+00 0.0081 *** *** *** *** *** t2 0.0822 *** -0.0333 0.0151 1.9E-04 * 2.9E-06 -0.0142 ** 0.1055 *** 0.1274 0.2107 -0.0814 -0.0268 4.8E-05 4.1E-06 -0.0105 *** *** t3 0.2307 *** -0.0782 *** -0.0138 -2.4E-05 9.6E-07 -0.0275 *** -0.2245 *** 0.2856 0.2005 0.0885 0.0789 -4.5E-05 -5.3E-06 0.0253 *** *** *** *** *** t4 0.2198 *** 0.0677 *** 0.0632 *** 9.3E-06 1.3E-06 -0.0088 -0.0271 ** 0.1094 ** ** 0.2440 *** 0.2032 -0.2375 *** 0.2780 -0.0407 *** 0.1241 50 Table 7 Time Invariant Alphas Regressions The dependent variable is the portfolio average time varying alpha. The alpha is calculated over all available dates, excluding backfill return dates from seven factor FLS regressions as in chart 3. Dum mf and dum inst are mutual and institutional fund dummies respectively, hedge fund dummy is omitted. Age is estimated in months over all reported return dates, the distribution is Windsorized at 1980. Average assets are in millions. Average revenues is an average of the product of monthly fees and assets in millions. Average fees are reported on monthly basis in decimals. For mutual funds expense ratios are used when available. For institutional funds estimates fees are used, while for hedge funds performance fees are added for periods of positive returns. Family assets are a sum of average portfolio assets in conglomerate in millions. Alpha st dev is standard deviation of time varying alpha. The equity dummy is one for equity. All regressions except for the first column are estimated with fixed effectsLSDV1 for industry dummies and LSDV3 for conglomerate dummies. Fees are normalized within portfolio type group around the mean of 1. *** ind All Variable Intercept Dum MF Dum Inst Age Since Start Avg Assets Avg Revenues Normalized Fees Family Assets St Dev Alpha Equity Dummy Adj R-Sq Simple OLS -0.3829 *** 0.1997 *** 0.7584 *** -0.0014 *** -1.6E-05 0.0564 ** 0.0000 0.5825 *** 0.0963 ** 0.0542 -0.3221 0.2253 0.7168 -0.0011 -1.9E-05 0.0587 *** *** *** *** ** ** -0.2762 0.1880 0.6940 -0.0010 -2.3E-06 *** *** *** *** Fixed Income Equity *** ** *** ** ** Fixed Income -0.0887 0.3911 *** 0.4965 *** -0.0017 *** 1.6E-06 -0.0986 *** 0.2579 ** 0.0991 0.6627 *** 0.1304 0.3500 *** 0.1073 Equity -0.4153 *** 0.1570 * 0.8830 *** -0.0004 -1.3E-05 0.0080 0.6596 *** 0.1280 Fixed Effects -0.2012 ** -0.4511 0.4308 *** 0.2021 0.4975 *** 0.9497 -0.0017 *** -0.0005 -1.1E-05 -5.0E-05 0.0603 * 0.0771 -0.0439 ** 0.5346 *** 0.0507 0.0932 0.5647 *** 0.0913 ** 0.0924 51 Table 8 Alphas Over Four Non Overlapping Time Periods This table reports results of pooled cross sectional fixed effects regressions. The dependent variable is the portfolio seven factor alpha. The time varying alpha is estimated over all but back filled return dates, but is average over one of the four periods. At least 8 quarterly or 12 monthly observations are used over non overlapping time periods for both estimation and averaging. The first time period is for 1993 through 1996; the second -through 1999, the third- through 2002, the fourth - through 2005. All regressions are estimated with fixed effects- LSDV1 for industry dummies and LSDV3 for conglomerate dummies. Age TV is age of all available returns to the end of fund life or period end, including the backfilled returns. Fees are normalized within portfolio type group around the mean of 1. Alpha st dev is standard deviation of time varying alpha. *** indicate significance at 1 %, while * at 10%. Variable Intercept Dum MF Dum Inst Age TV Avg Assets Avg Revenues Normalized Fees St Dev Alpha Equity Dummy Adj R-Sq t1 -0.1911 ** -0.0736 -0.0411 -5.80E-04 -1.25E-05 0.0184 0.2935 ** 0.2704 *** 0.1074 -0.1398 -0.0875 -0.0432 -6.75E-04 -7.11E-06 -0.0410 0.2989 ** 0.2921 *** 0.1100 -0.1654 0.2763 0.4986 -1.92E-03 -2.55E-05 0.0539 *** *** *** ** ** t2 -0.1046 0.2275 ** 0.4594 *** -1.85E-03 *** -6.70E-06 -0.0525 * -0.2510 0.4006 *** 0.1589 -0.1593 0.1915 0.8238 -1.68E-03 -5.10E-06 0.0078 * ** *** *** t3 -0.1036 0.1804 ** 0.8367 *** -1.76E-03 *** -3.87E-06 -0.0733 *** 0.7253 *** 0.1196 ** 0.1517 -0.1245 0.3677 0.9636 -9.22E-04 -6.79E-06 0.0216 * *** *** *** t4 -0.1038 0.3486 *** 0.9527 *** -8.85E-04 *** -1.27E-06 -0.0145 0.1378 -0.0886 ** 0.1237 -0.2829 * 0.3516 *** 0.1599 0.7051 *** 0.0730 0.1474 0.1279 -0.1049 ** 0.1242 52