Do Emerging Market Hedge Fund Managers Add Value
for Their Investors?
S. S. Mol
Student number: 1462180 MSc Business Administration Specialization: Finance University of Groningen
Faculty of Economics and Business Supervisor: Dr. B. A. Boonstra Submission
Table of Contents
1. Introduction ... 2
2. Literature review ... 5
2.1 Hedge funds ... 5
2.2 Hedge fund data and databases ... 6
2.2.1 Database measurement biases ... 6
2.2.2 Serial correlation issues... 7
2.3 Emerging market hedge funds ... 8
3. Methodology ... 11
3.1 Regression-based style analysis ... 11
3.2 Autocorrelation and return smoothing ... 13
3.3 Time-varying risk exposures ... 13
4. Data ... 14 4.1 Data selection ... 14 4.2 Summary statistics ... 15 4.3 Correlations ... 17 4.4 Sharpe ratios ... 18 5. Results ... 21
5.1 Static regression-based style analysis results ... 21
5.1.1 The basic emerging market regression-based style analysis ... 21
5.1.2 The smoothing emerging market regression-based style analysis ... 23
5.1.3 Robustness check: Departures from the normal distribution ... 24
5.2 Dynamic regression-based style analysis results ... 26
5.2.1 The basic emerging market regression-based style analysis ... 27
5.2.2 The smoothing emerging market regression-based style analysis ... 33
5.2.3 Robustness check: Using a 48-month rolling window ... 35
6. Conclusions ... 36
References ... 39
Do Emerging Market Hedge Fund Managers Add Value for
Their Investors?
Stefanie Suzanne Mol* January 2012
ABSTRACT
Emerging markets provide hedge fund managers with tremendous return opportunities and risks at the same time. Using a static as well as a dynamic emerging market regression-based style analysis, I find that between 1995 and 2011 emerging market hedge funds’ systematic market exposure (beta) has decreased. While this indicates that emerging market hedge fund managers were better able to cope with market risks, emerging market hedge funds seem to have been unable to add value (alpha) for their investors as high as before the Asian crisis. During crisis periods, emerging market hedge fund managers did not add value at all. When taking into account return smoothing and illiquidity related issues, alphas are reduced considerably. This raises questions on the performance of emerging market hedge funds and their ability to add value for their investors.
JEL classification: G2, G10, G11, G15, G29
Keywords: Hedge funds, emerging markets, performance, financial crisis
* S. S. Mol, student number: 1462180, e-mail: stefaniemol@gmail.com, University of Groningen, Faculty of Economics and Business, Master Thesis, MSc Business Administration specialization: Finance. First supervisor, prof. dr. B. A. Boonstra and second and third supervisor, prof. dr. R. A. H. van der Meer and dr. R. Salomons.
1.
Introduction
Do emerging market hedge funds1 add value? While prior literature extensively researched this question for hedge funds in advanced markets, the literature specially focusing on emerging market hedge funds only arose during the second part of the 2000s. Opportunities to use dynamic trading strategies (e.g. short selling, derivatives) were limited due to market illiquidity or were completely prohibited in emerging markets. Consequently, this led emerging market hedge funds in fact to behave like mutual funds. For this reason, Fung and Hsieh (1997, 1999) for example excluded emerging market hedge funds from their sample.
Recently, investors have increasingly been investing in emerging market hedge funds. Although hedge funds’ investments targeting emerging markets made up for only around 0.4% percent of the global amount of money under management in hedge funds two decades ago, this strategy has gained popularity during the 2000s as the global investor community changed its perception of emerging markets. According to HFR (Hedge Fund Research, Inc.), assets under management of hedge funds investing in emerging markets grew from 3.2% in 2003 to 7.3% in 2010 (see also Figure 1)(HSBC Group, 2011). During the financial crisis of 2008 emerging market hedge funds were not hit overwhelmingly more compared to the overall hedge funds industry. Investors in emerging market hedge funds redeemed 6.7% and 12.7% of their assets in 2008 and 2009 respectively (compared to 8.3% and 9.3% for the total hedge fund industry) (HFR, 2010). Subsequently, total assets of emerging market hedge funds grew with almost 1.4 billion dollars in the second quarter of 2011, thereby reaching levels – 123 billion dollars – which exceed the last record of 117 billion dollars in 2007 (HFR, 2011).
The growth of investments in emerging market hedge funds reflects the interest for this strategy by wealthy individuals and institutional investors: it points to their expectations of adding value over ordinary investing in emerging market securities and fixed-income instruments and diversification opportunities. From an investor’s point of view it is therefore useful to investigate whether emerging market hedge funds indeed show superior performance or not. What makes emerging market strategies so interesting is, among others, the growth prospects in emerging markets which sharply contrast the stagnating GDP numbers in the developed world.
1 An emerging market hedge fund is a hedge fund that primarily invests in assets of emerging market countries (e.g.
Figure 1
HFR Geographical investment focus by assets under management globally 2010.
Source: HSBC Group (2011).
Moreover, emerging market fundamentals have clearly improved over the last decade, reducing the impact of several risk factors in these regions. On the contrary, the deteriorating financial climate in the developed markets causes managers to search for new diversification possibilities. Despite the developments, the emerging market environment remains more challenging for investors as information asymmetries and illiquidity are more present due to less developed capital markets, less disclosure, less regulation, and weaker forms of corporate governance. Hedge funds have some specific features2 which make them better able to deal with these emerging market characteristics, and are therefore believed to perform better than other investment vehicles like mutual funds. Investors require a higher return on investments with higher risk. The performance of hedge fund returns can be divided into two main components: the systematic market exposure (beta) and the added value by the hedge funds (alpha). The literature on the value added, i.e. positive alpha, by emerging markets hedge funds shows mixed results; the same holds for research on hedge funds in advanced markets. Both Naik et al. (2007) and Fung and Hsieh (2004) find that alphas for hedge funds declined over the periods between 1995 and 2004 and between 1994 and 2002 respectively. Strömqvist (2007) analyzes the period between 1994 and 2004 and finds that, although emerging market hedge funds underperformed other hedge funds on average, a positive alpha is found in the last sub-period.
Using data from the Hedge Fund Research, Abugri and Dutta (2009) find that prior to 2007 emerging market hedge funds behaved in a similar way as mutual funds and did not show significant superior performance. From 2007 on, however, emerging market hedge funds started to behave more like advanced market hedge funds, and on a risk-adjusted basis, some emerging
2 I discuss these features in more detail in section 2.
Global 43.3%
Emerging Markets 7.3% US 40.9%
Japan 1.1%
market hedge funds outperformed the benchmark. Abugri and Dutta (2009) conclude that although they do not find any significant results on this superior performance, emerging market hedge funds are in a transition period. Eling and Faust (2010), using hedge fund return data from January 1995 to August 2008, analyze the alphas of emerging market hedge funds and mutual funds as well as the underlying factors. Using an emerging markets asset class factor model, they find that emerging market hedge funds outperform emerging market mutual funds and some of them even show superior performance compared to benchmark indices. Furthermore they show an upward development of alphas generated by emerging market hedge funds over time.
The most recent studies by Abugri and Dutta (2009) and Eling and Faust (2010) however exclude data after August 2008 from their sample to avoid abnormal market volatility caused by the bankruptcy of Lehman Brothers. This may give a distorted picture of reality, as the financial crisis of 2008-2009 contains valuable information on the performance on emerging market hedge funds. Hence, one can wonder whether the upward trend in alpha they found has been persistent during the financial crisis of 2008-2009. For this reason I believe it is interesting to investigate whether the developments in emerging markets and the wider range of dynamic trading strategies emerging markets hedge funds can pursue (as restrictions are omitted), lead to better performance during crisis periods.
Performance measurement on hedge funds entails some problems however, which might bias the results. Eling and Faust (2010) for example correct for several biases often present in the databases providing hedge fund data.3 Another problem with hedge fund performance measurement is that hedge fund returns are often serially correlated possibly due to managers smoothing their returns in the case of investments that are relatively illiquid (Getmansky et al., 2004). This return smoothing may have an impact on the performance measurement models.
Using a static as well as a dynamic regression-based style analysis, I investigate the factor exposures (betas) and value added by emerging market hedge fund managers (alpha) of emerging market hedge funds. This study contributes to the literature on emerging market hedge funds in several ways. Whereas this paper is mainly based the work of Abugri and Dutta (2009) and Eling and Faust (2010), it expands their studies in at least two important ways. First, in this study the financial crisis of 2008-2009 is included. Second, I account for the often observed serial correlation in hedge funds returns by incorporating lagged hedge funds returns.
3 Eling and Faust (2010) take survivorship, backfilling, multi-period sampling biases into account in their sample.
Analyzing emerging market hedge fund return data from January 1995 to June 2011, the main findings in this study can be summarized as follows: Although emerging market hedge funds main strategies are focused on equity and bonds, I do find evidence that the exposure of emerging market hedge funds to equity has declined over time. Furthermore, emerging market hedge funds do add value, but not as much as in the 1990s. When taking return smoothing into account, added value is considerably lower.
This paper is structured as follows. In section two I discuss the literature on hedge funds, in specific emerging market hedge funds. Section three and four discuss the data and methodology used in this study. I present the results and the corresponding robustness checks in section five. In the last section I conclude on my work by elaborating on the findings and the limitations of my research, and by giving some suggestions for further research.
2.
Literature review
In this section I provide an overview of the existing literature on hedge funds in general, the particularities of hedge fund return data and performance measurement which need to be taken into consideration, and emerging market hedge funds in specific.
2.1
Hedge funds
Hedge funds are pooled and professionally managed portfolios of investments, but have a couple of features that make them different from mutual funds (Ackermann et al., 1999). Hedge funds are governed by a relatively liberal regulatory framework4 that permits them to be more flexible in using dynamic trading strategies like short selling, derivatives, and leverage, which allows them to generate higher returns and reduce the associated risks. This flexibility is also reflected in the possibility to rather quickly shift their investment positions, which enables hedge fund managers to better react on changed market circumstances and implement new investment insights. Furthermore, hedge funds are characterized by strong performance incentives. Management fees are the rule rather than the exception and managers themselves have often invested a substantial amount in their own hedge fund. These incentive alignments create a positive effect on performance. Agarwal et al. (2009) find a relationship between the alignment of management
4 Hedge funds have investor net worth requirements and are often setup as limited partnerships and are therefore not
incentives5 and superior performance of hedge funds. Ackermann et al. (1999) also find that in their regressions, the incentive fee variable explains performance consistently. Another particularity of hedge funds is the lockup period. This is the time window during which investors are not allowed to redeem their investments, providing managers with better possibilities to deal with illiquid investments. Longer lockup periods as well as redemption periods are associated with better performance (Agarwal et al., 2009).
Hedge funds are often managed by highly experienced investment professionals. As investment skills or more market knowledge can lead to better performance, Bae et al. (2010) analyze stock holdings and find evidence that hedge fund managers have informational advantages over other institutional investors. Furthermore, the stocks that hedge funds choose to invest in are relatively more subject to information asymmetries. Moreover, there exists evidence that hedge fund managers were able to time the technology bubble, as their exposure to technology stocks peaked before prices collapsed and reduced when prices started to fall (Brunnermeier and Nagel, 2004). On the other hand, Griffin and Xu (2009) do not find evidence for market timing and stock-picking skills. Also, institutional investors use hedge funds with the purpose to realize lower portfolio volatility and accordingly more stable returns. Haglund (2010) finds that even when controlling for the non-normal return distributions and serial correlation, hedge fund can offer positive diversification in portfolios.
2.2
Hedge fund data and databases
2.2.1 Database measurement biases
Data on the hedge fund industry is rather limited. Due to a lack of regulation among hedge funds, disclosure on financial information occurs only on a voluntary basis. Moreover, the amount of information reported is often limited as hedge funds prefer to conceal their trading positions and models.
As advertising is prohibited by the SEC, self-reporting information on monthly performance to public accessible databases (for example the HFR, TASS, AltVest, and MAR) is believed to be one way in which hedge funds can market themselves to potential investors (Jorion and Schwarz, 2010). An extensive body of literature has been written on the problems with this self-reporting
5 These incentives are approximated by the delta of option-like incentive fee contracts, the level of managerial ownership,
leading to several potential biases of which survivorship bias and self-selection bias are considered the most important.6 Even though the Hedge Fund Research (HFR) and CSFB/Tremont (CT) attempt to mitigate these biases, measurement biases cannot be completely avoided and therefore one should take these biases into account when analyzing hedge funds’ performance (Fung and Hsieh, 2002).
Unfortunately, I do not have data on individual emerging market hedge funds, but I make use of aggregate indices instead. It is for that reason not possible to correct for these biases, as for example Eling and Faust (2010) do. One should keep in mind however, that these biases are present in the indices I use in this paper, and that the performance results might be positively biased.
2.2.2 Serial correlation issues
Besides the biases in hedge fund indices, hedge fund returns are often subject to high serial correlation (Asness et al., 2001,; Brooks and Kat, 2002; Getmansky et al., 2004; Bollen and Pool, 2008, 2009; Billio, 2009; Jordan and Simlai, 2010). According to Getmansky et al. (2004) serially correlated hedge fund returns can be caused in several ways, but they find that both illiquidity issues and return smoothing are the most likely explanations. Illiquidity issues arise when securities are not actively traded in the market, and consequently, security prices are not always available or do not completely reflect up-to-date market information. When hedge funds have such illiquid securities in their portfolios, reported returns of these hedge funds appear smoother than the actual returns since the variance of those securities will be downwardly biased. In line with this, Brooks and Kat (2002) show that it is difficult for hedge fund managers to value their positions in illiquid securities, and so use the last reported transaction price or an estimate of the actual price. Illiquidity issues are even more likely in emerging markets as capital markets are still less developed compared to the advanced world. Therefore, autocorrelation in emerging market hedge funds may be present as well and taking into account this autocorrelation thus might have a negative effect on performance.
The serial correlation in hedge fund returns may also have a behavioral aspect however. Agarwal et al. (2011) argue that hedge fund managers intentionally smooth their returns due to incentive fees. Hedge funds underreport returns during the calendar year, enabling them to report a substantial increase in returns in December. This increase in December cannot be explained using
6 See Fung and Hsieh (2002) for a more comprehensive discussion on these biases and its implications on the analysis of
risk-based factors, but the incentive fees managers receive once a year based on annual performance can explain its occurrence. Asness et al. (2001) states that hedge funds manage their reported returns (i.e. they smooth returns), in order to reduce the observed volatility and correlation with the general market, making them more attractive to investors. When not taking into account serial correlation, the performance of hedge funds may be overstated. Moreover, these practices can lead to underestimation of hedge funds’ actual market exposure. Lo (2002) finds an overstatement of 65 percent in the annual risk-adjusted performance measures for hedge funds in case of positive serial correlation. Brooks and Kat (2002) also argue that excess smoothness in the returns of hedge funds understates the true risks, and investors should take this into account when making their investment or portfolio decisions.
2.3
Emerging market hedge funds
Investing in emerging markets is one of the possible strategies that hedge funds can pursue and it gained popularity during the last decade (see Figure 1). Emerging markets are characterized by strong economic growth and high volatility. Therefore, potential returns are high and relatively risky at the same time. Emerging markets thus offer opportunities as well as threats for investors. On the one hand, emerging markets such as the BRIC countries7 have become more attractive to investors in terms of high economic growth and financial development. On the other hand, advanced economies such as the US, Europe and Japan are losing their attractiveness as GDP growth is stagnating, interest rates are rising, unemployment rates are high, and they are suffering from the burden on their shoulders called national debt, which in turn results in dropping consumption in order to pay off this debt. Future prospects for the emerging markets are also bright. The International Monetary Fund predicted 7.1 and 6.4 percent real GDP growth for emerging markets in 2010 and 2011 respectively, of which 10.5 percent is predicted for populous giant China and 9.7 percent for India. Future predictions on trading terms (i.e. imports and exports) and current accounts balances are also more favorable for emerging and developing countries (see Appendix A). China, which is in possession of 2.4 trillion dollar in foreign exchange reserves, should also be taken into account (Tanzer, 2010).
According to Schultz (2010) investors are searching for economic developments which give them the flexibility to gain from volatility and profit opportunities. Therefore, hedge funds managers
have entered the growing regions to profit from diversification, large economic growth and low interest rates.
Moreover, in 2007 the improvements in emerging market fundamentals seem to be persistent compared to the situation during the Asian financial crisis in 1997-1998: high profits, according to the Bank of International Settlements decreased default rates, tightened fiscal policy, floating currencies, trade surpluses, paid-off currency debt and improvements in corporate earnings and corporate governance (Credit Suisse Tremont Hedge Fund Index, 2007).
Even though the emerging markets are in a transition period, these markets are still behind the advanced market in several respects. Information asymmetries and illiquidity are more present due to less developed capital markets, less disclosure, less regulation, weaker forms of corporate governance. Hedge funds have better opportunities to deal with these challenges in emerging markets according to Eichengreen et al. (1998). Since they can use long and short positions, leverage, and derivatives, hedge funds are better able to gain from volatility. In line with this, HFR president Kenneth Heinz said that ‘…funds investing in growing regions had the “tactical flexibility” to withstand the challenges posed by inflation, sovereign debt and currency instability in 2011.’ (Chellel, 2011). The lockup periods of hedge funds facilitates dealing with more illiquid investments. Furthermore, Bae et al. (2010) argue that hedge funds select stocks with higher asymmetries, in order to exploit possible information advantages.
hedge funds decreased during the last decade. In other words, emerging market hedge funds have become less sensitive to the equity markets, i.e. have less market risk.
Although the literature on emerging market hedge funds is limited, there are some indications of the performance of by emerging market hedge funds. Naik et al. (2007) and Fung and Hsieh (2004) find that alphas for hedge funds declined over the periods between 1995 and 2004 and between 1994 and 2002 respectively. Strömqvist (2007) analyzes the period between 1994 and 2004 and finds that although emerging market hedge funds underperformed other hedge funds on average, they show a positive alpha in the last sub-period. Using data from the Hedge Fund Research, Abugri and Dutta (2009) find that prior to 2007 emerging market hedge funds match mutual funds behavior and did not show significant superior performance. From 2007 on, however, emerging market hedge funds started to behave more like advanced market hedge funds and, on a risk-adjusted basis, some emerging market hedge funds outperformed the benchmark. They conclude that although they do not find any significant results on this superior performance, emerging market hedge funds are in a transition period. Eling and Faust (2010), using hedge fund return data from January 1995 to August 2008, analyze the alphas of emerging market hedge funds and mutual funds as well as the underlying factors. Using an emerging markets asset class factor model, they find that emerging market hedge funds outperform emerging market mutual funds and some of them even show superior performance compared to benchmark indices. Furthermore they show the development of alphas generated by emerging market hedge funds over time which support the upward trend found by Strömqvist (2007).
If emerging market hedge funds started to behave like advanced market hedge funds as emerging markets started to provide more instruments needed to pursue dynamic trading strategies, this may have enabled them to even better deal with the typical risk factors (information asymmetries, illiquidity, asset volatility) of emerging markets.
3.
Methodology
In order to investigate whether emerging market hedge fund managers add value for their investors, a suitable performance measurement method is needed. In this section I present existing methodologies and the methods I use to answer the hypotheses. In this paper I use a static as well as a time-varying regression based style analysis, which will be explained. Furthermore, a modification is made to take into account possible autocorrelation effects.
3.1
Regression-based style analysis
The development of performance measurement models for hedge funds has been evolutionary. Almost all models have their origin in the classical CAPM model, where the expected return of a security is supposed to be risk-adjusted. Jensen’s alpha (1968) is based on this single-factor model and determines whether a security has a “positive alpha”, i.e. the actual return that is higher than the theoretical expected return. Fama and French (1993) extended this model by adding a market capitalization factor and book-to-market ratio factor to the overall market factor. Cahart (1997) added a momentum factor to the Fama and French three-factor model. Fung and Hsieh (1997, 2004) are the first to account for the fact that hedge funds use dynamic trading strategies and can invest in multiple securities, other than equities. Their model is based on the asset class factor model by Sharpe (1992), who shows that the performance of a sample of mutual funds can be analyzed by running a regression of the returns on only a few major asset classes.
In this study I use a static as well as a dynamic modified regression-based style analysis based on Sharpe (1992). A regression-based style analysis should contain the factors which are expected to be relevant in explaining the risk exposures and thus returns of emerging market hedge funds. I use the monthly returns of the four emerging market hedge fund indices, Emerging Markets Total, Emerging Markets Asia ex-Japan, Emerging Markets Latin America and Emerging Markets Russia/Eastern Europe, as the dependent variables. Based on the model of Eling and Faust (2010), the factors used in this model are a regional emerging market equity index, a regional emerging market bond index, and the credit spread.
According to Fung and Hsieh (2001, 2004) the credit spread is a relevant factor when hedge funds have invested in corporate bonds. Corporate bonds lose value when the credit spread rises, i.e. the lower quality bonds’ (Baa) yield rises more than the ‘risk-free’ bonds’ (10-year US Treasuries) yield, which has a negative impacts on hedge fund returns. Contrary to the equity and bond factor, the credit spread has no emerging market or regional focus. However, González-Rozada and Yeyati (2008) and Longstaff et al. (2011) show that most emerging market bond spreads can be better linked to global factors (US high-yield markets for instance) than to local economic factors. Since I have only emerging market hedge funds indices at my disposal, and these indices have a regional focus already, I only used the equity and bond variables in the regressions in which the regional classifications correspond with the dependent regional hedge funds variables.8
I start with a basic model. In the extended models, I will account for possible serial correlation (which enables me to better show the effect on the regression results). The basic model, which is applied per region (total emerging markets, Asia ex-Japan, Latin America, and Eastern Europe), is given by:
, , ∗ , ∗ , ∗ ! ", (1)9 where ,is the return of the hedge fund index for region i for month t, is the risk-free rate for month t, , is the excess monthly return on the MSCI equity index for region i for month t, , is the excess monthly return on the JP Morgan bond index for region i for month t, ! is the change in the credit spread for month t, and "is the error term. The # are the risk exposures, ,is the added value.
8 For example, only the Asian equity and bond benchmarks are used as regressors for the Asian emerging market hedge
fund returns.
3.2
Autocorrelation and return smoothing
To account for autocorrelation often observed in hedge fund returns, Eling and Faust (2010) included a lagged bond index factor in their model. This is called a distributed lag model, since only a lag of an explanatory variable is included, but not a lag of the explained variable (Brooks, 2008). Autocorrelation can also be due to the consciously smoothing of returns in order to make returns and thus hedge funds more attractive. I choose for an autoregressive distributed lag (ADL) model, which contains a lag of the dependent variable10, since the return smoothing explanation for autocorrelation is incorporated in the returns itself rather than in the explanatory variables. Therefore, a lagged hedge fund return factor is included in the smoothing model11:
, , ∗ , ∗ , ∗ ! $∗ ,% ", (2)12
where , is the return of the hedge fund index for region i for month t, is the risk-free rate for month t, , is the excess monthly return on the MSCI equity index for region i for month t, , is the excess monthly return on the JP Morgan bond index for region i for month t, ! is the change in the credit spread for month t, ,% is the lagged excess monthly return of the hedge fund index for region i for the prior month t-1, and " is the error term. The # are the risk exposures, ,is the added value.
3.3
Time-varying risk exposures
A key assumption of a regression is that the parameters, here style factors or risk exposures, are constant over time. However, since hedge funds can rather quickly change their investment positions, and thus risk profile, it may not be reasonable to assume that the parameters of a model for hedge funds are constant. A static model may therefore be not appropriate.
A technique to overcome this deficiency is to compute the factor estimates over a rolling window through the total sample (Zivot and Wang (2006). McGuire et al. (2005) use rolling regressions of a style analysis to capture the dynamic strategies, and there with the sensitivities to the risk factors (i.e. betas) of hedge funds through time. With a dynamic regression, the rolling alpha can also be
10 This violates the assumption that the explanatory variables are non-stochastic in nature, which may lead to biased
coefficients estimates. Hence, it is important to make a comparison of the coefficient estimates between the model without and with the lagged hedge fund return factor.
11 In the Results section I will also present two other models. The first with only a lagged bond index factor like Eling and
Faust (2010), and the second with both a lagged hedge fund return factor and a lagged bond index factor.
computed over time, which gives insight in the value added of emerging market hedge funds over time.
The greatest advantage of a rolling regression is thus that is picks up time-variation in exposure. Moreover, the possible problems of heteroskedasticity and non-normal distributed residuals are reduced as it concerns shorter time horizons. I use this rolling application for both the static basic and smoothing models presented above. An important disadvantage of a rolling regression is the increase in standard errors due to shorter subsamples.
4.
Data
This section start with a description of the dataset I use in this study, followed by summary statistics of and correlations between the variables I use in the regression-based style analysis.
4.1
Data selection
All the data I use in this study is obtained from DataStream and is available on a monthly basis13. Emerging market hedge fund total return index data is provided by the Hedge Fund Research (HFR). Data on hedge funds is available for four categories: one composite index HFR Emerging Markets Total Index, and three sub-indices with a specific regional focus, namely HFR Emerging Markets Asia ex-Japan Index, HFR Emerging Markets Latin America Index, and HFR Emerging Markets Russia/Eastern Europe Index (see Appendix B for the specific properties of these indices). I use data on equity indices from Morgan Stanley Capital International (MSCI) since these indices have again four categories with the same regional focus and are also used by Abugri and Dutta (2009) and Eling and Faust (2010). The emerging market equity indices are the MSCI Emerging Markets Total Index, the MSCI Asia Index, the MSCI Latin America Index, and the MSCI Eastern Europe Index (see Appendix B for the properties these indices). For the data on bonds I use the Emerging Market Bond Index + from JP Morgan, for which total as well as indices per region are available, and are also used by and Eling and Faust (2010). The emerging market bond indices are the JP Morgan Emerging Markets Total Index, the JP Morgan Asia Index, the JP Morgan Latin America Index, and the JP Morgan Eastern Europe Index (see Appendix B properties of these indices). The credit spread (Credit Spread) is month-end to month-end change in the difference between Moody's Baa yield and the Federal Reserve's 10-year constant-maturity yield. The risk-free rate is the three-month US Treasury Bill.
Since return data on MSCI EM Eastern Europe is only available from January 1995, I excluded the data before 1995 from the sample. The sample extends from January 1995 to June 2011, which includes 197 monthly observations. Furthermore, I transformed the total return indices into index returns in order to avoid the phenomenon of spurious regression since total return indices are often non-stationary in nature (Brooks, 2008).14
4.2
Summary statistics
Table 1 contains the descriptive statistics on the monthly return distributions of the emerging market hedge fund indices, emerging market equity and bond indices, and the credit spread; it shows the first four moments (mean, standard deviation, skewness, and kurtosis), the minimum, median, maximum, the number of observations, and the first-order autocorrelation coefficient. The last column provides information on the first-order autocorrelation coefficient (one-month lag). As is shown in Table 1, only the mean returns of the Eastern European emerging market hedge funds are higher than the equity and bond benchmarks with the same regional focus. While the mean return for emerging market hedge funds Asia are higher than the Asian equity benchmark, the Asian bond benchmark shows a higher return. The mean returns for the composite and Latin American emerging market hedge funds are lower than both the corresponding equity and bond benchmarks. Next, the emerging market hedge funds indices are less volatile than the corresponding equity indices, but more volatile compared to the corresponding bond indices. Table 1
Summary statistics for hedge funds, equity benchmarks, bond benchmarks and the credit spread the period from January 1995 to June 2011.
Mean SD Skew. Kurt. Min. Med. Max. Obs. (N) Autcor. (1 lag) Hedge fund indices
HFR Emerging Markets
Total Index 0.61 4.13 -1.07 7.48 -21.46 1.32 14.35 197 0.34*** HFR Asia ex-Japan Index 0.41 3.80 -0.11 3.37 -11.12 0.51 11.92 197 0.34*** HFR Latin America Index 0.55 4.70 -0.27 4.91 -16.07 1.13 18.81 197 0.17** HFR Russia/Eastern Europe
Index 1.33 7.66 -0.73 7.69 -39.03 1.56 29.67 197 0.37*** Table 1 continued on next page
14 Since all total return indices showed a trend upward through time (See also Figure 1 in Appendix C), this might lead to a
spurious regression. I performed Augmented Dickey-Fuller tests to see whether these indices are non-stationary, i.e. have a unit root. The results can be found in Tables 1 and 2 in Appendix C. The total return indices all have a unit root. This problem is completely solved when transforming the data into simple returns. Transforming total return index data into index returns by using the formula &'(
&'()*+ % &'()
&'()*+
Table 1 (continued) Equity benchmarks
MSCI Emerging Markets
Total Index 0.63 7.26 -0.56 4.81 -29.35 0.79 22.47 197 0.18** MSCI Asia Index 0.34 7.75 -0.06 3.44 -21.68 0.62 22.90 197 0.23*** MSCI Latin America 1.17 8.57 -0.65 4.90 -35.13 1.88 25.55 197 0.11 MSCI Eastern Europe Index 0.99 10.28 -0.57 5.62 -45.22 2.04 34.41 197 0.12* Bond benchmarks
JP Morgan Emerging
Markets Total 0.79 3.81 -1.97 14.70 -25.59 1.04 11.08 197 0.08 JP Morgan Asia Index 0.73 2.96 -1.36 14.88 -19.49 0.75 12.35 197 0.06 JP Morgan Latin America
Index 0.71 3.90 -1.26 8.35 -21.14 1.15 10.34 197 0.06 JP Morgan Eastern Europe
Index 1.17 6.03 -3.14 30.19 -50.77 1.24 18.10 197 0.22*** Other
Credit Spread* 0.01 0.25 0.99 10.13 -0.81 0.02 1.54 197 0.13*
Notes: all indices are analyzed on basis of excess returns, unless indicated with an asterisk (*). Returns are reported on a monthly basis and noted in percentages. *, **, and *** respectively indicate significance at the 10%-, 5%-, and 1%-level. Sources: All data is obtained from DataStream from Hedge Funds Research, Inc.15, Morgan Stanley Capital International16,
and J.P. Morgan17.
Furthermore, all indices are negatively skewed and show excess kurtosis. However, both negative skewness and excess kurtosis are more pronounced for the bond indices. Bond indices thus show relatively more often severe negative returns and are therefore less attractive to risk-averse investors. Furthermore, autocorrelation is observed for the emerging market hedge fund returns, as well as for the emerging market composite and Asian equity benchmarks and the Eastern Europe bond benchmark.
These summary statistics are also provided for an extended sample period from January 1990 to June 2011 (see Appendix D). Some points are worth noting. First, all emerging market hedge fund and equity benchmark indices have higher mean returns and slightly higher standard deviations compared to the sample starting from January 1995. Another striking difference is that the emerging market bond benchmark indices, although only available from January 1994, have lower mean returns and slightly higher standard deviations compared to the smaller sample used in this paper. The higher mean returns for emerging markets hedge fund indices can be partly explained by the higher maximum returns. On average, using data from January 1990 sketches a rosier picture of emerging market hedge funds’ return and risk characteristics when comparing to emerging market equity and bonds.18
15 See: www.hedgefundresearch.com 16 See: www.msci.com
17 See: www.jpmorgan.com
18 This might be explained by the newness of the industry in the beginning of the 1990s. Furthermore, the selection and
4.3
Correlations
Hedge funds are believed to be less correlated to the market and therefore offer investors diversification benefits. Table 2 presents the estimated correlations between the emerging market hedge fund indices, equity and bond benchmark indices and the credit spread. Emerging market hedge funds show higher correlation with the equity benchmarks compared to the bond benchmarks, which is in line with Abugri and Dutta (2009) and Eling and Faust (2010). The hedge fund indices are negatively correlated to the credit spread which supports the expectations and the reason for inclusion. Emerging market hedge funds show quite high correlations with traditional investments, hence this suggest that the diversification benefits of emerging market hedge funds are not very large. Although the correlations between the equity and bond benchmarks are not negligible, they are not alarmingly high, and therefore problems with multicollinearity19 can be ruled out.
Table 2
Correlations between emerging market hedge fund indices and the corresponding equity and bond benchmarks and the credit spread.
HFR Emerging Markets Total Index
MSCI Emerging Markets Total Index
JP Morgan Emerging
Markets Total Credit Spread HFR Emerging Markets
Total Index 1.00 MSCI Emerging Markets
Total Index 0.88 1.00 JP Morgan Emerging
Markets Total 0.69 0.66 1.00
Credit Spread -0.12 -0.03 -0.01 1.00 HFR Asia ex-Japan Index MSCI Asia Index JP Morgan Asia Index Credit Spread HFR Asia ex-Japan Index 1.00
MSCI Asia Index 0.87 1.00
JP Morgan Asia Index 0.40 0.51 1.00
Credit Spread -0.06 0.02 0.07 1.00 HFR Latin America Index MSCI Latin America
JP Morgan Latin America
Index Credit Spread HFR Latin America Index 1.00
MSCI Latin America 0.87 1.00 JP Morgan Latin America
Index 0.71 0.72 1.00
Credit Spread -0.10 -0.07 -0.03 1.00 HFR Russia/Eastern
Europe Index
MSCI Eastern Europe Index
JP Morgan Eastern
Europe Index Credit Spread HFR Russia/Eastern Europe
Index 1.00
MSCI Eastern Europe Index 0.76 1.00 JP Morgan Eastern Europe
Index 0.66 0.62 1.00
Credit Spread -0.17 -0.05 0.01 1.00
19 Two independent variables that are almost perfectly correlated can lead to a high R² while the individual coefficients
Whereas Abugri and Dutta (2009) and Eling and Faust (2010) provide correlations for different sub periods, I choose to roll the correlations between the hedge funds and equity indices and the correlations between hedge funds and bond indices using a 24-month window. The correlations between the hedge funds and equity indices are showed in Figures 1a-d in Appendix E. While they show some small fluctuations over time, they are all in the range between 0.7 and 0.95. One exception is the very low correlation in the Eastern Europe region. This can be explained by the equity index which was just launched in 1995. Furthermore, while the correlations in the Latin America and Eastern Europe regions were slightly decreasing in the period before the financial crisis in 2008-2009, a clear upsweep is showed during the crisis period. The correlations between the hedge funds and bond indices are showed in Figures 2a-d in Appendix E. Prior to the financial crisis, for all regions, except for Asia ex-Japan, the correlations decreased over time, and become even negative in Eastern Europe. This is in line with the findings of Abugri and Dutta (2009) and Eling and Faust (2010), who found that the correlations with bonds were much lower and insignificant. During the crisis period, the correlations between hedge funds and bonds rise sharply. The increased correlations can point to the fact that hedge funds have not been able to maintain their lower sensitivity to the market.
4.4
Sharpe ratios
To gather a first insight into the relative performance of emerging market hedge funds compared to the emerging market equity and bond benchmarks, Sharpe ratios are shown in Table 3. Individuals as well as institutions take expected return and volatility into consideration when deciding which securities to invest in. A very well-known (absolute) performance measurement ratio is the Sharpe ratio20, which measures risk-adjusted performance by looking at the relationship between the excess return of an asset and its standard deviation (Sharpe, 1966).
The Sharpe ratio is often criticized as this measure becomes difficult to interpret when excess returns become negative and assumes a symmetric distribution (commonly a normal distribution) which often does not hold for hedge fund returns. Eling and Schuhmacher (2007) evaluate other performance measures which for example do take into account the non-normal nature of hedge funds and conclude that the ranking of funds is practically similar for all measures. Moreover, Eling and Faust (2010) show that emerging market hedge funds outperform the benchmarks used in
20 The Sharpe ratio can be calculated by using the following formula: ,-./%.01
their study, and that this result is robust when using performance measures other than the Sharpe ratio.
The table confirms my previous findings when I analyzed the summary statistics. When looking at the period from January 1995 to June 2011, the emerging market hedge funds only have higher Sharpe ratios than the corresponding equity benchmarks, except for the Latin equity benchmark. However, when looking at the period from January 1990 to June 2011, the emerging market hedge funds have higher Sharpe ratios than both the corresponding equity and bond benchmarks, except for the Asian bond benchmark. Again, the sample period chosen appears to be important.
Table 3
Sharpe ratios for emerging market hedge funds, equity and bond benchmarks for the period from January 1995 to June 2011 and the period from January 1990-June 2011.
HFR MSCI JP Morgan E m e rg in g M a rk e ts T o ta l A si a e x-Ja p a n I n d e x L a ti n A m e ri ca I n d ex R u ss ia / E a st e rn E u ro p e In d e x E m e rg in g M a rk e ts T o ta l A si a I n d e x L a ti n A m e ri ca I n d ex E a st er n E u ro p e I n d ex E m e rg in g M a rk e ts T o ta l A si a I n d e x L a ti n A m e ri ca I n d ex E a st er n E u ro p e I n d ex 1995-2011 0.15 0.11 0.12 0.17 0.09 0.04 0.14 0.10 0.21 0.25 0.18 0.19 1990-2011 0.20 0.15 0.20 0.18 0.11 0.06 0.16 0.10 0.16 0.21 0.14 0.15
Notes: HFR are the hedge fund indices, MSCI are the equity benchmarks, JP Morgan are the bond benchmarks. All Sharpe ratios a computed by dividing excess return by the standard deviation.
Since the Sharpe ratio is a static measure it assumes that it has a stable value over time for the security in question. Consequently, its value depends on which time window is taken. A Sharpe ratio over rolling window might provide more information on the relative attractiveness of emerging market hedge funds over time. In Figure 2, the 24-month rolling Sharpe ratios21 show that the risk-adjusted performance of emerging market hedge funds returns has declined over time, while the Sharpe ratio for emerging market equity is trending upward. The Sharpe ratios are thus converging to each other. This is due to trends in both the numerator (excess return) and the
21 The rolling Sharpe ratio is based on monthly return data and calculated in a 24-rolling window. The rolling estimate for
denominator (standard deviation)22. Emerging market hedge funds have become less volatile over time, but at the same time their risk premium shows a declining trend23.
Figure 2
Sharpe ratios over a 24-month rolling window for the HFR Emerging Markets Total Index (hedge funds) and MSCI Emerging Markets Total Index (equity benchmark).
Note: “Linear” stands for the trend line for the rolling Sharpe ratio over time.
Moreover, from 2002 emerging market equity shows a higher rolling mean than emerging market hedge funds. The Credit Suisse Tremont Hedge Fund Index (2007) put forward that the declining volatility trend for emerging market hedge funds they find, might be explained by a changing investor profile from wealthy individuals to large-scale institutions.
When comparing rolling Sharpe ratios of emerging market hedge funds against emerging market bond indices (instead of equity indices), emerging market hedge funds’ risk-adjusted performance becomes less attractive over time when comparing to the bond indices (see also Graphs 3a-d in Appendix F) The only exception in this case is in the Eastern Europe region were emerging market hedge funds obtain higher Sharpe ratio values than the corresponding emerging market bond index over time.
Although the advantage of rolling Sharpe ratios is that using a shorter horizon increases the likelihood that the estimates will pick up time-variations, the disadvantage is the higher standard error of the estimate (Whitelaw, 1997). Therefore I also computed a rolling Sharpe ratio using a
22 See Graph 1a & b in Appendix F for the 24-month rolling average returns and standard deviations.
23 The same trends are found when looking at the regional indices (Asia ex-Japan, Latin America, and Eastern Europe). See
also Graphs 2a-c in Appendix F. -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1990 to 1992 1991 to 1993 1992 to 1994 1993 to 1995 1994 to 1996 1995 to 1997 1996 to 1998 1997 to 1999 1998 to 2000 1999 to 2001 2000 to 2002 2001 to 2003 2002 to 2004 2003 to 2005 2004 to 2006 2005 to 2007 2006 to 2008 2007 to 2009 2008 to 2010 2009 to 2011 HFR Emerging Markets Total Index
MSCI Emerging Markets Total Index
month window.24 The rolling Sharpe ratio of the emerging market hedge funds again shows a declining trend, and still converges to the Sharpe ratio of emerging market equity. The 48-month rolling means and standard deviation show the same trends as the 24-month rolling estimates. These observations may imply that emerging market hedge funds might have lost their attractiveness compared to the benchmark.
5.
Results
This section presents both the results for the static regression-based style analysis and the results for the dynamic regression-based style analysis.
5.1
Static regression-based style analysis results
In this first part, I present the results for the static basic and smoothing model which are applied for both the total emerging market hedge funds, as well as for the sub regions Asia ex-Japan, Latin America and Eastern Europe.25
5.1.1 The basic emerging market regression-based style analysis
Table 4 presents the results of the basic emerging market hedge funds regression-based style analysis. Before interpreting the results, the Durbin Watson statistic needs some attention. For all regions its value is lower than two (except Latin America where its value is higher than two), which may be an indication of positive (negative) autocorrelation in the error terms. To test for autocorrelation I used the Breusch (1978)-Godfrey (1978) serial correlation Lagrange multiplier test. Furthermore, I tested for heteroskedatisicity using a White’s test. To take into account autocorrelation and heteroskedasticity, I performed the regressions again using heteroskedasticy and autocorrelation consistent standard errors and covariance (Newey and West (1987).26 For all regions, hedge funds have highly significant exposure to the equity factors. Not only are these
24 Graphs 5a-c of the 48-month rolling Sharpe ratios, means and standard deviations for hedge funds and equity
benchmarks in the total emerging market area, can be found in Figure 4a-c in Appendix F. The results in the three sub-regions and when comparing to the bond benchmark instead of the equity benchmark using a 48-month rolling window are not presented in the Appendix, but show the same results as when using the 24-month rolling window.
25 Both models are also tested with a commodity factor (Thomas Reuters/Jefferies Commodity Research Bureau index)
included since according to the HFR Emerging Markets Hedge Fund Industry Report for Year End 2010 emerging market hedge funds are exposed to commodities. The commodity factor is removed from the analysis as it is only significant at the 10% level in the Asian region, and coefficients had values that are practically negligible.
26 The number of lags is automatically selected using the Akaike’s Information Criteria (AIC) (Brooks, 2008). For the Latin
America region, no serial correlation was found, and therefore the results are only determined using White’s
equity factors statistically significant, moreover they are of economic importance since the estimated coefficients range from 0.372 to 0.410 (for the regions Latin America and Asia ex-Japan respectively). Furthermore, hedge funds have significant exposure to the bond factors. One exception is the Asian region, which exposure to bonds is both statistically and economically insignificant. This is in line with Abugri and Dutta (2009) who also found the Barclays-Lehman Emerging Market Bonds insignificant for the Asian region for the whole period between January 1997 and August 2008 as well as for the sub periods. The credit spread is significant for all regions, and has the expected negative sign. Its economic importance is the strongest for the Eastern Europe region. This can be explained the fact that hedge funds in these regions are heavily invested in fixed-income instruments, which is in line with the finding of an estimated coefficient of nearly 0.4 for bonds. The intercept, i.e. alpha is positive for all regions except Latin America. Note however, that it is only significant for Asia. The adjusted R² ranges from 0.643 to 0.803.27
Table 4
Performance of emerging market hedge funds. Regression results for equation: , , ∗ ,
∗ , ∗ ! ", (1). HFR Emerging Markets Total Index HFR Asia ex-Japan Index HFR Latin America Index HFR Russia/ Eastern Europe Index α 0.212 0.331 ** -0.040 1.530 (1.054) (2.062) (-0.270) (1.658) (MSCI) 0.398 *** 0.410 *** 0.372 *** 0.391 *** (9.756) (15.210) (9.992) (9.389) (JP Morgan) 0.208 *** (-0.061) 0.235 *** 0.395 *** (3.982) -1.491 (3.530) (5.746) (Credit Spread) -2.229 *** (-2.320) *** -2.405 *** -4.101 *** (-2.950) (-4.188) (-2.903) (-2.997) Adjusted R-squared 0.803 0.772 0.778 0.643 Durbin-Watson stat. 1.508 1.769 2.195 1.666
Notes: The returns of the HFR Emerging Markets Total Index are regressed on the MSCI and JP Morgan Emerging Markets Total Index, i.e. equity and bond indices, whereas the returns of the regional emerging market hedge fund indices, i.e. HFR Asia ex-Japan, HFR Latin America, and HFR Russia/Eastern Europe, are regressed on the regional emerging market equity and bond indices. All based on excess returns. *, **, and *** respectively indicate significance at the 10%-, 5%-, and 1%-level. t-values are given in parentheses.
27 Eling and Faust (2010) find an adjusted R² of 0.8975 for their Emerging Market Model, which can be explained by the
5.1.2 The smoothing emerging market regression-based style analysis
The regression results for the smoothing emerging market hedge funds factor style analysis are presented in Table 5. The Durbin Watson statistic is now larger than two for all regions. Again I performed the Godfrey-Breusch serial correlation test and the White’s test to test for heteroskedatisicity. The autocorrelation has been eliminated for all regions except for Latin America. The results in the table are determined using White’s heteroskedasticy consistent standard errors and covariance, to correct for heteroskedasticity.28 Emerging market hedge funds’ exposure to the equity and bond factors is still highly significant and the coefficients only change marginally in magnitude. In the Asian region, hedge funds still do not have significant exposure to bonds. The credit spread is also still significant for all regions, although the coefficients have decreased somewhat in magnitude, which can be explained by the factor loadings of the lagged hedge fund return. The lagged hedge fund return is highly significant for all regions except Latin America which is a logic outcome as the Latin America hedge fund index is not facing autocorrelation. The most striking result is the smaller alpha, especially for the composite and Eastern European emerging market hedge funds. The adjusted R² for every region (except Latin America) improved slightly compared to the basic model analyzed in the previous subsection, which supports the significance of the lagged hedge fund return variable. Taking into account autocorrelation which is often present in hedge funds returns thus changes the extent to which emerging market hedge funds do add value.
The smoothing emerging market hedge fund factor analysis is also performed according to the methodology of Eling and Faust (2010). I replaced the lagged hedge fund return factor with a lagged bond factor. The regression results can be found in Appendix G. As shown in Table 1, the lagged bond factor is significant for both the analysis for the total emerging market and the Eastern Europe region. The Durbin Watson statistic, on the other hand, still indicates autocorrelation which was the reason for inclusion of this lagged bond factor in their model (Eling and Faust, 2010). Moreover, when the lagged hedge fund return factor is added to this model, the lagged bond factor becomes insignificant (see Table 2 in Appendix G). These results show that autocorrelation can be better captured by including a lag of the dependent variable, which captures both possible return smoothing and illiquidity issues.
28 For the Latin America region, serial correlation was found, and therefore the results are only determined using
Table 5
Performance of emerging market hedge funds. Regression results for equation: , , ∗ ,
∗ , ∗ ! $∗ ,% ", (2). HFR Emerging Markets Total Index HFR Asia ex-Japan Index HFR Latin America Index HFR Russia/ Eastern Europe Index α 0.069 0.278 ** -0.064 0.233 (0.561) (2.143) (-0.525) (0.732) (MSCI) 0.378 *** 0.393 *** 0.365 *** 0.376 *** (13.573) (15.034) (6.848) (9.553) (JP Morgan) 0.238 *** -0.047 0.246 *** 0.383 *** (4.389) (-1.183) (3.587) (5.892) (Credit Spread) -1.472 ** -2.046 *** -2.208 *** -2.886 *** (-2.288) (-3.598) (-2.702) (-2.200) $ (HFRI(-1)) 0.212 *** 0.148 *** 0.061 0.217 *** (5.220) (4.030) (1.603) (5.219) Adjusted R-squared 0.842 0.792 0.779 0.683 Durbin-Watson stat 2.085 2.157 2.305 2.108
Notes: The returns of the HFR Emerging Markets Total Index are regressed on the MSCI and JP Morgan Emerging Markets Total Index, i.e. equity and bond indices, whereas the returns of the regional emerging market hedge fund indices, i.e. HFR Asia ex-Japan, HFR Latin America, and HFR Russia/Eastern Europe, are regressed on the regional emerging market equity and bond indices. All based on excess returns. *, **, and *** respectively indicate significance at the 10%-, 5%-, and 1%-level. t-values are given in parentheses.
5.1.3 Robustness check: Departures from the normal distribution
An implicit assumption of performance evaluation models is that hedge funds returns are normally distributed. However, the returns of hedge funds frequently do not follow a normal distribution (Fung and Hsieh (1997), Brooks and Kat (2002), Agarwal and Naik (2004). In the section where the data was summarized the variables appeared to be negatively skewed and showed excess kurtosis, and the corresponding fat tails imply that more rare events occurred than the normal distribution would predict.
To test for departures from normality the Jarque Bera (1987) test is used. The outcome of this test using the complete sample is presented in columns 2 to 4 in Table 6.29 For all variables the
29 The residuals of the regressions are not normally distributed. The estimators are still unbiased and consistent but
hypothesis of normal distribution can be rejected since all t-statics are highly significant. The bond indices have extremely high Jarque Bera statistics, which is in line with the negative skewness of these variables in Table 1. Two remarkable exceptions are the emerging market hedge funds and equity index in the Asian region, which do not seem to depart from a normal distribution. As is often the case in financial modeling, the rejection of the normality assumption is caused by a few extreme outliers. The most extreme outliers are removed until the all variables for a specific region satisfy the normality assumption. The results of the Jarque Bera test using the adjusted sample is presented in columns 5 to 7 in Table 6. Herewith the abnormal market volatility caused by the bankruptcy of Lehman Brother is also avoided, which was the argument of Abugri and Dutta (2009) and Eling and Faust (2010) to exclude data after August 2008 from their sample.
Table 6
Jarque Bera statistic results.
All data included Outliers removed
Jarque Bera p-values N Jarque Bera p-values N HFR Emerging Markets
Total Index 202.164 *** 0.000 197 4.708 * 0.095 183 MSCI Emerging Markets
Total Index 37.209 *** 0.000 197 2.113 0.348 183 JP Morgan Emerging
Markets Total 1251.374 *** 0.000 197 0.205 0.903 183 Credit Spread 449.251 *** 0.000 197 0.946 0.623 183 HFR Asia ex-Japan Index 1.519 0.468 197 0.574 0.751 191 MSCI Asia Index 1.755 0.416 197 1.842 0.398 191 JP Morgan Asia Index 1218.505 *** 0.000 197 2.351 0.309 191 Credit Spread 449.251 *** 0.000 197 3.512 0.173 191 HFR Latin America Index 32.222 *** 0.000 197 2.207 0.332 185 MSCI Latin America 43.404 *** 0.000 197 0.129 0.937 185 JP Morgan Latin America
Index 287.335 *** 0.000 197 4.419 0.110 185 Credit Spread 449.251 *** 0.000 197 3.050 0.218 185 HFR Russia/Eastern Europe
Index 197.771 *** 0.000 197 4.664 * 0.097 182 MSCI Eastern Europe Index 67.160 *** 0.000 197 4.277 0.118 182 JP Morgan Eastern Europe
Index 6390.842 *** 0.000 197 5.436 * 0.066 182 Credit Spread 449.251 *** 0.000 197 2.995 0.224 182
As a robustness check, the basic and smoothing emerging market hedge funds factor style analysis are again computed using the sample from which the outliers are removed.30 The results from the regression analyses can be found in Appendix H, Tables 1-4. Although the results do not differ much from the analysis results using the complete sample, in the Eastern Europe region, the bond factor is less significant and the credit spread becomes even insignificant (only significant at the 10% level) in the basic (smoothing) emerging market hedge funds factor style analysis. This may point to the importance of the credit spread during crisis periods, when investors become more risk-averse and pull their sources out of relatively risky corporate bonds, i.e. emerging market bonds. Requested rate of return for emerging market bonds then increases more than that for ‘safer’ bonds, and these emerging market bonds lose value, which negatively impacts the performance of hedge funds who invested in these bonds. Furthermore, alpha in this region has decreased. Apart from this outcome, the results presented in the previous subsections are robust.
5.2
Dynamic regression-based style analysis results
In this second part, I present the results for the dynamic regression-based style analysis. In Figures 3a-h the results from the rolling regressions using a 24-month window are presented.31 Figures 3a,c,e,g show the estimated beta coefficients in a 90% confidence interval over time, which enables me to analyze the exposure of hedge funds to the equity benchmarks, i.e. the MSCI Emerging Markets Total Index, the MSCI Asia ex-Japan Index, the MSCI Latin America Index and the MSCI Eastern Europe Index. Figures 3b, d, f, h show the corresponding average returns of these equity benchmarks.
I start off with the main findings of the rolling regression using the basic model, based on equation (1)32. Thereafter I discuss the most important outcomes of the rolling regressions using the smoothing model, based on equation (2)33.
30 The number of outliers which removed from the analysis is 14, 6, 12, and 15 for respectively the total emerging
markets, Asia, Latin America and Eastern Europe region. The observations removed from the sample were mainly downside outliers.
31 All rolling regressions are performed using the specific features of the static regressions models. I checked several sub
periods to see whether the problem of heteroskedasticity had been disappeared due to the shorter horizon used. Although a substantial number of regressions using sub periods showed no heteroskedasticity problems anymore, during the crisis periods, for example January 2008 to December 2009, heteroskedasticity remained a problem. Therefore the corrections on the t-values (White’s and Newey-West) were still implemented.
32 Given by:
, , ∗ , ∗ , ∗ ! ", (1).
33 Given by:
5.2.1 The basic emerging market regression-based style analysis
During the Asian crisis, the total emerging market hedge fund industry reduced its exposure to the equity market when average returns were negative as well (see also Figure 3a-b). During 1999 hedge funds slightly increase their exposure before their exposure is reduced again from 2002 on. Emerging market hedge funds thus seem to be able to actively shift their positions. Even though equity returns become positive from 2004 on, exposure is not increased and stays on a lower level. During the financial crisis of 2008-2009, hedge funds further reduced their exposure to the equity market, and again stay relatively stable in the following years.
Figure 3a
MSCI Emerging Markets Total Index exposure of emerging market hedge funds (Estimated beta coefficients).
Figure 3b
Average returns of the MSCI Emerging Markets Total Index.
Note: Both the MSCI Emerging Markets Total Index exposure of emerging market hedge funds and average returns are estimated using a 24-month rolling window.
The results for the hedge funds in the Asian region are shown in Figure 3c-d. Although hedge funds show changing exposures over time, their exposure is increased when Asian equity shows positive returns from the second half of 2003 on and reduced from 2009 when returns became negative returns. These trends for the Asian region (both hedge fund exposure and average returns) are in line with the findings of Eling and Faust (2010), who also find periods of increasing and decreasing exposure to the Asian market which correspond with periods of positive and negative returns. Figure 3c
MSCI Asia Index exposure of emerging market hedge funds (Estimated beta coefficients).
Figure 3d
Average returns of the MSCI Asia Index.
Note: Both the MSCI Asia Index exposure of emerging market hedge funds and average returns are estimated using a 24-month rolling window.
In Figure 3e-f the results for the hedge funds in the Latin America region are shown. Remarkable is the remaining lower exposure to the Latin America equity market from 2004 on, even though returns were positive until the end of 2008.
Figure 3e
MSCI Latin America Index exposure of emerging market hedge funds (Estimated beta coefficients).
Figure 3f
Average returns of the MSCI Latin America Index.
Note: Both the MSCI Latin America Index exposure of emerging market hedge funds and average returns are estimated using a 24-month rolling window.
The exposure of hedge funds to and the return of the Eastern Europe equity markets are presented in Figure 3g-h. For this region, the graphs again shows changing exposures which occur at the same time as the changes in the attractiveness of this market. The highest exposures are obtained in the period right before and after the Asian crisis. From 2003, the exposure is to Eastern Europe equity is relatively stable. One notable observation is the increase in exposure during the second half of 2008. A possible explanation is that bonds in Eastern Europe lost their attractiveness during the crisis, and hedge funds shifted to equity in response.
Figure 3g
MSCI Eastern Europe Index exposure of emerging market hedge funds (Estimated beta coefficients).
Figure 3h
Average returns of the MSCI Eastern Europe Index.
Note: Both the MSCI Eastern Europe Index exposure of emerging market hedge funds and average returns are estimated using a 24-month rolling window.
A common result is the lower and relatively more stable exposure to equity market of hedge funds in the most recent period, with the exception of the Asian region.34 For the total emerging market area, the highest exposure35 is found in the period just before the crisis and lies over 3 standard
34 This can be explained by the fact that Asian focused hedge funds are mainly invested in stock, which is in line with the
results of the static model (see Table 4).
35 The highest and lowest (in the recent period)beta coefficients are encircled in the graphs with the hedge fund exposure
