Fund characteristics and their impacts on the performance of hedge funds an emperical evidence for european and north american hedge fund industries

FUND CHARACTERISTICS AND THEIR IMPACTS ON THE PERFORMANCE OF HEDGE FUNDS: AN EMPERICAL EVIDENCE FOR EUROPEAN AND NORTH AMERICAN HEDGE FUND In this paper, we investigate the impacts of fu

FUND CHARACTERISTICS AND THEIR IMPACTS ON THE PERFORMANCE OF HEDGE FUNDS: AN EMPERICAL EVIDENCE FOR EUROPEAN AND NORTH AMERICAN HEDGE FUND

In this paper, we investigate the impacts of fund characteristics on the performance

of North American hedge funds (NAHFs) and European hedge funds (EHFs) over a

period of 1st January 2001 - 31st December 2013 The study period is divided into

two sub-sample periods:(1) non-crisis sub-sample period (1st January 2001- 31st

December 2007) and (2) during-crisis and post-crisis sub-sample period (1st

January 2008 - 31st December 2013), to measure the impact of the global crisis on

the results of our study Our samples consist of 325 North American hedge funds

(NAHFs) and 166 European hedge funds (EHFs), which all survive through the

sample period Using the cross sectional regression method, the omega ratios of

hedge funds in each sample are regressed against their fund characteristics for a

full sample period and sub-sample periods Results show evidence of the significant

impacts of fund strategies on the performance of both NAHFs and EHFs More fund

characteristics are found to have significant impacts on the performance of NAHFs

and EHFs during the non-crisis sub-sample period As compared to NAHFs, the

performance of EHFs is rarely affected by theirs fund characteristics during the

crisis- and post-crisis sub-sample period

1 INTRODUCTION

Being well-known as an innovative financial product since the late 1990, a hedge fund often possesses many special characteristics designed by its managers to strike for an absolute performance rather than a relative performance as compared to the traditional investments such as mutual funds Knowing individual impacts of the designed fund characteristics of hedge funds is undoubtedly important to both hedge fund managers and investors in order to achieve better fund management and fund selection, respectively

Studies on the impact of hedge fund characteristics on their performance are often conducted with selective numbers of fund characteristics, namely incentive fees, management fees, fund age, fund strategies, fund size, lock-up period, leverage ratio, redemption period, redemption notification period, hurdle rate, high-water mark, and minimum investment However, it is found that results are often mixed and inconclusive In addition, it is also found that not all fund characteristics, that are believed to shed light on the way a hedge fund is managed, are often discussed together in a single study This could be due to the lack transparency or information provided by hedge fund managers

With an attempt to expand the study by Nguyen et al (2011), we investigate the impact of fund characteristics on the performance of hedge funds in three regions: Asian, Europe, and North America Thus, three regional samples of hedge funds are then selected for this study: (1) a sample of 325 North American hedge funds (NAHFs), and (3) a sample of 166 European hedge funds (EHFs) Information on fund characteristics and returns for all the three samples of hedge funds are fully available over the period of 1st January 2001 – 31st December 2013 A cross-sectional regression analysis is carried out, from which the omega ratios of hedge funds in each of the three samples are regressed against their own fund characteristics The omega ratios (Keating and Shadwick, 2002) are used to overcome the the possibly ambiguous results produced by the Sharpe ratio (Ackermann et al., 1999; Do et al, 2005) To our best knowledge, this is the first time the omega ratio is used as a performance measure in this kind of study It is also believed that the use of omega ratios will better reflect the actual performance of hedge funds during the study period Results show evidence of the significant impacts of fund strategies on the performance of both NAHFs and EHFs More fund characteristics are found to have significant impacts on the performance of NAHFs and EHFs during the non-crisis sub-sample period The performance of EHFs is found to be less predictable during-crisis- and post-crisis sub-sample period

The rest of the paper is organized as follows: related studies are reviewed in Part 2, followed

by Part 3 with hypothesis and testing method, main findings are presented in Part 4, and key conclusions are given in the final part, Part 5

2 LITERATURE REVIEW

In literature, fund characteristics which are mostly discussed are incentive fees, management fees, fund age, fund strategies, fund size, lock-up period, leverage ratio, redemption period, redemption notification period, hurdle rate, high-water mark, and minimum investment The impact of the incentive fee on the performance of hedge funds is found to be mixed in literature A number of studies find that the incentive fee has a significant positive impact on the performance of hedge funds (Ackermann, McEnally, Ravenscraft, 1999; Edwards and Caglayan, 2001; Fung, Xu, and Yau, 2002; Brown, Goetzmann, and Liang, 2004; and Do, Faff, and Wickramanayake, 2005) This implies that incentive fees do motivate fund managers to perform better On the other hand, a number of studies find no evidence of such significant positive impacts of incentive fees on the performance of hedge funds, namely Schneeweis, Kazemi, and Martin (2002), Brown, Goetzmann, and Liang (2004), and Fung,

Xu, and Yau (2004) Interestingly, some studies discover a negative impact of incentive fees

on the performance of hedge funds, i.e Koh, Koh, and Teo (2003), Kouwenberg & Ziemba (2004), and Steri, Giorgino, and Viviani (2009)

Unlike incentive fees, management fees are like fixed salaries paid to fund managers in order

to manage the funds Therefore, this fund characteristic is expected to be independent from the performance of a fund In fact, many studies find no significant evidence of the impact of this fund characteristic on fund returns (Liang, 1999; Edwards and Caglayan, 2001; Steri, Giorgino, and Viviani, 2009) However, a number of studies find negative impact of this fund characteristic on the performance of hedge funds, namely Ackermann, McEnally, and Ravenscraft (1999), Brown, Goetzmann, and Liang (2004), Kouwenberg & Ziemba (2004),

Do, Faff, and Wickramanayake (2005) The findings of these studies imply that funds with higher management fees tend to perform poorly

Findings on the impact of fund sizes on the performance of hedge funds are mixed Studies that find the significant positive impact of fund sizes on the performance of hedge funds are Liang (1999), Fung, Xu, and Yau (2002), Liang (2000), Brown, Goetzmann, and Liang (2004), Kouwenberg & Ziemba (2004), Do, Faff, and Wickramanayake (2005), and Nguyen

et al (2011) On the other hand, some studies find no relationship between fund size and the

performance of hedge funds (Regoriou and Rouah, 2002; Koh, Koh, and Teo, 2003) Moreover, a number of studies find that a fund’s size has a negative impact of on its performance (Schneeweis, Kazemi, and Martin, 2002; Harri and Brorsen, 2004; Kat and Paloro, 2005; Kat and Palaro, 2007)

For fund age, the relationship between this fund characteristic and the performance of hedge funds provides mixed results in literature In Liang (2000) and Edwards and Caglayan (2001), the fund age is found to have a significant positive impact on the returns of a hedge fund However, a contradicting result is found in Liang (1999), Fung, Xu, and Yau (2002), Brown, Goetzmann, and Liang (2004), Kouwenberg & Ziemba (2004) In terms of risks, Kouwenberg & Ziemba (2004) and Do, Faff, and Wickramanayake (2005) find that the older

a fund is, the higher standard deviation and systematic risk it seems to have Besides the above findings, some studies find no evidence at all that a fund’s age could have any significant impacts on performance of the fund (Ackermann, McEnally, Ravenscraft, 1999; Koh, Koh, and Teo, 2003; Kat and Palaro, 2007; Steri, Giorgino, and Viviani, 2009)

Most studies find that lock-up periods have significant positive impacts on the performance

of hedge funds (see Liang, 1999; Schneeweis, Kazemi, and Martin, 2002; Agarwal, Daniel, and Naik, 2007; Nguyen et al., 2011) However, the impact varies across performance measures and study periods For example, in Schneeweis, Kazemi, and Martin (2002), the positive effects of a lock-up period disappear when risk-adjusted returns are used for U.S equity hedge funds during 1996-2000 Moreover, Fung, Xu, and Yau (2002) find that the positive impact of lock-up periods on managers’ stock selection skills is not persistent throughout their sample period

In examining the impact of leverage ratios on the performance of hedge funds, Agarwal and Naik (2000) find some evidence although they are not statistically significant in most cases This finding implies that funds with leverage do not necessarily outperform or underperform those that do not employ leverage In another study by Fung, Xu, and Yau (2002), contrasting results are found, where the leverage ratio has a strong and positive impact on the Sharpe ratio as well as Jensen’s alpha, but not on the systematic risk The results testify to the misconception among investors that higher leverage employed by a hedge fund increases its risk Nguyen et al (2011) also find significant positive impact of leverage (0.046) on the annual returns of AHFs at the 5% level

For minimum investment, Koh, Koh, and Teo (2003) do not find any significant evidence that a fund’s minimum investment has any relation with its monthly returns For redemption period, Koh, Koh, and Teo (2003) find a positive correlation of the redemption period and the

returns of their funds, implying that a longer redemption period allows a fund to unwind from its position For redemption notification period, Steri, Giorgino, and Viviani (2009) find that the redemption notification period has a significant negative impact on the performance of Italian hedge funds at 5% level, implying that hedge funds that require longer redemption notification periods do not perform better than funds that require shorter ones

Clearly, from the above-reviewed literature, it is worth to note that (1) mixed results were found for individual fund characteristics, (2) not many hedge fund characteristics were explored in a single study, (3) none of similar studies has been conducted for hedge fund industries in Europe and North America Thus, we attempt to fill in these gaps

3 RESEARCH METHODOLOGY

3.1 The EurekaHedge Databases

EurekaHedge Inc is the world’s largest independent data provider and alternative research firm since 2001 EurekaHedge Inc specialises in hedge fund databases for four different regions: Asia, North America, Europe, and Latin America Besides performance data, EurekaHedge Inc., also records other useful information for hedge funds which are also useful for this study The additional information provided by EurekaHedge Inc are namely as company name, starting and ending dates, fund code, fund name, fund strategy, geographical market, size, managers' names, total asset under management, lockup period, penalty, redemption and notification period, minimum investment, listed in stock exchange, management fee and performance fee

Due to the comprehensive coverage of hedge funds offered by the EurekaHedge Inc., its three regional databases are then chosen for this study to establish two samples: (1) a sample

of North American hedge funds (NAHFs), and (23) a sample European hedge funds (EHFs)

By November 2013, the total numbers of funds present in the three databases for NAHFs and EHFs are 6629, and 7426, respectively The reported monthly returns in EurekaHedge are all net of fees, i.e management fees and performance fees Using the definition used in Ackermann et al (1999), a monthly return for a fund defined by EurekaHedge is the percentage of change in the fund’s net asset value during the month as compared to that at the beginning of that month, assuming that any distribution such as dividends are reinvested on

the reinvestment date The monthly returns are perceived to bring a greater accuracy in the estimates of risk (standard deviation)

3.2 The Study Period

The study period chosen for this study is from 1 January 2001 to 31 December 2013 This is due to the availability of the study at the time it was conducted This study period was then divided into two sub-periods: 1 January 2001 - 31 December 2007, 1 January 2008 - 31 December 2013, to observe if the results produced for this study may be affected by the 2008 global financial crisis

3.2 Two Regional Samples of Hedge Funds

To fulfil the objective set for this study - to examine the impact of characteristics of a hedge fund on its performance -, we excluded hedge funds without information on their fund characteristics, and/or hedge funds that were closed by 31st December 2013, and/or hedge funds that do not have full monthly return data for the entire period of 1st January 2001 – 31st December 2013 With the self-imposed criterion in this study, the three samples chosen consist of 325 NAHFs and 166 EHFs, respectively

3.3 Testing for Possible Data Bias for the Chosen Three samples

To ensure that the chosen samples can represent well the two hedge fund industries in North America and European, a two-sample t-test to test is carried out to test if there is any statistically significant difference between the average monthly returns of each sample of hedge funds and its population present in each of the regional databases

3.2 Performance Measure – Omega Ratio

In this study we adopt a new performance measure named “Omega ratio", which was first proposed by Keating and Shadwick (2002) According to Keating and Shadwick (2002), the Omega ratio takes into account the entire distribution of returns for individual hedge funds

As compared with other traditional performance measures such as alpha, average return, and sharp ratio, the omega ratio is therefore, perceived to be a better performance measure for hedge funds, whose return distributions are often non-normally distributed

The Omega function is constructed with regards to the quality of a bet that returns of a hedge fund would be above a predetermined threshold denoted as “r” percent The bet is successful

if returns of the fund are in fact above “r” percent and vice versa The formula for the Omega ratio at the threshold of “r” percent is given below

The above Omega function for the distribution F(x) is obtained by taking all possible returns,

r, between two boundaries of “a” and “b” This function is mathematically equivalent to the distribution itself as it contains all of the information entailed The interpretation for an Omega ratio is quite straight-forward At a given threshold of “r” percent, a hedge fund with

a higher Omega ratio is preferred over a hedge fund with a lower Omega ratio A flatter Omega curve implies a higher risk In this study, Omega ratios for all NAHFs and EHFs in the samples are computed at a chosen threshold of zero percent The use of zero percent as a threshold is based on the assumption that the targeted minimum return ought to be a positive return

For hedge funds in the study samples, omega ratios are computed for individual funds throughout the sample period of 1st January 2001- 31st December 2013 The threshold "r" of 0%

is chosen for this study based on a very basic rationality that positive returns are always preferred to negative returns for hedge funds

3.3 Hypothesis

The general hypothesis for testing in this study is that “Fund characteristics do not have any significant impacts on the performance of hedge funds in North America, and Europe”

3.4 Testing Methods

The Proposed Testing Model

Following Koh et al (2003) and Agarwal et al (2007), to test the impact of fund characteristics of NAHFs and EHFs on their performance, a cross-sectional regression model

is employed, where omega ratios of hedge funds from the three samples are regressed against their fund characteristics as mentioned above The cross-sectional regression method is the most appropriate method since most of the fund characteristics of hedge funds are fixed over the years

The dependent variable for the three regression models is the omega ratio Independent variables are 14 fund characteristics, i.e (1) fund size, (2) subscription frequency, (3) redemption frequency, (4) redemption notification frequency, (5) hurdle rate, (6) high-water mark provision, (7) management fee, (8) performance fee, (9) being listed in a stock market, (10) fund age, (11) leverage, (12) dividend, (13) penalty, and (14) fund strategies

The following models will be run for our samples of hedge funds

Where

is omega ratio for NAHF i;

is omega ratio for EHF i;

ln(FS i ) is the natural logarithm of fund size for fund i;

SUB i is the subscription frequency for fund i;

RED i is the redemption frequency for fund i;

REDNOT i is the redemption notification frequency for fund i;

HR i is the hurdle rate for fund i;

HM i is the high-water mark for fund i;

MF i is the management fee for fund i;

PF i is the performance fee for fund i;

LSE i is being listed in a stock exchange for fund i;

FA i is the fund age for fund i;

LE i is the leverage for fund i;

DIV i is dividend for fund i;

PE i is the penalty for fund i;

S ik is the dummy variable for strategies k for fund i, when k takes a value from 2 to n n is the number of strategies

G ik is the dummy variable for geographical mandates k for fund i, when k takes a value from

2 to m m is the number of geographical mandates

Si1 and G ik are the dummy variables that is omitted in the above equation to avoid dummy trap that causes perfect collinearity problem in the regression equation (Gujarati & Porter, 2009)1 In addition, they served as the terms of comparison for other strategies For the For the NAHF sample, Si1 and G ik are Arbitrage strategy and Asia including Japan geometrical mandate For the EHF sample, Si1 and G ik are Arbitrage strategy and Asia excluding Japan geometrical mandate

The above models are used to test for the full sample period of 1st January 2001 - 31st December 2013, and also for two sub-sample periods: (1) 1st January 2001 - 31st December

2007 and (2) 1st January 2008 and 31st December 2013 Test results produced for the two sub-sample periods may shed light on the effect of the 2008 global financial crisis on the performance of both NAHFs and EHFs

a Tests for Model Fit

The main potential problem in a cross-sectional model is correlation in space or spatial autocorrelation caused by possible misspecification in the model (See Gujarati and Porter, 2009)

Autocorrelation can occur as correlation in time series data and in space for cross-sectional data.2 In other words, autocorrelation is the lag correlation found in a given series with itself When the autocorrelation is believed to exist among the disturbances denoted as ui, it can be expressed as follows

Heteroscedasticity (ARHCH) Test, Breusch Godfrey (LM) Test, and Normality test, are carried out

4 FINDINGS

4.1 Data Bias Testing for the Three Samples

P-values of two-sample t-test statistics shown in Table 1 are all greater than 0.05, suggesting that the two selected samples of NAHFs and EHFs for this study are good representations for their respective hedge fund industries present in the EurekaHedge databases

Table 1: Results of Two Sample t-test for the Three Samples

In this table, results of two sample t-test are presented for the three samples: sample of 325 NAHFs, and sample of 166 EHFs The difference between each sample's and its population's average monthly returns is given in the second column of the table Results of the respective t-statistics and its p-value are given in the last two columns of the table

Pair Samples

Difference in sample average returns t-statistics p-value

Sample of 325 NAHFs & Population of

NAHFs Present in the Database 0.154 -0.757 0.449

Sample of 166 EHFs & Population of EHFs

Present in the Database 0.110 -0.947 0.3443

4.2 Cross Sectional Analysis

4.2.1 Results for the Sample of NAHFs

In Table 1 shows the results for only fund characteristics (subscription period and 12 fund strategies) that have significant impacts on the omega ratios of NAHFs over the full sample period of 1st January 2001 – 31st December 2013 Subscription period is found to have a significant positive impact (0.077) on the performance of the NAHFs at the 1% level, implying that NAHFs with longer subscription frequencies seem to have higher omega ratios

as compared to those with shorter ones

Interestingly, that all the twelve NAHF strategies, i.e (1) bottom up, (2) CTA/managed futures, (3) distressed securities, (4) event driven, (5) fixed income, (6) long short equities, (7) macro, (8) multi-strategies,(9) relative value, (10) top down, (11) value, and (12) others, are found to have significantly weaker impacts on the omega ratios of NAHFs as compared to that of the arbitrage strategy Surprisingly, many fund characteristics such as fund size,

management fee, performance fee, geographical mandate, leverage ratio, hurdle rate, water-mark provision, fund age, redemption period, redemption notification period, dividend payout are found to have no statistically significant impact on the omega ratio of NAHFs The adjusted R-square statistics shows that 49.7% of omega ratios of NAHFs in the sample can be explained by all their fund characteristics P-values of F, LM, and ARCH tests certify the fitness of the model

high-In Table 2, results for the sub-sample period of 1st January 2001 - 31st December 2007 show more fund characteristics with significant impacts on the omega ratios of NAHFs Fund size

is found to have a significant positive impact (0.212) on the omega ratio of NAHFs at the 1% level, implying that large funds may allow their managers to gain larger diversification effects, and thus perform better This finding is consistent with the findings in Liang (1999), Fung, Xu, and Yau (2002), Liang (2000), Brown, Goetzmann, and Liang (2004), Kouwenberg & Ziemba (2004), Do, Faff, and Wickramanayake (2005), and Nguyen et al (2011) In addition, performance fee is found to have a positive impact on the performance of NAHFs at the 10% level, suggesting that NAHF managers are motivated by the performance fee, and thus perform bettter In terms of liquidity, redemption period and redemption notification are also found to have positive impacts, i.e.0.003 and 0.016, on the performance

of NAHFs at the 1% level, respectively This suggests that longer redemption frequency and redemption notification period allow hedge fund managers to have more time to unwind their position so that further positive performance can be achieved before the fund is liquidated Similar to the results found for the full sample period, 12 above-mentioned strategies are also found to have significantly weaker impacts on the omega ratios of NAHFs as compared to that of the arbitrage strategy In terms of the goodness of fit, the adjusted R-square statistics suggest that 90.8% of the performance of NAHFs during 1st January 2001 – 31st December

2007 can be explained by all the fund characteristics P-values of F test, LM test, and ARCH test also certify the goodness of fit of the model

Findings for the crisis and post-crisis sub-sample period of 1st January 2008 – 31st December

2013 shown in Table 3, are somehow similar to those found for the full sample period, from which only subscription frequency and the 12 strategies are found to have significant impacts

on the performance of NAHFs The impact of longer subscription frequency during the crisis and post-crisis sub-sample period may suggest that a careful scrutiny over new subscription

to NAHFs, and thus leads to a longer subscription period, may generate positive impacts on the performance of NAHFs during this period The adjusted R-square statistics suggest that 100% of omega ratios of NAHFs in the sample can be explained by all their fund characteristics This may suggest that the internal structure of NAHFs allow them to be

“under-control” not only during the non-crisis period, but also during the crisis and post-crisis period P-values of F test, LM test, and ARCH test confirm the goodness of fit for the cross-sectional regression model run for this period