The geography of hedge funds melvyn teo

The Geography of Hedge Funds*

Melvyn Teo**

Abstract

This paper analyzes the relationship between the risk-adjusted performance of hedge funds and their proximity to investments using data on Asian hedge funds. We find, relative to an augmented Fung and Hsieh (2004) factor model, that hedge funds with a physical presence (head or research office) in their investment region outperform other hedge funds by 5.28 percent per year. The local information advantage is pervasive across all major geographical regions and investment styles, but is strongest for Emerging Market and Event Driven funds. These results are robust to adjustments for fund fees, serial correlation, backfill bias, and incubation bias.

JEL classifications: G11; G12; G23 Keywords: hedge fund performance; alpha; factor models; geography

______________________ * We are indebted to Vikas Agarwal, George Aragon, William Fung, Mila Getmansky, Robert Kosowski, and Bing Liang for many helpful discussions and to AsiaHedge and EurekaHedge for generously providing us with the data. This research is inspired by suggestions from William Fung. We alone are responsible for all errors. ** Address: Lee Kong Chian School of Business, Singapore Management University, 50 Stamford Road, #04-01, S(178899), Singapore. E-mail: [email protected], Tel: +65-6828-0735, and Fax: +65-6828-0777.

1

Does distance affect hedge fund performance? If not, why do some hedge funds locate

close to their investments while others prefer to monitor their investments from afar and from the

comfort of home? Anecdotal evidence1 suggests that proximity to investments may be helpful for

hedge funds. Being near their investments or target firms allows hedge funds to maintain close

contact with senior management and stay abreast of the latest developments. Fund managers can

learn valuable information from other local firms along the supply chain by being on the

ground.2 Nearby funds, who are also substantial stakeholders, can more easily engage in

constructive shareholder activism and exert pressure on management to make shareholder

friendly improvements.3

In this paper, we investigate the link between funds’ proximity to their primary

investment markets and hedge fund risk-adjusted returns.4 We show that funds that have a

physical presence (head office or research office) in their main investment region outperform

other funds on a risk-adjusted basis by on average 5.28 percent per year (t-statistic = 4.44). The

superior performance of funds with investment region presence cannot be attributed to lower

fees, greater backfill and incubation bias (Fung and Hsieh, 2000; Liang, 2000; Brown,

Goetzmann, and Ibbotson, 1999), or illiquidity (Getmansky, Lo, and Makarov, 2004). We obtain

similar results with fund alpha derived from pre-fee returns, from backfill and incubation bias

adjusted returns, and from unsmoothed returns adjusted for serial correlation. The hypothetical

strategy of buying nearby funds and shorting distant funds yields risk-adjusted returns of about

4.54 percent per year (t-statistic = 3.09). This spread alpha is robust across all major fund

investment styles and investment regions, and is substantially higher for styles/regions where the

local informational advantage is likely to be stronger or more relevant. For instance, the nearby

Emerging Markets fund portfolio outperforms the distant Emerging Markets fund portfolio by

1 In a Wall Street Journal report, Christophe Lee, chairman of the Alternative Investment Management Association chapter in Hong Kong, remarked that a distinct advantage for Hong Kong vis à vis Singapore as a hedge fund hub is its proximity to China, which is home to many investment opportunities (Raagas De Ramos, 2006). 2 For instance, Cohen and Frazzini (2006) show that a long/short strategy that takes advantage of information in consumer/supplier links can earn substantial risk-adjusted returns. 3 See Wighton (2006) for recent examples of investor activism by hedge funds. 4 To clarify, we do not endeavor to explain why hedge funds locate where they do, but rather we show how hedge fund geography affects hedge fund performance. To explain hedge fund location, one would have to take into account differences in air quality, income and corporate taxes, proximity to hedge fund clients, and the supply of prime brokerage services as well. Also, we note that differences in income and corporate tax regimes are not central to the arguments in the paper since income and corporate taxes are levied on manager income and fund profits, respectively, and not on the underlying portfolios themselves. Hence, the effects of income and corporate taxes should not show up in fund risk-adjusted performance.

2

32.06 percent per year. These results suggest that nearby hedge funds enjoy an informational

advantage. In addition, the results are particularly relevant to hedge fund managers, Fund of

Funds, and other fund investors.

Our study focuses on hedge funds that invest primarily in Asian financial markets. The

advantage of Asian hedge fund data is that we can derive a simple yet meaningful distance

measure by using fund location and fund investment region information. For instance, on one

hand, we classify funds investing in the Asia including Japan region but located in the United

States or United Kingdom as distant funds or funds without investment region presence. On the

other hand, we classify funds investing in the Asia including Japan region and located in

Singapore, Hong Kong, or Tokyo as nearby funds or funds with investment region presence. To

conduct the same analysis with the typical U.S.-centric fund database5 we will need information

on the location of fund holdings. Unfortunately, hedge fund holdings (including short positions)

are highly confidential and rarely, if at all, available to researchers.

To adjust for risk in Asian hedge funds, we employ an augmented Fung and Hsieh (2004)

factor model. Fung and Hsieh (1999, 2000, 2001), Mitchell and Pulvino (2001), and Agarwal

and Naik (2004) show that hedge fund returns relate to conventional asset class returns and

option-based strategy returns. Building on this pioneering work, Fung and Hsieh (2004) show

that their parsimonious asset-based style factor model can explain up to 80 percent of the

variation in fund portfolio returns. We augment the Fung and Hsieh (2004) factor model with

additional equity factors, which we derive from a principal components analysis approach

[following Fung and Hsieh (1997)]. The equity factors we use are the excess return on the Asia

ex Japan equity index and the excess return on the Japan equity index. The augmented factor

model explains up to 70 percent of the variation in Asian portfolio fund returns and allows us to

abstract from risk in our analysis of geography and hedge fund performance. As a robustness

check and in response to concerns that fund returns may resemble the payoffs from writing

options on the equity market (Agarwal and Naik, 2004), we also include in the factor model

additional factors derived from call and put options on the Japan Nikkei 225 index, but find

qualitatively similar results.6

5 Examples of such hedge fund databases include TASS, HFR, MSCI, and CISDM. 6 The results are also robust to the inclusion of an Asian size factor in the performance evaluation model.

3

This paper presents novel results linking distance to hedge fund alpha. In doing so, we

build on several themes. Fung and Hsieh (2004) and Agarwal and Naik (2004) show that fund

risk can be explained by loadings on option-like factors. By leveraging on these results,

Kosowski, Naik, and Teo (2006) show that top fund risk-adjusted performance cannot be

explained by luck, or sample variation. Consistent with this, they find that fund alpha persists

over annual horizons. Their results indicate that the top fund managers possess asset selection

skills. Our results suggest that part of this skill stems from the ability to take advantage of local

information. Funds that invest in nearby markets tend to exhibit greater skill (i.e., alpha) than

funds that invest in distant markets. French and Poterba (1991), Cooper and Kaplanis (1994),

and Tesar and Werner (1995) show that investors severely overweight their portfolios toward

domestic assets. Our results shed light on this international home bias puzzle. Related to this, we

also contribute to a growing literature on distance and investment performance.7 Coval and

Moskowitz (2001) demonstrate that U.S. mutual fund managers are better at picking stocks of

nearby firms than stocks of distant firms. Ivković and Weisbenner (2005) witness a similar

pattern with U.S. retail investors. At the same time, Malloy (2005) finds that U.S. equity analysts

issue more accurate earnings forecasts for nearby firms than for distant firms. On the

international front, Hau (2001) reports, using data on professional traders, that local traders

outperform foreign traders in the German market. Choe, Kho, and Stulz (2005) and Dvorak

(2005) provide evidence that foreigners underperform other investors in Korea and Indonesia,

respectively. Bae, Stulz, and Tan (2006) document, in their sample of 32 countries, that local

equity analysts issue more accurate earnings forecasts than do foreign analysts. In particular,

they find that the accuracy differential is stronger in Emerging Markets, where less information

is disclosed by firms.8

However, none of the aforementioned studies investigate the effects of distance on hedge

fund performance. By focusing on hedge funds instead of mutual funds, we can conveniently

abstract from diversification issues that affect mutual funds, since hedge funds are typically pure

7 See also Krugman (1991), Lucas (1993), Audretsch and Feldman (1996), Loughran and Schultz (2005), and Becker (2006) for related studies on the economic importance of geography. 8 The international evidence is mixed however. Seasholes (2000) and Grinblatt and Keloharju (2000) argue that foreign investors have access to better resources and expertise than local investors, which may allow them to overcome any local informational advantage. Consistent with this, they show that foreigners outperform locals in Taiwan and Finland, respectively. To the extent that this is true in our sample (i.e., foreign funds are more sophisticated than local funds), it biases our results downwards since we compare the performance of nearby local and foreign hedge funds with the performance of distant foreign hedge funds.

4

alpha bets. Further, unlike mutual funds, hedge funds can short sell and take advantage of

negative information. Hence, an analysis of hedge fund performance allows for a more powerful

test of the local information advantage hypothesis. Consistent with this, at the fund level, mutual

funds that invest more in local stocks do not significantly outperform mutual funds that invest

more in distant stocks [see pg. 823 of Coval and Moskowitz (2001)]. In contrast, our results

indicate that one can earn significant economic rents by investing in hedge funds that operate

near their investment markets. Moreover, in line with the view that hedge funds are better able to

take advantage of local information via short selling and the use of derivatives, the risk-adjusted

spread between nearby and distant hedge fund portfolios (4.54 percent per year) is more than

three times the risk-adjusted spread between the local and non-local stock portfolios of mutual

funds (1.18 percent per year) and of households (1.08 percent per year).9

The rest of this paper is as follows. Section 1 describes the data while Section 2 lays out

the framework for analyzing fund risk-adjusted performance. Section 3 presents the results

linking distance to fund risk-adjusted performance. Section 4 concludes.

1. Data

We hand collect data on Asian hedge funds from two hedge fund databases:

EurekaHedge and AsiaHedge. EurekaHedge is maintained by EurekaHedge Advisors Pte Ltd

while AsiaHedge is maintained by Hedge Fund Intelligence. Both databases include funds which

invest a significant portion of their assets in Asian markets. The sample consists of monthly fund

return data from January 1999 to December 2003. While AsiaHedge started collecting fund

return data from 1999, EurekaHedge only started collecting fund return data from 2000 (but

includes fund returns since inception). Hence, we focus on the period from 2000 onwards to

ameliorate survivorship bias.

Our sample also includes data on a host of fund characteristics including management

fee, performance fee, redemption period, investment style, investment geographical region, fund

location, fund size, fund family size, and fund minimum investment. These fund characteristics

are recorded in year 2003 and, for our purposes, we take these characteristics as constant over the

9 See the alpha spreads in Table 1 of Coval and Moskowitz (2001) and Table IX of Ivković and Weisbenner (2005).

5

sample period.10 Since minimum investments are sometimes quoted in currencies other than US

dollar, we convert all minimum investments to US dollars using exchange rates on Dec 31st 2003

so as to facilitate meaningful comparison. AsiaHedge records fund and family size in distinct

ranges. Hence, we convert fund and family size values to size categories, which condense size

information into a number from one to ten. Details on the size categories are available in

Appendix A.

Naturally, the hedge fund sample includes both dead and live funds. Of the 325 funds

with at least one month of return data in the sample period, 23 are dead funds. As such, our dead

funds to total funds ratio of 23 to 325 is higher than the dead funds to total funds ratio of 27 to

746 for the Hedge Fund Research (henceforth HFR) database used by Agarwal and Naik (2000).

To the extent that the true survival rates are comparable across different databases, this suggests

that survivorship issues11 (Brown, Goetzmann, and Ibbotson, 1999) are mitigated in our dataset.

To verify the accuracy of the return data in our dataset, we collect data on hedge fund

returns from TASS, a popular database used in hedge fund research. Next, we match by hand

funds from the combined AsiaHedge and EurekaHedge database with those from TASS using

fund name. We are able to obtain matches for 74 funds that have at least 24 months of common

return data in the January 2000 to December 2002 period (we are only able to obtain TASS

return data prior to 2003). Then, we calculate the mean correlation in returns for these funds. We

find that the mean correlation in fund returns from the two databases is 0.97, adding comfort to

our use of the return data.

As mentioned, the data contains information on fund investment region and fund

location. In addition, some funds also provide information on the location of their research

offices. This allows us to determine whether funds have a physical presence (head or research

office) within their investment region. In Table 1, we breakdown the funds in our sample by

investment style and investment region, and report the following summary statistics: the total

10 This assumption seems reasonable for all the fund characteristics except fund size and fund family size. Our use of broad size categories in place of size values ameliorates the problem. But nonetheless, we will avoid making any inferences on fund and family size. We shall revisit this issue later in the paper. 11 One caveat is that any residual survivorship bias is still likely to be greater in our dataset than in the latest versions of the HFR, TASS, or CISDM datasets where more dead funds have been added [see Kosowski, Naik, and Teo (2006)]. The underlying assumption we make is that survivorship bias does not systematically differ between funds with investment region presence and funds without investment region presence. Such an assumption seems reasonable given that the ratio of dead funds to the total number of funds is 7.1% both for funds with investment region presence and for funds without investment region presence.

6

number of funds, the number of dead funds, the number of return months, as well as the number

of funds with and without investment region presence.

[Please insert Table 1 here]

From the summary statistics, we can see that the sample spans a broad range of

investment styles and investment regions. While it is not surprising that the majority of funds

engage in the Equity Long/Short strategy, we also have Relative Value, Event Driven, Macro,

Directional, Fixed Income, and Managed Futures funds in our sample. Further, the funds invest

in several geographical regions including Korea, Greater China, Taiwan, India, Asia excluding

Japan, Asia including Japan, Australia/New Zealand, Emerging Markets, Global, and Japan only.

More importantly, for all the major investment styles and regions with at least 20 funds (except

for the Global12 and Australia/New Zealand regions) there are both funds with investment region

presence and funds without investment region presence. The variation in investment region

presence within styles and regions facilitates our tests of hedge fund geography.

We recognize that hedge fund data are susceptible to many biases (Fung and Hsieh,

2000). These biases stem from the fact that, due to the lack of regulation amongst hedge funds,

inclusion in hedge fund databases is voluntary. As a result, there is a self-selection bias. For

instance, funds often undergo an incubation period where they rely on internal funding before

seeking capital from outside investors. Incubated funds with successful track records then go on

to list in various hedge fund databases while the unsuccessful funds do not, resulting in an

incubation bias. Separate from this, when a fund is listed on a database, it often includes data

prior to the listing date. Again, since successful funds have a strong incentive to list and attract

capital inflows, these backfilled returns tend to be higher than the non-backfilled returns. In the

analysis that follows, we will repeat the tests after dropping the first 12 months of return data

from each fund so as to ensure that the results are robust to backfill and incubation bias.

2. Asset based style factors for hedge fund returns

Fung and Hsieh (1999, 2000, 2001), Mitchell and Pulvino (2001), and Agarwal and Naik

(2004) show that hedge fund returns relate to conventional asset class returns and option-based

12 This necessarily holds by construction for funds investing in the Global region.

7

strategy returns. Building on this pioneering work, Fung and Hsieh (2004) propose an asset

based style (henceforth ABS) factor model that can explain up to 80 percent of the monthly

variation in hedge fund portfolios. Their ABS model avoids using a broad based index of hedge

funds to model hedge fund risk since a fund index can inherit errors that were inherent in hedge

fund databases. These problems have been noted by Brown, Goetzmann, and Ibbotson (1999),

Fung and Hsieh (2000), and Liang (2000), and they include survivorship bias, backfill or instant

history bias, selection bias, sampling differences across fund databases, lack of transparency in

the construction of fund indexes, and the choice of index weights.13 Fung and Hsieh (2004)

choose their ABS factors by extracting common return components from sub-groups of hedge

funds that were classified by data vendors as having similar styles. Then, they link those

common sources of risk in hedge fund returns to observable market prices.

In this section, we adopt a similar methodology to augment the Fung and Hsieh (2004)

model so as to better explain risk in Asian hedge funds. By measuring fund performance relative

to the augmented Fung and Hsieh (2004) model, we can abstract from risk when analyzing the

relationship between fund performance and fund proximity to investments. To identify additional

ABS factors in Asian hedge fund space, we first perform principal components analysis of hedge

funds grouped by investment style and investment region. We do so for all style/region

intersections with at least five funds. Table 2 reports the R-squares from the regression of fund

style/region returns on each of the top ten principal components as well as the percentage of the

total variation explained by each principal component. It indicates that the first principal

component alone can account for about 51 percent of the variation in fund style/region returns.

This compares favorably with Fung and Hsieh (1997) who find that the top five principal

components can account for 43 percent of the variation in offshore and onshore US fund returns.


The R-squares in Table 2 reveal that the first principal component is likely to be an Asian

equity factor since it is explains much of the return variation for funds in the Equity Long/Short

style (although it is also explains the returns of Event Driven funds in the Asia excluding Japan

region and Fixed income funds that invest in Emerging Markets). The second principal

component is most likely a Global Macro factor given the high R-square with Global Macro

funds (R-square = 0.71). Since the second principal component explains only 16 percent of the

13 We refer the reader to Fung and Hsieh (2004) for an excellent review.

8

variation in fund style/region returns, we focus on linking the first principal component to

observable market prices.

In this effort, we compute the correlation between the first principal component and

various Asian equity market indices from Datastream, including the Asia Ex Japan, Asia, Pacific

Basin, and Japan market indices. We find that the correlation with the Asia Ex Japan market

index is the highest at 0.82. In contrast, the correlations with the Asia, Pacific Basin, and Japan

market indices are 0.64, 0.65, and 0.43, respectively. Hence, we augment the Fung and Hsieh

(2004) model with two additional factors:14 the excess return on the Datastream Asia Ex Japan

equity market index (henceforth ASIAMRF) and the excess return on the Datastream Japan

equity market index15 (henceforth JAPMRF), to better explain variation in Asian hedge fund

returns. We evaluate the efficacy of the augmented factor model by regressing the Asian hedge

fund portfolio on the factors from the augmented and non-augmented models, and comparing the

adjusted R-squares.

The augmented factor model is as follows:

immiM

miMmiMmiMmiM

miMmiMmiMmiMiMim

JAPMRFkASIAMRFjPTFSCOMhPTFSFXgPTFSBDf

BAAMTSYeRETBDdSCMLCcSNPMRFbar

ε++++++++++= 10

(1)

where m = 1,…,M, imr is the monthly return on portfolio i in excess of the one-month T-bill

return, SNPMRF is the S&P 500 return minus risk free rate, SCMLC is the Wilshire small cap

minus large cap return, BD10RET is the change in the constant maturity yield of the 10 year

treasury, BAAMTSY is the change in the spread of Moody's Baa - 10 year treasury, PTFSBD is

the bond PTFS, PTFSFX currency PTFS, PTFSCOM is the commodities PTFS, where PTFS is

primitive trend following strategy [see Fung and Hsieh (2004)].


The results in Panel A of Table 3 indicate that the augmented model well explains

variation in Asian hedge fund returns. The adjusted R-square for the augmented model is 63

percent or a meaningful 11 percent more than that for the non-augmented model. Also, it is not

14 We deliberately avoid adding even more factors so as to maintain the parsimony of the factor model and keep the degrees of freedom high in the regressions. 15 The Japan equity factor (JAPMRF) is included to help explain the returns of Japan only funds.

9

surprising that the factor loading on ASIAMRF is statistically significant at the 1 percent level (t-

statistic = 2.88), given the high correlation of ASIAMRF with the important first principal

component. Moreover, when we estimate similar regressions for fund portfolios stratified by

investment style and investment region (for styles and regions with at least 20 funds), we find

that the augmented factor model can explain more than 30 percent of the variation in portfolio

returns for all styles and regions except for Relative Value funds. Consistent with this, we find

that for two of the three styles and for four of the six regions, the loadings on ASIAMRF are

statistically significant at the 10 percent level. As a comparison, the loadings on SNPMRF are

statistically significant at the 10 percent level for only three regions. Similarly, the loadings on

SCMLC are statistically significant at the 10 percent level for only two regions. None of the other

factors are statistically significant for more than one region or style. Clearly the ASIAMRF factor

does a fairly decent job of explaining Asian hedge fund returns. Not surprisingly, the Japan only

fund portfolio loads significantly on the JAPMRF factor. In results not reported, we find that the

R-square with the augmented factor model for the all funds portfolio is 70 percent. This

compares favorably with the R-squares in Table 2 of Fung and Hsieh (2004) where the returns on

fund indices are evaluated relative to their factor model.

As a prelude to analyzing the effects of hedge fund location on fund alpha, we investigate

the statistical significance of the top hedge fund alphas. If we find that the risk-adjusted

performance of the top hedge funds cannot be explained by luck or sampling variation, then it

makes the analysis that follows particularly relevant for investors since location may hold the

key to distinguishing genuinely good funds from the pack. If on the other hand, we find that the

performance of the top hedge funds can be explained by luck, then analyzing how location

affects the cross-section of fund alpha will be of little interest to fund investors since even the top

funds are just riding on a lucky streak.

To evaluate the significance of hedge fund alpha, while taking into account the cross-

sectional distribution of hedge fund alpha as well as non-normalities in hedge fund returns, we

employ the bootstrap methodology used in Kosowski, Timmermann, Wermers, and White

(2006). Details of the bootstrap procedure are available in Kosowski, Timmermann, Wermers,

and White (2006). The basic idea is that by re-scrambling fund residuals across time and sorting

fund alphas (and alpha t-statistics) across funds, we can empirically bootstrap the distributions of

the top and bottom hedge fund alphas (and alpha t-statistics). The bootstrap p-values of the top

10

alpha and alpha t-statistic funds are displayed in Panel A of Table 4 and indicate that the

performance of the top funds cannot be attributed to luck alone.16 Of the 198 funds in our

bootstrap sample, by chance we expect ten funds to attain alphas greater than 12 percent per

year. In reality, 69 funds achieve annual alphas greater than 12 percent.

One may quibble that these results are driven by backfill bias, incubation bias, or

illiquidity. Hence, we remove the first 12 months of return data for each fund to mitigate backfill

and incubation bias, and redo the bootstrap on the remaining sample of fund returns. We also

unsmooth the returns of the hedge funds using the Getmansky, Lo, and Makarov (2004)

correction for serial correlation. In particular, we map the fund categories in Table 8 of

Getmansky, Lo, and Makarov (2004) to our fund categories (see Appendix B) and use the 10 ,θθ ,

and 2θ estimates for each fund category from their Table 8 to unsmooth fund returns. Then, we

re-do the bootstrap on the unsmoothed sample of fund returns. The bootstrapped p-values in

Panels B and C of Table 4 indicate that none of our inferences change with these adjustments.


3. Geography and hedge fund risk-adjusted performance

The discussion thus far lays out the framework for analyzing fund alpha. Also, according

to our bootstrap estimates, managers of the top alpha funds seem to possess skill. But are these

top alpha funds also funds with investment region presence? Does part of their “skill” stem from

a local informational advantage? It remains to investigate the relationship between hedge fund

geography and risk-adjusted return or alpha.

3.1. The cross-section of hedge fund alpha

To this end, we estimate cross-sectional regressions of hedge fund alpha, measured

relative to the augmented Fung and Hsieh (2004) model, on an indicator variable for investment

region presence. We set investment region presence equal to one for hedge funds with a head

office or research office in their respective investment regions. Otherwise, we set investment

16 We perform the bootstrap analysis on funds with at least 24 months of return information. Similar results obtain for funds with at least 36 months of returns.

11

region presence equal to zero. We estimate both a univariate regression, and a multivariate

regression, where we control for hedge fund characteristics that may affect fund performance.

To be precise, following Carhart (1997), we first calculate monthly fund abnormal return

or alpha as fund excess returns minus the factor realizations times loadings estimated over the

entire sample period [see Eq 5. of Carhart (1997)]. Hence, we have

)

10(

miM

miMmiMmiMmiM

miMmiMmiMmiMimim

JAPMRFkASIAMRFjPTFSCOMhPTFSFXgPTFSBDfBAAMTSYeRETBDdSCMLCcSNPMRFbrALPHA

++++++++−≡

(2)

where i = 1, …, nfunds, m = 1,…,M, imALPHA is the abnormal return of fund i for month m, imr

is fund return in excess of the riskfree rate, and the other variables are as defined in Section 2.

Then, we estimate the following pooled OLS regressions:

im

Y

y

ym

yS

s

si

siim YRDUMdSTYLEDUMcbPRESENCEaALPHA ε++++= ∑∑

−

=

−

=

1

1

1

1

, and (3)

im

Y

y

ym

yS

s

si

sii

iiiiiim

YRDUMkSTYLEDUMjhFAMSIZEgFUNDSIZE

fMININVeREDEMPdMGTFEEcPERFFEEbPRESENCEaALPHA

ε+++++

+++++=

∑∑−

=

−

=

1

1

1

1

(4)

where PRESENCE is investment region presence, PERFFEE is fund performance fee, MGTFEE

is fund management fee, REDEMP is fund redemption period, MININV is minimum investment

amount, FUNDSIZE is fund size category, FAMSIZE is fund family size category, STYLEDUM

is fund investment style dummy, and YRDUM is year dummy. Fund fees and minimum

investments may proxy for managerial ability since anecdotal evidence suggests that skilled

managers often demand high fees and minimum investments17 so as to extract rents from

investors. Fund and family size may proxy for the level of resources available to managers.

However, the decreasing returns to scale argument advanced in Goetzmann, Ingersoll, and Ross

(2003), Berk and Green (2004), and Fung, Hsieh, Naik, and Ramadorai (2006) may apply

17 For instance, James Simons’ extremely successful Renaissance Technologies Medallion fund charges a management fee of 5 percent and a performance fee of 44 percent (“Really Big Bucks” Alpha Magazine, May 2006).

12

instead. Finally, as in Aragon (2006) funds with longer redemption periods may take on greater

liquidity risk and achieve higher expected returns.18 To facilitate the estimation of fund alpha, we

only include results for funds with at least 24 months of return data.19


The results presented in columns one and two of Panel A, Table 5 are striking. The

coefficient estimate on the investment region presence variable in the univariate regression

(leftmost column of Panel A, Table 5) suggests that funds with a physical or research presence in

their investment region outperform other funds by 44 basis points per month or 5.28 percent per

year. Even after adjusting for the other hedge fund characteristics, funds with investment region

presence still outperform funds without investment region presence by 4.68 percent per year.

Both these effects are statistically significant at the 1 percent level (t-statistics = 4.44 and 3.59,

respectively). In contrast, the other fund characteristics, save the size category variables, do not

seem to explain fund alpha.

There are concerns that the results may be due to differences in the way backfill and

incubation bias affects funds with and without investment region presence. If funds with

investment region presence backfill or incubate their returns more often than funds without

investment region presence, it may explain the overperformance of the former group of funds.

Also, there are concerns that funds with investment region presence may trade more illiquid

securities. As a result, for these funds, reported returns tend to be smoother than true economic

returns, which understates volatility and overstates the statistical significance of risk-adjusted

measures like alpha. To cater for such concerns, we re-estimate the regressions with backfill and

incubation bias adjusted alpha and with alpha derived from unsmoothed returns using the

Getmansky, Lo, and Makarov (2004) correction. As in the previous section, to adjust for backfill

and incubation bias, we simply remove the first 12 months of return data from each fund and

redo the alpha estimation for funds with at least 24 months of remaining return data. Lastly, to

ensure that the overperformance of funds with investment region presence is not simply due to

lower fees, we also re-do the analysis on pre-fee20 alpha. The results from this robustness

analysis are presented in columns three to eight of Panel A, Table 5 and suggest that the

18 Nonetheless, this may not translate to higher alphas since the factor model takes into account liquidity risk as well. 19 Similar results obtain for funds with at least 36 months of return data. Results are available upon request. 20 Pre-fee alphas are computed from pre-fee returns. We derive pre-fee returns by taking the high watermark and hurdle rate as the T-bill, and assuming that the returns accrue to a first-year investor in the fund.

13

overperformance of hedge funds with investment region presence is not simply an artifact of

backfill bias, incubation bias, illiquidity, or lower fund fees. The coefficient estimates on the

investment region presence variable are statistically and economically significant (greater than

4.30 percent per year) for all specifications and for both the univariate and multivariate

regressions.21

Yet another concern is that the pooled OLS regressions do not take into account cross-

correlation in residuals. To alleviate this problem, we also report as robustness checks

regressions run using the Fama and MacBeth (1973) method. Specifically, we first run cross-

sectional regressions for each month. Then, we report the time series averages of the coefficient

estimates, and use the time-series standard errors of the average slopes to draw inferences. The

Fama-MacBeth methodology is a convenient and conservative way of accounting for potential

cross-correlation in residuals. According to Fama and French (2002), Fama-MacBeth standard

errors are often two to five times the OLS standard errors from pooled panel regressions that

ignore cross-correlation. The coefficient estimates and t-statistics from the Fama and MacBeth

(1973) regressions reported in Panel B of Table 5 strongly corroborate our findings with the

pooled OLS regressions.

3.2. Sorts on hedge fund geography

To further gauge the economic relevance of the apparent local informational advantage,

we construct portfolios of funds based on fund investment region presence and compare their

risk-adjusted performance. Specifically, we form an equal-weighted portfolio of funds with

investment region presence (portfolio A) and an equal-weighted portfolio of funds without

investment region presence (portfolio B). Then, we measure performance of the spread between

portfolio A and B relative to the augmented Fung and Hsieh (2004) model. The alpha of the

spread portfolio represents the value added to hedge fund investors (including Fund of Funds) of

21 While most of the other fund characteristics do not explain the cross-sectional variation in fund alpha for all regression specifications, there seems to be a positive relationship between fund size and alpha. However, we caution against making any inferences from this relationship as the fund size variables are taken at the end of the sample period and may simply reflect the fact that successful funds garner more inflows (Agarwal, Daniel, and Naik, 2006).

14

investing in funds with investment region presence and avoiding funds without investment

region presence.

The results from this exercise are displayed in Panel A of Table 6 and show that the

strategy of buying funds with investment region presence and avoiding funds without investment

region presence yields a risk-adjusted return of 4.54 percent per year (t-statistic = 3.09). This

suggests that astute hedge fund investors can benefit from the apparent local informational

advantage of funds with investment region presence. Consistent with the cross-sectional

regression results, our findings are robust to adjustments for backfill and incubation bias, serial

correlation in returns, and fund fees (see Panels B, C, and D of Table 6). All the spread alphas in

Panels B, C, and D are comfortably above 4 percent per year and statistically significant at the 5

percent level. It is reassuring to note that the portfolios returns (portfolio A and B) are well

explained by the augmented Fung and Hsieh (2004) factor model. All but one of the adjusted R-

squares for portfolios A and B are greater than or equal to 50 percent. In addition, the loadings

on the Asia ex Japan equity market factor are economically and statistically significant (at the 5

percent level)22 for all A and B portfolios attesting to the explanatory power of the Asia ex Japan

equity market factor.


Figure 1 complements the results from Table 6. It illustrates the monthly cumulative

average residuals (henceforth CARs) from the portfolio of funds with investment region presence

(portfolio A) and the portfolio of funds without investment region presence (portfolio B). CAR is

the cumulative difference between a portfolio's excess return and its factor loadings (estimated

over the entire sample period) multiplied by the augmented Fung and Hsieh (2004) risk factors.

The CARs in Figure 1 indicate that portfolio A consistently outperforms portfolio B over the

entire sample period and suggest that the effects of investment region presence are not peculiar

to a particular year.

[Please insert Figure 1 here]

One concern is that the option-based factors of Fung and Hsieh (2004) may not capture

the return payoffs of funds engaged in option-based strategies on Asian equity market indices.

For instance, Agarwal and Naik (2004) show that the payoffs of many equity hedge funds

resemble that from writing naked out-of-the-money (henceforth OTM) put options on the equity

22 The t-statistics on the Asian equity factor loadings are omitted for brevity and are available upon request.

15

market. Yet another concern is that the SCMLC factor may not adequately capture the risk

premium of small versus large stocks in Asia. If, in addition, nearby funds load more on small

stocks than do distant funds, then this may explain the apparent overperformance of the former.

In response to these concerns, we replicate the analysis in Table 5 after supplementing

the augmented model with two option-based factors on the Japan Nikkei 225 index and an Asian

size factor. Following Agarwal and Naik (2004), we use OTM European call and put options on

the Nikkei 225 traded on the Singapore Stock Exchange. The OTM call option strategy is as

follows. On the first day of January, buy an OTM call option on the Nikkei 225 that expires on

February. On the first day of February, sell the option bought a month ago (i.e., in January) and

buy another OTM call option on the Nikkei 225 that expires in March. Repeat this process every

month and record the returns from this strategy. The payoffs for the OTM put option-based

strategy are derived using an analogous procedure. We select the OTM call (put) option to be the

one with the next higher (lower) strike price relative to the at-the-money call (put) option whose

present value of strike price is closest to the current index value. For the Asian size factor, we

take the return difference between the Dow Jones Asia Small Cap Index and the Dow Jones Asia

Large Cap Index. The results from this analysis are virtually identical to those from Table 6 and

indicate that our inferences are robust to controlling for return covariation with the Nikkei 225

index option-based factors and the Asian size factor.23

Another issue with our results is that they may be driven by pseudo cross-region effects.

Suppose funds investing in region A consistently outperform, on a risk-adjusted basis, funds

investing in region B. In addition, funds investing in region A all have offices in region A while

funds investing in region B do not set up local offices. Then, this could generate the results in

Tables 5 and 6 even though no local informational advantage exists. Related to this, Table 1

indicates that all funds investing in the Global and Australia/New Zealand regions also have a

presence in those markets. Instead of a local informational advantage, our results may well be

capturing the overperformance of funds investing in the Global and Australia/New Zealand

regions, and the underperformance of funds investing elsewhere.

23 We also replicate the cross-sectional regression analysis in Table 5 after supplementing the augmented factor model with the two option-based factors (derived from OTM options on the Nikkei 225) and the Asian size factor. To conserve the degrees of freedom (since some funds only have 24 months of data) we adopt the stepwise regression approach of Agarwal and Naik (2004). We find that the results are virtually identical with this alternative specification.

16

Moreover, if the overperformance of funds with investment region presence stems from a

local informational advantage, then it must be that the overperformance is greater for regions

where firms disclose less public information and where the local informational advantage is

large. Extant research has shown that the local/foreign informational asymmetry is particularly

severe in Emerging Markets [see Dovrak (2005), Choe, Kho, and Stulz (2005), and Bae, Stulz,

and Tan (2006)]. Hence, we expect the overperformance of funds with investment region

presence to be greater for Emerging Markets than for other regions.

To address these issues, we break down the analysis in Table 6 by geographical region

for regions (excluding the Global and Australia/New Zealand regions) where we have at least 20

funds. The results in Table 7 clearly indicate that the overperformance of funds with investment

region presence is not driven by pseudo cross-region effects. For each major geographical

region, funds with investment region presence outperform funds without investment presence.

The alpha for the spread portfolio is consistently above 4 percent per year for all four investment

regions. Naturally with the reduced sample size, some of the within region spread alphas are no

longer statistically significant. Nonetheless, the spread alpha for the Asia including Japan region

and for Emerging Markets are statistically significant at the 1 percent level. Indeed, the spread

alpha is highest and most reliably different from zero for Emerging Market funds (t-statistic =

2.77). This dovetails with our a priori intuition that the local informational advantage should be

strongest in Emerging Markets.

[Please insert Table 7 here] As a final robustness check, we also break down the analysis in Table 6 by investment

style for styles with at least 20 funds. The results displayed in Table 8 complement those of

Table 7. They demonstrate that the overperformance of funds with investment region presence

holds within each of the three major investment styles. The alpha for the within style spread is

above 5 percent per year for all three styles. It is comforting to note that the alpha spread is

statistically significant at the 5 percent level (t-statistic = 2.47) for the most popular style: Equity

Long/Short. Out of the three styles, the spread alpha is largest (8.19 percent per year) and most

statistically significant (t-statistic = 3.32) for Event Driven funds. This is not surprising since

Event Driven funds employ strategies involving investments in opportunities created by

significant transactional events such as spinoffs, mergers and acquisitions, bankruptcy

reorganizations, recapitalizations, and share buybacks. The informational demands of such an

17

investing strategy are likely to be onerous. Hence, relative to other funds, Event Driven funds

have much to gain from a local presence.


4. Conclusion

In summary, the results in this paper tell a consistent story. Hedge funds with a physical

presence in their investment region outperform funds without a physical presence on a risk-

adjusted basis. The difference in performance manifests in the cross-section of fund alpha and in

fund portfolio sorts. The superior performance of nearby funds is not simply a by-product of

fund fees, backfill bias, incubation bias, or illiquidity. This overperformance is also not due to

cross-region effects. Within each major investment region and investment style, funds with a

local presence outperform other funds. Moreover, the overperformance is exceptionally strong

for investment regions (e.g., Emerging Markets) where the local/foreign information asymmetry

is likely to be acute and for investment styles (e.g., Event Driven) where an informational

advantage is likely to matter most. Collectively, these results suggest that nearby hedge funds

enjoy a local informational advantage. These results also contribute to our understanding of

hedge fund alpha, and are particularly relevant to fund managers and to fund investors like Fund

of Funds.

References

Audretsch, D.B., Feldman, M.P., 1996. R & D spillovers and the geography of innovation and production. American Economic Review 86, 630-640.

Agarwal, V., Daniel, N., Naik, N.Y., 2006. Flows, performance, and managerial incentives in the

hedge fund industry. Unpublished working paper. Georgia State University. Agarwal, V., Naik, N.Y., 2000. Multi-period performance persistence analysis of hedge funds.

Journal of Financial and Quantitative Analysis 35, 327-342. Agarwal, V., Naik, N.Y., 2004. Risk and portfolio decisions involving hedge funds. Review of

Financial Studies 17, 63-98.

18

Aragon, G., 2006. Share restrictions and asset pricing: evidence from the hedge fund industry. Journal of Financial Economics, forthcoming.

Bae, K.H., Stulz, R., Tan, H., 2006. Do local analysts know more? A cross-country study of the

performance of local analysts and foreign analysts. Unpublished working paper. Ohio State University.

Becker, Bo., 2006. Geographical segmentation of U.S. capital markets. Journal of Financial

Economics, forthcoming. Berk, J., Green, R., 2004. Mutual fund flows and performance in rational markets. Journal of

Political Economy 112, 1269 -1295. Brown, S., Goetzmann, W., Ibbotson, R., 1999. Offshore hedge funds: survival and performance,

1989-95. Journal of Business 72, 91-117. Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52, 57-82. Choe, H., Kho, B.C., Stulz, R., 2005. Do domestic investors have an edge? The trading

experience of foreign investors in Korea. Review of Financial Studies 18, 795-829. Cohen, L., Frazzini, A., 2006. Economic links and predictable returns. Unpublished working

paper. University of Chicago. Cooper, I., Kaplanis, E., 1994. Home bias in equity portfolios, inflation hedging, and

international capital market equilibrium. Review of Financial Studies 7, 45-60. Coval, J., Moskowitz, T., 2001. The geography of investment: informed trading and asset prices.

Journal of Political Economy 4, 811-841. Dvorak, T., 2005. Do domestic investors have an information advantage? Evidence from

Indonesia. Journal of Finance 60, 817-839. Fama, E., French, K., 2002. Testing trade-off and pecking predictions about dividends and debt.

Review of Financial Studies 15, 1-33. Fama, E., MacBeth, J.D., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political

Economy 81, 607-636. French, K., Poterba, J., 1991. Investor diversification and international equity markets. American

Economic Review 81, 222-226. Fung, W., Hsieh, D., 1997. Empirical characteristics of dynamic trading strategies: the case of

hedge funds. Review of Financial Studies 10, 275-302. Fung, W., Hsieh, D., 1999. A primer on hedge funds. Journal of Empirical Finance 6, 309-331.

19

Fung, W., Hsieh, D., 2000. Performance characteristics of hedge funds and CTA funds: natural

versus spurious biases. Journal of Financial and Quantitative Analysis 35, 291-307. Fung, W., Hsieh, D., 2001. The risk in hedge fund strategies: theory and evidence from trend

followers. Review of Financial Studies 14, 313-341. Fung, W., Hsieh, D., 2004. Hedge fund benchmarks: a risk based approach. Financial Analyst

Journal 60, 65-80. Fung, W., Hsieh, D., Naik, N., Ramadorai, T., 2006. Hedge funds: performance, risk, and capital

formation. Unpublished working paper. London Business School. Getmansky, M., Lo, A., Makarov, I., 2004. An econometric model of serial correlation and

illiquidity of hedge fund returns. Journal of Financial Economics 74, 529-610. Goetzmann, W., Ingersoll, J., Ross, S., 2003. High water marks and hedge fund management

contracts. Journal of Finance 58, 1685-1717. Grinblatt, M., Keloharju, M., 2000. The investment behaviour and performance of various

investor types: a study of Finland’s unique data set. Journal of Financial Economics 55, 43-47.

Hau, H., 2001. Location matters: an examination of trading profits. Journal of Finance 56, 1959-

1983. Ivković, Z., Weisbenner, S., 2005. Local does as local is: Informational content of the geography

of individual investors’ common stock investments. Journal of Finance 60, 267-306. Kosowski, R., Timmermann, A., Wermers, R., White, H., 2006. Can mutual fund “stars” really

pick stocks? New evidence from a bootstrap analysis. Journal of Finance, forthcoming. Kosowski, R., Naik, N., Teo, M., 2006. Do hedge funds deliver alpha? A Bayesian and bootstrap

analysis. Journal of Financial Economics, forthcoming. Krugman, P., 1991. Increasing returns and economic geography. Journal of Political Economy

99, 483-499. Liang, B., 2000. Hedge funds: the living and the dead. Journal of Financial and Quantitative

Analysis 35, 309-327. Loughran, T., Schultz, P., 2005. Liquidity: urban versus rural firms. Journal of Financial

Economics 78, 341-374. Lucas, R.E. Jr., 1993. Making a miracle. Econometrica 61, 251-272.

20

Malloy, C., 2005. The geography of equity analysis. Journal of Finance 60, 719-755. Mitchell, M., Pulvino, T., 2001. Characteristics of risk in risk arbitrage. Journal of Finance 56,

2135-2175. Ragaas De Ramos, R., 2006. Hong Kong outpaces Singapore in snaring hedge fund assets. The

Wall Street Journal, May 15 2006. Seasholes, M., 2000. Smart foreign traders in emerging markets. Unpublished working paper.

Harvard University. Tesar, L., Werner, I., 1995. Home bias and high turnover. Journal of International Money and

Finance 14, 467-492. Wighton, D., 2006. Runners and raiders. Financial Times. June 10 2006. Appendix A: Fund size category definitions Fund size category Fund size (in millions US$) 1 0-25 2 26-100 3 101-250 4 251-500 5 501-750 6 751-1,000 7 1,001-2,500 8 2,501-5,000 9 5,001-10,000 10 10,001+ Appendix B: Fund investment style mapping Asian investment styles Getmansky, Lo, and Makarov (2004) categories Relative value Non-directional/Relative value Managed futures Pure managed futures Directional Pure emerging market Event driven Event driven Fixed income Fixed income directional Equity long/short Asian equity hedge Macro Global macro Unknown Not categorized

Figure 1: Cumulative average residuals of hedge funds with presence in investment region versus hedge funds without presence in investment region. Equal-weighted portfolios of hedge funds are constructed by sorting funds on whether they have a physical / research presence in the geographical region they invest in.Cumulative average residual (CAR) is the difference between a portfolio's excess return and its factor loadings multiplied by the augmented Fung and Hsieh (2004)risk factors. Factor loadings are estimated over the entire sample period. The sample period is from January 2000 to December 2003.

2000 2001 2002 20030

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

year

cum

ulat

ive

aver

age

resid

uals

portfolioA (with presence in investment region)portfolio B (without presence in investment region)

Investment strategy Total funds Dead funds With presence Without presence Return monthsEquity Long/Short 214 16 108 106 6264Relative Value 26 5 12 13 820Event Driven 38 0 29 9 1158Macro 16 1 16 0 464Directional 1 0 0 1 39Fixed Income 12 0 6 6 407Managed Futures 10 1 10 0 311Unknown 8 0 3 5 259

Investment region Total funds Dead funds With presence Without presence Return monthsAsia excluding Japan 38 0 27 11 1183Asia including Japan 63 8 38 25 1991Australia/New Zealand 23 0 23 0 622Emerging Markets 33 0 8 25 1121Global 52 7 52 0 1479Japan only 108 8 30 77 3046Korea 2 0 1 1 96Greater China 2 0 2 0 42Taiwan 2 0 1 1 88India 2 0 2 0 54

Panel B: By investment region

The sample period is from January 2000 to December 2003. Funds are grouped according to their primary investmentstrategy (Panel A) or according to their primary investment geographical region (Panel B). Funds without investmentstrategy information are classfied as "unknown." Funds with presence are funds with a physical presence (head office orresearch office) in their investment geographical region. The total number of funds is 325. The total number of deadfunds is 23. And the total number of fund months with return information is 9722.

Table 1Summary Statistics

Panel A: By investment strategy

NumberInvestment style Investment region of funds PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9 PC10Equity Long/Short Asia excl Japan 25 0.84 0.04 0.02 0.01 0.02 0.01 0.01 0.02 0.01 0.01Equity Long/Short Asia incl Japan 44 0.78 0.02 0.01 0.01 0.10 0.01 0.02 0.00 0.02 0.00Equity Long/Short Australia/NZ 15 0.58 0.17 0.04 0.11 0.01 0.06 0.00 0.00 0.00 0.01Equity Long/Short Emerging Markets 16 0.86 0.01 0.01 0.06 0.00 0.00 0.00 0.03 0.01 0.00Equity Long/Short Global 19 0.44 0.14 0.01 0.03 0.23 0.13 0.01 0.01 0.00 0.00Equity Long/Short Japan only 92 0.22 0.05 0.06 0.00 0.14 0.00 0.33 0.00 0.00 0.01Relative Value Asia incl Japan 6 0.01 0.05 0.02 0.05 0.00 0.07 0.04 0.01 0.00 0.20Relative Value Japan only 11 0.01 0.00 0.00 0.01 0.00 0.06 0.48 0.03 0.06 0.24Event Driven Asia excl Japan 8 0.45 0.07 0.07 0.13 0.00 0.08 0.00 0.00 0.09 0.03Event Driven Asia incl Japan 8 0.05 0.04 0.00 0.00 0.00 0.17 0.19 0.34 0.00 0.08Event Driven Australia/NZ 6 0.30 0.02 0.07 0.05 0.08 0.08 0.00 0.00 0.23 0.00Event Driven Global 6 0.17 0.01 0.01 0.34 0.09 0.00 0.01 0.00 0.03 0.03Managed Futures Global 9 0.01 0.21 0.63 0.09 0.02 0.01 0.01 0.00 0.01 0.00Fixed Income Emerging Markets 7 0.42 0.04 0.05 0.08 0.15 0.12 0.00 0.09 0.02 0.01Macro Global 13 0.18 0.71 0.06 0.03 0.00 0.00 0.00 0.00 0.01 0.00

50.54 16.40 7.77 5.74 4.49 3.53 2.58 2.22 1.87 1.34Percentage of total variance explained

Table 2Explaining hedge fund returns: A principal components analysis

R-squares from regressions of hedge fund style/region returns on their principal components. For each style and region intersection with at least 5 funds, the equal-weighted portfolio of the style/region returns is constructed. Principal components analysis is used to break the returns of the style/region portfolios intoorthogonal principal components. The returns of each style/region portfolio are then regressed on each principal component, and the r-squares from theregressions recorded. PC1 denotes the top principal component, PC2 denotes the second principal component, etc. For brevity, only the r-squares with the top tenprincipal components are presented. The sample period is from January 2000 to December 2003.

Top ten principal components

Portfolio Alp

ha (p

ct/

year

)

SN

PMR

F

SC

MLC

BD

10R

ET

BA

AM

TSY

PTF

SBD

PTF

SFX

PTF

SCO

M

ASI

AM

RF

JAPM

RF

Adj. R2

All funds 8.96 0.21 0.12 0.08 0.10 0.00 0.00 0.05 0.52(3.96) (5.40) (2.72) (0.96) (0.77) (0.20) (-0.05) (2.88)

All funds 9.78 0.10 0.09 0.05 -0.05 0.00 0.00 0.03 0.13 0.04 0.63(4.81) (2.01) (2.11) (0.74) (-0.36) (-0.06) (-0.36) (1.90) (2.88) (1.23)

Equity Long/Short 9.50 0.09 0.06 0.02 -0.03 0.00 -0.01 0.04 0.15 0.04 0.62(4.03) (1.65) (1.34) (0.29) (-0.20) (-0.09) (-1.14) (2.09) (2.99) (1.23)

Relative Value 7.62 0.10 0.05 -0.08 -0.22 0.00 0.03 -0.03 -0.05 0.06 0.00(2.49) (1.36) (0.86) (-0.76) (-1.18) (-0.07) (1.99) (-1.00) (-0.77) (1.36)

Event Driven 11.88 0.03 0.05 -0.01 0.06 -0.01 0.00 0.04 0.12 0.01 0.38(4.9) (0.55) (1.07) (-0.15) (0.39) (-0.61) (0.09) (1.68) (2.25) (0.35)

Asia excluding Japan 13.27 0.03 -0.02 -0.06 0.17 0.00 0.02 0.02 0.29 0.03 0.47(3.27) (0.32) (-0.30) (-0.43) (0.70) (0.23) (1.03) (0.48) (3.23) (0.54)

Asia including Japan 8.72 0.02 0.09 0.09 -0.05 0.00 -0.01 0.04 0.24 0.04 0.51(2.72) (0.22) (1.36) (0.81) (-0.27) (-0.31) (-0.82) (1.41) (3.40) (0.99)

Australia/New Zealand 11.07 0.17 0.05 0.05 0.08 0.02 -0.02 -0.02 0.11 0.01 0.53(3.76) (2.40) (0.77) (0.46) (0.46) (1.26) (-1.45) (-0.78) (1.73) (0.14)

Emerging Markets 15.03 0.27 0.13 0.18 0.02 0.00 0.00 0.05 0.18 -0.01 0.54(3.76) (2.86) (1.60) (1.28) (0.07) (-0.21) (-0.12) (1.55) (2.07) (-0.12)

Global 9.43 0.18 0.09 0.30 0.15 0.00 0.01 0.04 0.10 -0.04 0.49(3.45) (2.84) (1.71) (3.05) (0.89) (-0.16) (0.73) (1.54) (1.63) (-0.92)

Japan only 8.13 0.01 0.11 -0.13 -0.21 0.00 -0.01 0.03 -0.02 0.10 0.31(3.59) (0.27) (2.33) (-1.62) (-1.52) (-0.01) (-0.74) (1.52) (-0.33) (3.04)

Panel B: By investment style

Panel C: By investment region

Table 3Explaining hedge fund returns: An augmented Fung and Hsieh (2004) factor model

Hedge fund portfolio performance is estimated relative to the augmented Fung and Hsieh (2004) factors. The augmentedFung and and Hsieh (2004) factors are S&P 500 return minus risk free rate (SNPMRF), Wilshire small cap minus largecap return (SCMLC), change in the constant maturity yield of the 10 year treasury (BD10RET), change in the spread ofMoody's Baa - 10 year treasury (BAAMTSY), bond PTFS (PTFSBD), currency PTFS (PTFSFX), commodities PTFS(PTFSCOM), Asia Ex Japan index return minus risk free rate (ASIAMRF), and Japan index return minus risk free rate(JAPMRF), where PTFS is primitive trend following strategy. The sample period is from January 2000 - December 2003.

Panel A: All funds

Bottom 2. 3. 4. 5. 1% 3% 5% 10% 10% 5% 3% 1% 5. 4. 3. 2. Top

monthly alpha (%) -1.64 -1.48 -1.34 -1.17 -1.17 -1.48 -0.97 -0.79 -0.47 1.71 2.67 2.99 4.06 3.52 3.71 4.06 4.39 4.59p -value (bootstrapped) 0.95 0.90 0.90 0.94 0.86 0.90 0.97 0.96 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02t -alpha -3.72 -3.25 -2.93 -2.75 -2.35 -3.25 -2.29 -1.74 -1.04 3.89 4.76 5.68 8.19 6.96 7.58 8.19 8.63 9.69p -value (bootstrapped) 0.77 0.77 0.80 0.79 0.95 0.77 0.93 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00



Panel A: Full sample of funds

Panel B: Adjusting for incubation and backfill bias

Panel C: With unsmoothed returns using the Getmansky, Lo and Makarov (2004) correction

Table 4Statistical significance of the best and worst hedge funds' performance

Panel A reports the statistical significance of performance measures for all funds. Panel B reports the statistical significance of performance measures after correcting forincubation and backfill bias by removing the first 12 months of data for each fund. Panel C reports analogous estimates after correcting for serial correlation using theGetmansky, Lo and Makarov (2004) correction. In each panel, the first and second rows reports the ranked OLS estimate of alpha in percent per month and thebootstrapped p -value of alpha. Alpha is estimated relative to the augmented Fung and Hsieh (2004) factors. The third and fourth rows report the ranked t -statistic ofalpha based on heteroskedasticity and autocorrelation consistent standard errors as well as the bootstrapped p -value of the t -statistic. The first column on the left (right)reports funds with the lowest (highest) alpha and t -statistic followed by results for the funds with the second lowest (highest) alpha and t -statistic and marginal funds atdifferent percentiles in the left (right) tail of the distribution. The p -value is based on the distribution of the best (worst) funds in 1000 bootstrap resamples. The augment-ed Fung and Hsieh (2004) factors are S&P 500 return minus risk free rate (SNPMRF), Wilshire small cap minus large cap return (SCMLC), change in the constantmaturity yield of the 10 year treasury (BD10RET), change in the spread of Moody's Baa - 10 year treasury (BAAMTSY), bond PTFS (PTFSBD), currency PTFS(PTFSFX), and commodities PTFS (PTFSCOM), Asia Ex Japan index return minus risk free rate (ASIAMRF), and Japan index return minus risk free rate (JAPMRF),where PTFS is primitive trend following strategy. The sample period is from January 2000 to December 2003.

Independent variables

investment region presence 0.44** 0.39** 0.43** 0.36** 0.43** 0.38** 0.54** 0.49**(4.44) (3.59) (3.98) (3.07) (3.99) (3.22) (5.23) (4.40)

management fee (%) 0.07 -0.28* 0.09 0.16(0.64) (-2.50) (0.81) (1.49)

performance fee (%) 0.00 -0.03 0.01 0.02(0.31) (-1.39) (0.43) (1.51)

redemption period (days) 0.00 0.01 0.00 0.00(0.44) (1.72) (0.23) (0.69)

minimum investment (US$m) -0.06 -0.05 -0.07 -0.07(-1.67) (-1.23) (-1.91) (-2.00)

fund size category 0.13* 0.23** 0.12* 0.13*(2.26) (3.60) (2.13) (2.32)

family size category -0.05* -0.06 -0.05* -0.07**(-2.12) (-1.91) (-2.03) (-2.64)

investment region presence 0.47** 0.39** 0.45** 0.42** 0.46** 0.38** 0.57** 0.50**(3.83) (3.65) (3.94) (3.88) (3.41) (3.30) (4.60) (4.55)

management fee (%) 0.09 -0.30* 0.11 0.18(0.64) (-2.24) (0.76) (1.31)

performance fee (%) 0.01 -0.03 0.01 0.03(0.58) (-1.60) (0.68) (1.49)

redemption period (days) 0.00 0.01* 0.00 0.00(1.00) (2.01) (0.76) (1.30)

minimum investment (US$m) -0.07* -0.04 -0.09* -0.09*(-2.09) (-1.16) (-2.46) (-2.58)

fund size category 0.13* 0.22** 0.13* 0.13*(2.34) (3.09) (2.24) (2.48)

family size category -0.06* -0.04 -0.06* -0.07*(-2.08) (-1.27) (-2.00) (-2.55)

unsmoothed returnsalpha from

Dependent variablepre-fee alpha

Panel A: Pooled OLS regressions

Panel B: Fama-MacBeth (1973) cross-sectional regressions

Table 5Cross-sectional regressions on hedge fund risk-adjusted performance

Pooled OLS regressions and Fama MacBeth (1973) regressions on the cross-section of hedge fund alphas. The dependentvariable is hedge fund monthly alpha estimated relative to the augmented Fung and Hsieh (2004) factors. The augmentedFung and and Hsieh (2004) factors are S&P 500 return minus risk free rate (SNPMRF), Wilshire small cap minus large capreturn (SCMLC), change in the constant maturity yield of the 10 year treasury (BD10RET), change in the spread ofMoody's Baa - 10 year treasury (BAAMTSY), bond PTFS (PTFSBD), currency PTFS (PTFSFX), and commodities PTFS(PTFSCOM), Asia Ex Japan index return minus risk free rate (ASIAMRF), and Japan index return minus risk free rate(JAPMRF), where PTFS is primitive trend following strategy. The independent variables are hedge fund characteristicssuch as management fee, performance fee, redemption notification period, minimum investment, fund size category, familysize category, investment region presence, and style dummies. Investment region presence is an indicator variable thatequals 1 when the fund has a physical or research presence in the geographical region it invests in, and equals 0 otherwise.

bias-adjusted alphamonthly alpha backfill and incubation

The t -statistics derived from White (1980) standard errors, are in parentheses. The OLS regressions (Panel A) also includeyear dummies. The coefficient estimates on the year and style dummies are omitted for brevity. The sample period is fromJanuary 2000 to December 2003. * Significant at the 5% level; ** Significant at the 1% level.

PortfolioMean Ret. (pct/ year) Std. Dev.

Alpha (pct/ year)

t -stat of alpha S

NPM

RF

SC

MLC

BD

10R

ET

BA

AM

TSY

PTF

SBD

PTF

SFX

PTF

SCO

M

ASI

AM

RF

JAPM

RF

Adj. R2

Portfolio A (with presence) 13.33 6.10 11.88 5.75 0.09 0.09 0.06 0.05 0.00 0.00 0.02 0.15 0.03 0.67Portfolio B (no presence) 7.14 5.58 7.34 3.20 0.10 0.08 0.04 -0.17 0.00 -0.01 0.04 0.10 0.04 0.51Spread (A-B) 6.19 3.01 4.54 3.09 -0.01 0.00 0.02 0.22 0.00 0.01 -0.02 0.05 -0.01 0.27

Portfolio A (with presence) 13.38 6.93 11.45 5.09 0.10 0.11 0.07 0.08 0.01 0.00 0.02 0.18 0.03 0.70Portfolio B (no presence) 6.43 6.14 7.25 2.76 0.10 0.07 0.03 -0.16 -0.01 -0.01 0.05 0.12 0.03 0.48Spread (A-B) 6.95 4.29 4.20 1.98 0.00 0.04 0.03 0.24 0.01 0.01 -0.04 0.06 0.00 0.25

Portfolio A (with presence) 13.36 6.50 11.91 5.41 0.11 0.10 0.08 0.05 0.00 0.00 0.03 0.16 0.03 0.67Portfolio B (no presence) 6.98 6.14 7.27 2.85 0.12 0.09 0.08 -0.17 0.00 -0.01 0.05 0.11 0.03 0.50Spread (A-B) 6.38 3.30 4.64 2.78 -0.02 0.01 -0.01 0.22 0.00 0.01 -0.02 0.04 0.00 0.22

Portfolio A (with presence) 17.59 6.30 16.33 7.72 0.09 0.09 0.05 0.07 0.00 0.00 0.03 0.15 0.04 0.68Portfolio B (no presence) 10.08 5.74 10.46 4.49 0.10 0.08 0.04 -0.16 0.00 -0.01 0.05 0.11 0.04 0.53Spread (A-B) 7.51 3.01 5.87 3.98 -0.01 0.00 0.02 0.23 0.00 0.01 -0.02 0.05 -0.01 0.27

Panel B: Adjusted for backfill and incubation bias

Panel C: Adjusted for serial correlation

Panel D: Pre-fee returns

Table 6Sorts on hedge fund investment region presence

Hedge funds are sorted based on whether they have a presence in the geographical region they invest in. Portfolio A is the equal-weighted portfolio of funds withinvestment region presence. Portfolio B is the equal-weighted portfolio of funds without investment region presence. Alpha is estimated relative to the augmented Fungand Hsieh (2004) factors. The augmented Fung and and Hsieh (2004) factors are S&P 500 return minus risk free rate (SNPMRF), Wilshire small cap minus large capreturn (SCMLC), change in the constant maturity yield of the 10 year treasury (BD10RET), change in the spread of Moody's Baa - 10 year treasury (BAAMTSY), bondPTFS (PTFSBD), currency PTFS (PTFSFX), and commodities PTFS (PTFSCOM), Asia Ex Japan index return minus risk free rate (ASIAMRF), and Japan index returnminus risk free rate (JAPMRF), where PTFS is primitive trend following strategy. The sample period is from January 2000 - December 2003.

Panel A: Raw returns


Alpha (pct/ year)

t -stat of alpha S

NPM

RF

SC

MLC

BD

10R

ET

BA

AM

TSY

PTF

SBD

PTF

SFX

PTF

SCO

M

ASI

AM

RF

JAPM

RF

Adj. R2

Portfolio A (with presence) 15.61 9.62 15.72 3.46 0.05 -0.03 -0.11 0.19 0.01 0.03 0.01 0.21 0.06 0.35Portfolio B (no presence) 7.60 11.92 7.58 1.65 -0.02 -0.02 0.09 0.16 -0.01 -0.01 0.03 0.49 -0.06 0.56Spread (A-B) 8.01 8.87 8.15 1.72 0.07 -0.01 -0.20 0.03 0.02 0.04 -0.02 -0.28 0.12 0.14

Portfolio A (with presence) 11.78 8.14 11.53 3.45 -0.01 0.09 0.09 -0.01 -0.01 -0.02 0.03 0.25 0.07 0.51Portfolio B (no presence) 4.93 7.70 4.81 1.37 0.05 0.09 0.08 -0.12 0.00 -0.01 0.05 0.22 0.00 0.41Spread (A-B) 6.86 4.30 6.72 2.71 -0.06 0.00 0.01 0.11 -0.01 -0.01 -0.02 0.03 0.07 -0.01

Portfolio A (with presence) 12.80 6.91 11.38 3.19 -0.01 0.16 -0.21 -0.18 0.01 -0.01 0.00 0.03 0.12 0.19Portfolio B (no presence) 6.37 4.73 6.92 3.09 0.03 0.08 -0.10 -0.23 -0.01 -0.01 0.05 -0.04 0.09 0.32Spread (A-B) 6.43 5.80 4.46 1.42 -0.04 0.08 -0.11 0.05 0.02 0.00 -0.05 0.07 0.03 0.11

Portfolio A (with presence) 36.53 23.69 42.70 3.70 0.40 0.19 -0.12 0.73 0.04 0.03 0.24 0.14 0.00 0.29Portfolio B (no presence) 11.68 9.92 10.64 2.61 0.28 0.13 0.25 -0.14 -0.01 -0.01 0.03 0.19 -0.02 0.50Spread (A-B) 24.84 20.98 32.06 2.77 0.12 0.06 -0.37 0.87 0.05 0.04 0.21 -0.05 0.02 0.07

Panel B: Asia including Japan

Panel C: Japan only

Panel D: Emerging Markets

Table 7Sorts on hedge fund investment region presence by investment region


Panel A: Asia excluding Japan


Alpha (pct/ year)

t -stat of alpha S

NPM

RF

SC

MLC

BD

10R

ET

BA

AM

TSY

PTF

SBD

PTF

SFX

PTF

SCO

M

ASI

AM

RF

JAPM

RF

Adj. R2

Portfolio A (with presence) 12.76 7.19 10.63 2.58 -0.01 0.01 -0.14 -0.07 -0.01 0.05 -0.01 0.07 0.03 0.04Portfolio B (no presence) 6.82 5.60 3.41 1.12 0.17 0.07 -0.02 -0.30 0.01 0.01 -0.04 -0.13 0.07 0.06Spread (A-B) 5.94 8.11 7.22 1.67 -0.17 -0.07 -0.12 0.23 -0.02 0.05 0.03 0.20 -0.05 0.14

Portfolio A (with presence) 14.02 5.35 13.82 5.45 0.01 0.04 -0.06 0.05 -0.01 0.00 0.04 0.12 0.02 0.36Portfolio B (no presence) 9.00 6.05 5.63 1.93 0.09 0.09 0.12 0.10 0.00 0.01 0.01 0.11 0.00 0.34Spread (A-B) 5.02 4.44 8.19 3.32 -0.08 -0.05 -0.18 -0.05 -0.01 -0.01 0.04 0.01 0.01 0.06

Portfolio A (with presence) 12.52 7.42 12.45 4.81 0.14 0.05 0.01 0.09 0.00 -0.01 0.03 0.14 0.06 0.64Portfolio B (no presence) 5.58 6.38 6.47 2.39 0.04 0.07 0.04 -0.14 -0.01 -0.02 0.06 0.17 0.02 0.48Spread (A-B) 6.94 4.91 5.98 2.47 0.10 -0.02 -0.03 0.23 0.01 0.00 -0.04 -0.03 0.03 0.26

Panel B: Event Driven

Panel C: Equity Long/Short

Table 8Sorts on hedge fund investment region presence by investment style


Panel A: Relative Value

The geography of hedge funds melvyn teo

Economy & Finance

Transcript of The geography of hedge funds melvyn teo