Inspirar para Transformar

Performance and Persistence

of Brazilian Hedge Funds
During the Financial Crisis

Marcelo Moura
Gustavo P. Joaquim

Insper Working Paper

WPE: 234/2011
 Inspirar para Transformar

Copyright Insper. Todos os direitos reservados.

É proibida a reprodução parcial ou integral do conteúdo deste

documento por qualquer meio de distribuição, digital ou im-
presso, sem a expressa autorização do
Insper ou de seu autor.

A reprodução para fins didáticos é permitida observando-sea

citação completa do documento
 P ERFORMANCE AND P ERSISTENCE OF B RAZILIAN
H EDGE F UNDS D URING THE F INANCIAL C RISIS

G U S TA V O P. J O A Q U I M †

I N S P ER – I N S TI TU TO DE E N S IN O E P ES Q U IS A

M A R C E LO L. M O U R A ‡

I N S P ER – I N S TI TU TO DE E N S IN O E P ES Q U IS A

† Rua Quatá, 300. São Paulo/SP, BRASIL. CEP 04546-042. Email: gustavopgj@al.insper.edu.br.; tel.:

55-11-4504-2430.

‡ Corresponding Author: Insper, Rua Quatá, 300. São Paulo/SP, BRASIL. CEP 04546-042. Email:

marcelom@insper.org.br; tel.: 55-11-4504-2435.

We thank Anbima (Associação Brasileira das Entidades dos Mercados Financeiro e de Capitais –

Brazilian Association of Financial and Capital Market Entities) for providing us with the dataset for this

study and also to useful comments of seminar participants at Anbima.

 P ERFORMANCE AND P ERSISTENCE OF B RAZILIAN
H EDGE F UNDS D URING THE F INANCIAL C RISIS

Abstract

This study investigates the performance and persistence of the Brazilian hedge fund market

using daily data from October 2007 to February 2011. A period containing what was

characterized by many as the worst world financial crisis since the great depression of the

1930’s. Despite the financial turmoil, results indicate the existence of a representative group

of funds with abnormal returns as well as evidence of a joint persistence of funds with time

frames of 1 to 3 months. Individual evaluations of the funds, however, indicate a reduced

number of persistent funds.

Keywords: multimarket funds, hedge funds, alpha calculation, performance evaluation,

persistence, Brazilian funds.

JEL: G11

1
I. Introduction

The Brazilian hedge fund industry, also known as the multimarket fund industry, has

shown rapid growth in recent years. In April 2010, Brazilian multimarket funds represented

around R$539 billion (approximately $315 billion).1 Three years ago, in 2007, this value was

R$295 billion.2 This increase depends on various factors in addition to internal factors related

to the sector itself3 but also the decline in the real interest rate in Brazil during this period

amidst the increase in investors’ attention toward this category of funds. Using daily data

from October 2007 to January 2010, this study investigates the performance and persistence

of the Brazilian hedge fund market. The main objective is to search for evidence of solid

hedge fund management in relation to market benchmarks and to test whether this

relationship is consistent throughout time, that is, whether the past performance of one fund

may be a good indicator of future performance. Finally, we seek to analyze the main

determinants of performance and persistence based on data on management fees,

performance fees and adopted strategies.

In the analysis of performance, it is possible to use a very large variety of performance

indicators. According to Eling (2009), these measures may be classified into five groups,

namely, return, risk, greatest moments, correlation and performance adjusted for risk. In this

study, the Sharpe ratio (1966) and Jensen’s alpha are used, which are performance measures

weighted by risk. According to Eling (2009), these types of performance measures are the

most important.

Regarding persistence, authors such as Park and Staum (1998) use the TASS database

(Tremont Advisory Shareholders Services) from 1986 to 1997, the contingency tables method

and Spearman’s rank correlation coefficient to conclude that there is persistence with periods

2
of one year. However, another work with similar specifications (Malkiel and Saha, 2005)

analyzes the period from 1996 to 2003 to conclude that there is no persistence over time

frames of one year. Furthermore, works such as Brown and Goetzman (2003) and Capocci,

Corhay and Hubner (2005), which use regressions as statistical methods and the Jensen’s

alpha as the best indicator of performances, did not find persistence for annual periods either,

with robust results given the database used: CISDM (Center for International Securities and

Derivatives Markets), HFR (Hedge Fund Research) and TASS.

Similar to the objectives of the present study, Agarwal and Naik (2000b) analyze a

time horizon of three years (1995-1998), subdivided into time frames of three months. In this

period, 167 funds are covered and analyzed in terms of performance according to the alpha

and information ratio. With respect to persistence, Agarwal and Naik (2000b) use the cross-

product ratio and regressions, reaching the conclusion that there is strong evidence to believe

quarterly persistence exists in fund performance.

It is noteworthy that according to Eling (2009), there are no clear guidelines in the

literature for the length of the period necessary to measure persistence. From one point of

view, larger samples would certainly provide better data, yet the hedge fund manager market

is dynamic, which means that very large samples are not appropriate to measure the quality of

managers, as they frequently change funds. In this study, we did not run into this second

problem, as the availability of data is small due to the recent organization of these funds in

Brazil.

Another important question is the choice of the analysis time horizon, that is, how

long the analyzed time frames are. Harri and Brorsen (2004) and Henn and Meier (2004)

show that this choice leads to large differences in the levels of persistence, as there is a

smoothness of returns for shorter periods. According to Henn and Meier (2004), this effect is

3
due to the presence of non-liquid assets in the portfolio of many funds as well as to the

administered return. That is, many managers try to maximize the smoothness of the return

time series. Another hypothesis, which is defended by Barès, Gibson and Gyger (2003) and

Jagannathan, Malakhov and Novikov (2006), is the effect known in the literature as hot-hand,

which assumes that assets under fund management that yield excellent returns in one period

will also do so in the following period.

Studies focused specifically on the Brazilian fund market are still relatively scarce.

Recently, Gomes and Cresto (2010) studied the performance of long-short multimarket funds

in Brazil, estimating the CAPM model with market timing via GMM. The results indicate

that some funds generate statistically significant Jensen’s alpha but are not robust for

different time frames, and in most cases, the market-timing coefficient is negative, thus

reducing the returns. Xavier (2008) analyzes the persistence of performance for 44

multimarket funds with equity and leverage, finding evidence of persistence in Sharpe ratio.

Summing up, with the improvement in the quality of databases in recent years, we

have observed an increasing number of works using quantitative finance techniques in the

evaluation of hedge funds (Eling and Faust, 2010). However, the results obtained in this

literature up until now are very contradictory, in large part due to the use of different time

frames, the application of different statistical methods and the use of different databases.

This article draws on existing literature by analyzing a representative emerging

market in the light of the recent financial turmoil from 2008/2009. We explore a unique data

set of Brazilian Hedge Funds with daily observation. It is worthwhile to point that due to the

Brazilian regulation, this data is audited by third parties and reported daily to the Brazilian

Securities Commission (CVM – Comissão de Valores Mobiliários).

4
 In order to achieve this goal, we first analyze performance through the accumulated

return, the Sharpe ratio (1966) and Jensen’s alpha (1968). The latter is analyzed with three

models using alternative factors. Second, we evaluate persistence using the contingency

tables method, Spearman’s rank correlation coefficient and a simple parametric regression.

Finally, we examine the relationship between the characteristics of the funds (i.e.,

management fees, performance fees and management strategy) and performance and

persistence. In addition, we study the influence of persistence on performance or, rather,

whether more persistent funds tend to present better or worse performance.

This work is divided into four sections, including this introduction. The next section

presents the pricing models, the performance indicators and the statistical methods used in the

study of performance. The third part describes the data used in this study and presents the

results. In the final section, the conclusions of this work are presented, along with limitations

and possible extensions.

II. Methodology

A. Performance Indicators

The first performance indicator used was the net return of management and

performance fees. Because we use daily data (for working days), the average monthly

cumulative returns of a fund i of a period of T working days is calculated by the following

equation:

22/T
 T 
Ri ac
   (1  Rit )  1
 t 1  ,

where Rit denotes the daily return of the fund. The Sharpe ratio was selected as one of

the methods of evaluating performance for its appeal and because it provides performance

rankings that are identical to those of most modern indices, such as the Modigliani index.

5
The Sharpe ratio represents the risk premium given one additional unit of total risk for the

fund. This can be calculated from the following formula:

E ( Rit )  R f
Si 
i ,

where Si is the Sharpe ratio of fund i during that period, Rf is the return of the risk-free asset

during that period and σi is the standard error of the returns of the fund i during that period.

The use of the Sharpe ratio is not unproblematic in the case of multimarket funds, as

this category of funds involves significant investments in the derivatives market, making the

return structure non-linear and thus leading to a lack of normality across the returns.

However, the Sharpe ratio is still widely applied in the literature (Eling, 2009) and is used by

many fund managers.

Additionally, to evaluate performance, Jensen’s alpha (1968) was also used, which is

the intercept of the regressions performed by the pricing models; it represents the capacity of

a manager to obtain abnormal returns, which are not inherent to the risk exposure factors of

the model, which in this case are two versions of CAPM.

This model is based on assumptions that no investor is large enough to change market

prices, all investors possess the same expectations and the same investment time horizon, all

parties use the Markovitz optimization process based on the risk-return criterion, all investors

have access to the same universe of investments (limited to assets negotiated in the market),

all investors can apply or borrow at the same rate, and there are no transaction or information

costs.

Mathematically, the equilibrium ratio of the model can be described as follows:

Rit  R f   i   i ( Rmt  R f )
,

6
where Rf is the risk-free asset return rate and βi is the beta coefficient, which measures the

correlation of the portfolio risk denoted by Rmt. This model can be estimated econometrically

using the index model represented by the following equation for the return series for asset i:

(1) Rit  R f   i   i ( Rmt  R f )   it

,

where is the noise estimation. Although many of the hypotheses described above

are difficult to verify empirically, and despite the impossibility of testing with real data, the

model is still widely applied in practice, especially as a guideline for investment decisions.

One critique of the CAPM specification in (1) is that the market index, which in our

case is Ibovespa (the main stock market Brazilian index), is not representative for the hedge

fund industry given that hedge funds allow short positions and investments in other assets not

listed on the market. Therefore, in an attempt to provide more consistency to our conclusions,

we opted to use a specific hedge fund index reported by Anbima, Associação Brasileira das

Entidades dos Mercados Financeiro e de Capitais – Brazilian Association of Financial and

Capital Market Entities, the IHFA (Anbima Hedge Fund Index) market index as well, which

will be explained below, as a representation of the market portfolio such that

(2) Rit  R f   i   i ( RIHFA,t  R f )   it

,

where RIHFA,t is the daily return of the IHFA. The inclusion of an index in pricing

models is widely used in the literature, as in Agarwal and Naik (2000a).

Authors such as Fung and Hsieh (1997) and Brown, Goetzmann and Ibbtson (1999)

stress the importance of the inclusion of factors specific to hedge funds, such as characteristic

indices (Brown et al., 1999). In the Brazilian case, the expression “Brazil kit” is common in

the multimarket fund market; it consists of taking positions bought on the market and sold in

7
American dollars at interest. To represent differences in styles, we propose the style factor

model, which includes the return of public title indices IRF-M and IMA-B4 as well as

variations in the exchange rate:

Rit  R f   i  1,i ( RIHFA,t  R f )   2,i ( RIRF  M ,t  R f )

(3)
 3,i ( RIMA B,t  R f )   2,i ( RE ,t  R f )   it
,

where , , and represent, respectively, the daily return of

investments in the IRF-M and IMA-B indices and the Brazilian Real/US dollar exchange rate,

R$/US$.

In specifications (2) and (3), coefficient i indicates the abnormal returns that fund i

obtains, which is a return obtained by a risk factor that is not taken into account in the

respective model. In other words, coefficient i indicates the ability of the managers of each

fund.

B. Persistence Indicators

To measure persistence, we used Spearman’s correlation coefficient, the contingency

table methods (i.e., the cross-product ratio and chi-square test) for two periods and the

parametric method based on the regression of present values with past values for the

performance indicators.

Spearman’s rank correlation coefficient is calculated for all prior and subsequent

periods. That is, if we are analyzing a time frame of 66 working days (or 3 months), the

coefficient captures the relationship between the performance ranking of the funds in the 66

prior days (in relation to some performance indicator) and the 66 days after a given date. To

analyze the entire period, this technique is used repeatedly. It is worth noting that unlike a

linear correlation, Spearman’s correlation seeks only a monotonic relationship among the

8
rankings in different periods. Persistence is observed when this coefficient is positive and

significant. This method is important because it allows us to identify the intensity and

direction of the relation and because it is non-parametric. That is, this method does not

require the assumption of a certain probability distribution. Because there are no ties between

classifications, Spearman’s rank correlation coefficient (SPR) between period X and period Y

is given by

6 ( R( X i )  R(Yi ))2
 ( X i , Yi )  1  i
,
n(n2  1)

where R(Xi) is the position of fund i in list X (prior period), R(Yi) is the position of

fund i in list Y (subsequent period), and n is the number of funds. According to Eling (2009),

the significance of this coefficient can be tested by Fisher’s T statistic, which follows a t-

student distribution with n – 2 degrees of freedom, which is given by

 SPR 
TSPR  n  2  
 1  SPR  .

Under the contingency table methods, analyses are based on the establishment of

winners and losers. A fund is a winner (W) or a loser (L) in relation to the median of funds

for a performance-specific measure. Because the approach assumes two periods, the funds

that are above the median in the two periods under analysis are considered WW (winners),

whereas those that are consistently below the median are LL (losers). The funds that improve

or worsen in their comparative performance over time respectively are WL (declining in their

relative performance over time) and LW (improving their relative performance over time).

This method, as with Spearman’s rank correlation coefficient, has the advantage of being a

non-parametric test.

9
 We can consider two test statistics using this approach. The first is the cross-product

ratio (CPR), which is the quotient of persistent funds with non-persistent ones given by

WW  LL
CPR 
WL  LW .

Under the null hypothesis that there is no persistence, it is expected that this coefficient is

close to one, that is, that the number of funds that persist in the categories of winners or losers

is similar to the number of funds that do not persist. The significance of this coefficient can

be tested by a chi-square test. The χ² test statistic is given by comparing the expected and

observed distributions for WW, WL, LW and LL; it follows a χ² distribution with one degree

of freedom (Eling, 2009).

(4) ,

where D1 = (WW + WL)(WW + LW)/n , D2 = (WW + WL)(WL + LL)/n, D3 = (LW +

LL)(WW + LW)/n, D4 = (LW + LL)(WL + LL)/n, and n is the total number of funds.

However, neither of these two methods manages to capture fund-to-fund performance;

that is, we only obtain a vision of persistence specific to the funds of the sample. One

possible use of the χ2 indicator is to perform fund-to-fund analysis. For this, it is sufficient to

use the percentage of times that the fund was above and below the median (i.e., W and L,

respectively) and construct respective sequences for each fund with respect to WW, WL, LW

and LL. The indicator for individual persistence is then calculated using (4).

Another indicator of individual persistence for each fund is obtained by the parametric

regression method. This indicator is obtained by the regression of the performance indicator

during period t in relation to period t-1, that is,

10
 I it  a  bI i ,t 1   it .

In this regression, a positive and significant angular coefficient (b > 0) provides

evidence of persistence. The significance of this coefficient may be tested through a student’s

t test, where the null hypothesis indicates the lack of persistence. Because this is a parametric

method, we know that many of the hypotheses in the regression may not be verified in

practice, such as normal errors and non-correlations over time, which are characteristics

empirically observed in the literature on mutual funds (Eling, 2008). In addition, the

smoothness of the returns of multimarket funds throughout time makes the presence of serial

autocorrelation a factor that leads to the observation of persistence through these methods,

mainly those of correlation. Taking this fact into account, this study is based on parametric

regression methods with a Newey and West (1987) correction, which provides robust

estimators for serial correlation and heteroscedasticity.

III. Data Description

In this study, we used the Anbima database. The period of analysis is from October 1,

2007 to February 8, 2010, with 593 observations for daily returns. Obviously, this time frame

has a certain limitation for definitive conclusions due to the small period of time, but there

are two relevant factors that justify this choice. First, the IHFA, which is one of the

benchmarks used, only began to be calculated in June 2007. Furthermore, this period consists

of 593 daily observations for each fund, which is greater than the number used in the majority

of studies in the literature that use monthly, quarterly or annual returns and includes

important periods of high volatility and financial stress (June 2008 to November 2008) and

also in the scenario of very rapid recovery rally (December 2008 to February 2010). Finally,

the period of observation allows us to disregard the dynamics of the manager market, as

highlighted by Eling (2009).

11
 A total of 161 funds were chosen, which consist of funds that were present, at some

point, in the IHFA as well as those possessing a complete sample for the studied period.

Despite the relatively small number of funds in relation to American studies, which use

thousands of funds, this sample is representative, as it contains funds from the main

institutions in the Brazilian market as well as funds that, according to the selection criteria of

Anbima,5 possess de facto hedge fund characteristics. These choices are based on two

concerns. First, because the analysis period covers the financial crisis, funds that emerged

after the end of the crisis or left the market before the onset of the crisis may render the

results less consistent. Second, because one of the objectives is to compile a ranking in order

to evaluate the funds, they should still be active so that, from the point of view of the

investor, the results can be applied (Andaku and Pinto, 2003).

The exclusion of funds that are no longer active generates bias for the survival rate.

Malkiel and Saha (2005) find evidence that upon discarding non-operative funds, the level of

persistence increases. However, Eling (2009) concludes that approaches with biases for

survival do not systematically lead to higher or lower levels of persistence. Therefore, we

believe that the existence of this problem in the analysis does not invalidate the results but

rather provides the results with greater applicability from the point of view of the investor.

In the Brazilian market, there is no consensus regarding which risk-free rates should

be used. This study uses the daily rate based on the Interbank Certificate Of Deposit (ICD), as

the analyzed funds use this rate as the basis to calculate the performance fee. In turn, two

indicators were used for the market, namely, returns from the Ibovespa index and the IHFA.

The IHFA is calculated daily by Anbima as follows. All funds that fit the criteria are

selected to not be exclusive and are classified as multimarket funds for more than one year.

This information is updated with daily quotas. Then, a theoretical portfolio is calculated at the

12
beginning of each quarter, where the weight as a percentage of each of the funds corresponds

to its net worth divided by the total net worth of all of the funds present in the index. The

number of points that each fund has in the IHFA is given by the weight multiplied by the

index of the day prior to rebalancing, which is done quarterly. Therefore, the calculation of

the index becomes the sum of the multiplication of the theoretical quantity of the quotas of

each fund (i.e., index points) and the value of their quotas for each fund.6 The data on the

ICD rates and of Ibovespa were obtained from the Thomson/Reuters Datastream database,

whereas the IHFA data were obtained from Anbima.

Table 1 presents the descriptive statistics of the funds, grouped by strategy, together

with the benchmarks used. We observed that the return on the funds was greater than the

returns from the Ibovespa and the ICD during the period analyzed. As expected, the

hypothesis regarding the normality of returns is rejected in all cases. It is worth noting that

during the period analyzed, there was a strong fall in the markets due to the worldwide

financial crisis beginning in September 2008, and there was also a strong recovery phase

beginning in December 2009. Therefore, the data consist of bearish and bullish phases.

[INSERT TABLE 1 AROUND HERE]

In Figure 1 we observe the accumulated returns for the Ibovespa, the ICD and the

Hedge Fund index IHFA. It is interesting to notice that volatility is much lower in the IHFA

than in the Ibovespa. It also evident that the IHFA returns follows very closely the ICD

returns, probably due to the fact that Hedge Funds were strongly positioned in fixed income

instruments with the goal of lowering volatility and taking advantage of high local interest

rates.

[INSERT FIGURE 1 AROUND HERE]

13
IV. Results

Table 2 presents the estimations of the performance indicators explained in the

methodology section. The estimations of models (1), (2) and (3) were performed by ordinary

least squares corrected by the covariance matrix of Newey and West (1987), which corrects

problems of heteroscedasticity and autocorrelation of the errors. The results show that a high

percentage of funds generate the alpha, with the greatest value found for the CAPM model

(48% of the funds), followed by the CAPM-IHFA model (37% of the funds) and the Style

model (34% of the funds). In an initial analysis, the performance indicators for this select

group of funds present satisfactory performance for the period analyzed.

[INSERT TABLE 2 AROUND HERE]

In Table 3, we evaluate the exposure of funds to risk factors in the three studied

models. We conclude that the funds exhibit an elevated index of positive exposure to the

market indicators, that is, Ibovespa and IHFA. In relation to other risk factors, the majority of

the funds do not demonstrate exposure to public titles and currency exchange. For those that

present exposure to these factors, a greater number of funds are exposed to public titles than

to currency exchange, and in general, positive exposures overcome negative exposures to

these factors. These results show that the studied risk factors systematically capture the

exposure to risk factors presented by the funds.

[INSERT TABLE 3 AROUND HERE]

The evaluation of persistence is presented in Table 4. Various interesting factors can

be observed. First, for the joint return test and the Sharpe’s ratio test, there were a greater

number of persistence periods than the Jensen’s alphas. Second, in the analysis of joint

persistence, timeframes of three months generally exhibited greater persistence than those of

14
one month. Third, upon observing individual tests, the persistence index is far lower in

comparison to the joint persistence tests of the funds. Fourth, the parametric tests present

greater evidence of persistence than shown by the non-parametric tests. Finally, all of these

factors lead to the conclusion that although there is evidence of joint persistence, in

individual terms, only a select group of funds is persistent.

[INSERT TABLE 4 AROUND HERE]

After separately analyzing performance and persistence, we turned our attention to the

relationship between the two. Table 5 presents various regression specifications among the

performance indicators and test statistics for individual persistence. The results point to a

weak relationship between performance and persistence; when it is statistically significant, it

is not always positive. Therefore, according to the evidence found, we could not confirm that

performance and persistence are related.

[INSERT TABLE 5 AROUND HERE]

Finally, we evaluated whether the fund characteristics can serve as explanatory

variables for performance and returns. Table 6 presents the results for the performance

indicators. The most interesting result is the relationship between performance and

management fees, which presents positive and statistically significant coefficients in all

cases, except in the case of Sharpe’s test, where this value is negative and significant. In our

evaluation, this result indicates that the funds with the best performance tend to charge higher

management fees, but the negative Sharpe ratio indicates that the management fee does not

reflect the ratio of returns per unit risk. The analysis of the strategy appears to be of little

relevance. In relation to the determinants of persistence, Table 7 presents the regressions for

the parametric indicators for durations of one and three months. In this case, the results found

15
for the Style model are noteworthy, as persistence appears to be positively related to the

strategies, with the exception of multi-strategy multimarkets.

[INSERT TABLE 6 AROUND HERE]

V. Conclusion

The Brazilian hedge fund market represents an extremely dynamic industry segment

within the Brazilian fund industry. This study sought to evaluate this market from the point of

view of performance and persistence. To this end, we used daily data and analyzed a select

group of funds classified as authentic hedge funds by the Brazilian financial association

Anbima (Brazilian Association of Financial and Capital Market Entities).

In some aspects, this analysis plays the role of a stress test, given the financial turmoil

observed from October 2007 to January 2010, the time period of our dataset. A period

characterized by many as the worst world financial crisis since the great depression of the

1930’s. In practice, hedge funds should provide hedge in moments like this. To test if this is

in fact true is one of the goals of this study, at least, for an Emerging Market as the Brazilian.

Through a robust but not exhaustive analysis of performance and persistence

indicators, we demonstrated that a good number of the funds present abnormal indicators of

returns and persistence at a combined level. However, only a few funds exhibit persistence of

performance in individual terms. The charging of management fees appears to be closely

related to the performance of funds, but other characteristics are apparently not relevant as

drivers of performance and persistence. This result matches the finding of Chen and Liang

(2007) who found that, for the US market, market timing hedge funds performance is

relatively stronger in bear and volatile market conditions.

Inevitably, this study presents limitations that may be overcome in future studies. One

possibility is to estimate alternative factor models through other techniques that does not

16
assume a static Jensen’s alpha and market beta. In addition, new factor models may be

estimated and presented, including the influence of factors external to the Brazilian economy,

for example, through factor models based on arbitrage pricing theory (APT). Finally, using

out-of-sample exercises, the economic value of these results can be evaluated by comparing

the performance of fund portfolios selected on the basis of performance and persistence

indicators.

References

Agarwal, V., and N. Y. Naik. “Multi-Period Performance Persistence Analysis of Hedge

Funds.” Journal of Financial and Quantitative Analysis, 35 (2000a), 327–342.

Agarwal, V., and N. Y. Naik. “On Taking the “Alternative” Route: The Risks, Rewards, and

Performance Persistence of Hedge Funds.” Journal of Alternative Investments, 2 (2000b), 6-

23.

Andaku, F. T. A., and A. C. F. Pinto. “A persistência de desempenho dos fundos de

investimento em ações no Brasil.” Revista de Economia e Administração, Rio de Janeiro, 2

(2003), 23-33.

Barès, P.-A.; R. Gibson.; and S. Gyger. “Performance in the Hedge Funds Industry: An

Analysis of Short and Long-Term Persistence.” Journal of Alternative Investments, 6 (2003),

25–41.

Brown, S. J., and W. N. Goetzmann. “Hedge Funds with Style.” Journal of Portfolio

Management, 29 (2003), 101–112.

Brown, S. J.; W. N. Goetzmann; and R. G. Ibbtson. “Offshore Hedge Funds: Survival and

Performance 1989–1995.” Journal of Business, 72 (1999), 91–117.

17
Capocci, D.; A. Corhay; and G. Hubner. “Hedge Fund Performance and Persistence in Bull

and Bear Markets.” European Journal of Finance, 11 (2005), 361–392.

Chen, Y., and B. Liang. “Do Market Timing Hedge Funds Time the Market?” Journal of

Financial and Quantitative Analysis, 42 (2007), 827–856.

Eling, M. “Does the Measure Matter in the Mutual Fund Industry? Financial Analysts

Journal, 64 (2008), 54–66.

Eling, M. “Does Hedge Funds Performance Persist? – Overview and Empirical Evidence.”

European Financial Management, 15 (2009), 362–401.

Eling, M., and R. Faust. “The Performance of Hedge Funds and Mutual Funds in Emerging

Markets.” Journal of Banking & Finance, 34 (2010), 1993-2009.

Fung, W., and D. A. Hsieh. “Empirical Characteristics of Dynamic Trading Strategies: The

Case of Hedge Funds.” Review of Financial Studies, 10 (1997), 275–302.

Gomes, F., and V. Cresto. “Avaliação de desempenho de fundos long-short no Brasil.”

Revista Brasileira de Finanças, 8 (2010), 505–529.

Harri, A., and B. W. Brorsen. “Performance Persistence and the Source of Returns for Hedge

Funds.” Applied Financial Economics, 14 (2004), 131–141.

Henn, J., and I. Meier. “Performance Analysis of Hedge Funds.” In Handbuch Hedge Funds,

H. Dichtl, J. M. Kleeberg, and C. Schlenger, eds. Bad Soden, Germany: Uhlenbruch (2004).

Jensen, M. C. “The Performance of Mutual Funds in the Period 1945-1964.” Journal of

Finance, 23 (1968), 389-416.

18
Jagannathan, R.; A. Malakhov; and D. Novikov. “Do Hot Hands Persist Among Hedge Fund

Managers? An Empirical Evaluation.” The Journal of Finance, 65 (2010), 217–255.

Malkiel, B. G., and A. Saha. “Hedge Funds: Risk and Return.” Financial Analysts Journal,

61 (2005), 80–88.

Newey, W. K., and K. D. West. “A Simple, Positive Semi-Definite, Heteroskedasticity and

Autocorrelation Consistent Covariance Matrix.” Econometric Society, Econometrica, 55

(1987), 703-708.

Park, J. M., and J. C. Staum. “Performance Persistence in the Alternative Investment

Industry.” Available at SSRN: http://ssrn.com/abstract=139068 or doi:10.2139/ssrn.139068.

Sharpe, W. F. “Mutual Fund Performance.” Journal of Business, 39 (1966), 119–138.

Xavier, A. “Persistência de Performance: Fundos Multimercado Com Renda Variável e

Alavancagem, Dissertação de Mestrado, Ibmec-RJ, 2008.

19
Footnotes

1
This includes funds classified by Anbima (Associação Brasileira das Entidades dos

Mercados Financeiro e de Capitais – Brazilian Association of Financial and Capital Market

Entities) as multi-strategy multimarkets, interest and currency multimarkets, specific strategy

multimarkets, multi-manager multimarkets, long and short neutral multimarkets and long and

short directional multimarkets.

2
Data are obtained from the Economatica fund database, with the CVM (Comissão de

Valores Mobiliários - Securities and Exchange Commission) database as the primary source.
3
The regulation of funds varies greatly between countries, mainly in areas such as the use of

the fund manager’s own capital, different types of management strategies, the minimum

capital necessary for participation and so on.

4
More information on the methodology used to construct these indices can be found on the

Anbima website at http://www.anbima.com.br.

5
These criteria will be explored below.
6
For details on the methodology for selecting IHFA funds, see

http://www.andima.com.br/ihfa/ihfa_cartilha.asp.

20
Figure 1 – Accumulated Returns of Brazilian Stock Market (Ibovespa), the Hedge Fund
Index (IHFA) and Fixed Income (Short term risk-free, ICD)

1.3

1.2

1.1

1.0

0.9

0.8

0.7

0.6

0.5

0.4
IV I II III IV I II III IV I
2007 2008 2009

Stock Market (Ibovespa)

Hedge Fund Index (IHFA)
Fixed Income (risk-freeICD rate)

21
Table 1 - Descriptive statistics - IHFA member hedge funds and market benchmarks
Returnsa
Number Standard Kurtosis Jarque-
a
of funds Mean (%) deviation (%)a Asymmetry (Excess) Bera(%)
Multi-strategy multimarkets 80 0,930 1,03 -0,209 15,550 0%
Macro Multimarkets 39 1,223 2,16 -0,264 11,742 0%
Long/Short - Neutral 22 0,956 1,12 0,055 6,043 0%
Long/Short - Directional 9 0,886 1,98 -0,444 9,537 0%
Interest and Currency Multimarkets 4 0,871 0,19 -1,113 39,155 0%
Specific Strategy Multimarkets 4 0,927 1,14 0,085 7,293 0%
Multi-manager Multimarkets 3 0,991 1,11 -0,449 14,089 0%
All Funds 161 1,002 1,35 -0,219 13,346 0%
Ibovespa 0,165 12,27 0,322 4,530 <0.0001
ICD 0,640 0,02 -0,114 -1,226 <0.0001
IHFA 0,958 0,96 -0,557 9,276 <0.0001
IRF-M 1,031 0,86 -1,485 19,818 <0.0001
IMA-B 1,117 1,45 -0,527 8,564 <0.0001
a
Monthly Values

22
Table 2 - Performance indicators: Sharpe and Jensen's Alpha

% Positive % Positive % Positive

Number Mean Alpha and Mean Alpha and Mean Alpha and
a a a
of funds Sharpe CAPM (%) Significant IHFA(%) Significant Style (%) Significant
Multi-strategy
multimarkets 80 0,454 0,290 58% 0,139 45% 0,119 41%
Macro Multimarkets
39 0,361 0,600 41% 0,258 26% 0,256 31%
Long And Short -
Neutral 22 0,343 0,330 32% 0,263 32% 0,278 32%
Long And Short -
Directional 9 0,179 0,264 22% 0,051 11% 0,089 11%
Interest and Currency
Multimarkets 4 1,238 0,228 100% 0,212 100% 0,189 100%
Specific Strategy
Multimarkets 4 0,509 0,263 50% 0,056 50% 0,075 50%
Multi-manager
Multimarkets 3 0,393 0,336 33% 0,019 33% 0,037 33%
All Funds 161 0,421 0,368 48% 0,177 37% 0,171 34%
a
Monthly values

23
Table 3 - Estimation of betas
CAPM Model IHFA Model
% positive % positive
Number CAPM and IHFA and
of funds Beta significant Beta significant
Multi-strategy multimarkets 80 0,031 76% 0,506 89%
Macro Multimarkets 39 0,078 87% 1,147 85%
Long And Short - Neutral 22 0,004 41% 0,210 59%
Long And Short - Directional 9 0,029 56% 0,689 78%
Interest and Currency Multimarkets 4 0,003 50% 0,053 50%
Specific Strategy Multimarkets 4 0,048 50% 0,698 50%
Multi-manager Multimarkets 3 0,065 100% 1,054 100%
All Funds 161 0,039 72% 0,635 81%

Style Model
% positive % positive % negative % positive % negative % positive % negative
Number IHFA and IMA-B and and IRF-M and and Exchange and and
of funds Beta significant Beta significant significant Beta significant significant rate Beta significant significant
Multi-strategy multimarkets 80 0,458 89% 0,008 28% 11% 0,092 38% 9% 0,003 15% 5%
Macro Multimarkets 39 1,117 82% 0,044 46% 23% -0,016 28% 13% 0,002 0% 13%
Long And Short - Neutral 22 0,252 77% -0,036 0% 5% -0,025 0% 5% 0,005 9% 5%
Long And Short - Directional 9 0,808 78% -0,088 0% 22% -0,097 0% 11% 0,002 11% 0%
Interest and Currency Multimarkets 4 0,013 25% -0,014 0% 50% 0,117 75% 25% 0,001 0% 0%
Specific Strategy Multimarkets 4 0,759 50% -0,027 25% 25% -0,085 0% 25% -0,005 0% 0%
Multi-manager Multimarkets 3 1,117 100% -0,044 33% 33% -0,056 0% 0% -0,002 0% 0%
All Funds 161 0,618 83% 0,003 26% 16% 0,033 27% 10% 0,003 9% 6%

24
Table 4 - Performance Persistence Indicators
1 month 3 months
Number of CAPM IHFA Style CAPM IHFA Style
funds Return Sharpe Alpha Alpha Alpha Return Sharpe Alpha Alpha Alpha
A. Joint Tests
Spearman's Correlation Coefficient 84% 84% 72% 60% 28% 86% 86% 86% 86% 43%
CPR 72% 60% 52% 40% 32% 71% 86% 71% 71% 43%
χ² 72% 64% 44% 36% 28% 86% 86% 86% 86% 43%
B. Individual Tests - Regression
Multi-strategy multimarkets 80 33% 30% 28% 19% 19% 14% 15% 9% 8% 18%
Macro Multimarkets 39 59% 38% 44% 13% 10% 15% 13% 13% 13% 8%
Long And Short - Neutral 22 14% 9% 14% 5% 5% 27% 0% 14% 18% 0%
Long And Short - Directional 9 22% 22% 44% 33% 22% 11% 0% 11% 11% 33%
Interest and Currency Multimarkets 4 50% 0% 0% 75% 0% 75% 0% 25% 0% 25%
Specific Strategy Multimarkets 4 50% 25% 25% 25% 0% 0% 25% 0% 25% 0%
Multi-manager Multimarkets 3 67% 67% 67% 0% 33% 0% 0% 0% 0% 0%
All Funds 161 37% 29% 30% 17% 14% 17% 11% 11% 11% 13%
C. Individual Tests - χ²
Multi-strategy multimarkets 80 5% 9% 15% 14% 5% 11% 13% 13% 14% 9%
Macro Multimarkets 39 8% 3% 5% 10% 5% 15% 13% 13% 13% 0%
Long And Short - Neutral 22 23% 0% 9% 0% 0% 9% 5% 18% 23% 5%
Long And Short - Directional 9 22% 0% 33% 33% 0% 11% 11% 22% 11% 11%
Interest and Currency Multimarkets 4 0% 0% 75% 0% 0% 0% 0% 25% 0% 0%
Specific Strategy Multimarkets 4 0% 0% 0% 0% 50% 0% 0% 0% 0% 0%
Multi-manager Multimarkets 3 33% 33% 0% 0% 0% 0% 0% 0% 0% 0%
All Funds 161 9% 6% 14% 11% 5% 11% 11% 14% 14% 6%
Note: Joint tests indicate the percentage of time frames analyzed in which the ranking (based on the respective indicator) proved to be persistent. Individual tests evaluate the percentage of
funds that presented an indication of persistence in the time frame analyzed. In both cases, we adopted a significance level of 10%.

25
Table 5 - Relationship between performance and performance persistence
Dependent Variable
Returns
Constant 0,995111 0,900701 1,050012 1,002016
(0.046932)*** (0.046329)*** (0.051247)*** (0.035606)***
a 0,455117
b (1 month)
(0.139304)***
χ²(1 month) 0,003725
(0.017462)
a -0,016232
b (3 months)
(0.107363)
χ²(3 months) -0,031299
(0.022694)
R² 0,062908 0,000288 0,000144 0,012689
Observations 161 160 161 150
Sharpe
Constant 0,466684 0,359885 0,429135 0,415464
(0.044317)*** (0.042243)*** (0.031262)*** (0.041273)***
b a(1 month) -0,212926
(0.149104)
χ²(1 month) 0,043015
(0.02097)**
b a(3 months) -0,097906
(0.082704)
χ²(3 months) -0,01772
(0.01769)
R² 0,012663 0,025781 0,008737 0,007066
Observations 161 161 161 143
Ibovespa Alpha
0,000151 0,0002 0,000167 0,000217
Constant
(0.0000232)*** (0.0000222)*** (0.000018)*** (0.0000257)***
a 0,0000748
b (1 month)
(0.0000716)
-0,000017
χ²(1 month)
(0.0000069)**
-0,00000511
b a(3 months)
(0.0000588)
-0,0000296
χ²(3 months)
(0.0000116)**
R² 0,006819 0,036907 0,000047 0,042097
Observations 161 160 161 151

26
Table 5 - Relationship between performance and performance persistence (cont.)
IHFA Alpha
Constant 0,000103 0,0000879 0,0000823 0,000101
(0.0000251)*** (0.0000248)*** (0.0000205)*** (0.0000279)***
a -0,000137
b (1 month)
(0.0000937)
χ²(1 month) -0,00000459
(0.00000853)
a 0,0000217
b (3 months)
(0.0000691)
χ²(3 months) -0,0000128
(0.0000127)
R² 0,013298 0,001831 0,000622 0,006632
Observations 161 160 161 154
Style Alpha
Constant 0,0000793 8,22E-05 8,31E-05 7,13E-05
(0.0000209)*** (0.0000273)*** (0.0000214)*** (0.0000292)**
b a(1 month) -0,0000635
(0.0000969)
χ²(1 month) -4,29E-06
(0.000015)
a 5,71E-05
b (3 months)
(0.0000543)
χ²(3 months) 2,89E-06
(0.0000155)
R² 0,002697 0,000516 0,006908 0,000232
Observations 161 160 161 152
*,**,*** represent significance of 10%, 5%, and 1%, respectively.
a
Coefficient from the parametric regression of persitence.

27
Table 6 - Relationship among performance and fund characteristics
Dependent Variable
Returns Sharpe Ibovespa Alpha IHFA Alpha Style Alpha
0,564 0,563885 -0,0000684 -0,000189 -0,000196
Constant
(0.22512)** (0.185236)*** (0,00011) (0,00013) (0,00014)
Management fee 17,320 -14,17265 0,00987 0,010162 0,011231
(6.596452)*** (5.427792)*** (0.003321)*** (0.00377)*** (0.003984)***
Performance fee -0,076 -0,546635 -0,000019 0,0000742 9,19E-05
(0,50151) (0,41266) (0,00025) (0,00029) (0,00030)
Multi-strategy Multimarketsa 0,093 0,246555 0,0000396 0,0000661 4,27E-05
(0,15352) (0.126318)* (0,00008) (0,00009) (0,00009)
Macro Multimarketsa 0,346 0,176771 0,000157 0,0000988 8,09E-05
(0.160207)** (0,13182) (0.0000806)* (0,00009) (0,00010)
Long And Short - Neutrala 0,074 0,167991 0,0000316 0,0000965 8,61E-05
(0,17148) (0,14110) (0,00009) (0,00010) (0,00010)
Interest and Currency Multimarketsa 0,164 0,912459 0,0000855 0,000178 0,000161
(0,26904) (0.22137)*** (0,00014) (0,00015) (0,00016)
Specific Strategy Multimarketsa 0,120 0,265728 0,0000449 0,0000485 4,48E-05
(0,26202) (0,21560) (0,00013) (0,00015) (0,00016)
a 0,148 0,138157 0,0000588 0,0000184 1,36E-05
Multi-manager Multimarkets
(0,29153) (0,23988) (0,00015) (0,00017) (0,00018)
R² 0,120864 0,191794 0,122707 0,061894 0,065556
Observations 161 161 161 161 161
*,**,*** represent significances of 10%, 5%, and 1%, respectively.
a
Dummy Variables (1 if it belongs to the category, 0 otherwise). The benchmark group are the long and short directional funds.

28
Table 7 - Relationship among performance persistence indicators (b coefficient, parametric regression) and fund characteristics
Dependent Variable
a a
b -Returns b -Sharpe b -Ibovespa Alphaa b -Ihfa Alphaa b - Style Alphaa
1 month 3 months 1 month 3 months 1 month 3 months 1 month 3 months 1 month 3 months
Constant 0,393661 0,12256 0,235154 -0,091736 0,376491 -0,004258 0,259887 -0,097593 0,029456 -0,396439
(0.12256)*** (0,16339) (0.10502)** (0,18813) (0.122106)*** (0,15579) (0.104555)** (0,14892) (0,11205) (0.193843)**
Management fee -4,429395 -4,260882 -1,183047 8,484406 -1,35154 0,277307 -1,78759 1,244665 -2,593306 -0,153363
(4,26088) (4,78759) (3,07731) (5,51255) (3,57794) (4,56488) (3,06366) (4,36374) (3,28335) (5,67998)
Performance fee -0,215898 0,231443 -0,015744 0,046996 -0,17361 0,515522 0,06191 0,526389 0,313639 -0,158839
(0.231443)** (0,36399) (0,23396) (0,41911) (0,27202) (0,34706) (0,23292) (0,33177) (0,24963) (0,43184)
Multi-strategy -0,113242 -0,078319 -0,007506 0,040925 -0,171333 -0,131755 -0,109084 -0,113849 -0,060449 0,276614
Multimarketsb (0.078319)** (0,11142) (0,07162) (0,12829) (0.083267)** (0,10624) (0,07130) (0,10156) (0,07641) (0.132187)**
Macro 0,107427 -0,060656 0,069181 0,078059 0,015478 -0,100535 0,007303 -0,107166 -0,006455 0,444809
Multimarketsb (0.060656)* (0,11628) (0,07474) (0,13388) (0,08690) (0,11087) (0,07441) (0,10598) (0,07974) (0.137948)***
Long And Short - -0,197591 -0,274469 -0,067217 -0,13999 -0,18391 -0,259506 -0,177248 -0,181712 0,061307 0,478745
Neutralb (0.274469)*** (0.124455)** (0,08000) (0,14330) (0.09301)** (0.118665)** (0.079641)** (0,11344) (0,08535) (0.147653)***
Interest and 0,153413 0,490814 -0,118825 -0,213771 -0,295611 0,214526 -0,069532 -0,033741 -0,015954 0,511905
Currency (0.490814)*** (0.195261)** (0,12551) (0,22483) (0.145925)** (0,18618) (0,12495) (0,17797) (0,13391) (0.231656)**
Specific Strategy 0,155117 -0,160591 0,032084 0,356541 0,063427 -0,193395 0,0942 -0,195464 -0,105477 0,40457
Multimarketsb (0,16059) (0,19017) (0,12224) (0,21897) (0,14212) (0,18132) (0,12169) (0,17333) (0,13042) (0.225616)*
Multi-manager 0,069491 -0,265103 0,202055 -0,163416 0,155496 -0,292294 0,121603 -0,2908 -0,038107 0,418832
Multimarketsb (0.265103)*** (0,21159) (0,13600) (0,24363) (0,15813) (0,20175) (0,13540) (0,19286) (0,14511) (0.251029)*
R² 0,239291 0,151136 0,069907 0,085375 0,165123 0,088542 0,123792 0,046016 0,050616 0,102257
Observations 161 161 161 161 161 161 161 161 161 161
*,**,*** represent significances of 10%, 5%, and 1%, respectively.
a
Coefficient from the parametric regression of persitence.
b
Dummy Variables (1 if it belongs to the category, 0 otherwise). The benchmark group are the long and short directional funds.

29