Offshore Investment Funds: Monsters in Emerging Markets? Woochan Kim and Shang-Jin Wei. CID Working Paper No. 69 June 2001

Transcription

1 Offshore Investment Funds: Monsters n Emergng Markets? Woochan Km and Shang-Jn We CID Workng Paper No. 69 June 2001 Copyrght 2001 Woochan Km and Shang-Jn We and the Presdent and Fellows of Harvard College Workng Papers Center for Internatonal Development at Harvard Unversty

2 Offshore Investment Funds: Monsters n Emergng Markets? Woochan Km and Shang-Jn We Abstract The fnancal crses n the emergng markets have brought to the foreground the concern about offshore nvestment funds and ther possble role n exacerbatng fnancal market volatlty. Offshore nvestment funds are alleged to engage n tradng behavors that are dfferent from ther onshore counterparts. Because they are less moderated by tax consequences and are subect to less supervson and regulaton the offshore funds may trade more ntensely. They could also engage more aggressvely n certan tradng patterns such as postve feedback tradng or herdng that could contrbute to greater market volatlty. Usng a unque data set we compare the tradng behavor n Korea by offshore funds wth three sets of onshore funds as control groups. There are a number of nterestng fndngs. Frst the offshore funds do trade more ntensely than ther onshore counterparts. Second however the offshore funds do not engage n postve feedback tradng n a sgnfcant way. In contrast there s strong evdence that the onshore funds from the U.S. and U.K. do. Thrd whle offshore funds herd they do so sgnfcantly less than the onshore funds durng the crss. In sum the offshore funds are not especally worrsome monsters. Keywords: offshore funds foregn nvestment crss feedback tradng herdng JEL Codes: F21 F3 G15 Woochan Km s currently a faculty member at the School of Publc Polcy and Management at the Korean Development Insttute (KDI). Dr. Km receved hs Ph.D. from the John F. Kennedy School of Government at Harvard Unversty. Prevously he also served at the Korean Mnstry of Fnance and Economy as Deputy Drector n charge of foregn exchange ssues. Hs research area s n behavoral fnance nternatonal captal markets and corporate governance. Shang-Jn We s the New Century Char n Internatonal Economcs and Senor Fellow at the Brookngs Insttuton a Research Fellow at Harvard Unversty Center for Internatonal Development and a Faculty Research Fellow at the NBER. He was an Assocate Professor of Publc Polcy at the Kennedy School of Government and an Advsor at the World Bank. * We thank Chul-Hee Park for the data Rchard Zeckhauser semnar partcpants at Harvard Unversty Unversty of Maryland and Brandes Unversty and an anonymous referee for helpful comments and Greg Dorchak and Mke Prosser for edtoral assstance. The vews n the paper are the authors own and may not be shared by any organzaton they are or have been assocated wth.

3 Offshore Investment Funds: Monsters n Emergng Markets? Woochan Km and Shang-Jn We 1. Introducton The fnancal crses n the emergng markets have brought to the foreground the concern about offshore nvestment funds and ther possble role n exacerbatng volatlty n the markets they nvest n. Offshore funds are collectve nvestment funds regstered n tax havens typcally small slands n the Carbbean Europe and Asa Pacfc. Many or probably most offshore funds are so-called hedge funds. 1 Celebrated offshore funds nclude George Soros Quantum Fund and Julan Robertson s Tger Fund. Note however hedge funds can also choose to locate onshore (e.g. n the U.S.) partcularly f they ntend to trade prmarly on the securtes of maor onshore markets (e.g. the U.S. stocks). The regulatory and nsttutonal envronment faced by offshore funds can be qute dfferent from onshore funds. The host countres/terrtores of the offshore funds typcally do not collect captal gans tax. More often than not they typcally do not forward the fnancal nformaton to other tax and fnancal authortes ether (even f the ultmate owners of the funds are located elsewhere). Furthermore the regulaton on these funds n the tax havens s often less strngent than that of maor ndustralzed countres where most of the onshore nvestment funds are located. Helm (1997 p414) lsted seven areas n whch offshore funds face less regulatons as compared wth ther counterparts n the U.S. For example offshore funds would have greater flexblty and less procedural delays n changng the nature structure or operaton of ther products and they would face fewer nvestment restrctons short-term tradng lmtatons captal structure requrements governance provsons and restrctons on performance-based fees. Whle onshore mutual funds are generally prohbted from leveragng ther postons (.e. 1 Fnancal market partcpants and the IMF economsts who worked on hedge funds confrm to us that many f not most offshore funds are hedge funds. However the reverse s not true: hedge funds could also choose to locate onshore (e.g. n the U.S.) partcularly those that choose to trade actvely on the stocks n the maor onshore (.e. U.S.) markets. 1

4 borrow money to nvest) offshore funds face no such restrctons unless they elect to do so themselves. As a consequence offshore funds may trade more ntensely or aggressvely than onshore funds because the zero or lower captal gans tax reduces the requred expected gans from them to trade. They may also engage n tradng behavors that are dfferent from ther onshore counterparts. 2 For example t has been alleged that foregn portfolo nvestors may engage n postve feedback tradng (e.g. rushng to buy when the market s boomng and rushng to sell when the market s declnng) and are eager to mmc each other s behavor whle gnorng nformaton about the fundamentals. There s a concern that offshore funds may be more prone to ths knd of tradng pattern than ther onshore counterparts ether due to the nature of ther nvestment styles or due to lower regulatory constrants they face at home. Behavors such as these by offshore funds could exacerbate a fnancal crss n a country to an extent not otherwse warranted by economc fundamentals. A better understandng of the offshore funds behavor s hghly relevant for the renewed debate on captal controls on short-term portfolo captal flows. Asde from outrght captal controls mposed by captal recevng countres one may magne better supervson and rsk regulaton by the governments of the captal-exportng countres as another way to regulate nternatonal captal flows. Indeed many may prefer ths approach to outrght captal controls mposed by captal-mportng countres. However the presence of offshore funds adds challenges to ths approach. Even when the G7 2 The actual dfference n tax oblgaton between the offshore and onshore funds can be complcated. In our sample those offshore funds that come from a ursdcton that does not have a tax treaty wth the Korean government are subect to wthholdng taxes mposed by the Korean government. There s a 25% wthholdng tax on dvdends but no tax on the captal gans f the nvestor owns less than 25% of the outstandng shares. A 10% surcharge (called nhabtant tax ) s added to the ncome tax but could be waved by a blateral treaty. However snce most Korean stocks tradtonally gves out only small amounts of dvdends t s possble for offshore funds to face no wthholdng tax at all. For onshore funds the wthholdng taxes mposed by the Korean government depend on the blateral treaty (f any) between the domcle of the funds and Korea. For example for onshore funds from the U.S. there s a 15% wthholdng tax on dvdends but no tax on captal gans. The 10% surcharge ( nhabtant tax ) s not waved for Amercan nvestors. See the Korea Stock Exchange Webste Of course offshore funds are not subect to any captal gans tax at home but non-tax-exempted U.S. nvestors face a federal ncome tax (ncludng dvdend and short-term captal gans) at a rate between 15 to 39.6% and a long-term captal gans tax at a 20% rate. They are subect to addtonal tax at a state level. However the U.S. nvestors receve a tax credt for the wthholdng tax that they pay to foregn governments (up to the amount of what ther U.S. tax oblgaton would have been f the dvdends and captal gans had been derved from a U.S. source). The tax oblgaton for non-u.s. nvestors could be 2

5 governments and the IMF can agree on a partcular regulatory structure t may not apply to the offshore centers. Moreover many current onshore funds could mgrate offshore as a result of changes n the regulatons n ther onshore domcles. The hypothess that offshore funds may pursue destablzng tradng strateges can be connected wth an emergng lterature on behavoral fnance mostly n the domestc fnance context. For example usng evdence from domestc market data t has been argued that nsttutonal nvestors often exhbt herdng behavor though the tendency s quanttatvely small (see Lakonshok Shlefer and Vshny 1992). There are also theoretcal models n whch ratonal nvestors may pursue postve feedback strateges destablzng prces n the process (De Long Shlefer Summers and Waldmann 1990). A number of authors have emprcally examned the behavor of foregn nvestors n emergng markets. Frankel and Schmukler ( ) have used the data on closedend country funds to nvestgate whether foregn nvestors move out of a country wth an mmnent currency crss ahead of domestc resdents. They reached a negatve fndng suggestng that domestc resdents are on average better nformed than nternatonal nvestors. Choe Kho Stulz (1999) have examned the effects of foregn nvestors as a whole on the Korean stock prces. They concluded that foregn nvestors are unlkely to have played a bg role n the market downturn partly because they had been a relatvely small part of the market. Froot O Connell and Seasholes (1998) have examned the relatonshp between aggregate portfolo flows nto varous countres and the returns of the stock market n these countres. They found evdence of a two-way mpact: portfolo flows affect the returns n the emergng market and vce versa. Dfferent types of foregn nvestors may behave dfferently. Km and We (1999) have compared tradng behavor between ndvdual versus nsttutonal foregn nvestors and between foregn nvestors who resde n Korea versus those who nvest from outsde the country. The evdence does suggest that t may be msleadng to lump all foregn nvestors nto one basket. For example foregn ndvduals tend to herd more than foregn nsttutonal nvestors and non-resdent nvestors herd more than resdent foregn nvestors. Kamnsky Lyons and Schmukler (2000) examned the tradng behavor of the mutual funds that nvest n Latn Amerca. They found evdence of postve feedback dfferent dependng on ther own home-country tax laws. 3

6 tradng both among the managers of the mutual funds and among the ultmate nvestors n the mutual funds. As far as we know none of the papers n the lterature has compared the behavor between offshore and onshore funds. As stated earler ths paper compares offshore versus onshore funds rather than hedge versus non-hedge funds. Nonetheless snce most offshore funds are hedge funds the lterature on hedge funds s also relevant for our dscusson. Fung and Hseh (1997) Brown Goetzmann and Ibbotson (1999) and Brown Goetzmann and Park (1999) poneered the examnaton of tradng strateges of hedge funds many of them located offshore. They found that hedge funds appear to shft weghts on dfferent assets very frequently. The last paper found that the currency hedge funds were unlkely to have trggered the Asan currency crss. Lackng the data on actual poston holdngs of the funds these papers utlze return nformaton to nfer tradng strateges a la Sharpe s (1995) style analyss. Ths s clever and very useful but there can be errors f certan assets that the funds have actually traded on are not ncluded n the analyss by the econometrcans and the omtted and ncluded assets have correlated returns. In ths paper we utlze a unque data set on actual month-end tradng postons of foregn funds n Korea to study the behavor of the (non-penson) offshore funds. 3 To put the results n context we compare them wth three control groups. The frst s a group of mutual funds/unt trusts that are regstered n the Unted States and Unted Kngdom. The second s a group of mutual funds regstered n eght contnental European countres. Fnally the thrd control group conssts of mutual funds/unt trusts from Sngapore and Hong Kong. All three control groups have well-regarded securtes and mutual fund laws and competent regulatory agences. Ths s partcularly true for the four countres n the frst and the thrd group. 4 We nclude the funds from contnental Europe to see f any nferences that we obtan regardng the behavor of the funds from the U.S. and U.K. are 3 Relatvely few offshore funds are penson funds whch we have excluded to mantan comparablty wth the onshore mutual funds. 4 In a survey of frms reported by the Global Compettveness Report 1998 (World Economc Forum 1998) the respondents were asked to rate the perceved adequacy of fnancal regulaton. On a 1 (least adequate) to 7 (most adequate) scale the Unted States and Unted Kngdom receved an average of 6.53 and 6.36 scores respectvely. Both of them are among the top fve most adequate countres n the sample of 53 countres. In addton Sngapore and Hong Kong (wth the scores of 6.29 and 5.72 respectvely) are also among the top ffteen countres n the country n terms of regulatory adequacy of fnancal nsttutons. 4

7 not pecular to these two countres. We make Sngapore and Hong Kong a separate control group because they lke the offshore centers have zero captal gans tax on ther funds but unlke the offshore centers do have a well-regarded regulatory system. It s useful to note that the effect of foregn nvestors as a group was found to be small on the Korean market volatlty n 1997 n part because foregn nvestors were not a large part of the market (Choe Kho and Stulz 1999). We stll would want to know f the offshore funds engage n tradng patterns potentally more destablzng than ther onshore counterparts. If the answer s yes then n markets where they have a larger presence (that s n smaller and/or more open markets than Korea n 1997 whch may nclude Korea tself n the future) they could stll contrbute to the market volatlty n a sgnfcant way. The paper s organzed as follows. Secton 2 descrbes our data sets. Sectons 3 4 and 5 examne three aspects of foregn nvestor behavor respectvely: tradng ntensty feedback tradng and herdng. Secton 6 offers some concludng remarks. 2. Data Offshore and onshore funds and ther postons Our nvestor poston data set dentfes each foregn nvestor by a unque ID code and reports the domcle of each fund and ts month-end holdng of every stock lsted n the Korean stock exchange. Our sample covers the perod from the end of 1996 to end of Ths propretary data set was kndly provded to us by the Korea Securtes Computer Corporaton (KOSCOM) an afflate to the Korea Stock Exchange (KSE). Our set of offshore funds are mutual funds or unt trusts that report ther domcle to the Korean government as ether Bahamas Bahran Bermuda Cayman Islands Channel Islands Guernsey Jersey Lechtensten Panama or the Brtsh Vrgn Islands. There are 133 such funds that own some stocks at least sometme durng the sample. It s nterestng to note that almost every sngle such domcle has a current or hstorcal Anglo-Saxon connecton. Accordng to anecdotal evdence many of the nvestors n the 5

8 offshore funds are current or past natonals of the Unted States Unted Kngdom or other G7 countres. For comparson we construct three control groups. They are mutual funds or unt trusts from (a) the Unted States and Unted Kngdom (as a group) the two largest homes of the onshore nvestment funds n the world; (b) eght contnental European countres (Belgum Denmark France Germany Italy Netherlands Portugal and Span); and (c) Sngapore and Hong Kong respectvely. There are a maxmum of 838 funds n the U.S./U.K. group 85 funds n the contnental Europe group and 64 funds n the Sngapore/HK group n the sample. We exclude funds from many other domcles such as Luxembourg from the analyss because we cannot separate offshore from onshore funds regstered n the same country. We also exclude penson funds commercal banks nvestment banks or nsurance companes from our analyss because relatvely few of them from the offshore centers were actve n Korea durng our sample. Table 1 reports the number of funds n each category. We see that the average poston of an offshore fund n Korea s a lot smaller than the average of an Amercan or Brtsh fund though slghtly larger than that of a Sngapore or Hong Kong fund. There s no category labeled as hedge funds n our sample. Our understandng from communcatng wth KOSCOM s that they would regster themselves ether as mutual funds unt trusts or as others. Notce that a hedge fund can ether be an onshore or offshore fund. Our presumpton would be that a greater fracton of the funds from our offshore group are hedge funds or pursue hedge-fund-lke strateges than those from the U.S. and U.K. The poston data by nvestor and by stock s not generally avalable as they are not always collected. In our case the Korean government s restrcton on foregn ownershp of Korean stocks and the need to enforce t helps to make ths data avalable. 5 5 For example between May and November 1997 foregn nvestors n aggregate could not own more than 23% of the outstandng shares per company and foregn nvestors ndvdually could not own more than 6%. Snce May 1998 generc restrctons on foregn ownershp are removed. However as of December 1999 the end of our sample perod there s stll a 30% aggregate foregn ownershp restrcton on two companes (Pohang Iron and Steel or POSCO and Korea Electrc Power or KEPCO) and some ndustryspecfc restrctons. The generc upper celng on foregn nvestors n aggregate changed from 10% (Jan 1992) 12% (Dec 1994) 15% (Jul 1995) 18% (Apr 1996) 20% (Oct 1996) 23% (May 1997) 26% (Nov 1997) 55% (Dec 1997) 100% (May 1998). As for ndvdual foregn nvestor 6

9 Stock Data For each stock we collect nformaton on () month-end prce () month-end number of shares outstandng and () whether the nvestment celng s bndng n that month. In addton we also collect nformaton on the Korea Composte Stock Prce Index (KOSPI) from KSE and month-end Won/dollar exchange rate from the Federal Reserve Board s web ste. 6 Fgures 1 and 2 plot the exchange rate (U.S. dollar/1000 Won) and the stock market prce ndex (KOSPI) respectvely. Combnng the two peces of nformaton Fgure 3 traces the dollar value of a $100 nvestment n KOSPI on January throughout the sample (to December ). Sub-perods n the sample We dvde our sample nto four sub-perods. (a) December 1996 May 1997 tranqul perod. Ths was the tme when Korea was regarded as one of the mracle economes n East Asa and foregn nvestors were enthusastc about nvestng n Korea. (b) June 1997 October 1997 pre-currency crss perod. Whle Korea s own currency crss would come later n November of that year the currency of Thaland Baht (and maybe other currences n Asa) were under several speculatve attacks n June. The Tha Baht collapsed at the begnnng of July markng the begnnng of what we now call the Asan fnancal crss. The Tha crss sent repercussons throughout the regon. The Korean stock market also started ts slde n June and contnued more or less durng the perod. (c) November 1997 June 1998 crss perod. On November 18 the Bank of Korea gave up defendng the Korean Won. On November 21 the Korean government asked the IMF for a bal-out. The crss began n November 1997 and contnued to around md when the currency market began to be stablzed. There were also some nstances of labor unrest and maor bankruptces durng the perod. the upper celng changed from 3% (Jan 1992) 4% (Apr 1996) 5% (Oct 1996) 6% (May 1997) 7% (Nov 1997) 50% (Dec 1997) 100% (May 1998)

10 (d) July 1998 December 1999 recovery perod. From July 1998 the Korean stock market started to rebound and contnued throughout ths sub-perod. The Korean exchange rate had started a reversal a bt earler (around February 1998) but snce July to October 1998 ts value became also relatvely stable. 3. Intensty of Tradng Not havng to pay captal gans tax at home and facng less supervson and regulaton from home governments may nduce offshore funds to trade more ntensely than ther onshore counterparts. In addton nvestment funds that prefer to trade more actvely may self-select to locate n the offshore centers. In ths secton we examne whether offshore funds actually trade more ntensely or not. Because our data does not record wthn-month transactons we cannot compute an accurate measure of turnover. However we observe the total changes n the weghts allocated to dfferent stocks on a monthly bass. Our presumpton s that across nvestor groups the total changes n the month-to-month weghts are hghly correlated wth the true turnovers. We wll use the term tradng ntensty n subsequent dscussons to denote the changes n the weghts on all the stocks. Let w(k t) denote the value of stock held by nvestor k at the end of month t dvded by the total value of all stocks held by the same nvestor at the same tme. We compute the sum of the absolute values of the changes n the weghts across all stocks for nvestor k at tme t usng the followng defnton: (1) TN ( k t) = w( k t) w( k t 1) The average tradng ntensty (weght changes) for nvestor k defned as: T 1 (2) TN ( k) = TN( k t) T 1 t= 2 8

11 where T s the total number of months n the sample. Let K() be the total number of nvestors n nvestor group ( = offshore funds US/UK funds etc). The average tradng ntensty for nvestors n a gven group s then the average of all TN(k) over all nvestors n the group (subscrpt- omtted): 1 (3) TN = TN( k) K( ) k Under the central lmt theory the TN measure s asymptotcally normal. Table 2 reports for each of the four groups of the funds the tradng ntensty measured n ths way. We see that for each of the four sub-perods offshore funds ndeed trade more ntensely than the onshore funds from the U.S. and U.K. The dfference s statstcally sgnfcant at the 5% level for all sub-perods. Moreover for the offshore funds the tradng ntensty s the hghest durng the pre-crss and the crss perods. We can perform a smlar comparson of the offshore funds wth the onshore funds from contnental Europe. Ths tme the tradng ntensty s hgher for the offshore funds n three out of four perods but the dfference s statstcally sgnfcant only n one perod. The comparson wth the funds from Hong Kong and Sngapore s nterestng. In each of the four sub-perods there s no statstcally sgnfcant dfference between the two groups. Together ths suggests that the more ntense tradng by the offshore funds (relatve to the U.S. and European funds) that we observe probably comes from the waver of captal gans tax that ther funds enoy rather than the laxty of regulaton. Future research s needed to confrm ths conecture. The defnton of tradng ntensty has an unattractve feature: f the prces of the dfferent stocks fluctuate by a dfferent amount the value of ntensty ndex changes even f no tradng takes place. As a robustness check we also mplement a dfferent defnton of tradng ntensty n terms of changng weghts n the physcal shares of stocks. To be more precse we let w(k t) be the number of stock held by nvestor k at the end of month t dvded by the total number of shares of all stocks that she held at the same tme. 9

12 Then TN(k) and TN are defned n the same way as before. The results are reported n Table 3. We can see clearly that all the qualtatve results from Table 2 reman to be true here. Thus the offshore funds do trade more ntensely than onshore funds (especally compared wth those from the U.S. and U.K.) both before the crss and even more so durng the crss. 4. Postve Feedback Tradng There are concerns that offshore funds may engage n postve-feedback tradng more aggressvely than onshore funds and that postve feedback tradng could destablze the market. A postve-feedback-tradng pattern s when one buys securtes when the prces rse and sells when the prces fall. Ths tradng pattern can result from extrapolatve expectatons about prces from stop-loss orders --automatcally sellng when the prce falls below a certan pont from forced lqudatons when an nvestor s unable to meet her margn calls or from a portfolo nsurance nvestment strategy whch calls for sellng a stock when the prce falls and buyng t when the prce rses. Postve feedback tradng can destablze the market by movng asset prces away from the fundamentals. At least snce Fredman (1953) many economsts beleve that postve feedback traders cannot be mportant n market equlbrum as they are lkely to lose money on average. Ths vew has been challenged n the last decade or so. De Long Shlefer Summers and Waldmann (1990) argued that n the presence of nose traders even ratonal nvestors may want to engage n postve feedback tradng and n the process destablze the market. Emprcal examnaton of ths ssue has emerged recently. Usng quarterly data on U.S. penson funds n the U.S. market Lakonshok Shlefer and Vshny (1992 LSV for short n later reference) dd not fnd strong evdence of sgnfcant feedback tradng. On the other hand and Grnblatt Ttman and Wermers (1995) dd fnd evdence of postve feedback tradng wth ther sample of 274 U.S. mutual funds durng Usng transacton-level data Choe Kho and Stulz (1999) also fnd evdence that foregn nvestors as a group engage n postve feedback tradng n Korea. No paper that we are aware of compares the postve tradng tendences of offshore versus onshore tradng strateges. 10

13 To examne whether nvestors engage n postve feedback tradng we need to measure the connecton between ther tradng on partcular stocks and the pror performance of the stocks. Followng a metrc proposed n Grnblatt Ttman and Wermers (1995) and modfed by Kamnsky Lyons and Schmukler (2000) we adopt the followng measure of momentum tradng for nvestor group k: Q( k t) Q( k t 1) (4) M ( k t) = [ ] R( t 1) Q *( k t) where Q (k t) s the number of shares of stock held by nvestor (or nvestor group) k at tme t Q* (k t) s the average of Q (k t) and Q (k t-1) and R ( t-1) s the return on stock from t-2 to t-1. The momentum measure for a partcular nvestor (or nvestor group) k over a gven sample perod s 1 (5) M ( k) = M ( k t) JT t where J s the total number of stocks traded by k and T s the total number of tme perods under consderaton. Under the null of no feedback tradng (n ether drecton) the mean value of M (k) s zero. Furthermore M (k) s asymptotcally normal (as J and T approach nfnty). If there s systematc postve feedback tradng then M (k) would be postve. On the other hand f there s systematc negatve feedback tradng then M (k) would be negatve. 7 To avod possble bases n quantfyng the tradng behavor we exclude certan observatons (nvestors or stock-month). Frst nvestors who declare ther purpose of the stock purchase as drect nvestment are excluded because they do not engage n actve 7 Our data does not allow us to examne a portfolo rebalancng effect. Portfolo rebalancng normally calls for sellng apprecatng stocks and buyng deprecatng stocks the opposte of postve feedback tradng. So the presence of a portfolo rebalancng effect would mply that postve feedback tradng may be stronger than our statstc suggests (but negatve feedback tradng may be weaker). 11

14 tradng. Second stocks not owned by any foregn nvestor n the prevous month are excluded. Snce short-sellng s not permtted any change n poston n these stocks can only be a buy by foregners. Thrd stocks that have reached foregn ownershp lmt are dropped because any change n the net poston of the foregn nvestors as a whole has to be a sell to Korean nvestors. The last two crtera are meant to mnmze possble bases n computed momentum. Table 4 reports the basc fndng on momentum tradng. Let us start wth the offshore funds (Column 1 of Table 4). For the frst three sub-perods ncludng the crss epsode there s no statstcally sgnfcant evdence that they engage n ether postve or negatve feedback tradng. The excepton s the recovery phrase when the offshore funds engage n contraran tradng. The U.S./UK funds make an nterestng comparson (Column 2 A). Ther momentum tradng statstcs are always sgnfcantly dfferent from zero n all subperods. Whle they may engage n contraran tradng n the pre-crss and recovery stages t s precsely durng the crss when they adopt a buy-hgh-sell-low postvefeedback-tradng pattern. Ther tendency to engage n the postve feedback tradng strategy durng the crss s sgnfcantly greater than the offshore funds at the fvepercent level (reported n Column 3 A). The momentum statstcs for the funds from contnental Europe (reported n B of Table 4) are very smlar to ther counterparts from the U.S. and U.K. In partcular whle they may engage n contraran tradng n non-crss perods they pursue postve feedback tradng strategy durng the crss. The funds from Hong Kong and Sngapore dsplay a weaker tendency to engage n momentum tradng. However they do engage n postve feedback tradng durng the crss whch s smlar to the funds from the U.S. and Europe but dfferent from the offshore funds. To summarze to the extent that postve feedback tradng may be destablzng n the emergng markets the offshore funds n our sample are unque n our sample by not engagng n t. All three control groups demonstrate a statstcally sgnfcant tendency to engage n postve feedback tradng durng the crss (though contraran tradng durng some other tmes). 12

15 pattern. 8 A possble defense of postve feedback tradng s that foregn nvestors (resdng In Table 4 the returns on the stocks are measured n unts of the local currency the won. Snce the nvestors n the sample are all nternatonal maybe a more relevant measure of the return should be based on the U.S. dollar whch allows the mpact of the exchange rate movement to be taken nto account. In Table 5 we re-compute the statstcs of the momentum tradng by usng the U.S.-dollar returns. Whle the numercal values of the statstcs vary from those n Table 4 the qualtatve features are very smlar. Most mportant we fnd that the funds n all the three control groups engage n postve feedback tradng durng the crss but the offshore funds are an excepton to ths abroad) may be nformatonally dsadvantaged relatve to domestc nvestors. They may take a (relatvely greater) declne n the prce of a partcular stock as unfavorable news revealed by domestc nvestors and may therefore ratonally choose to sell t (more aggressvely relatve to other stocks) (See Brennan and Cao 1997 for such a model). It may be useful to check f the postve-feedback-tradng pattern n our sample s ex post proftable. We do t n two steps. Frst n each month we form an equally-weghted portfolo of the ten best performng stocks and another equally-weghted portfolo of the ten worst performng stocks based on the prevous month s return defned n won. The results are reported n Table 6. 9 The average returns of the two portfolos n the prevous months are reported n the frst row of each of the four panels (representng the four dfferent sub-perods) n Table 6 (labeled as horzon -1 ). Second we track ther performances over the subsequent sx months. The results are reported n the other rows of Table 6 (labeled as horzons 1-6 ). We perform a dfference n mean test (mean return of the past wnners mnus that of the past losers). Durng the tranqul or pre-crss perod there s no sgnfcant dfference between the past wnners and past losers n terms of ther subsequent returns. However durng 8 In Appendx Table 1 we re-do the momentum tradng calculatons for the full sample (.e. wthout excludng the observatons dscussed n ths secton). The dfference n momentum tradng between the offshore and onshore funds becomes statstcally nsgnfcant n all sub-perods and for all par-wse comparsons. 9 We have performed calculatons usng the U.S.-dollar returns. The results are qualtatvely very smlar to Table 6 wth won-returns. They are not reported to save space. 13

16 the crss (as well as the recovery) stages the relatve rankng of the wnners and losers reverses tself: on average past wnners tend to do worse than past losers n terms of the subsequent returns. Ths s true at all horzons from one to sx months. The dfference n performance s sgnfcant statstcally at horzons over 4 months durng the crss. In other words f one has to choose between a postve and a negatve feedback tradng strategy durng ths sub-perod the negatve feedback strategy would have done better. As a robustness check we also form equally weghted portfolos of the 30 best performng and the 30 worst performng (based on prevous-month s returns) stocks. The results are reported n the rght half of Table 6. For these enlarged portfolos agan there s reversal n the rankng of relatve performance durng the crss. Agan a contraran tradng strategy rather than a postve feedback one would have been more proftable for ths sub-perod. As qualfcatons we note that our thought experments above have not adusted for rsk levels of the stocks and do not preclude the possblty that a postve feedback tradng strategy could be proftable wthn a day or for horzons longer than sx months. We also make an attempt to compare the rsk-adusted performance of the postve and negatve feedback tradng strateges as actually pursued by some funds n our sample. We focus on the group of the U.S. and U.K. funds as they are the largest group. Usng a technque proposed by Grnblatt and Ttman (1993) we adust for rsk by comparng the returns of the new and the old portfolos of the nvestor. In other words the rsk levels on the new and the old portfolos are assumed to be smlar so that the return on the new portfolo n excess of the old s naturally adusted for ts rsk level. We proceed n two steps. Frst we classfy all the nvestor-month pars nto two categores postve versus negatve feedback traders dependng on whether an nvestor s momentum measure M s postve or negatve n a gven month. Second for each category we compute the followng rsk-adusted returns averaged over all traders n the same group. 1 Q( k t) Q( k t 1) (6) Performance( n) = [ ] R( t + n) KJT Q *( k t) k t 14

17 where K J and T are number of nvestors n the group number of stocks and number of months n the perod respectvely. Lower case n n Performance(n) and R( t+n) denotes return horzon. For example R( t+1) and R( t+3) are the returns for stock over 1-month and 3-month horzons respectvely. Under the assumpton that that the systematc rsks for the old and new portfolos are (approxmately) the same Performance(1) and Performance(3) measure the rsk-adusted return for the new portfolo over one and three month horzons respectvely. 10 Table 7 reports the proftablty calculatons for the two tradng strateges. Usng ths new defnton of ex post proftablty the postve feedback tradng looks less terrble. In partcular t appears to do better than a contraran strategy before the crss (at the one-month horzon) and durng the recovery perod. However t s precsely durng the crss durng whch most funds were engagng n postve feedback tradng when such a strategy turns out to be unproftable. To summarze on the bass of the mpled ex post proftablty (wthout adustng for rsk) a contraran strategy seems to domnate a postve feedback strategy. On the bass of a rsk-adusted measure of proftablty the postve feedback strategy looks better though contnues to be nferor to a contraran strategy durng the crss epsode. 5. Herdng Herdng s the tendency that nvestors of a partcular group mmc each other s tradng. Portfolo nvestors may herd ratonally or rratonally. Informatonal asymmetry may cause unnformed but ratonal speculators to choose to trade n the same way as nformed traders (Bkhchandan Hrshlefer and Welch 1992; and Baneree 1992). Snce the nformatonal problem may be more serous when t comes to nvestng n a foregn market than the domestc one herdng may also be more severe. Whether offshore funds herd more or less than the onshore funds depends on ther relatve capacty n collectng and processng nformaton about the emergng market n queston. 10 Grnblatt and Ttman (1993) provde some evdence that the betas are the same for the two portfolos n ther sample. 15

18 There s an alternatve explanaton for herdng among nsttutonal nvestors. Unlke ndvdual nvestors fund managers face regular revews (e.g. quarterly for mutual funds and annually for penson funds) on ther performance relatve to a benchmark and/or to each other. Ths may nduce them to mmc each other s tradng to a greater extent than they otherwse would (See Scharfsten and Sten 1990). By ths logc whether the offshore funds herd more or less than the onshore funds depends on whether nformatonal asymmetry s greater or less for them. By ths logc there mght be less herdng among offshore funds f they are subect to ether fewer or less frequent performance revews. There have been several emprcal papers that quantfy herdng behavor. Usng data on nsttutonal nvestors the poneerng paper by Lakonshok Shlefer and Vshny (or LSV 1992) followed by Grnblatt Ttman and Wermers (1995) and Wyle (1997) all report evdence of herdng among U.S. or U.K. nsttutonal nvestors. Usng data on foregn nvestors (or U.S. nvestors) n Korea as a sngle group Choe Kho and Stulz (1999) fnd evdence of herdng. None of the prevous papers that we are aware of compares dfferent herdng tendences by dfferent nvestor types on data from a sngle source whch s the central focus of ths secton of our paper. We employ the herdng ndex measure proposed by LSV (1992). Whle we refer to the LSV measure as a herdng ndex as they do t s useful to remember that what t measures s the correlaton n tradng patterns among members of a group (the tendency to whch nvestors buy or sell the same subset of stocks). Obvously herdng leads to correlated tradng but the reverse may not be true. Let B ( t) be the number of nvestors n group that have ncreased the holdngs of stock n month t (.e. number of net buyers) and S ( t) the number of nvestors n group that have decreased the holdngs of stock n month t (number of net sellers). Let p ( t) be the number of net buyers n group aggregated across all stocks n month t dvded by the total number of actve traders (number of net buyers plus number of net sellers) n group aggregated across all stocks n month t. Then H ( t) s defned as the herdng ndex for nvestors n group on stock n month t. 16

19 17 (7) ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( t p t S t B t B E t p t S t B t B t H + + = (8) = = = + = N N N t S t B t B t p ) ( ) ( ) ( ) ( (9) = = N t H N t H 1 ) ( 1 ) ( (10) = = = T t N t H NT H 1 1 ) ( 1 ) ( H( t) s the herdng ndex for group n month t averaged across all stocks. H() s the herdng ndex for group averaged across all months n the sample. In the defnton of H( t) ) ( t p s subtracted to make sure that the resultng ndex s nsenstve to general market condtons (.e. a bull or bear market). By takng absolute values the frst term n equaton (7) captures how much of the nvestment s polarzed n the drecton of ether buyng or sellng. The second term n equaton (7) also called as adustment factor s subtracted to correct for the mean value of the frst term under the assumpton of no herdng. The second term can be computed under the assumpton that ) ( t B follows a bnomal dstrbuton. Note that for large N and T ) ( t H and ) ( H follow normal dstrbutons by the central lmt theorem. To avod any possble bas n computng the herdng ndces we exclude certan nvestors and observatons (stock-month) from our sample. Lke the sample we have constructed to examne postve feedback tradng we exclude here (1) drect nvestors (2) stock-months for whch the foregn ownershp lmt s reached and (3) stock-months for whch the stocks are not owned by foregn nvestors n the prevous month. The last excluson s motvated by the short-sellng constrant. When short sellng s not allowed any trade on that stock would have to frst show up as a buy thus basng the herdng ndex upward (Wyle 1997). Fnally f a stock n a gven month s traded by only one foregn nvestor n that group that observaton s dropped.

20 The basc results are presented n Table 8. For each fund group and sample perod we report the correspondng herdng statstcs H() wth standard errors n the parenthess below. Then we perform a sequence of dfference-n-mean tests between offshore and onshore funds (reported n Columns 3 5 and 7). The most mportant fndngs are the followng. Frst for both offshore funds as well as the three groups of onshore funds there s clear evdence of herdng: the herdng measure s statstcally dfferent from zero for all funds n each sub-perod except for the Hong Kong/Sngapore funds n the pre-crss epsode. Second and most mportantly the evdence suggests that to the extent that there s a dfference n the herdng tendency the U.S./U.K. funds herd sgnfcantly more than ther offshore counterparts n two of the four sub-perods (and are comparable wth the offshore funds n the other two subperods). The offshore funds do herd statstcally sgnfcantly more than the European onshore funds durng the crss epsode. But ths does not generalze to other sub-perods or to comparsons wth other onshore funds. In Appendx Table 2 we re-do the same calculatons for the whole sample (rather than the restrcted sample reported n Table 8). Broadly smlar results are obtaned. One notable excepton s that n the full sample the offshore funds no longer herd more durng the crss sub-perod than the European onshore funds. Collectvely the evdence reects the presumpton that offshore funds would generally herd more aggressvely than ther onshore counterparts. If anythng there s a bt of evdence that the U.S. and U.K. onshore funds can sometmes herd sgnfcantly more than the offshore funds. So far we have seen evdence that nvestment funds do engage n herdng though offshore funds do not necessarly do so more than ther onshore counterparts. It may be useful to nvestgate whether herdng has actually been proftable for the funds at least on an ex post bass. Let R( t n) denote the return of stock from t to t+n (n won). Let H(k t) denote LSV herdng ndex for stock n month t. For each nvestor group we run the followng fxed effects regresson: 18

21 (11) R( t n) = α + stock dummes + tme dummes + β 1 D ( t) H ( t) + β 2 [1 - D ( t)] H ( t) A buy dummy s defned as D ( t) = 1 f B ( t) / [B ( t) + S ( t)] > P (t) and 0 otherwse. P(t) s the fracton of all trades that s a buy. 1-D(k t) s effectvely a sell dummy. The buy and sell dummes are used to measure possble proftablty of buyherdng and sell-herdng separately. If the stocks that nvestors herd to buy (or sell) tend to apprecate (or deprecate) more than the market average we would expect to see â 1 > 0 and â 2 < 0. We perform ths regresson for both the one-month and three-month nvestment horzons (.e. n = 1 and 3). The results are reported n Table 8. In overwhelmng number of groups we see that the estmates of β 1 and â 2 are not dfferent from zero. Ths means herdng s not generally assocated wth abnormal returns. There are nne pont estmates that are statstcally sgnfcant. Among these nne however seven are of the wrong sgn. In other words the stocks that are herded to buy often experence a declne n value rather than an ncrease whereas those stocks that are herded to sell often apprecate n subsequent perods. As another way to summarze ths table we observe that there s no sngle group of funds that have managed to earn a proft from herdng n more than one sub-perod. There s no sngle sub-perod n whch more than one group of funds earns a proft from herdng. One possblty for nvestors to trade n a smlar drecton s that they all respond to common sgnals. Under the ont hypotheses that the nvestment funds respond to common sgnals and that the sgnals are payoff-relevant we would expect that those stocks that the nvestors herd more aggressvely should yeld abnormal returns (relatve to those stocks they do not herd as much). not supported n the data for most funds and most sub-perods. Accordng to Table 9 ths ont hypothess s 19

22 6. Concludng Remarks In ths paper we study the behavor of offshore nvestment funds as compared wth three groups of onshore counterpart (a) from the U.S. and U.K. (b) from contnental Europe and (c) from Sngapore and Hong Kong. Ths s made possble by a unque data set that detals the monthly stock postons of foregn nvestors n Korea as well as the home domcle of the nvestment funds. There are a number of fndngs that are worth hghlghtng here. Frst there s evdence that offshore funds ndeed trade more aggressvely than ther onshore counterparts udgng from the average turnover. Ths extra aggressveness by the offshore funds s more pronounced when compared wth the onshore funds from the U.S. and Europe whch are well-regulated and subect to a captal gans tax. But t s broadly smlar to the funds from Hong Kong and Sngapore whch are well-regulated but not subect to a captal gans tax. Ths suggests that zero tax rather than lax regulaton enoyed by the offshore funds may be more responsble for ther extra ntensty of tradng. Second there s no sgnfcant evdence to support the allegaton that the offshore funds engage n postve feedback tradng. In contrast there s strong evdence that funds from the U.S. and U.K. (and from the other onshore control groups ) exhbt a tendency to do so durng the crss perod. To the extent that a postve feedback tradng strategy by foregn nvestors may have exacerbated the volatlty n the emergng markets offshore funds are probably the wrong group to blame. Thrd offshore funds do herd. However they do not necessarly herd more than onshore funds. Indeed the evdence suggests that they often herd less than the funds from the U.S. or U.K. Agan f herdng by foregn nvestors s consdered undesrable offshore funds would not stand out as the greater culprt. A drawback of our data set s that we do not observe the asset holdngs of the funds outsde Korea. So we cannot make sweepng statements regardng the funds overall tradng patterns. However the evdence so far suggests that the offshore funds are not the partcularly worrsome monsters n the emergng markets. 20

23 References Baneree Abht (1992) A Smple Model of Herd Behavor. Quarterly Journal of Economcs 107 pp Bekaert Greet and Campbell Harvey 1998 Captal Flows and the Behavor of Emergng Market Equty Returns NBER Workng Paper No (1998) Bkhchandan Sushl Davd Hrshlefer and Ivo Welch (1992) A Theory of Fads Fashon Custom and Cultural Change as Informaton Cascades. Journal of Poltcal Economy 100 pp Brennan M. J. and H. Cao 1997 Internatonal Portfolo Investment Flows Journal of Fnance 52: Brown Stephen J. Wllam N. Goetzmann and Roger G. Ibbotson 1999 Offshore Hedge Funds: Survval and Performance Journal of Busness 72(1) January. Brown Stephen J. Wllam N. Goetzmann and James Park 1999 Hedge Funds and the Asan Currency Crss of 1997 forthcomng Journal of Portfolo Management. Choe Hyuk Bong-Chan Kho and Rene M. Stulz 1999 "Do Foregn Investors Destablze Stock Markets? The Korean Experence n 1997" Journal of Fnancal Economcs 54: De Long J. Bradford Andre Shlefer Lawrence H. Summers and Robert J. Waldmann (1990) Postve Feedback Investment Strateges and Destablzng Ratonal Speculaton. Journal of Fnance Vol. 45 No. 2 pp Fnancal Stablty Forum 2000 Report of the Workng Group on Offshore Centres Fnancal Supervsory Servce (Korea) 2000 Foregn Investors Trend of Frankel Jeffrey A. and Sergo L. Schmukler (1996) Country Fund Dscounts Asymmetrc Informaton and the Mexcan Crss of 1994: Dd Local Resdents Turn Pessmstc Before Internatonal Investors? NBER Workng Paper No Frankel Jeffrey A. and Sergo L. Schmukler (1998) Country Funds and Asymmetrc Informaton. Polcy Research Workng Paper No The World Bank. 21

24 Fredman Mlton (1953) The Case for Flexble Exchange Rates n Mlton Fredman ed. Essays n Postve Economcs (Unversty of Chcago Press Chcago IL). Froot Kenneth A. Paul G.J. O'Connell and Mark S. Seasholes 1998 "The Portfolo Flows of Internatonal Investors I." NBER Workng Paper No Forthcomng Journal of Fnancal Economcs. Fung Wllam and Davd A. Hseh 1997 Emprcal Characterzaton of Dynamc Tradng Strateges: The Case of Hedge Funds Revew of Fnancal Studes 10: Goldfan Ilan and Rodrgo Valdes 1997 Captal Flows and Twn Crses: The Role of Lqudty IMF Workng Paper WP/97/87. Grnblatt Mark and Sherdan Ttman 1993 Performance Measurement wthout Benchmarks: An Examnaton of Mutual Fund Returns Journal of Busness 66: Grnblatt Mark Sherdan Ttman and Russ Wermers1995 Momentum Investment Strateges Portfolo Performance and Herdng: A Study of Mutual Fund Behavor. Amercan Economc Revew Vol. 85 pp Helm Robert W Offshore Investment Funds Chapter 17 n Clfford E. Krsch ed. The Fnancal Servces Revoluton: Understandng the Changng Role of Banks Mutual Funds and Insurance Companes Chcago London and Sngapore: Irwn Henry Peter 2000 Stock Market Lberalzaton Economc Reform and Emergng Market Equty Prces Journal of Fnance 55: Kamnsky Gracela Rchard Lyons and Sergo Schmukler 2000 Managers Investors and Crses: Mutual Fund Strateges n Emergng Markets NBER Workng Paper 7855 and the World Bank Workng Paper Km Woochan and Shang-Jn We 1999 Foregn Portfolo Investors Before and Durng a Crss NBER Workng Paper 6968 February. [Also released as OECD Economcs Department Workng Paper No. 210 February 1999.] Forthcomng Journal of Internatonal Economcs. Lakonshok Josef Andre Shlefer and Robert Vshny (1992) The Impact of Insttutonal Tradng on Stock Prces. Journal of Fnancal Economcs Vol.32 pp

25 Mlroy Robert 1998 Standard & Poor s Mcropal Gude to Offshore Investment Funds Edton. Internatonal Offshore Publcatons Lmted Guernsey Channel Islands. Scharfsten Davd S. and Jeremy C. Sten (1990) Herd Behavor and Investment. Amercan Economc Revew 80 pp Sharpe Wllam 1995 The Styles and Performances of Large Seasoned U.S. Mutual Funds Workng paper Stanford Unversty Busness School. World Economc Forum (1998) The Global Compettveness Report 1998 Geneva: Swtzerland 1988 Wermers Russ 1999 Mutual Fund Herdng and the Impact on Stock Prces Journal of Fnance 54: Wyle Samuel (1997) Tests of the Accuracy of Measures of Herdng. Unpublshed Paper. 23