Emerging market hedge funds in the United States

Emerging Markets Review 22 (2015) 25–42

Contents lists available at ScienceDirect

Emerging Markets Review journal homepage: www.elsevier.com/locate/emr

Emerging market hedge funds in the United States Hyuna Park ⁎ College of Business, Minnesota State University, Mankato, 150 Morris Hall, Mankato, MN 56001, United States

a r t i c l e

i n f o

Article history: Received 24 July 2014 Received in revised form 2 November 2014 Accepted 24 November 2014 Available online 3 December 2014 JEL classification: G11 G15 G23 Keywords: Geography of emerging market funds Share restrictions Flow-performance relation

a b s t r a c t This paper compares the performance and capital flow to emerging market hedge funds located in the US with those of the funds located in other countries. I find that the US funds on average provided neither a positive risk-adjusted return nor a liquidity premium to compensate for the stronger share restrictions they imposed than the other funds during 1994–2012. Quarterly flow to the US funds was 1.25 to 7.88 times higher but it was less sensitive to past performance. Smart money effect was observed among the Caribbean funds, Hong Kong funds and Singapore funds, but not among the US funds. Published by Elsevier B.V.

1. Introduction How well have US investors done in emerging markets? Do past performance influence capital flows to emerging markets? Do flows have forecasting power for future returns in emerging markets? The main objective of this paper is to answer these questions empirically using hedge fund data while different data have been utilized in previous research. For example, Bekaert and Harvey (2000, 2003) use accumulated capital flow data from the US Treasury to analyze the performance of US investors in emerging equity markets. That is, their definition of US investors is comprehensive, including all US investments covered by the aggregate equity flow statistics. Their finding is mixed; they show some evidence of US investors' ability to choose the right countries but the overall US allocation performance is quite similar to the performance that would have been obtained from market capitalization weighting. In contrast, Froot et al. (2001) define emerging market investors more specifically and focus only on institutional investors using international settlement data from a large custodian bank. Following the daily

⁎ Tel.: +1 507 389 5406; fax: +1 507 389 5497. E-mail address: [email protected].

http://dx.doi.org/10.1016/j.ememar.2014.11.004 1566-0141 Published by Elsevier B.V.

26

H. Park / Emerging Markets Review 22 (2015) 25–42

international settlements of institutional investors in 44 countries during 1994–1998, they find that flows to emerging markets are strongly influenced by past returns and inflows have positive forecasting power for future returns. However, their sample of institutional investors combines many different types of institutions such as pension funds, endowments, mutual funds, and governments. Kaminsky et al. (2001) focus on analyzing mutual funds in emerging markets especially during the financial crisis of 1997–1998. They show that there were large withdrawals from emerging market mutual funds during the crisis. Both mutual funds and hedge funds invest in emerging markets, but hedge funds have more flexibility. While most mutual funds invest only in stocks and bonds, hedge funds can use real estates, commodities, currencies, and derivatives. As emerging market hedge funds can also use leverage, expected return as well as risk can become significantly higher. Eling and Faust (2010) compare hedge funds and mutual funds active in emerging markets and find that hedge funds outperform mutual funds especially in bad or neutral market environments during the sample period of Jan 1995–Aug 2008. Using the Chow test on structural breaks, they show that hedge funds adapt to changing market environments while mutual funds do not. As shown in Fig. 1, emerging market hedge funds have grown very rapidly during the past decade especially in the US. According to Tremont Advisory Shareholders Service (TASS) database, the assets under

18000 16000 14000

AUM (US$mm)

12000 10000 8000 6000 4000 2000

Jan94 Sep94 May95 Jan96 Sep96 May97 Jan98 Sep98 May99 Jan00 Sep00 May01 Jan02 Sep02 May03 Jan04 Sep04 May05 Jan06 Sep06 May07 Jan08 Sep08 May09 Jan10 Sep10 May11 Jan12 Sep12

0

US % of Total AUM by Management Company Locaon US Caribbean UK HKSP EM

Caribbean

UK

January 1994 Total AUM ($US 2.32bn) 5 43 20 0 17

HKSP

EM

June 2007 ($US 51.4bn) 32 10 8 5 10

December 2012 ($US 26.8bn) 24 13 30 10 18

Fig. 1. Total AUM of emerging market hedge funds by the location of fund management company. US means the management company is located in one of the 50 states of the United States of America. Caribbean means the management company is located in Cayman Islands, British or US Virgin Islands, Bahamas, Bermuda, Anguilla, or Saint Kitts and Ne. UK means the management company is in United Kingdom. HKSP means the management company is in Hong Kong or Singapore. EM means the management company is located in emerging markets: Argentina, Brazil, Chile, China, Czech Republic, India, Indonesia, Malaysia, Mauritius, Russia, South Africa, Thailand, Turkey, or United Arab Emirates.

H. Park / Emerging Markets Review 22 (2015) 25–42

27

management (AUM) of emerging market hedge funds grew from US$2.3 billion in Jan 1994 to its peak at US $51.4 billion in June 2007 before going down to US$26.8 billion in Dec 2012.1 Note that the geography of emerging market hedge funds has changed with the growth of the AUM; 43% of the total AUM was managed by the funds located in the Caribbean such as Cayman Islands in 1994, but the proportion is only 13% in 2012. In contrast, emerging market hedge funds located in the United States were managing only 5% of the AUM in 1994, but they are managing 24% of the AUM in 2012.2 Did the rapidly growing US emerging market hedge funds provide a high risk-adjusted return? Kotkatvuori-Örnberg et al. (2011) show that emerging market hedge funds on average do not outperform the market benchmarks, and the funds with a geographical focus outperform the funds that invest globally. Teo (2009) analyzes the risk-adjusted performance of Asia-focused hedge funds using three databases: EurekaHedge, AsiaHedge, and HFR (Hedge Fund Research). He finds that the US funds and the UK funds that focus on Asia underperform the funds that are physically located in Asia. However, previous research does not analyze how US funds perform compared to the funds in other countries when the data include not only Asia-focused funds but all emerging market hedge funds. Prior research does not shed light on how performance affects flow to emerging market hedge funds, either. Therefore, this paper compares the risk-adjusted performance and flow-performance relation of all emerging market hedge funds in the United States with otherwise similar funds located in other countries. Using 881 emerging market hedge funds located in 40 countries during 1994–2012, I find that US funds on average impose stronger share restrictions than the other funds but did not provide a higher risk-adjusted return.3 For example, the average lockup period of US funds is about two times longer than that of the other funds (5 vs. 2.6 months), but the average alpha of US funds is negative and lower than that of the other funds (−0.31 vs. −0.19% per month). That is, the positive relation between lockup provision and performance of hedge funds reported in the hedge fund literature does not hold for emerging market hedge funds. I also find that the positive relation between lockup and asset illiquidity reported in the hedge fund literature does not hold for emerging market hedge funds either. For example, the average asset illiquidity of UK funds is significantly higher than that of US funds (0.21 vs. 0.14, t-statistic: 2.42), but the average lockup period of UK funds is significantly shorter than that of US funds (1.6 vs. 5 months, t-statistic: 5.12).4 At a portfolio level test, I find that emerging market hedge funds on average failed to generate a positive risk-adjusted return across all manager locations during 1994–2012. At an individual fund-level test, I find that 12% of emerging market hedge funds generated a significantly positive risk-adjusted return during the sample period, but the proportion was much higher for the funds that are located in emerging markets than in the US (20.34 vs. 9.72%). This finding is consistent with Teo (2009) who shows that local information advantage is strong for emerging market hedge funds and thus the funds with a physical presence in their investment region outperform the other funds. I also find that US funds are geographically less focused than the other funds, especially the ones in Hong Kong or Singapore.5 As previous research shows that geographically focused emerging market hedge funds outperform the ones that invest globally, I test whether the focused US funds outperform the global US funds. After controlling the impact of the lockup provision, I find that neither the US focused fund portfolio nor the US global fund portfolio has a significantly positive alpha. That is, in case of the US funds, I find no evidence of geographical focus affecting risk-adjusted performance. Another important difference between the US funds and the other funds is found in capital flow. The US funds on average have a higher quarterly net flow but they show a weaker flow-performance relation than the other funds (Adjusted R-square: 0.87% vs. 6.33%). The test of smart money effect as in Grinblatt and 1 I define emerging markets as the constituent countries underlying the MSCI Emerging Market Index or Frontier Market Index as of December 31, 2013 (source: http://www.msci.com/products/indices/country_and_regional/em/ and http://www.msci.com/products/ indices/country_and_regional/fm/). 2 In this paper, the location of an emerging market hedge fund is defined by the address of its management company reported to TASS. For example, if the address of a fund's management company is in New York, it is classified as a US fund. 3 See Aragon (2007) for details on share restrictions such as a lockup provision, and the relations among share restrictions, asset illiquidity, and risk-adjusted performance of hedge funds. 4 Following Lo (2001), Getmansky et al. (2004), and Khandani and Lo (2007), I use the first order serial correlation coefficient of a fund's returns as a measure of asset illiquidity. 5 While 95% of the HKSP funds focus on investing in Asia, 51% of the US funds invest globally instead of focusing on a specific region.

28

H. Park / Emerging Markets Review 22 (2015) 25–42

Titman (hereafter GT, 1993) and Zheng (1999) also confirms a weaker relation between capital flow and performance in the US funds than in the other funds (Newey-West t-statistic: −0.43 for US funds vs. 2.13 and 1.86 for Caribbean and HKSP funds, respectively). When examining these empirical results, note that the US funds on average have a longer history than the other funds (for example, average age of US vs. HKSP funds: 70 vs. 55 months, t-statistic for the difference: 2.31) and this age difference may affect the estimated risk and performance as discussed in Stambaugh (1997). Therefore, as a robustness check, I adjust the estimated risk of short history funds using the information on the long history emerging market index, and show that age difference does not have a significant impact on the evaluation of the risk-adjusted performance of emerging market hedge funds. Overall these results show that the emerging market hedge funds located in the US on average provided neither a positive risk-adjusted return nor a liquidity premium to compensate for the strong share restrictions they impose. Investors of the US funds were slower in withdrawing capital out of poor performers compared to the investors of the emerging market hedge funds located in other countries. The rest of the paper is organized as follows. Section 2 describes data and Section 3 explains methodology with details on the robustness test provided in Appendix A and description of variables included in Appendix B. Section 4 presents empirical results and Section 5 concludes.

2. Data Monthly returns, assets under management (AUMs), and other characteristics of emerging market hedge funds are obtained from TASS, the most widely used database in the hedge fund literature. Fund characteristics provided by TASS include fee structure, address of fund management company, share restrictions such as a lockup period, geographical focus, and so on. In order to mitigate survivorship bias, I include both live and defunct funds in the analyses. TASS includes information on defunct funds as well as live funds, but the defunct funds database does not retain funds that dropped out of the live fund database before 1994. Therefore, the sample period starts in January 1994 as in prior studies in the hedge fund literature and it ends in December 2012.6 Among the 19,003 funds that report monthly net of fee returns to TASS, 1057 are emerging market hedge funds. Among the 1057 funds, 881 funds use US dollar as their currency, 105 funds use Euro, 38 funds use British pounds, and 33 funds use other currencies such as Swiss Franc, for example. As shown in Table 1, 66% of the 881 funds are defunct funds but the proportion varies across different manager locations. For example, among HKSP funds only 38% are defunct funds but the proportion is 69% for US funds. Table 1 also shows the variation in geographical focus across different manager locations; more than half of US and UK funds invest globally while most HKSP funds focus on Asia.7 Before comparing the risk and return of these funds, back-fill bias needs to be adjusted. Following Aggarwal and Jorion (2010a, 2010b), I delete return observations of each fund dated before the date when the fund was entered into the database. After this adjustment, there remain 727 funds that have at least 12 monthly return observations. These funds are used to evaluate the risk-adjusted performance of emerging market hedge funds.

3. Methodology 3.1. Risk-adjusted performance The standard approach to evaluate fund performance, dating back to Jensen (1968), is to regress fund returns on systematic risk factors. Fung and Hsieh (hereafter FH, 2004) identify seven asset-based style factors 6 See Fung and Hsieh (2000, 2002) for details on biases in hedge fund databases and measures to mitigate them. For reasons why hedge funds drop out of the live fund database and move to the defunct fund database, see Liang and Park (2010). 7 In Tables 1 and 6, “Global” means GF_Global dummy variable in TASS of 1 and “Focus” is for GF_Global dummy of 0. The sum of the percent of focused funds for each region may exceed “100% — percent of global funds” because some funds may focus on more than one region. For example, among the 49% (=100%–51%) of US funds that have a geographical focus, some funds may focus on both Africa and Eastern Europe.

H. Park / Emerging Markets Review 22 (2015) 25–42

29

Table 1 Manager location and geographical focus. This table compares the total AUM and the geographical focus of emerging market hedge funds by the location of the fund management company. “Global” means emerging market funds that invest globally instead of focusing on specific regions. Location of fund manager

US Others Caribbean UK HKSP EM All

Number of funds

173 708 167 93 91 67 881

Number and proportion of defunct funds

120 (69%) 459 (65%) 57 (34%) 58 (62%) 35 (38%) 32 (48%) 579 (66%)

Total AUM ($bn as of Dec 2012)

6.53 20.31 3.51 8.15 2.65 4.88 26.84

Geographical focus (% of funds) Global

51 31 34 52 5 12 35

Focused Asia

Latin America

Eastern Europe

Africa

27 37 27 17 95 37 35

22 15 19 10 0 27 16

17 19 27 31 0 14 18

12 7 7 6 0 12 8

to explain the variation in returns on portfolios of hedge funds. To identify the best factor model, I first construct an equally weighted portfolio of emerging market hedge funds after adjusting for backfill bias. Then I apply the seven factors of FH (2004) to the portfolio: i) the excess return on the market defined as the value-weighted return on all NYSE, AMEX, and NASDAQ stocks minus the one-month Treasury bill rate (MFA), ii) the size factor (SMB) as in Fama and French (1993), iii) the excess return on Fama Treasury bond portfolio with maturities greater than ten years (TERM), iv) the excess return on the Citi group corporate BBB 10+ year index minus TERM (CREDIT) as in Jagannathan et al. (2010), v–vii) the excess returns on the portfolios of look back straddle options on bonds, currencies, and commodities, respectively (TFBD, TFFX, and TFCOM) as in Fung and Hsieh (2001).8 As shown in Panel A of Table 2, the seven-factor model explains only 48% of the variation in the emerging market fund portfolio return. When the excess return on MSCI (Morgan Stanley Capital International) emerging market monthly total return index (EMF) is added to the model, the explanatory power increases sharply; the adjusted R-square of the eight factor model is 78%. I also add two lags of EMF to the model in order to adjust for illiquid holdings and serial correlation in hedge fund returns.9 I examine different combinations of these eight factors and find that a four factor model (EMF, SMB, TERM, and CREDIT) with two EMF lags (EL1 and EL2) has the highest adjusted-R2. Therefore, I use this four-factor model to measure the riskadjusted performance of emerging market hedge funds. 3.2. Serial correlation and conditional return smoothing Getmansky et al. (2004) document substantial positive serial correlation in reported monthly hedge fund returns. They point out that serial correlation can be caused by purposeful managerial smoothing of contemporaneous and lagged asset returns as well as illiquidity of assets in the fund portfolio. Bollen and Pool (2008) build on these findings and develop a model to distinguish purposeful managerial smoothing from other reasons of serial correlation. They argue that reported hedge fund returns will feature conditional serial correlation if a hedge fund manager fully reports gains but delays reporting losses. During periods of large positive returns, managers will fully report returns not to be behind their competitors. During periods of huge negative returns, they may only partially report fund returns to mitigate capital flight. They suggest using conditional serial correlation (CSC) as an indicator of manipulated returns. The CSC of a fund is estimated using the following regression: þ

−

r t ¼ a þ b1 r t−1 þ b1 It−1 r t−1 þ εt

ð1Þ

8 I thank Kenneth French and David Hsieh for providing downloadable data on their websites. The size factor is from http://mba.tuck. dartmouth.edu/pages/faculty/ken.french/data_library.html and the trend-following factors are from http://faculty.fuqua.duke.edu/ ~dah7/HFRFData.htm. 9 See Asness et al. (2001) and Getmansky et al. (2004) for details on illiquidity and serial correlation in hedge fund returns.

30

H. Park / Emerging Markets Review 22 (2015) 25–42

Table 2 Factor models for emerging market funds. Panel A compares regression models using the portfolio of all emerging market funds. Panel B presents results from a four factor model and the portfolios formed by the location of fund management company. ***, **, and * denote statistical significances at the 1%, 5%, and 10% levels, respectively. Panel A: model comparison Model

Beta EMF

Alpha EL1

EL2

0.09 (4.96)*** 0.09 (5.13)***

0.45 0.18 −0.50 (8.51)*** (3.03)*** (−0.49) −0.06 0.08 −0.02 (−1.32) (1.95)* (−0.24) 0.04 −0.01 0.06 0.06 (2.30)** (−0.30) (1.58) (0.92) 0.04 0.06 0.05 (2.44)** (1.54) (0.84)

1 2 3 4

0.52 (17.26)*** 0.51 (17.56)*** 0.50 (23.53)***

MFA

SMB

TERM

CREDIT

TFBD

TFFX

TFCOM

0.37 (3.31)*** 0.08 (1.05) 0.00 (0.04) 0.00 (0.03)

−0.02 (−1.33) −0.01 (−1.05) −0.01 (−0.65)

0.00 (0.02) 0.00 (0.06) 0.00 (0.41)

0.01 (0.47) 0.01 (0.57) 0.01 (0.63)

Adj-R2

0.06 0.48 (0.29) 0.13 0.78 (0.97) 0.05 0.80 (0.38) 0.05 0.81 (0.41)

Panel B: alpha and beta by the location of fund manager Location of manager

(Jan 1994–Dec 2012) US Caribbean UK HKSP EM

(Jan 1994–June 2007) US Caribbean UK HKSP EM

(July 2007–Dec 2012) US Caribbean UK HKSP EM

Beta

Alpha

Adj-R2

0.73

EMF

EL1

EL2

SMB

TERM

CREDIT

0.54 (18.27)*** 0.58 (15.85)*** 0.49 (14.80)*** 0.38 (18.14)*** 0.60 (16.32)***

0.11 (4.40)*** 0.20 (6.32)*** 0.11 (4.07)*** 0.05 (2.79)*** 0.11 (3.46)***

0.07 (2.70)*** 0.00 (0.03) 0.04 (1.60) 0.04 (2.30)** 0.05 (1.67)*

0.09 (1.80)* −0.07 (−1.10) 0.07 (1.25) −0.02 (−0.45) 0.03 (0.52)

0.09 (1.08) 0.04 (0.35) 0.02 (0.20) 0.00 (0.02) 0.22 (2.07)**

−0.06 (−0.63) −0.07 (−0.64) −0.17 (−1.63) 0.09 (1.38) 0.14 (1.22)

0.07 (0.42) −0.08 (−0.39) 0.08 (0.44) −0.03 (−0.21) 0.00 (0.00)

0.60 (14.73)*** 0.62 (12.15)*** 0.59 (13.14)*** 0.36 (13.55)*** 0.64 (12.43)***

0.14 (3.85)*** 0.27 (6.04)*** 0.15 (3.85)*** 0.09 (3.78)*** 0.13 (2.96)***

0.06 (1.79)* −0.03 (−0.74) 0.07 (1.74)* 0.04 (1.90)* 0.03 (0.59)

0.08 (1.33) −0.09 (−1.19) 0.03 (0.43) −0.02 (−0.48) 0.03 (0.36)

0.18 (1.44) 0.14 (0.91) 0.12 (0.92) −0.03 (−0.36) 0.28 (1.86)*

0.02 (0.15) 0.06 (0.33) 0.04 (0.22) 0.15 (1.39) 0.24 (1.23)

0.07 (0.30) 0.11 (0.37) 0.12 (0.47) −0.09 (−0.61) −0.01 (−0.03)

0.70

0.43 (16.18)*** 0.50 (17.82)*** 0.30 (18.00)*** 0.41 (13.28)*** 0.52 (14.36)***

0.06 (2.76)*** 0.08 (3.61)*** 0.05 (4.07)*** −0.01 (−0.05) 0.06 (2.15)**

0.07 (3.36)*** 0.06 (2.80)*** 0.02 (1.40) 0.03 (1.42) 0.09 (3.12)***

−0.09 (−1.18) −0.13 (−1.62) −0.09 (−1.81)* −0.09 (−0.99) −0.08 (−0.73)

−0.04 (−0.49) −0.04 (−0.45) −0.09 (−1.86)* 0.03 (0.30) 0.13 (1.19)

0.10 (1.44) 0.03 (0.35) 0.03 (0.76) 0.09 (1.11) 0.20 (2.06)**

0.07 (0.43) −0.57 (−3.14)*** −0.12 (−1.12) 0.12 (0.60) −0.01 (−0.03)

0.90

0.68 0.63 0.72 0.69

0.63 0.66 0.66 0.63

0.91 0.92 0.85 0.89

where rt is the demeaned fund return at time t and It − 1 is an indicator variable that equals one if the systematic component of the fund return at time t − 1 from the 4-factor model is less than the mean systematic return. If b− 1 of a fund is significantly different from zero, the fund has a conditional serial correlation.

H. Park / Emerging Markets Review 22 (2015) 25–42

31

3.3. The relation between past performance and flow After analyzing the risk-adjusted performance and conditional smoothing of emerging market hedge funds, I test the relation between performance and flow. Sirri and Tufano (1998) study the flow of funds into and out of equity mutual funds, and find that investors disproportionately flock to high performing funds while failing to flee lower performing funds at the same rate. Using a similar methodology, I analyze the flow to emerging market hedge funds. The capital flow into a fund during a quarter is measured by the growth rate of net new money, which is defined as follows.

 Flowi;t ¼ TNAi;t −TNAi;t−1 1 þ Ri;t =TNAi;t−1

ð2Þ

TNAi,t is fund i's total net assets at the end of quarter t, and Ri,t is the fund's return during the quarter. That is, Flowi,t represents the percentage growth of a fund during the quarter in excess of the growth that would have occurred if no new money had flowed in. As in previous research, the top and bottom 1% of the flows are winsorized to mitigate the effect of outliers. First, I compare the mean and standard deviation of quarterly flows across different locations of fund management company. Then, I group the funds into “US funds” and “Others” and run a piecewise linear regression of investor flows on relative performance variables Lowt, Midt, and Hight. These variables are defined using a fractional rank (FRANK) that represents a fund's percentile performance that ranges from 0 to 1. The bottom performance tercile (Lowt) is defined as Min (1/3, FRANKt − 1), the middle performance tercile (Midt) is Min (1/3, FRANKt − 1 − Lowt), and the highest performance tercile (Hight) is defined as Min (1/3, FRANKt − 1 − Lowt − Midt). For example, if a fund's FRANKt − 1 is 0.80, its Lowt is 0.33, its Midt is 0.33, and its Hight is 0.14. The regression includes the logarithm of the size in the previous period (Log (TNAt − 1)) as a control variable because an equal dollar flow has a larger percentage impact on smaller funds. I also include the standard deviation of monthly returns during the previous year, lockup dummy, incentive fee, and high-water mark (HWM) dummy as control variables. I run the regressions quarterly, then calculate the Fama and MacBeth (1973) coefficients as well as the t-statistics. As I require at least 24 funds in the cross-sectional regression for each quarter, the sample period for the flow-performance regression starts in Oct 1999, not in Jan 1994, and ends in Dec 2012. Grouping all the funds whose management company is not in the US into one group (Others) in Panel B of Table 7 is also due to the sample size restriction. That is, the sample sizes for Caribbean, UK, HKSP, and EM funds are not large enough to generate time series of separate cross-sectional regressions, and thus they are combined into “Others” group and analyzed together. 3.4. Smart money effect: the test of Investors' fund selection ability After analyzing the relation between flow and past performance, I test the relation between flow and future performance using Grinblatt and Titman (GT, 1993) performance measure. Previous research used the GT measure to test the fund selection ability of mutual fund investors.
 GT ¼ ∑i Ri;tþ1 wi;t –wi;t−1

ð3Þ

where Ri,t + 1 is the return of fund i between time t and t + 1 and wi,t is the weight for fund i at time t. Following Zheng (1999), I calculate quarterly portfolio weights as the AUM of each fund divided by the sum of AUM of all funds for each quarter. The weights for quarter t minus weights for quarter t − 1 are multiplied by the corresponding future monthly returns of the funds. For example, the Jan 2010 through Mar 2010 returns are multiplied by the difference between portfolio weights on Dec 31, 2009 and the weights on Sep 30, 2009. Under the null hypothesis that investors have no selection ability, the GT measures are serially uncorrelated with mean zero. To adjust for autocorrelation, I present the t-statistic based on the Newey and West (1987) heteroskedasticity and autocorrelation consistent covariance matrix as well as the traditional t-statistic. As the sample period is set to include at least 10 funds in the cross-section of each region when the GT measure is estimated, it is shortest for the HKSP funds (Nov 2007–Dec 2012, 62 months) and the longest

32

H. Park / Emerging Markets Review 22 (2015) 25–42

for the US funds (Dec 1996–Dec 2012, 193 months). As HKSP funds are younger than US funds on average, the difference in sample periods may affect the estimated performance. Therefore, in the next section I test the impact of short history on the estimated performance of emerging market hedge funds as a robustness check.

3.5. Analyzing the impact of short history on the evaluation of performance Stambaugh (1997) analyzes investments whose histories differ in length and shows that the longer histories provide greater information about risk and return, not only for the longer-history assets, but for the shorter-history assets as well. He presents two methods to adjust the standard deviation (STD) and Sharp ratio (SR) of short history assets using the returns of longer history assets: i) an adjustment based on maximum-likelihood estimation and ii) an adjustment using the Bayesian approach. Following his approach, I define STDAdj and STDB as in Eqs. (A.1)–(A.5) in Appendix A. The subscript “Adj” is for the adjustment based on maximum-likelihood estimation and “B” is for the Bayesian approach. Using these equations, I adjust estimated risk of short-history funds to examine whether the differences in length of return history of emerging market hedge funds affect estimated risk-adjusted performance.

Table 3 Risk and return of emerging market funds by manager location. Panel A presents the cross-sectional average values of the size, age, average return, risk, and risk-adjusted performance of emerging market hedge funds by the location of the fund management company. For example, “Average AUM” reported in the second column is the cross-sectional average value of the time-series average of month-end AUM of each fund during the fund life. “4-factor Model Alpha” and “EMF Beta” are cross-sectional average values of alpha and the emerging market factor beta from the four-factor model that includes EMF, SMB, TERM, and CREDIT with two lags of EMF. Panel A also reports the t-statistics for the differences between US funds and funds in other locations. ***, **, and * denote statistical significances at the 1%, 5%, and 10% levels, respectively. Panel B shows the cross-sectional distribution of the t-statistic of alpha and the average alpha of funds with a positive and significant alpha. Panel A. Risk and return: cross-sectional average Location of manager

Average AUM ($mm)

US 110 Others 92 Caribbean 58 UK 163 HKSP 71 EM 117 All EMF 96 t-statistic for US — Caribbean 2.59*** US — UK −1.38 US — HKSP 1.63 US — EM −0.23

Age Average (months) return (%) 70 64 66 70 55 81 65

Standard deviation (%)

0.20 0.23 0.09 0.39 0.31 0.62 0.22

0.72 0.00 2.31** −1.42

Sharpe ratio

6.12 5.42 6.13 4.49 4.66 6.79 5.56

0.07 0.09 0.06 0.07 0.10 0.32 0.08

−0.02 3.48*** 2.92*** −1.07

0.52 −0.95 −0.58 −1.27

0.15 −0.13 −0.53 −1.61

4-Factor-model alpha (%)

EMF beta

−0.31 −0.19 −0.41 0.01 −0.18 0.17 −0.22

0.46 0.43 0.46 0.34 0.44 0.56 0.43

0.44 −1.66* −0.61 −1.30

0.00 2.47** 0.36 −1.39

Panel B. Cross-sectional distribution of the 4-factor-model alpha Location of manager

Number of funds

% of funds with a significantly negative or positive alpha t ≤ −2.326 (one-sided 1%)

t ≤ −1.960 (2.5%)

t ≤ −1.645 (5%)

t ≥ 1.645 (5%)

t ≥ 1.960 (2.5%)

t ≥ 2.326 (1%)

Average alpha of the funds with a significantly positive alpha (t ≥ 1.645)

US Caribbean UK HKSP EM All EMF

144 149 77 66 59 727

2.78 12.75 1.30 4.55 6.78 6.60

6.25 13.42 5.19 4.55 8.47 9.35

9.72 15.44 14.29 9.09 10.17 12.79

9.72 10.74 10.39 9.09 20.34 12.10

6.94 8.05 7.79 6.06 10.17 8.25

2.08 6.04 5.19 4.55 8.47 5.64

1.17% 1.03% 1.16% 0.84% 2.80% 1.39%

H. Park / Emerging Markets Review 22 (2015) 25–42

33

4. Empirical results 4.1. Risk-adjusted performance In order to measure the risk-adjusted performance of emerging market hedge funds, I use the four factor model (EMF, SMB, Term, and Credit) with two EMF lags (EL1 and EL2) as developed in Section 3.1 at a portfolio level as well as an individual fund level. In the portfolio-level test, equally-weighted portfolios are formed by the location of fund management company. As shown in Panel B of Table 2, the explanatory power of EMF is high and alpha is not significantly different from zero across all manager locations. That is, the portfolios of emerging market hedge funds did not generate a positive risk-adjusted return during 1994–2012. I also do a sub-period analysis in order to examine if the poor performance of emerging market hedge funds was driven by the global financial crisis of 2007–2009. I find that the financial crisis is not the main reason for the poor performance. Emerging market hedge fund portfolios failed to generate a positive riskadjusted return during Jan 1994–June 2007 (before the crisis) and July 2007–Dec 2012 (after the crisis). Note that the explanatory power of the four-factor model is uniformly higher during the second sample period across all manager locations (Adj-R2: 63–70% vs. 85–92%). That is, emerging market hedge funds are following their market benchmark index (EMF) more closely after the global financial crisis. After finding the poor performance of emerging market hedge funds at the portfolio level, I examined the risk-adjusted performance of individual funds. As shown in Table 3, 12.1% of emerging market hedge funds (88 out of the 727 funds) have a positive and significant alpha at the 5% level. The proportion is lower than the average in the US funds (9.7%) and higher among EM funds (20.3%). This finding is consistent with Teo (2009) who shows that emerging market hedge funds that are located closer to their investments perform better. 4.2. Share restrictions and asset liquidity Previous research in the hedge fund literature shows that share restrictions are related to performance. Ackermann et al. (1999) and Liang (1999) find a positive relation between aggregate-level hedge fund returns and lockup provisions. Aragon (2007) finds lockup funds have a higher alpha than non-lockup funds and argues that share restrictions allow hedge funds to efficiently manage illiquid assets, and these benefits are Table 4 Share restrictions, asset illiquidity, and fee structure. This table compares share restriction, asset illiquidity measured by the first order serial correlation, and the fees of emerging market funds across different manager locations. “US Onshore” means the fund management company is located in the United States and the fund is registered in the United States. “US Caribbean” means the fund management company is located in the United States but the fund is registered in Cayman Islands, British or US Virgin Islands, Bahamas, Bermuda, Anguilla, or Saint Kitts and Ne. Bottom rows present t-tests for the differences between US funds and other funds. ***, **, and * represent the statistical significances at the 1%, 5%, and 10% levels, respectively. Location of manager

Lock-up period (months)

Minimum investment ($mm)

Redemption notice period (days)

Redemption frequency (days)

Subscription frequency (days)

US US Onshore US Caribbean Others Caribbean UK HKSP EM All t-statistic for US — Caribbean US — UK US — HKSP US — EM

5.06 5.79 4.96 2.62 2.78 1.58 2.82 1.78 3.09

2.34 0.61 3.30 0.60 0.41 0.95 0.39 0.70 0.94

42.94 41.05 45.38 35.45 34.01 41.98 24.96 40.42 36.87

72.10 78.46 69.67 44.90 39.35 44.92 37.93 48.94 50.05

35.05 40.49 32.51 28.23 26.47 28.32 29.13 30.24 29.58

2.91*** 5.12*** 2.69*** 4.63***

4.04*** 2.43** 4.10*** 3.19***

2.91*** 0.17 5.69*** 0.53

6.98*** 3.85*** 6.92*** 4.01***

5.67*** 4.47*** 3.60*** 2.73***

Asset illiquidity 0.1395 0.1378 0.1364 0.1564 0.1388 0.2128 0.1384 0.1468 0.1514 0.03 −2.42** 0.04 −0.22

Management fee (%) 1.56 1.49 1.59 1.62 1.69 1.63 1.43 1.55 1.61 −2.69*** −1.24 2.07** 0.15

Incentive fee (%) 17.64 17.00 18.02 16.76 17.29 16.40 14.23 17.84 16.94 0.53 1.52 3.54*** −0.25

34

H. Park / Emerging Markets Review 22 (2015) 25–42

captured by investors as a share illiquidity premium. However, these authors analyze the data of all hedge funds that include many different investment styles and thus it is not clear whether these findings would still hold when we analyze emerging market hedge funds exclusively. I find that the positive relations among share restrictions, asset illiquidity, and risk-adjusted performance reported in the hedge fund literature do not hold for emerging market hedge funds. For example, Tables 3 and 4 show that UK funds on average invest in more illiquid assets than US funds (serial correlation: 0.2128 vs. 0.1395), but their average lockup period is shorter than US fund's (1.58 vs. 5.06 months) and they have a higher average alpha than US funds (0.01% vs. −0.31% per month during 1994–2012). Note that US funds on average impose a longer lockup period and a longer redemption notice period, process redemption and subscription less frequently, and require a higher minimum investment than the other funds even though they invest in more liquid assets and have a lower alpha. That is, the stronger share restrictions used by emerging market hedge funds in the United States are neither attributable to a higher asset illiquidity nor providing share illiquidity premium for their investors. As asset illiquidity does not explain the strong share restriction of US funds, I divide the US funds sample into “US Onshore” and “US Caribbean” groups based on the place of fund registration in order to test whether regulation in the US explains the stronger share restriction of US funds. Both “US Onshore” and “US Caribbean” funds have their management company located in the US but “US Onshore” funds are registered in the US while “US Caribbean” funds are registered in the Caribbean. Aragon et al. (2014) show that onshore hedge funds impose stronger share restrictions than hedge funds registered in the Caribbean because of regulatory constraints on raising capital such as restrictions on the number and type of investor accounts and strict marketing prohibitions.10 Consistent with previous research, Table 4 shows that US onshore funds have a longer lockup period than US Caribbean funds on average (5.79 vs. 4.96 months). However, even US Caribbean funds on average have stronger share restrictions than funds in all other locations. That is, neither asset illiquidity nor regulatory constraints on raising capitals explain the difference in share restrictions between US funds and other funds. 4.3. Conditional serial correlation (CSC) and return smoothing Bollen and Pool (2008) use CSC as a measure of conditional return smoothing. They argue that CSC is a leading indicator of fraud because reported returns of hedge funds will have CSC if true returns of hedge funds are independently distributed and a hedge fund manager fully reports gains but delays reporting losses. In order to examine if the potential for fraud and return smoothing of emerging market funds is affected by the location of fund management company, I compare the CSCs of US funds and other funds. As was explained in Section 3.2, a fund has a CSC if b− 1 of the regression in Eq. (1) is significantly positive. To make the results comparable to Bollen and Pool's, two-sided t-tests are used. They find that only 4.83% of the hedge funds they examined have a CSC at the two-sided 5% level, and I find a similar result. As shown in Table 5, 4.13% of emerging market funds (30 out of 727 funds) have a CSC at the two-sided 5% level. Note that the proportion of funds with a CSC is lower than the significance level across most locations at the significance levels of 5% and 10%. That is, there is no significant evidence of conditional return smoothing observed in emerging market hedge funds across all locations. Note that “Unknown” group has the highest proportion of funds with a CSC at all significance levels. That is, we are more likely to find a fund that has conditionally smoothed returns among the funds that do not report the location of the fund management company to the database. 4.4. The impact of geographical focus on the performance of US funds Why do emerging market hedge funds managed by a company located in the United States perform poorly on average even though they impose stronger share restrictions than the funds located in other countries? Is it because they are less focused geographically when selecting their investments? In order to answer these questions, I construct portfolios of US funds and other funds based on geographical focus and compare the four-factor model alpha after controlling the impact of a lockup provision. That is, 10 Cumming et al. (2014) find that US hedge funds that are registered in Delaware impose stronger share restrictions than the funds registered in other states, and the proportion of Delaware funds has increased during 1994–2010.

H. Park / Emerging Markets Review 22 (2015) 25–42

35

Table 5 Conditional serial correlation (CSC) by manager location. This table presents the proportion of emerging market hedge funds that have a − positive and significant CSC. The regression to find CSC is rt = a + b+ 1 rt − 1 + b1 It − 1rt − 1 + εt, where rt is the demeaned fund return between time t − 1 and t and It − 1 is an indicator variable that equals one if the fund's systematic return from the 4-factor model at month t − 1 is less than the mean systematic return. If b− 1 is significantly positive, the fund has a CSC. The first column shows the location of the fund management company and “unknown” means that the address of the fund management company was not reported to TASS. To be included in the analysis, a fund should have at least twelve monthly return observations after the adjustment for the backfill bias. Location of manager

US Caribbean UK HKSP EM Unknown All

Number of funds with 12 or more monthly returns after adjusting for backfill bias 144 149 77 66 59 188 727

% of Funds with a CSC t-stat of b− 1 ≥ 1.645 (two-sided 10%)

t-stat of b1 ≥ 1.960 (two-sided 5%)

t-stat of b1 ≥ 2.576 (two-sided 1%)

8.33 6.04 5.19 4.55 5.08 9.57 7.29

4.17 2.68 2.60 0.00 5.08 6.91 4.13

1.39 1.34 1.30 0.00 1.69 2.66 1.65

eight portfolios are formed based on the manager location (US vs. Others), geographical focus (Global vs. Focus), and the lockup provision (Lockup vs. No Lockup) as shown in Table 6. Note that US funds are geographically less focused than other funds (Global funds' proportion: 51% vs. 31%) as shown in Table 1, and previous research shows that Global funds underperform geographically focused funds (KotkatvuoriÖrnberg et al. (2011)). Consistent with previous research, I find that the lowest alpha (−0.21% per month during '94–12, t-statistic: −2.04) is observed in the portfolio of funds that have neither a lockup provision nor a geographical focus in case of the other funds. However, among the US funds, I do not find evidence that geographical focus or a lockup provision affects performance. The alphas of all the four US portfolios (Global Lockup, Global No Lockup, Focus Lockup, and Focus No Lockup) are not significantly different from zero. That is, neither geographical focus nor lockup explains the performance of emerging market hedge funds located in the United States.

Table 6 The impact of focus and share restriction on performance: US vs. other funds. This table compares the alphas of portfolios formed by manager location, geographical focus and lockup provision. A four factor model with two EMF lags is used. “Global” means the portfolio of funds with the GF_Global dummy variable in TASS of 1. “Focus” is for GF_Global dummy of 0. ***, **, and * denote statistical significances at the 1%, 5%, and 10% levels, respectively. Manager Focus location US

Lockup

Global Lockup No lockup Focus

Lockup No lockup

Others

Global Lockup No lockup Focus

Lockup No lockup

Number of Beta funds (%) EMF

EL1

EL2

SMB

TERM

CREDIT

20 (29%) 53 (71%) 36 (51%) 35 (49%) 27 (15%) 151 (85%) 62 (16%) 327 (84%)

0.06 (1.44) 0.07 (3.55)*** 0.18 (3.87)*** 0.09 (2.81)*** 0.10 (2.68)*** 0.08 (5.24)*** 0.05 (1.99)** 0.10 (4.53)***

0.06 (1.46) 0.04 (2.14)** 0.09 (2.01)** 0.04 (1.13) 0.08 (2.17)** 0.04 (2.88)*** 0.07 (3.20)*** 0.02 (1.18)

0.15 (1.85)* 0.09 (2.30)** −0.02 (−0.20) 0.12 (1.76)* 0.04 (0.62) 0.04 (1.21) −0.02 (−0.36) 0.04 (1.01)

0.01 (0.06) 0.10 (1.46) −0.04 (−0.23) 0.08 (0.70) −0.08 (−0.64) 0.07 (1.28) 0.10 (1.36) 0.04 (0.61)

−0.09 (−0.61) 0.05 (0.67) −0.31 (−1.83)* −0.04 (−0.33) −0.10 (−0.81) 0.10 (1.77)* 0.25 (3.24)*** −0.03 (−0.34)

0.57 (11.70)*** 0.41 (16.88)*** 0.75 (13.11)*** 0.65 (16.81)*** 0.28 (6.13)*** 0.39 (22.35)*** 0.43 (14.47)*** 0.53 (21.67)***

Alpha

Adj-R2

0.29 (1.03) 0.07 (0.51) −0.09 (−0.27) 0.19 (0.85) 0.11 (0.45) −0.21 (−2.04)** 0.24 (1.42) 0.25 (1.75)*

0.59 0.72 0.61 0.69 0.34 0.82 0.79 0.79

36

H. Park / Emerging Markets Review 22 (2015) 25–42

4.5. The relation between fund performance and capital flow After finding the poor average performance of emerging market hedge funds in the United States, I analyze how investors of the US funds responded to the poor performance. Previous research finds that investors of equity mutual funds chase high performers but they are slow to flee poor performers (Sirri and Tufano (1998)). Fung et al. (2008) analyze flow to funds of hedge funds (FOFs) and find that alpha-producing FOFs experience far greater and steadier capital inflows than past poor performers. I find that capital flow to emerging market hedge funds located in the US was much higher but it was less sensitive to past performance than capital flow to the funds in other countries. For example, Panel A of Table 7 shows that the average quarterly flow to US funds is over 4% while the flow to Caribbean funds is only 0.51% during 1994–2012, and the difference is significant at the 1% level. Note that the coefficient of bottom performance tercile is highly significant for “Others”, but not for “US” in Panel B of Table 7. That is, investors of emerging market hedge funds located in other countries are faster to flee poor performers than investors of US funds. For example, a drop in rank from the 20th percentile to

Table 7 The effect of performance on capital flows: US vs. Other Funds. This table presents quarterly capital flows to emerging market hedge funds. Quarterly flows are measured by using the growth rate of net new money, which is defined as Flowi,t = (TNAi,t − TNAi,t − 1 ∗ (1 + Ri,t)) / TNAi,t − 1. TNAi,t is fund i's total net assets at the end of quarter t, and Ri,t is the fund's return during the quarter. Panel A compares US and other funds in terms of mean and standard deviation of quarterly flows. Panel B presents the effect of relative performance on capital flows to US and other funds. In Panel B, the coefficient estimates are presented from the piece-wise linear regression of investor flows on relative performance variables, Low, Mid, and High. These variables are defined using a fractional rank (FRANK) that represents a fund's percentile performance during the previous year. FRANK ranges from 0 to 1. The bottom performance tercile (Lowt) is defined as Min (1/3, FRANKt − 1), the middle performance tercile (Midt,) is defined as Min (1/3, FRANKt − 1 − Lowt), and the top performance tercile (Hight) is defined as Min (1/3, FRANKt − 1 − Lowt − Midt). As control variables, risk, size, lockup dummy (dlock), incentive fee, and highwater mark (HWM) dummy variables are included. The regressions are run quarterly, and standard errors and t-statistics are calculated from the quarterly results as in Fama and MacBeth (1973). As I require at least 24 funds in the cross-sectional regression for each quarter, the sample period for Panel B is Oct 1999–Dec 2012 (53 quarters).***, **, and * denote significances at the 1%, 5% and 10% levels, respectively. Panel A: mean and standard deviation of quarterly flows Cross-sectional average value of

US

Average quarterly flow (%) (t-statistics for the difference between the US and other regions) Standard deviation of quarterly flow (%) (t-statistics for the difference between the US and other regions)

4.02 18.76

Caribbean

UK

HKSP

EM

0.51 (3.08)*** 16.21 (1.49)

1.06 (2.32)** 16.13 (1.58)

3.21 (0.48) 16.21 (1.06)

2.54 (1.40) 13.34 (3.52)***

US — other difference

t-statistic

Panel B: the effect of performance on capital flows US coefficient Intercept Relative Performance Bottom Performance Tercile (low) Middle Performance Tercile (mid) Top Performance Tercile (high) Std. dev. of monthly returns Log (TNAt − 1) High Water Mark (HWM) dlock Incentive fee Adj-R2 (%)

t-statistic

Other coefficient

t-statistic

0.0958

1.08

0.0383

0.88

0.0575

0.58

−0.0152

−0.17

0.1660

2.71***

−0.1812

0.0399

0.43

−1.70*

0.1715

2.06**

0.1316

3.36***

0.0260

0.33

0.0760

1.62

−0.0500

−0.55

−0.0086

−2.73***

−0.0064

−3.41***

−0.0022

−0.60

−0.0063 −0.0337 0.0033 0.0020 0.87

−1.65 −2.33** 0.22 0.79

−0.0036 0.0232 0.0097 −0.0015 6.33

−1.65 2.60** 1.10 −2.86***

−0.0027 −0.0569 −0.0064 0.0034

−0.61 −3.35*** −0.42 1.35

H. Park / Emerging Markets Review 22 (2015) 25–42

37

10th percentile decreases the quarterly net flow to other funds by 1.66% and the decrease is significant at the 1% level, but the same drop in rank does not reduce the flow to US funds. Another important finding is that investors of emerging market hedge funds do not chase high performing funds unlike investors of equity mutual funds; the coefficient of top performance tercile is not significantly different from zero for both US funds and other funds. Overall these results show that past performance does not explain the high average net flow to emerging market hedge funds in the United States. Only 0.87% of the variation in quarterly net flow to US funds is explained by performance and other control variables while over 6% of variation in flow to the other funds is explained by the same variables. 4.6. Smart money effect After finding that past performance does not explain the flow to emerging market hedge funds in the US, I test whether the flow can predict future performance using the GT measure that was developed to examine mutual fund investors' fund selection ability. As explained in Section 3.4, the GT measure is serially uncorrelated with mean zero if investors have no fund selection ability. That is, if investors have a fund selection ability, funds that receive more money will subsequently perform significantly better than those that lose money and thus the GT measure will be significantly positive. As shown in Table 8, the GT measure for US funds is not significantly different from zero for all sample periods tested while the GT measures for Caribbean funds and HKSP funds are significantly positive after adjusting for autocorrelation. That is, US fund investors do not display fund selection ability while the ability is observed among the investors of Caribbean funds, Hong Kong funds, and Singapore funds. 4.7. The impact of short history on the evaluation of performance When the smart money effect was tested in Table 8, a long history (193 months) was used for US funds but the sample period for HSKP funds was short (62 months). The reason is, HKSP funds are relatively younger as shown in Fig. 2 and thus there was not enough funds in the cross-section of HSKP funds at earlier months. As such a short history of some funds in the sample may lead to a bias in estimated risk and return as pointed out by Stambaugh (1997), I apply a similar method to emerging market funds to examine the impact of short history on the estimated risk and risk-adjusted performance.

Table 8 GT performance estimates by manager location. This table presents the GT performance measure as in Grinblatt and Titman (1993) and Zheng (1999). GT = ∑iRi,t + 1 (wi,t − wi,t − 1), where Ri,t + 1 is the return of fund i between time t and t + 1 and wi,t is the weight for fund i at time t. Under the null hypothesis that investors have no selection ability, the GT measures are serially uncorrelated with mean zero. As the sample period is set to include at least 10 funds in each region, it is Nov 2007–Dec 2012 (62 months) for HKSP funds. For EM funds, the 10-fund requirement is satisfied during Apr 2004–Dec 2012 (105 months) and the 62-month period (Nov 2007–Dec 2012) is also analyzed to make the results comparable to HKSP. The same approach is applied to US, Caribbean, and UK funds. That is, the US has the longest sample period when there is at least 10 funds in the cross-section, Dec 1996–Dec 2012 (193 months). ***, **, and * denote statistical significances at the 1%, 5% and 10% levels, respectively. To adjust for autocorrelation, I present the Newey and West (1987) tstatistic as well as the traditional t-statistic. Location of manager

Number of months

GT measure

t-statistic

Newey-West t-statistic

US

193 132 105 62 132 105 62 132 105 62 62 105 62

0.0029 −0.0012 −0.0055 −0.0235 0.0439 0.0405 0.0814 0.0525 0.0606 0.0830 0.1999 0.0039 0.0275

0.11 −0.06 −0.25 −0.99 2.62*** 2.45** 4.27*** 2.20** 2.33** 3.13*** 4.62*** 0.18 1.28

0.09 −0.04 −0.15 −0.43 1.84* 1.53 2.13** 1.46 1.37 1.41 1.86* 0.14 0.66

Caribbean

UK

HKSP EM

38

H. Park / Emerging Markets Review 22 (2015) 25–42

As shown in Table 9, the short history of some emerging market hedge funds does not cause a significant bias in evaluation of risk-adjusted performance. The Sharpe ratio rankings adjusted for a short history using maximum-likelihood estimation (Rankadj) as well as the Bayesian approach (RankB) are highly correlated to the ranking without any adjustment (Rank); the rank correlation coefficients are above 0.9. That is, during the sample period of 1994–2012 there is no evidence of significant bias in the estimated performance of emerging market hedge funds that have a short history. 5. Conclusion This paper analyzes the risk-adjusted performance and flow-performance relation of US portfolio investment in emerging markets using hedge fund data while previous research in the emerging markets finance literature use more aggregate data and find mixed results. By comparing the emerging market hedge funds in the US with otherwise similar funds in other countries in portfolio level tests, I find that alpha is not significantly positive across all manager locations, and the emerging market factor explains most of the variation in

(a) US Funds

(b) HKSP Funds

Fig. 2. Histograms for the age of emerging market hedge funds: US funds vs. HKSP funds.

Table 9 The impact of short history on the measured performance of emerging market hedge funds. This table shows standard deviation (STD) and Sharp Ratio (SR) adjusted for short history as in Stambaugh (1997) for the 727 emerging market funds that have at least 12 monthly return observations after adjusting for the back-fill bias. The subscript “Adj” stands for the adjustment using maximum-likelihood estimation (Eq. (A.3) in Appendix A) and “B” is from the Bayesian approach (Eq. (A.4) in Appendix A). “Number of Months” in the sixth column is the number of remaining return observations after adjusting for the backfill bias. “N/A” means the location of the fund management company was not reported to the database. The Spearman rank correlation coefficient is 0.9050 for Rank and Rankadj and 0.9046 for Rank and RankB. Manager location

Registration

Status

Age

Number of months

STDadj

STDB

Mauritius Mauritius Singapore Cayman Islands Cayman Islands N/A Bermuda United States United States N/A

Cayman Islands Mauritius Cayman Islands Curacao Cayman Islands Cayman Islands Bermuda Cayman Islands United States United States

Live Live Live Live Defunct Defunct Live Live Live Defunct

47 41 76 86 14 56 12 40 40 56

47 41 75 78 12 48 12 40 40 48

0.16 0.13 0.09 0.29 0.96 1.12 1.11 0.99 0.99 1.24

0.16 0.14 0.09 0.29 1.32 1.12 1.19 1.09 1.09 1.24

0.17 0.15 0.10 0.30 1.66 1.19 1.46 1.12 1.12 1.32

7.82 4.73 3.22 1.49 1.02 0.99 0.97 0.96 0.92 0.92

7.76 4.71 3.22 1.49 0.51 0.97 0.82 0.81 0.78 0.90

7.30 4.37 3.10 1.44 0.40 0.91 0.66 0.79 0.75 0.84

1 2 3 4 5 6 7 8 9 10

1 2 3 4 24 5 9 10 11 8

1 2 3 4 29 5 11 9 10 6

Defunct Defunct Live Live Defunct Live Defunct Defunct Defunct Defunct

49 29 64 64 63 64 23 64 19 13

14 16 64 64 14 64 17 18 12 13

20.56 2.67 6.81 6.82 11.49 6.82 4.65 2.25 6.97 3.15

20.48 2.65 6.74 6.75 10.05 6.75 3.64 2.25 5.07 3.11

25.20 3.11 7.04 7.05 11.91 7.05 4.12 2.70 5.67 3.85

−0.52 −0.52 −0.55 −0.56 −0.57 −0.57 −0.59 −0.59 −0.62 −0.76

−0.53 −0.51 −0.56 −0.57 −0.20 −0.57 −0.57 −0.59 −0.15 −0.82

−0.43 −0.44 −0.53 −0.54 −0.16 −0.55 −0.50 −0.49 −0.13 −0.66

718 719 720 721 722 723 724 725 726 727

721 720 722 724 645 725 723 726 618 727

717 718 724 725 636 726 723 722 614 727

Funds 11–717 are not shown to save space 718 United States Bahamas 719 N/A Cayman Islands 720 Cayman Islands Cayman Islands 721 Cayman Islands Cayman Islands 722 Cayman Islands Mauritius 723 Cayman Islands Cayman Islands 724 United States Cayman Islands 725 N/A Cayman Islands 726 Thailand British Virgin Islands 727 N/A Mauritius

STD

SR

SRadj

SRB

Rank

Rankadj

RankB

H. Park / Emerging Markets Review 22 (2015) 25–42

Fund 1 2 3 4 5 6 7 8 9 10

39

40

H. Park / Emerging Markets Review 22 (2015) 25–42

the portfolio returns especially after the recent financial crisis. In the individual fund level test, I find that the proportion of funds with a significantly positive alpha is higher when the fund is located in an emerging market than in the US, and the average alpha of the US funds was not positive. Despite the poor performance, however, the US funds had much higher net flow than the other funds. The investors of the US funds were not as fast as the investors of the other funds in withdrawing capital from poor performers. Another important finding is that flow to top performing emerging market funds did not increase with performance across all manager locations. That is, the convex flow-performance relation reported in the equity mutual fund literature does not apply to emerging market hedge funds. In future research it would be interesting to analyze the flow-performance relation of emerging market mutual funds. If the flow to emerging market mutual funds is not sensitive to past performance either, we may find that macroeconomic factors play a more important role than performance in the flow to emerging market funds. For example, when the US Federal Reserve publicly described conditions for scaling back its bond buying program in May 2013, some emerging markets experienced sharp reversal of capital inflows as discussed in Nechio (2014). It would be an interesting future research project to analyze the impact of the US Fed's tapering on the flows to emerging market hedge funds and mutual funds. Acknowledgements Financial support from Minnesota State University Mankato through Faculty Research Grant (FRG) (211756) is gratefully acknowledged. I thank Bing Liang and Session 158 (Hedge Funds in Asia) participants at 2014 FMA (Financial Management Association) Annual Meeting for helpful comments and suggestions. Appendix A. Adjusting the risk of short history assets When we analyze the return history of investments that are differ in length, the longer history asset may provide greater information about risk and return for the shorter-history assets. Stambaugh (1997) presents two methods to adjust the standard deviation (STD) and Sharp ratio (SR) of short history assets using the returns of longer history assets: i) an adjustment based on maximum-likelihood estimation and ii) an adjustment using the Bayesian approach. Following his approach, I define STDAdj and STDB as in Eqs. (A.1)–(A.5). The subscript “Adj” is for the adjustment based on maximum-likelihood estimation and “B” is for the Bayesian approach. I use the MSCI emerging market monthly total return index (R1,t) as the long history asset where t is observed for periods 1,…, T and let R2,t denote the return of an emerging market hedge fund, a short history asset, where t is observed for periods s,…, T. Let S = T − s + 1 denote the number of observation in the fund return. For any period t ≥ s, we have both R1,t and R2,t and thus can regress R2,t on R1,t to obtain the coefficients, Ĉ. h i
 ^ ¼ α; ^ ’ ¼ X’ X −1 X’ Y ^ β C

ðA:1Þ

where X = [1S R1,S], 1S is an S-vector of ones, R1,S is an S-vector of the long history asset return during the period when the fund return is observed, t = s,…,T. Y = (R2,s, R2,s + 1 …, R2,T)′ is the S vector of the fund return. The sample residual-covariance matrix from the regression is 

c ¼ 1 Y−XC ^ ^ ’ Y−XC ∑ S

ðA:2Þ

Let V1 = variance of R1,t where t = 1,…,T, E1 = average of R1,t where t = 1,…,T, V1,S = variance of R1,t where t = s,…,T, E1,S = average of R1,t where t = s,…,T, and V2,S = variance of R2,t where t = s,…,T. STDAd j ¼

qffiffiffiffiffiffiffi
 c ; where V2 c ¼ V ‐β ^ V ‐V β ^ V2 2;S 1;S 1

ðA:3Þ

H. Park / Emerging Markets Review 22 (2015) 25–42

STDB ¼  k¼

qffiffiffiffiffiffiffiffiffi   c þ T þ 1 βV d ; where V2B d ¼k∑ ^ ^ β V2B 1 T−4

   
2 S 1 T þ1
ð1 þ ½1 þ V1 =V1;S þ E1 –E1;S =V1;S S−3 S T−4

41

ðA:4Þ

ðA:5Þ

I use these five equations to adjust estimated risk of short-history funds. Then I examine whether the differences in length of return history of emerging market hedge funds affect the ranks in estimated riskadjusted performance. Appendix B. Description of variables • Flowi,t: the percentage growth of a fund during the period between time t − 1 and time t in excess of the growth that would have occurred if no new money had flowed in • FRANK: a fund's percentile performance that ranges from 0 to 1 • GT: Grinblatt and Titman's measure of investors' fund selection ability as defined in Eq. (3) • Lowt: the bottom performance tercile defined as Min (1/3, FRANKt − 1) • Midt: the middle performance tercile defined as Min (1/3, FRANKt − 1 – Lowt) • Hight: the highest performance tercile is defined as Min (1/3, FRANKt − 1 − Lowt − Midt) • It − 1: indicator variable that equals one if the systematic component of the fund return at time t − 1 from the 4-factor model is less than the mean systematic return • Ri,t: fund return during the period between time t − 1 and time t • rt: demeaned fund return during the period between time t − 1 and time t • TNAi,t: fund i's total net assets at time t References Ackermann, C., McEnally, R., Ravenscraft, D., 1999. The performance of hedge funds: risk, return, and incentives. J. Financ. 54, 833–874. Aggarwal, R., Jorion, P., 2010a. The performance of emerging hedge funds and managers. J. Financ. Econ. 96, 238–256. Aggarwal, R., Jorion, P., 2010b. Hidden survivorship bias in hedge fund returns. Financ. Anal. J. 66, 69–74. Aragon, G.O., 2007. Share restrictions and asset pricing: evidence from the hedge fund industry. J. Financ. Econ. 83, 33–58. Aragon, G., Liang, B., Park, H., 2014. Onshore and offshore hedge funds: are they twins? Manag. Sci. 60 (1), 74–91. Asness, G., Krail, R., Liew, J., 2001. Do hedge funds hedge? J. Portf. Manag. 28, 6–19. Bekaert, G., Harvey, C.R., 2000. Foreign speculators and emerging equity markets. J. Financ. 55 (2), 565–613. Bekaert, G., Harvey, C.R., 2003. Emerging markets finance. J. Empir. Financ. 10, 3–55. Bollen, N., Pool, V., 2008. Conditional return smoothing in the hedge fund industry. J. Financ. Quant. Anal. 43 (2), 267–298. Cumming, D., Dai, N., Johan, S., 2014. Are hedge funds registered in Delaware different? Working Paper (http://www.fma.org/Nashville/ Papers/CDS_HFdelaware_FMA2014.pdf). Eling, M., Faust, R., 2010. The performance of hedge funds and mutual funds in emerging markets. J. Bank. Financ. 34, 1993–2009. Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. J. Financ. Econ. 33, 3–56. Fama, E., MacBeth, J., 1973. Risk, return and equilibrium: empirical tests. J. Polit. Econ. 38, 607–636. Froot, K., O'Connell, P.G.J., Seasholes, M.S., 2001. The portfolio flows of international investors. J. Financ. Econ. 59, 151–193. Fung, W., Hsieh, D., 2000. Performance characteristics of hedge funds and CTA funds: natural versus spurious biases. J. Financ. Quant. Anal. 35, 291–307. Fung, W., Hsieh, D., 2001. The risk in hedge fund strategies: theory and evidence from trend followers. Rev. Financ. Stud. 14, 313–341. Fung, W., Hsieh, D., 2002. Benchmarks of hedge fund performance: information content and measurement biases. Financ. Anal. J. 58, 22–34. Fung, W., Hsieh, D., 2004. Hedge fund benchmarks: a risk based approach. Financ. Anal. J. 60, 65–80. Fung, W., Hsieh, D., Naik, N., Ramadorai, T., 2008. Hedge funds: performance, risk and capital formation. J. Financ. 63, 1777–1803. Getmansky, M., Lo, A., Makarov, I., 2004. An econometric model of serial correlation and illiquidity in hedge fund returns. J. Financ. Econ. 74, 529–610. Grinblatt, M., Titman, S., 1993. Performance measurement without benchmarks: an examination of mutual fund returns. J. Bus. 66, 47–68. Jagannathan, R., Malakhov, A., Novikov, D., 2010. Do hot hands exist among hedge fund managers? An empirical evaluation. J. Financ. 65 (1), 217–255. Jensen, M., 1968. The performance of mutual funds in the period 1945–1964. J. Financ. 23, 389–416. Kaminsky, G., Lyons, R.K., Schmukler, S.L., 2001. Mutual fund investment in emerging markets: an overview. World Bank Econ. Rev. 15 (2), 315–340. Khandani, A., Lo, A., 2007. What happened to the quants in August 2007? J. Invest. Manag. 5, 5–54. Kotkatvuori-Örnberg, J., Nikkinen, J., Peltomäki, J., 2011. Geographical focus in emerging markets and hedge fund performance. Emerg. Mark. Rev. 12, 309–320.

42

H. Park / Emerging Markets Review 22 (2015) 25–42

Liang, B., 1999. On the performance of hedge funds. Financ. Anal. J. 55, 72–85. Liang, B., Park, H., 2010. Predicting hedge fund failure: a comparison of risk measures. J. Financ. Quant. Anal. 45 (1), 199–222. Lo, A., 2001. Risk management for hedge funds: introduction and overview. Financ. Anal. J. 57, 16–33. Nechio, F., 2014. Fed Tapering News and Emerging Markets. Federal Reserve Bank San Francisco (FRBSF) Economic Letter (http://www. frbsf.org/economic-research/publications/economic-letter/2014/march/federal-reserve-tapering-emerging-markets/). Newey, W., West, K., 1987. A simple positive semi-definite heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55, 703–708. Sirri, E., Tufano, P., 1998. Costly search and mutual fund flows. J. Financ. 53, 1589–1622. Stambaugh, R., 1997. Analyzing investments whose histories differ in length. J. Financ. Econ. 45, 285–331. Teo, M., 2009. The geography of hedge funds. Rev. Financ. Stud. 22 (9), 3531–3561. Zheng, L., 1999. Is money smart? A study of mutual fund investors' fund selection ability. J. Financ. 54 (3), 901–933.