(PDF) International equity flows and country funds - DOKUMEN.TIPS (2024)

Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

Contents lists available at ScienceDirect

Journal of International FinancialMarkets, Institutions & Money

journal homepage: www.elsevier.com/locate/ intf in

International equity flows and country funds

Pei-Jung Tsai ∗

Department of Finance, National Chung Cheng University, 168 University Rd., Min-Hsiung, Chia-yi 621, Taiwan, ROC

a r t i c l e i n f o

Article history:Received 22 October 2008Accepted 9 July 2009Available online 17 July 2009

JEL classification:E44F02G15

Keywords:Country fundsEquity flowsVolatility effectsAsian crisis

a b s t r a c t

This study investigates the relationships between U.S. equity flowsin foreign countries and returns of closed-end country fundsfor emerging Latin American markets, emerging Asian marketsand developed markets. The major issues addressed are (1) rela-tionships between flows and fund returns based on two basicmodels—information contribution and feedback trading effects, (2)the role of volatility in these relationships, and (3) the effects of theAsian crisis. Basic findings include: (1) information contribution(past flows affect returns) and feedback trading arguments (pastreturns affect flows) are supported; (2) strong evidence is foundfor the market segmentation argument rather than the investorsentiment argument; (3) there exists strong evidence of significantvolatility effects under information contribution and feedback trad-ing; (4) the Asian crisis effects are important but limited to Asianfunds.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

As emerging countries continue to open their markets to attract foreign capital and investors areeager to diversify their portfolios internationally, the role of international equity flows becomes inter-estingly important. Policymakers of emerging markets are especially concerned about the impact ofinternational equity flows on their domestic markets, because, during recessionary periods, unex-pected reversals of international equity flows may destabilize domestic financial markets.

As is well known, underlying foreign assets in country funds are packaged into shares and tradedon U.S. stock exchanges. Country funds trade at market prices that are determined in the U.S. markets,whereas their underlying net asset values (NAVs) are determined in the foreign equity markets. The

∗ Tel.: +886 52720411x34205; fax: +886 52720818.E-mail addresses: [emailprotected], [emailprotected].

1042-4431/$ – see front matter © 2009 Elsevier B.V. All rights reserved.doi:10.1016/j.intfin.2009.07.003

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 863

dynamic pricing relationships between fund share prices and NAVs allow the distinction between localand U.S. investor expectations in linking international asset prices.

The linkages between international equity flows and country fund returns are important to investorsand fund managers for diversification decisions and tactical trading strategies. Of specific interest iswhether new market information incorporated in international equity flows is efficiently reflected infund prices and NAVs, and whether the flows are appreciably affected by prices and NAVs. Past studiesshow that international equity flows positively impact country fund returns (e.g., Nishiotis, 2006;Froot and Ramadorai, 2008). However, the impact of country fund returns on international equityflows generally has been ignored in analyzing the dynamic pricing relationships for country funds.And, little work has been done concerning effects of volatility on the relationships between flows andfund returns. This paper presents a more comprehensive specification of the relationship betweeninternational equity flows and country fund returns (prices and NAVs) by combining volatility andmean effects.

Several definitions of the relationship between international equity flows and country fund returnshave appeared in the literature. This paper examines two identified relationships: (1) feedback trading(current international equity flows are affected by past returns) and (2) information contribution (cur-rent international equity flows impact current and future returns). Information contribution effectsinclude long-term effects, viewed as fundamental information revelation, and short-term effects,designated as price pressure (noise effect).

Country funds’ prices typically diverge from their NAVs; that is, country funds trade, on average,at discounts (where fund prices are less than NAVs). Eun and Janakiramanan (1986) argue that globalcapital markets are somewhat segmented due to various market imperfections. In segmented markets,local and U.S. investors might interpret price and/or fundamental information differently, therebyleading to a differential price-flow and NAV-flow relationship. On the other hand, Hardouvelis et al.(1994) argue that U.S. investor sentiment affects country fund price returns, but not NAV returns. U.S.investor sentiment effect potentially could be the factor that drives these differential effects betweenprice returns and flows relative to NAV returns and flows. This paper extends the work by separatingthe two explanations for differential effects between price-flow and NAV-flow relationships.

International equity flows may change during periods of turbulence and crisis. Although a vastliterature has shown that the strength of inter-market relationships is reduced during financial crises,little has been done dealing with a crisis effect on the existing relationships between internationalequity flows and a fund’s return-generating process. The dangers of adverse international capital flowsmay grow stronger during a crisis period, and heterogeneous information between U.S. and foreigninvestors may lead to different market behaviors from increased risk during periods of crisis. Thus, thechanges of the relationship between flows and fund returns become especially important as financialmarkets experience various degrees of uncertainty.

This paper addresses these issues and includes both emerging and developing markets. Over-all, the paper (1) investigates the relationships between flows and fund returns within two basicframeworks-information contribution and feedback trading, (2) distinguishes between the two dif-ferent explanations (market segmentation versus investor sentiment) for differential price-flow andNAV-flow relationships, (3) examines the dynamics between fund returns and flow volatility andbetween fund return volatility and flows, and (4) analyzes the impact of the 1997 Asian crisis onthe structure of the relationships.

The results contribute to the literature in three primary respects. First, this paper complementsthe findings of earlier papers by investigating and identifying the differentials between price-flowand NAV-flow relationships, an area where little work has been done. Second, while previous studiesfocused only on possible mean impacts, this study extends existing literature by combining volatilityand mean effects in the empirical investigation of interrelationships between fund returns and inter-national equity flows. Last, it incorporates the crisis effects into the dynamic framework. While thereis a vast literature of contagion and financial market crises, little has been done dealing with effects ofa crisis on existing relationships between flows and country fund returns.

The remainder of the study proceeds as follows. Section 2 reviews previous research and litera-ture and develops a set of empirical hypotheses. Section 3 describes the data. Section 4 sets out themethodologies, and Section 5 presents and interprets empirical results. Section 6 concludes.

864 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

2. Related literature and empirical hypotheses

Past studies concerning the information contribution concept suggest that local equity returnsare related to current and past foreign equity flows.1 More specifically, Gemmill and Thomas (2002)show that monthly flows by retail investors into UK closed-end funds forecast movements in discounts.Nishiotis (2006) finds a long-run relationship between country fund discounts and international equityflows. Froot and Ramadorai (2008) present evidence that foreign equity flows help forecast both priceand NAV returns in emerging market country funds.

Prior studies of feedback trading show that current equity flows are positively affected by past equityreturns.2 However, the empirical literature on the impact of country fund returns on internationalequity flows is limited to papers such as Froot and Ramadorai (2008) who show foreign equity flowsdisplay trend-following behavior in response to movements in price and NAV returns.

The results of the aforementioned studies can lead to the potential link between international equityflows and country fund returns. A country fund’s price and NAV movements are closely connected,driven by the same return-generating process. Thus, the information contribution concept proposesthat fund price and NAV returns should respond to new market information (fundamentals or non-fundamentals) incorporated in international equity flows. Similarly, feedback trading hypothesizesthat high local returns (proxied by fund price and NAV returns) are expected to attract internationalequity flows into local markets.

However, various market imperfections would generate divergent expectations and asymmetricmarket information across local and U.S. investors (e.g., Errunza et al., 1998; Hardouvelis et al., 1994;Bodurtha et al., 1995; Patro, 2005). In this regard, there exist differential responses between countryfund price and NAV returns for international equity flows. Bekaert (1995) and Nishiotis (2004) find thatindirect investment barriers have significant impacts on asset pricing differences across countries. Thatis, in segmented markets, local and U.S. investors might interpret price and/or fundamental informationdifferently, thereby leading to a differential price-flow and NAV-flow relationship. On the other hand,Bodurtha et al. (1995) argue that small investors are usually noise traders whose investment strategiesrely on sentiment rather than on fundamental information about asset valuation. Hardouvelis et al.(1994) find empirical evidence that a change in U.S. investor sentiment is reflected in country fundprices, whereas a change in foreign investor sentiment is mirrored in NAVs. Thus, such differentialresponses between price returns and flows relative to NAV returns and flows could be attributed toU.S. investor sentiment effects.

Based on the information contribution hypothesis, cross-border equity flows are linked to newmarket information revelation, and would be expected to impact positively on both country fund priceand NAV returns. An important consideration in the analysis of the effect of flows on fund returns is thatdifferential responses between price and NAV returns for flows could be attributed to either marketsegmentation or investor sentiment effects. If the differential sensitivity between fund price returnsand NAV returns persists even after controlling for U.S. investor sentiment effect, this may providesubstantial support for the interpretation of market segmentation. On the contrary, if no significantdifferential reactions between fund price returns and NAV returns are found after controlling for U.S.investor sentiment effect, then the interpretation of investor sentiment is favored against marketsegmentation argument.

Similarly, based on the feedback trading hypothesis, international equity flows are affected by coun-try fund price and NAV returns. The impact differences between price and flows relative to NAV and

1 Research concerning the information contribution concept attempts to address whether the impact of equity flows on returnsare temporary or permanent. One part of the information contribution concept argues that foreign equity flows incorporatemarkets’ fundamentals, thereby making the impact of foreign equity flows on returns permanent (Scholes, 1972; Kraus and Stoll,1972; Dann et al., 1977). The other part, price pressure explanations, suggests that foreign equity flows incorporate not market’sfundamentals but noise, thereby making the impact of foreign equity flows on returns temporary (Froot and Ramadorai, 2001;Harris and Gurel, 1986).

2 Various studies of feedback trading document that foreign equity flows are positively affected by equity returns (Barberiset al., 1998; Jegadeesh and Titman, 1993, 2001; Shleifer, 2000). Similarly, Froot and Ramadorai (2001), and Griffin et al. (2004)point out that high equity returns attract foreign equity inflows while low equity returns drive foreign capital outflows.

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 865

flows should also be found due to either market segmentation or investor sentiment effects. The inter-national equity flow data used in this paper are more likely driven by U.S. institutional investors. Ifinstitutional investors are value-chasing, then the flows are linked to rational asset valuations andshould not be subject to the variation in irrational investor sentiment. In this case, if the impact dif-ferences between price returns and NAV returns are found after controlling for a possible investorsentiment effect, this should be the evidence in favor of the market segmentation argument. Onthe other hand, if no significant differential reactions exist after controlling for U.S. investor senti-ment effect, this would support investor sentiment hypothesis that international equity flows aredriven by noise trading. This analysis leads to Granger causality tests, which allows for tests of mar-ket segmentation versus investor sentiment arguments for the differential price-flow and NAV-flowrelationship.

Financial market crises, such as the 1994 Mexican peso crash and the 1997 Asian financial crisis,have received more attention than lesser volatility effects. Studies that examine the crisis dynamicsof the information transmission process include Frankel and Schmukler (1996) who employ threeMexican country funds during the 1994 Mexican peso crisis and find that Mexican investors turnpessimistic before international investors. On the contrary, Pan et al. (2001), using six Asian countryfunds, find that foreign investors reacted to the Asian financial crisis earlier than local investors. Boweand Domuta (2001) find that both local and foreign investor expectations are important in determiningthe pricing behavior of Asian assets trading in Asian and U.S. equity markets, and that the impact offund price returns on local Asian asset returns is strengthened during the period of the Asian financialcrisis. These studies find significant crisis effects on country fund pricing and premiums/discounts.However, relatively little work has been done on how the Asian crisis affected the sensitivity of priceand NAV returns to equity flows and whether this effect was different for price versus NAV returns.

While numerous papers have examined the impact of foreign equity flows on equity returns, littleattention has been given to the role of volatility on the existing relationships between flows and equityreturns. Related volatility studies include Pontiff (1997) for U.S. domestic closed-end funds and Lee andHong (2002) for non-U.S. country funds, and both find that unconditional variance of fund price returnsis significantly larger than the variance of returns on NAVs. Lee and Hong (2002) find that much ofcountry fund price return variance appears to be due to non-fundamentals. This paper extends existingliterature by combining volatility and mean effects, and by incorporating crisis effects into both theinformation contribution and the feedback trading framework.

3. Data

The data set consists of 35 closed-end single country funds, encompassing 22 countries. The datasetbegins January 1994 and ends December 2007.3 For each fund, share prices, net asset values (NAVs)and dividends are collected from Lipper Analytical Services.4 Both fund prices and NAVs are reportedin U.S. dollars. The U.S. equity market is proxied by the S&P 500 index.

Monthly fund returns (adjusted for stock splits, dividends and rights offerings) are calculated asfirst differences of the logarithm.5 U.S. market returns are calculated as first differenced logs of U.S.equity market index. The use of returns computed on a monthly basis may partially relieve the issues ofshorter observation periods, because daily data compound non-synchronous trading hours via timingof quote computations and conversion into dollars, and weekly data may confound day of week effectswhen markets are not always closed/open on the same day(s).6

3 Several country funds (e.g., Brazil, Brazilian Equity, First Philippine, India Growth, J. F. India, Templeton China, Austria, FranceGrowth, Italy and Portugal) open-ended or liquidated since 2002 or 2003.

4 A special thanks to Greg Jacyszyn at Lipper Analytical Services for his assistance in providing data for this research project.5 Fund price returns are computed as log(Pt + DIVt) − log(Pt−1), where Pt is the price of the fund at period t and DIVt is the

dividend and capital gain distributions paid by the fund at period t. Fund NAV returns are computed as first differenced logs ofmonthly individual fund NAVs.

6 Delcoure and Zhong (2007) present a very clear and complete explanation of the problem and apply both Goetzmann et al.(2001) and Engle and Sarkar’s (2002) suggested corrections for their daily data. Interestingly, the results are similar for the twoadjusted and for the non-adjusted measures.

866 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

Lastly, monthly data on international capital flows are collected from the U.S. Department of theTreasury. The Treasury Department records the value of U.S. purchases and sales of domestic and for-eign securities (bonds and equities) on a bilateral basis for developed and emerging markets. These datareflect the location of the security transactions, that is, equity and bond outflows and inflows betweenthe U.S. and the individual countries. International equity flows are defined as net U.S. purchases offoreign securities relative to each individual country where net U.S. purchases are computed as grosssales by foreigners of foreign securities to U.S. residents minus gross purchases by foreigners of foreignsecurities from U.S. residents. The net flows are done on a country by country basis vis-à-vis the U.S. Toscale the flows, net equity flows are divided by market capitalization. Without scaling, it is problematicto compare flows across countries or even across time within a country. Market capitalization data forindividual markets are from Datastream, and are denominated in U.S. dollars. Following Bodurtha etal. (1995), premiums/discounts of an equally weighted index of U.S. domestic closed-end funds arecalculated as a control for possible U.S. investor sentiment in the pricing of closed-end country fundsand their underlying assets.7

Descriptive statistics for the emerging market funds and developed market funds are set forth inTable 1. Shown are correlation coefficients of international net equity flows (F) with country fundprice returns (PR), NAV returns (NAVR), and U.S. market returns (USR). As indicated, all fund priceand NAV returns are significantly positively correlated with flows except three China-invested funds.Interestingly, the correlation between flows and U.S. market returns is significantly negative only for20 of 35 funds.8 Premiums/discounts of an equally weighted index of U.S. domestic closed-end funds(PDW) show significant positive correlation with fund price returns in 12 cases, but no significantcorrelation with either NAV returns or flows.

4. Methodology

The primary methodology used to investigate the relationships between international equity flowsand country fund returns is Granger causality tests in a vector autoregression model (VAR). Preliminarystationarity and cointegration tests are conducted.

4.1. Stationarity tests, Granger causality and cointegration

Because Granger causality tests require that all data series in the system be stationary, diagnosticprocedures are conducted as a preliminary step. Otherwise inferences from test statistics may be spu-rious because of nonstandard distributions. To test for nonstationarity (or a unit root) of the includedvariables in the sample, both the Augmented Dickey-Fuller test (1979) and the Phillips-Perron test(1988) are employed. Lag length is determined using the Akaike Information Criterion (AIC) and theSchwartz Bayesian Criterion (SBC). Results identify a unit root in the level series of both fund shareprices and NAVs, indicating that the series of both fund share prices and NAVs are nonstationary. But,the flows series are stationary. After taking first differences for fund share prices and for NAVs, bothdata series are stationary.9

If two time series are not stationary in levels but are stationary in first differences and are coin-tegrated, there is a long-run equilibrium between the two series. And, Engle and Granger (1987)argue that if cointegration exists but is ignored, Granger causality tests will be over-differencedand misspecified without taking into account possible long-term relationships. In other words, inthe presence of cointegration, the standard Granger causality tests become inappropriate, since onlyshort-run effects will be captured when all the series are simply first differenced. Since fund share

7 The data for the 60 domestic closed-end funds in existence during the sample period are obtained from The Wall StreetJournal. The list of funds is available from the author upon request.

8 The moving variance of flows and returns (VF, V PR, V NAVR, V USR, and V PDW) is not significantly correlated with the flowsand returns (F, PR, NAVR, USR, and PDW) for any fund. Detailed results are available from the author upon request.

9 Detailed results are available from the author upon request.

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 867

Table 1Correlations.

Fund name Pearson correlation coefficientsof flows (F) with

Pearson correlation coefficients ofpremiums/discounts of U.S. domesticclosed-end funds (PDW) with

PR NAVR USR PR NAVR F

Emerging MarketsLatin American

Brazil 0.3165** 0.3650** −0.2084** 0.0249 0.0200 0.0843Brazilian equity 0.3488** 0.4022** −0.1002 0.0442 0.0011 0.0648Chile 0.1382* 0.1623* −0.0910 0.0654 0.0346 0.0027Mexico equity & income 0.2299** 0.1291* −0.1007 0.1260* 0.0543 0.0123Mexico 0.2204** 0.1294* −0.1175* 0.1234* 0.0604 0.0125

AsianChina 0.0418 0.1083 −0.0816 0.0902 0.0504 0.0449First Philippine 0.1459* 0.1688** −0.1201* 0.0612 0.0553 0.0677Greater China 0.0134 0.1038 −0.0816 0.0740 0.0395 0.0449India 0.1790** 0.1955** −0.0079 0.0688 0.0182 0.0989India Growth 0.1434* 0.1651* −0.1388* 0.0535 0.0063 0.0786Indonesia 0.1460* 0.1587* −0.1548** 0.1272* 0.1114 0.0350J. F. China 0.0836 0.1044 −0.0816 0.0848 0.0533 0.0449J. F. India 0.1558* 0.1780* −0.0826 0.0521 0.0133 0.0838Korea equity 0.1494* 0.1777* −0.1381* 0.1256* 0.0326 0.0622Korea 0.1486* 0.1682* −0.1100* 0.1247* 0.0187 0.0639Malaysia 0.1464** 0.1795** −0.1320* 0.1283* 0.0958 0.1260Morg Stan India 0.1398* 0.1434* −0.0875 0.0960 0.0037 0.0362ROC Taiwan 0.1523* 0.1705** −0.1268* 0.0335 0.0028 0.0026Singapore 0.1476* 0.1626* −0.1434* 0.1019 0.0609 0.0458Taiwan 0.1580* 0.1877** −0.1418* 0.0502 0.0007 0.0027Templeton China 0.1389* 0.1448* 0.0832 0.1254 0.0636 0.0067Thai Capital 0.1686** 0.2528** −0.1775** 0.1119* 0.0913 0.0899Thai 0.2080** 0.2201** −0.1399* 0.1365** 0.1243 0.0895

Developed markets (Asian)Aberdeen Australia 0.1630** 0.1734** −0.1028 0.1429** 0.1293 0.0949Japan equity 0.3043** 0.4547** −0.1518** 0.0844 0.0569 0.1000Japan small cap 0.3456** 0.4553** −0.1508** 0.0615 0.0190 0.0971

Developed markets (European)Austria 0.1499** 0.1628** −0.1090 0.0829 0.0559 0.1392France Growth 0.2053** 0.2343** −0.1215* 0.0293 0.0256 0.0363Germany 0.1592** 0.2253** −0.1498* 0.0533 0.0348 0.2483Italy 0.1278* 0.1727* −0.1001 0.0516 0.0106 0.0483New Germany 0.1239* 0.1824** −0.1352* 0.0466 0.0136 0.0484New Ireland 0.1423* 0.1840* −0.1409* 0.1451* 0.0704 0.0870Portugal 0.1247* 0.1433* −0.1053 0.1011 0.0979 0.0461Spain 0.1323* 0.1638* −0.1594** 0.1243** 0.1083 0.0289Swiss Helvetia 0.1352* 0.1601* −0.1034 0.1121* 0.0261 0.0071

The sample period covers 14 years, January 1994 through December 2007. PR is the price returns on the fund; NAVR is thereturns on the fund’s net asset value; F is international net equity flows; USR is the U.S. market index returns; PDW is thepremiums/discounts on an equally weighted index of U.S. domestic closed-end funds. Several country funds (e.g., Brazil, BrazilianEquity, First Philippine, India Growth, J. F. India, Templeton China, Austria, France Growth, Italy and Portugal) open-ended orliquidated since 2002 or 2003. Significance at the 5% level is denoted by **, and significance at the 10% level is denoted by *.

price and NAV movements are driven by the same underlying assets, there exists a strong connec-tion between the two variables. If the two series (e.g., fund share price and NAV) are cointegrated,vector error correction models (VECMs) are appropriate. Thus the trace test and the maximumeigenvalue (�max) test (Johansen and Juselius, 1990) are used to test for cointegrating relation-ships.

868 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

The trace test and �max test statistics indicate that fund share prices and NAVs are cointegratedfor all funds except J.F. India, Morg Stan India, and Templeton China.10 Thus the VECM becomes theappropriate formulation for making causality inferences between flows and fund returns for all butthese three funds.

4.2. Overall relationships between international equity flows and fund returns

4.2.1. Entire period relationships between flows and fund returnsThe results of earlier studies show a potential link between international equity flows and country

fund returns (prices and NAVs). Granger causality analysis will be conducted to examine the dynamicrelationships between flows and returns within two basic hypotheses (feedback trading and infor-mation contribution) developed above. Specifically, feedback trading hypothesizes that internationalequity flows are affected by local equity returns; that is, fund returns will Granger-cause interna-tional equity flows. Conversely, information contribution argues that local equity returns are affectedby international equity flows; that is, international equity flows will Granger-cause fund returns.Because of cointegration, Granger causality tests will incorporate an error-correction term to capturethe long-term equilibrium relationship (Engle and Granger, 1987; Granger, 1988).

Country fund returns have been shown to be sensitive to U.S. market returns (e.g., Hardouveliset al., 1994; Bodurtha et al., 1995). Since flows to and from the U.S. market are the measures ofinternational equity flows being considered, U.S. market returns would be expected to influence theinterrelationships of flows and fund returns. Thus, to control for the possible effect of U.S. marketreturns on both flows and fund returns, U.S. market returns (proxied by the S&P 500) are incorpo-rated into the three-variable VECM methods. Premiums/discounts of an equally weighted index of U.S.domestic closed-end funds as a proxy for U.S. investor sentiment are added in the model to test mar-ket segmentation versus investor sentiment arguments for the differential price-flow and NAV-flowrelationship:

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i

+L∑

i=1

a1i PDWt−i + e1t (1)

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i

+L∑

i=1

a2i PDWt−i + e2t (2)

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i

+L∑

i=1

a3i PDWt−i + e3t (3)

10 The cointegration relationships between fund share prices and flows and between fund NAVs and flows are not found for anyfund. To conserve space, the estimated results of the cointegration relationships between fund share prices and NAVs, betweenfund share prices and flows, and between fund NAVs and flows are omitted and are available from the author upon request.

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 869

where PRt is the price returns on the fund at period t; NAVRt is the returns of the fund NAVs at periodt; Ft is international net equity flows at period t; ECTt−1 is the error-correction term at period t − 1derived from the long-term cointegration relationship; USRt is the returns of the U.S. market index atperiod t; PDWt is the premiums/discounts on an equally weighted index of U.S. domestic closed-endfunds at period t.

The vector error-correction model (VECM) allows analysis of both long-run and short-run relation-ships between international equity flows and fund returns. The usual version of the Granger causalitytest does not restrict the sign of coefficients. This paper uses a modified version of the Granger causalitytest to test both the existence and the direction of predictability, following the procedure of Chordiaand Swaminathan (2000).

A significant error correction term coefficient indicates that the dependent variable adjusts bymoving toward long-run equilibrium. Short-term effects are conveyed through the remaining inde-pendent variables. Ft Granger-causes PRt (NAVRt) if the sum of the �1i (�2i) coefficients is statisticallydifferent from zero, indicating that Ft can improve the forecast performance of PRt (NAVRt). Grangercausality from flows to fund returns will be evidence in favor of the information contribution hypoth-esis.

Furthermore, if the sum of the �1i is significantly different from the sum of the �2i even aftercontrolling for U.S. investor sentiment effect, this will provide strong evidence of market segmentationover and above investor sentiment. On the contrary, if the sum of the �1i is not significantly differentfrom the sum of the �2i after controlling for U.S. investor sentiment effect, then the interpretation ofinvestor sentiment is favored against the market segmentation effect.

In addition, if the information contribution argument holds, the impact of international equityflows on fund returns may be either short-term or long-term. The long-term explanation (informationrevelation) is based on changes in fundamentals while the short-term explanation (price pressure) isbased on trading noise. Impulse response functions allow a means of testing the information revelationversus price pressure hypotheses. Choleski decomposition is used where the contemporaneous valueof international equity flows has a contemporaneous effect on fund returns (prices and NAVs), butthe contemporaneous value of fund returns does not have a contemporaneous effect on internationalequity flows.

Similarly, PRt (NAVRt) Granger-causes Ft if the sum of the ˇ3i (�3i) coefficients is statistically dif-ferent from zero, indicating that PRt (NAVRt) can improve the forecast performance of Ft. Grangercausality from fund returns to flows will support the feedback trading hypothesis. Moreover, if thesum of the ˇ3i coefficients is significant and different from the sum of the �3i coefficients even whenU.S. investor sentiment effect is controlled, this provides evidence consistent with the presence ofmarket segmentation in the differential price-flow and NAV-flow relationship. Otherwise, if the sumof the ˇ3i coefficients is not significantly different from the sum of the �3i coefficients even when U.S.investor sentiment is controlled, this can be interpreted as favoring the investor sentiment hypothesis.F tests are used to examine the relative predictability as reflected in the magnitude of the �1i and �2icoefficients, and of the ˇ3i and �3i coefficients.

The optimal lag length for the models is determined by AIC and SBC. Newey-West and Whitemethodologies are utilized, as appropriate, to address serial correlation and heteroscedasticity.

4.2.2. Financial crisis effectsTo investigate the effects of the crisis on the relationships between international equity flows and

fund returns, dummy variables are incorporated to compare crisis and non-crisis periods. The proxyfor measuring crisis effects is the Asian crisis period defined as July 1997 through December 1998. TheThai baht devaluation in July of 1997 is used as the trigger for the Asian crisis.11 The dummy variable, D,is set equal to one during the financial crisis period and zero otherwise. The inclusion of both constantdummy variables and interactive dummy variables in Eqs. (1)–(3) yields (4)–(6):

11 Froot et al. (2001) and Nagayasu (2001) concluded that the Asian crisis period was from July 1997 to December 1998, aperiod consistent with the largest fluctuations in exchange rates during the sample period.

870 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i

+L∑

i=1

a1i PDWt−i + d10D +L∑

i=1

d11iD × NAVRt−i +L∑

i=1

d12iD × Ft−i +L∑

i=1

d13iD × USRt−i

+L∑

i=1

d14iD × PDWt−i + e1t (4)

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i

+L∑

i=1

a2i PDWt−1 + d20D +L∑

i=1

d21iD × PRt−i +L∑

i=1

d22iD × Ft−i +L∑

i=1

d23iD × USRt−i

+L∑

i=1

d24iD × PDWt−i + e2t (5)

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i

+L∑

i=1

a3i PDWt−i + d30D +L∑

i=1

d31iD × PRt−i +L∑

i=1

d32iD × NAVRt−i +L∑

i=1

d33iD × USRt−i

+L∑

i=1

d34iD × PDWt−i + e3t (6)

where d10, d20 and d30 are the intercept dummy variables, and d11i, d12i, d13i, d14i, d21i, d22i, d23i, d24i, d31i,d32i, d33i and d34i are the interactive dummy variables. If d10, d20 and d30 are not significantly differentfrom zero, no structural intercept change from the crisis is indicated. Similarly, if the d11i, d12i, d13i,d14i, d21i, d22i, d23i, d24i, d31i, d32i, d33i and d34i coefficients are equal to zero, no structural interactionchange occurred. Increased uncertainties resulting from the existence of a crisis should reduce bothinternational equity flows and fund returns, making the intercept dummy variables negative (d10 < 0,d20 < 0, and d30 < 0). The expected signs of the interactive dummy variable coefficients, however, areindeterminate.

In addition, if the sign of the d12i and d22i coefficients is positive (negative), the crisis effect wouldintensify (weaken) the information contribution effects. And, if the sign of the d31i and d32i coefficientsis positive (negative), the crisis effect would amplify (reduce) the feedback trading effects. F testsare used to examine the relative magnitude of the d12i and d22i coefficients, and of the d31i and d32icoefficients.

4.3. Relationships between mean and volatility linkages

Uncertainty and information flows have been widely recognized as factors contributing to the pric-ing of financial assets. Kim and Singal (1997) and Richards (1996) conclude that returns are associatedwith past returns and with volatility. Because of different information available to U.S. and foreigninvestors, they may not react in the same manner to changes in returns or to changes in volatility(risk).

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 871

To analyze the relative effects of mean connections and volatility connections, volatilities of theincluded variables are measured by the unconditional variance using a conventional rollover process.12

In order to capture both expected and unexpected changes in volatility, a “moving-average” varianceis calculated as the average sum of squares of the monthly returns for the first five months, thendropping the first month data and adding the sixth month data to obtain a new estimate (see, Officer,1973; Merton, 1980; Swanson, 2004). This process is followed until the last month of data is included.The moving (rollover) variance is calculated as:

v2 =∑n

i=1(xi − x̄)2

n − 1

where n is the sample size, and x̄ is the sample mean.Results of the Augmented Dickey–Fuller test and Phillips–Perron test show that the volatility series

of both flows and fund returns (prices and NAVs) are stationary. No cointegrating relationship can existbetween fund returns and volatilities series, and between flows and volatility series.

4.3.1. Mean and volatility relationshipsThe dynamic relations between flows and volatility of returns and between returns and volatil-

ity of flows help better understand the transmission of information across national markets becausevolatility is a key ingredient of risk. Tests of volatility effects on international equity flows and fundreturns are based upon the basic causality estimations of Eqs. (1)–(3) when volatility measures forcore independent variables are added, yielding (7)–(9):

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i

+L∑

i=1

a1i PDWt−i +L∑

i=1

�1iV NAVRt−i +L∑

i=1

ω1iVFt−i +L∑

i=1

�1iVUSRt−i

+L∑

i=1

b1iV PDWt−i + e1t (7)

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i

+L∑

i=1

a2i PDWt−i +L∑

i=1

�2iV PRt−i +L∑

i=1

ω2iVFt−i +L∑

i=1

�2iVUSRt−i

+L∑

i=1

b2iV PDWt−i + e2t (8)

12 Alternatively, volatility effects on fund returns and on flows are tested using the conditional variance generated from aGARCH (1,1) process. However, the inclusion of a GARCH variable introduces the issue of generated regressors. In addition, whenGARCH volatility is substituted in the equation, there is no significant GARCH volatility result for any fund. A likely explanationis that GARCH effects are less informative than moving volatility effects because the monthly observation period is too long tocapture much of the volatility clustering.

872 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i

+L∑

i=1

a3i PDWt−i +L∑

i=1

�3iV PRt−i +L∑

i=1

ω3iV NAVRt−i +L∑

i=1

�3iVUSRt−i

+L∑

i=1

b3iV PDWt−i + e3t (9)

where VPRt is the moving variance of fund price returns at period t; VNAVRt is the moving varianceof fund NAV returns at period t; VFt is the moving variance of international net equity flows at periodt; VUSRt is the moving variance of the U.S. market index returns at period t; VPDWt is the movingvariance of the premiums/discounts of an equally weighted index of U.S. domestic closed-end fundsat period t.

The expectation is that ω1i < 0 and ω2i < 0 as uncertainties created by increasing volatility of flowsnegatively affect fund returns (both prices and NAVs). Similarly, volatility measures of fund returnswould be expected to diminish the flows into local markets, assuming that investors are risk averse(�3i < 0, ω3i < 0). Again, F tests are used to examine the relative predictability as reflected in the mag-nitude of the ω1i and �2i coefficients, and of the �3i and ω3i coefficients. Newey-West and Whitemethodologies are utilized, as appropriate, to address serial correlation and heteroscedasticity.

4.3.2. Mean and volatility financial crisis effectsFinancial crisis effects are investigated by adding both a constant dummy variable and interactive

dummy variables for core independent variables to Eqs. (10)–(12):

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i

+L∑

i=1

a1i PDWt−i +L∑

i=1

�1iV NAVRt−i +L∑

i=1

ω1iVFt−i +L∑

i=1

�1iVUSRt−i

+L∑

i=1

b1iV PDWt−i + d10D +L∑

i=1

d11iD × NAVRt−i +L∑

i=1

d12iD × Ft−i +L∑

i=1

d13iD × USRt−i

+L∑

i=1

d14iD × PDWt−i +L∑

i=1

d15iD × V NAVRt−i +L∑

i=1

d16iD × VFt−i

+L∑

i=1

d17iD × VUSRt−i +L∑

i=1

d18iD × V PDWt−i + e1t (10)

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i

+L∑

i=1

a2i PDWt−i +L∑

i=1

�2iV PRt−i +L∑

i=1

ω2iVFt−i +L∑

i=1

�2iVUSRt−i +L∑

i=1

b2iV PDWt−i

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 873

+ d20D +L∑

i=1

d21iD × PRt−i +L∑

i=1

d22iD × Ft−i +L∑

i=1

d23iD × USRt−i +L∑

i=1

d24iD

× PDWt−i +L∑

i=1

d25iD × VPRt−i +L∑

i=1

d26iD × VFt−i +L∑

i=1

d27iD × VUSRt−i

+L∑

i=1

d28iD × V PDWt−i + e2t (11)

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i

+L∑

i=1

a3i PDWt−i +L∑

i=1

�3iV PRt−i +L∑

i=1

ω3iV NAVRt−i +L∑

i=1

�3iVUSRt−i

+L∑

i=1

b3iV PDWt−i + d30D +L∑

i=1

d31iD × PRt−i +L∑

i=1

d32iD × NAVRt−i +L∑

i=1

d33iD × USRt−i

+L∑

i=1

d34iD × PDWt−i +L∑

i=1

d35iD × V PRt−i +L∑

i=1

d36iD × V NAVRt−i +L∑

i=1

d37iD × VUSRt−i

+L∑

i=1

d38iD × V PDWt−i + e3t (12)

where D is as defined earlier. The anticipated sign for the intercept dummy variables d10, d20 and d30is negative, indicating that the mere existence of the crisis reduces both international equity flowsand fund returns. The expected signs of the interactive dummy variable coefficients, however, areindeterminate. Again, F tests are used to examine the relative magnitude of the d16i and d26i coefficients,and of the d35i and d36i coefficients.

5. Empirical results

5.1. Flows and fund returns relationships

Table 2 sets forth estimation results of Granger causality tests between international equity flowsand fund returns (both prices and NAVs) with premiums/discounts on an index of U.S. domestic closed-end funds as the proxy variable for U.S. investor sentiment. After filtering out these effects, the resultsare essentially the same as the results without control for U.S. investor sentiment effect.13

Tests based on AIC and SBC criteria indicate that a first-order autoregressive model is appropriatefor the analysis. The error correction term for each price return equation is negatively and statisticallysignificant for all funds except Chile, India and India Growth, indicating that price returns adjust toachieve long-run equilibrium. NAV returns, on the other hand, are not responsive to disequilibriumsituations for all emerging market funds except Singapore. For emerging markets, fund share priceswill eventually adjust to NAVs that reflect long-term market fundamentals. Error correction terms fordeveloped market NAV returns are positive and significant for nine of twelve cases, suggesting that, in

13 To conserve space, the VECM results without control for premiums/discounts on an index of U.S. domestic closed-end fundsare available from the author upon request.

874P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 2Estimation results of Granger causality tests controlling for investor sentiment effects.

Entire sample period

Long-run adjustments Flows do not Granger-cause price returns

Flows do not Granger-cause NAV returns

Price returns do notGranger-cause flows

Eq. (1) Eq. (2) Eq. (3) Eq. (1) Eq. (2) Eq. (3)

Null hypothesis H0: ı1 = 0 H0: ı2 = 0 H0: ı3 = 0 H0:∑

�1i = 0 H0:∑

�2i = 0 H0: �11 = �21 H0:∑

ˇ3i = 0ı1 ı2 ı3 �11 �21 F-Value ˇ31

Emerging MarketsLatin American

Brazil −0.0071* 0.0398 −0.0195 0.0383* 0.0048 2.31 0.1158Brazilian equity −0.0412* 0.0462 −0.0537 0.0807* 0.0339* 2.73* 0.1142Chile −00110 0.0374 0.0062 0.0262 0.0030 2.25 0.0951Mexico equity & income −0.0482* 0.0162 −0.0281 0.0224** 0.0165* 1.56 0.0849Mexico −0.0266* 0.0788 −0.1656 0.1028** 0.0574** 2.68* 0.1244

AsianChina −0.0403** 0.0091 −0.0997 0.0401** 0.0168 2.26 0.2811First Philippine −0.1036* 0.0234 −0.0538 0.2310** 0.1268** 8.34** 0.1102Greater China −0.0476* 0.0201 −0.0594 0.2112* 0.1029 8.39** 0.1028India −0.0602 0.1537 −0.0665 0.1500 0.1367 1.75 0.0012India Growth −0.0033 0.0258 0.0780 0.1908 0.1844 1.58 0.0009Indonesia −0.0558** 0.0128 −0.0711 0.2219** 0.1163* 8.36** 0.1117J. F. China −0.0281* 0.0246 −0.0386 0.3531** 0.2706 5.56** 0.1079J. F. India na na na 0.0855 0.0664 1.89 0.0006Korea equity −0.0326** 0.0170 −0.215 0.2090** 0.1205** 7.21** 0.1054Korea −0.0127** −0.0219 −0.1720 0.1206** 0.1002** 1.94 0.1380Malaysia −0.0569** 0.0092 −0.0104 0.1432** 0.0459 8.25** 0.1183Morg Stan India na na na 0.1452 0.1430 1.34 0.1008ROC Taiwan −0.0262* 0.0417 −0.1512 0.2088** 0.1267* 5.52** 0.1187Singapore −0.0436** 0.0456** −0.0278 0.2357* 0.2309* 1.47 0.1658Taiwan −0.0047* 0.0503 −0.1162 0.1624** 0.1316* 2.28 0.1090Templeton China na na na 0.2383** 0.1253 8.45** 0.1086Thai Capital −0.0683** 0.0011 −0.0133 0.2481** 0.1372** 8.42** 0.1214Thai −0.0291** 0.0158 −0.0046 0.2120** 0.1264** 6.76** 0.1475

Developed markets(Asian)Aberdeen Australia −0.0072** 0.0577** −0.0130 0.1536* 0.1430* 1.72 0.1330**Japan equity −0.0298* 0.0326* −0.0341 0.2470** 0.2538** 1.59 0.1316**Japan small cap −0.0227* 0.0275** −0.0108 0.2773** 0.3129** 2.33 0.2037**

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894875

Developed markets (European)Austria −0.0236* 0.0469* 0.0528 0.1975 0.1137* 6.70** 0.1393France Growth −0.0122** 0.0240* −0.0461 0.1448** 0.1486** 1.40 0.2133*Germany −0.0619** 0.0088* −0.0126 0.1280** 0.1545** 2.27 0.1600**Italy −0.0763* 0.0035 −0.0649 0.1691* 0.1206 3.79** 0.1116**New Germany −0.1229** 0.0170** −0.0223 0.1002** 0.1229** 2.23 0.1349**New Ireland −0.0843* 0.0280* −0.1269 0.1488* 0.1365** 1.73 0.1130**Portugal −0.0066* 0.0164 −0.1023 0.0149 0.0128 1.33 0.1041Spain −0.0535* 0.0429* −0.0602 0.1576** 0.1363** 2.21 0.1367**Swiss Helvetia −0.1047** 0.0105 −0.1077 0.1411 0.1252** 1.81 0.1016

Entire sample period

NAV returns do notGranger-cause flows

U.S market impacts/U.S. investor sentiment impacts R2

Eq. (3) Eq. (1) Eq. (2) Eq. (3) Eq. (1)/Eq. (2)/Eq. (3)

Null hypothesis H0:∑

�3i = 0 H0: �31 = �31 H0:∑

�1i = 0/∑

a1i = 0 H0:∑

�2i = 0/∑

a2i = 0 H0:∑

�3i = 0/∑

a3i = 0�31 F-Value �11/a11 �21/a21 �31/a31

Emerging MarketsLatin American

Brazil 0.2057** 7.22** 0.3031**/0.1674 0.1648**/0.0936 −0.0246*/0.0803 0.1108/0.1110/0.1115Brazilian equity 0.1464** 2.29 0.2548**/0.1653 0.2179**/0.0795 −0.0604**/0.0821 0.0939/0.0952/0.0978Chile 0.1260 2.05 0.3276**/0.1726* 0.1856**/0.1008 0.0218/0.0995 0.1041/0.1067/0.1082Mexico equity & income 0.1226* 2.35 0.3522**/0.2035* 0.3008**/0.1002 −0.0337*/0.0986 0.1172/0.1176/0.1099Mexico 0.1936** 4.42** 0.3534**/0.2047* 0.2590**/0.0994 −0.0580*/0.1005 0.0949/0.0938/0.0983

AsianChina 0.3455* 4.22** 0.2797**/0.1865* 0.2692**/0.0955 −0.0503/0.0771 0.0958/0.0950/0.0931First Philippine 0.1618* 3.95* 0.2366**/0.1724* 0.1801*/0.0739 −0.0216*/0.0806 0.1026/0.1048/0.1053Greater China 0.1317* 2.01 0.2714**/0.2001** 0.2058**/0.1022 −0.0458/0.0923 0.1093/0.1088/0.1094India 0.0023 1.01 0.3059**/0.1293 0.2651**/0.0737 0.0249/0.0702 0.0799/0.0805/0.0793India Growth 0.0019 0.99 0.3654**/0.1278 0.3136**/0.0698 −0.0391/0.0924 0.0812/0.0814/0.0811Indonesia 0.1867* 5.26** 0.5198**/0.2068** 0.3990/0.1034 −0.0246**/0.0916 0.0840/0.0832/0.0837J. F. China 0.1351* 1.93 0.2922**/0.1804 0.2107**/0.1106 −0.0202/0.0920 0.0843/0.0837/0.0841J. F. India 0.0018 0.94 0.2873**/0.1176 0.2538**/0.1027 0.0138/0.0879 0.0796/0.0785/0.0792Korea equity 0.1526** 2.79* 0.3660**/0.2236** 0.1705**/0.1148 −0.0145*/0.0838 0.1294/0.1289/0.1288Korea 0.1913* 3.98** 0.3055**/0.2308** 0.2448**/0.1150 −0.0381**/0.0906 0.1255/0.1247/0.1251Malaysia 0.1864 4.39** 0.3728**/0.2315** 0.1544**/0.1131 −0.0130**/0.1005 0.1044/0.1034/0.1029Morg Stan India 0.1013 0.96 0.2551*/0.1819 0.1667*/0.0778 0.0055/0.0706 0.0783/0.0790/0.0693Taiwan, ROC 0.1513** 2.31 0.3028**/0.1774 0.1563**/0.0805 −0.0137/0.0883 0.1208/0.1211/0.1204Singapore 0.1784** 1.56 0.4855**/0.1816* 0.4110**/0.0903 −0.0106/0.0793 0.1310/0.1319/0.1283Taiwan 0.1311** 1.73 0.3770**/0.1645 0.1647**/0.0861 −0.0103/0.0625 0.1238/0.1235/0.1114Templeton China 0.1729 4.21** 0.2511**/0.1953** 0.1506**/0.0945 0.0218/0.0861 0.0716/0.0752/0.0694

876P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 2 (Continued )

Entire sample period

NAV returns do notGranger-cause flows

U.S market impacts/U.S. investor sentiment impacts R2

Eq. (3) Eq. (1) Eq. (2) Eq. (3) Eq. (1)/Eq. (2)/Eq. (3)

Null hypothesis H0:∑

�3i = 0 H0: �31 = �31 H0:∑

�1i = 0/∑

a1i = 0 H0:∑

�2i = 0/∑

a2i = 0 H0:∑

�3i = 0/∑

a3i = 0�31 F-Value �11/a11 �21/a21 �31/a31

Thai Capital 0.1885** 4.28** 0.4087**/0.2017** 0.3165**/0.1038 −0.0370**/0.0904 0.1017/0.01033/0.0976Thai 0.1532** 4.00** 0.4865**/0.2052** 0.4556**/0.1022 −0.0406**/0.0881 0.1046/0.1039/0.0965

Developed markets(Asian)Aberdeen Australia 0.1800** 2.78* 0.2716**/0.2030** 0.1642**/0.0776 −0.0506**/0.0703 0.1290/0.1289/0.1171Japan equity 0.1415** 1.52 0.3220**/0.1216 0.3003**/0.0732 −0.1335**/0.0716 0.1278/0.1271/0.1062Japan small cap 0.2268** 1.78 0.2638**/0.1223 0.1174**/0.0725 −0.1402**/0.0720 0.1218/0.1237/0.1205

Developed markets (European)Austria 0.1882 2.81* 0.3022**/0.1781* 0.2130**/0.0874 0.0192/0.0736 0.1210/0.1209/0.1193France Growth 0.2343* 1.68 0.1418**/0.1478 0.1077**/0.0833 −0.0227/0.0782 0.1272/0.1264/0.1112Germany 0.1786** 1.67 0.2909**/0.0935 0.2857**/0.0761 −0.0301/0.0755 0.1293/0.1285/0.1246Italy 0.1694** 4.15** 0.4871**/0.1823* 0.2303**/0.0739 0.0101/0.0724 0.1017/0.1023/0.0949New Germany 0.1437** 1.47 0.2866**/0.1047 0.2846**/0.0755 −0.0216/0.0808 0.1311/0.1316/0.1268New Ireland 0.1466** 2.33 0.2926**/0.1906* 0.2762**/0.0801 −0.0131/0.0773 0.1015/0.1019/0.0940Portugal 0.1053 1.02 0.2719**/0.1801* 0.2677**/0.0816 −0.0128/0.0782 0.1039/0.1056/0.0981Spain 0.1679** 2.20 0.2908**/0.1955** 0.2858**/0.0753 −0.0120/0.0729 0.1038/0.1047/0.0970Swiss Helvetia 0.1044 1.08 0.2645**/0.1971** 0.2613**/0.0725 −0.0215/0.0704 0.0962/0.0968/0.0833

Test results for Granger causality are based on the following equations:

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i +L∑

i=1

a1i PDWt−i + e1t

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i +L∑

i=1

a2i PDWt−i + e2t

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i +L∑

i=1

a3i PDWt−i + e3t

where PRt is the price returns on the fund at period t; ECTt−1 is the error-correction term at period t − 1 derived from the long-term cointegration relationship; NAVRt is the returns of thefund NAVs at period t; Ft is international net equity flows at period t; USRt is the returns of the U.S. market index at period t; PDWt is the premiums/discounts on an equally weighted indexof U.S. domestic closed-end funds at period t. The test statistics are corrected for heteroskedasticity and serial correlation. Significance at the 5% level is denoted by **, and significance atthe 10% level is denoted by *.

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 877

the more integrated developed markets, fund share prices and NAVs appear to share the momentumof achieving equilibrium. However, there is no significant coefficient for the error correction term forthe flow equation for any funds. And, the addition of the flow measure does not change the price andNAV relationship.

Based on Eqs. (1) and (2), flows Granger-cause fund NAV (price) returns for 12 (18) emerging marketfunds and for 10 (9) developed market funds.14 Consistent with Froot and Ramadorai (2008), theevidence strongly supports information contribution argument that cross-border equity flows reflectnew market information and positively affect both fund price and NAV returns.

The effects of flows on price returns differ from the corresponding effects on NAV returns for allemerging market funds as anticipated, and significant effects are found in half emerging market funds(12). Developed market funds reveal different results. The coefficient of flows for price returns isappreciably different from that of flows for NAV returns only in two developed market funds (Austriaand Italy). These variables previously significant in the results without control for investor sentimenteffects still remain significant with essentially the same magnitudes after controlling for possibleU.S. investor sentiment effects. Thus, such differences between price-flow and NAV-flow relationshipsappear to be consistent with market segmentation effects and inconsistent with investor sentimentarguments. Asymmetric information of local and U.S. investors generated by market segmentationleads to divergent pricing of assets in emerging markets. For developed markets, the fact that bothprice and NAV returns respond similarly to the new market information obtained in internationalequity flows in most cases reveals the closer integration between U.S. market and developed markets.

A further analysis of the information contribution is required to identify whether or not the effectsare long-term or short-term. Using impulse response functions to test this dynamic interrelationshipbetween flows and fund returns, in all cases, fund return responses to flow innovations decay slowly,with the effects being almost complete after about five months. This long-term response of fund returnsto flow innovations should be consistent with the information revelation hypothesis.15

Based on Eq. (3), NAV returns Granger-cause flows for 16 emerging market funds, and none offlows are significantly affected by price returns. The return impacts are somewhat more importantfor developed market funds. Both price and NAV returns Granger-cause flows for all cases except forAustria, Portugal, and Swiss Helvetia. The finding of Granger causality from fund returns to flows isconsistent with the feedback trading argument that U.S. institutional investors make their tradingdecisions based on local market returns.

The effects of NAV returns on flows differ appreciably from the effects of price returns on flowsin 11 emerging market funds and only 3 developed market funds (Australia, Austria, and Italy). Sincethe result of all the variables is qualitatively the same as the result without control for U.S. investorsentiment effects, this presents evidence in support for the market segmentation hypothesis. Thebasic explanation for the major variation in the empirical result between emerging market and devel-oped market funds is that the economic and market fundamentals of local markets may be reflectedmore precisely by fund NAVs than by fund prices in emerging market funds. U.S. institutional partici-pants who are value-chasing will make investment decisions in accordance with market fundamentalsrevealed by fund NAVs. On the other hand, in the more closely integrated developed markets, price andNAV returns appear to be sharing quick dissemination of fundamental information across markets.

The U.S. market effects on fund price and NAV returns are positive and significant for all fundsexcept Indonesia where no short-term relationship between the U.S. and NAV returns is found. Thereare significant effects on flows from U.S. returns for 11 emerging market funds, plus Aberdeen Australia,Japan Equity and Japan Small Cap, and the signs are negative, as expected. Premiums/discounts of anindex of U.S. domestic closed-end funds (PDW) show a significant positive correlation with fund pricereturns in 14 emerging market funds and 7 developed market funds, and no significant correlationwith both NAV returns and flows. This provides some evidence that U.S. investor sentiment affectscountry fund price returns, consistent with the findings of Hardouvelis et al. (1994), and Bodurtha etal. (1995).

14 For Latin America, insignificant effects for Chile may be reflecting stronger restrictions on foreign equity investments, therebyreducing the impact of foreign capital inflows on local markets.

15 Detailed results of the impulse response function tests are available from the author.

878 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

The fact that 27 of the 35 significant flow coefficients are positive generally supports the infor-mation contribution arguments that increasing international equity flows, which incorporate newmarket information, push prices up in both local markets and corresponding country funds. Feed-back trading arguments are also supported in that the trading activities of institutional investors areclosely associated with country fund returns. Furthermore, the analysis on differential sensitivitiesbetween price-flow and NAV-flow relationships provides strong evidence in favor of market segmen-tation arguments. Noteworthy are the differences in the results for emerging markets relative to thosefor developed markets. Specifically, the effects of flows on price returns differ significantly from thecorresponding effects on NAV returns for emerging market funds, but not for developed market funds.For emerging markets, only NAV returns Granger-cause flows, whereas for developed market funds,both price and NAV returns Granger-cause flows. These differential results would be interpreted asfavoring market segmentation arguments, because investor sentiment effects should be documenteduniformly across all funds and not in selected emerging market funds.

5.2. Effects of Asian crisis

Results slightly differ when consideration of the Asian crisis effect is incorporated as reported inTable 3. The intercept dummy variables for fund price returns are significant and negative in 14 of the18 emerging market Asian funds, but not significant for any other funds. The finding shows strongevidence of structural changes in Asian fund performance, reducing fund price returns. Only seven ofthe NAV return intercept dummy variables are significantly negative, suggesting that fund price returnsare affected more strongly than NAV returns. In addition, the intercept dummy variables for flows aresignificant (and again negative) for nine emerging market Asian funds and two Japan-invested funds.As expected, crisis effects are strong for Asian countries.

As described in Eqs. (4) and (5), the Asian crisis positively affected Brazilian Equity fund (for NAVreturns) and Mexico Equity & Income fund (for price returns) for Latin American countries. All Asiancountries except Singapore (negatively) and four India-invested funds were significantly and positivelyaffected by the crisis. The results indicate that the impact of flows on both fund price and NAV returns isamplified in Asian countries but is weakened in Singapore during the crisis period, clearly supportingamplified influence of foreign investors on local markets during the Asian crisis. Most affected wasThailand. Significant crisis effects on Thailand are not unexpected because the crisis began in Thailand.The unique role of Singapore as a financial center for international portfolio management may partiallymitigate the impact of the crisis. There is no evidence of crisis effects for the interaction dummy variableof flows for developed market funds, the anticipated result.

Noteworthy is the fact that the magnitudes of significant flow interactive dummies for price returnsare significantly different from those of significant flow interactive dummies for NAV returns in mostemerging market Asian funds, but not significant for any other funds. A possible explanation is that insegmented markets, asymmetric information between U.S. and local investors would lead to differentmarket behavior toward increased risk.

Based on Eq. (6), the Asian crisis affected Asian funds most. For emerging Asian markets, the sig-nificant interaction effects are in 14 funds, and these effects are negative, indicating that the effects offund returns on flows were weakened during the crisis period. None of the effects are positive and sig-nificant as hypothesized. One possible explanation will be that local market returns would be reducedand become less attractive during the crisis period because of ongoing risk in these markets. The strongeffects in Malaysia may be reflecting Malaysia’s imposed constraints on foreign capital flows duringthe crisis period. The Asian crisis effects in developed markets were minimal.

The magnitudes of NAV return interactive dummies for flows are appreciably different from thoseof price return interactive dummies for flows in several emerging market Asia funds, suggestingthat crisis effects amplified asymmetric information between local and U.S. investors in emergingmarkets.

As can be seen, the crisis intensifies the influence of U.S. institutional investors on local mar-kets, thereby strengthening information contribution effects. On the contrary, feedback trading of U.S.institutional investors is diminished since potential higher crisis-related risk during the crisis periodwould reduce local market returns and drive foreign capital outflows. Importantly, these uncertainties

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894879

Table 3Estimation results of Granger causality tests incorporating Asian crisis effects.

Asian crisis period

Structural intercept changes Flows do not Granger-cause price returns

Flows do not Granger-cause NAV returns

Price returns donot Granger-cause flows

Eq. (4) Eq. (5) Eq. (6) Eq. (4) Eq. (5) Eq. (6)

Null hypothesis H0: d10 = 0 H0: d20 = 0 H0: d30 = 0 H0:∑

d12i = 0 H0:∑

d22i = 0 H0: d121 = d221 H0:∑

d31i = 0Fund name d10 d20 d30 d121 d221 F-Value d311

Emerging marketsLatin American

Brazil −0.0271 −0.0073 −0.0094 0.1011 0.0693 2.10 −0.0735Brazilian equity −0.0242 0.0405 0.0328 0.1158 0.0868** 2.03 −0.0804**Chile −0.0612 −0.0227 −0.0444 0.0901 0.0817 1.03 −0.1025Mexico equity & income −0.0278 −0.0168 −0.2115 0.1079* 0.0835 1.95 −0.0769Mexico −0.0316 −0.0065 −0.1903 0.0702 0.0633 0.97 −0.1063

AsianChina −0.0480* −0.0388* −0.0384* 0.1740* 0.1282* 2.70* −0.1248*First Philippine −0.0068* −0.0045* −0.0076* 0.1532** 0.0836** 4.58** −0.0671Greater China −0.0873** −0.0558* −0.0327* 0.1831** 0.1167* 4.33** −0.1122India −0.0092* 0.0023 −0.0940 −0.1100 −0.0885 1.84 −0.0777India Growth −0.0112* 0.0017 −0.0601 −0.0537 −0.0300 1.93 −0.0618Indonesia −0.0536** −0.0405** −0.0133* 0.1866** 0.1345** 3.89* −0.0625J. F. China −0.0255 0.0022 −0.0306* 0.1842* 0.1357 3.17* −0.1019J. F. India −0.0118* −0.0035 −0.0025 −0.0228 −0.0158 0.98 −0.0295Korea equity −0.0165** −0.0076 0.1958 0.2308** 0.1769* 4.00** −0.0829**Korea −0.0140* −0.0018 0.2003 0.1693* 0.1240* 2.68* −0.1003**Malaysia −0.0810** −0.0579** −0.0682** 0.1867** 0.1248 4.04** −0.1401**Morg Stan India -0.0101* 0.0002 −0.0015 −0.0061 −0.0042 0.63 −0.0021ROC Taiwan −0.0292 −0.0266 −0.0399 0.1783** 0.1380* 2.37 −0.1227Singapore −0.0255 −0.0170 0.0203 −0.0376 −0.0355* 0.64 −0.0749Taiwan −0.0247 −0.0125 −0.0307 0.1702* 0.1231** 3.12* −0.1290*Templeton China −0.0443* −0.0207 −0.0275* 0.1863** 0.1506 2.29 −0.1043Thai Capital −0.0148** −0.0104** −0.0068** 0.2250** 0.1462** 5.56** −0.1235Thai −0.0216** −0.0174* −0.0180** 0.2053** 0.1369** 4.52** −0.1202

880P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 3 (Continued )

Asian crisis period

Structural intercept changes Flows do not Granger-causeprice returns

Flows do not Granger-causeNAV returns

Price returns donot Granger-causeflows

Eq. (4) Eq. (5) Eq. (6) Eq. (4) Eq. (5) Eq. (6)Null hypothesis H0: d10 = 0 H0: d20 = 0 H0: d30 = 0 H0:

∑d12i = 0 H0:

∑d22i = 0 H0: d121 = d221 H0:

∑d31i = 0

Fund name d10 d20 d30 d121 d221 F-Value d311

Developed markets (Asian)Aberdeen Australia −0.0362 −0.0154 0.0532 0.0967 0.0914 0.81 −0.1365Japan equity −0.0140 −0.0092 −0.0184* −0.1246 −0.1372 1.27 −0.1368Japan small cap −0.0127 −0.0116 −0.0195* −0.1103 −0.1299 1.80 −0.1382

Developed markets (European)Austria 0.0213 −0.0024 0.0005 0.0061 −0.0054 1.24 −0.0263France Growth 0.0188 0.0105 −0.0068 0.1153 0.1185 0.70 −0.1671Germany 0.0065 −0.0012 −0.0177 0.0168 0.0202 0.72 −0.1440Italy 0.0214 0.0201 −0.0142 0.1382 0.1194 1.73 −0.1109New Germany −0.0002 −0.0034 −0.0161 0.1005 0.1096 1.12 −0.1316New Ireland 0.0031 0.0092 −0.0418 0.0981 0.0726 1.97 −0.1343Portugal −0.0250 −0.0182 0.0387 0.1709 0.1671 0.75 −0.1200Spain 0.0112 0.0014 −0.0293 0.1342 0.1067 2.00 −0.0788Swiss Helvetia 0.0009 −0.0011 −0.0534 0.0226 −0.0053 2.01 −0.1469

Asian crisis period

NAV returns do not Granger-cause flows U.S market impacts/U.S. investor sentiment impacts R2

Eq. (6) Eq. (4) Eq. (5) Eq. (6) Eq. (4)/Eq. (5)/Eq. (6)

Null hypothesis H0:∑

d32i = 0 H0: d311 = d321 H0:∑

d13i = 0/∑

d14i = 0 H0:∑

d23i = 0/∑

d24i = 0 H0:∑

d33i = 0/∑

d34i = 0Fund name d321 F-Value d131/d141 d231/d241 d331/d341

Emerging marketsLatin American

Brazil −0.0800 0.97 −0.3046**/−0.1036 −0.1752*/−0.0703 −0.3111/−0.0697 0.1119/0.1112/0.1123Brazilian equity −0.1215* 2.42 −0.2702*/−0.0972 −0.1384/−0.0784 −0.1080/−0.0705 0.1014/0.1016/0.1020Chile −0.1149 1.27 −0.1006/−0.0915 −0.0269/−0.0776 0.0053/0.0036 0.1063/0.1075/0.1089Mexico equity & income −0.1073 2.08 −0.1904/−0.0903 −0.0387*/−0.0802 0.0056/0.0225 0.1218/0.1226/0.1154Mexico −0.1222* 1.47 −0.1641/−0.0957 0.0161/0.0806 −0.1138/−0.0307 0.1009/0.1006/0.1018

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894881

AsianChina −0.1816* 4.04** −0.1811**/−0.1124 −0.1288**/−0.0776 −0.1655/−0.0723 0.0984/0.0968/0.0947First Philippine −0.1237* 4.03** −0.2358/−0.1053 −0.1945/−0.0684 −0.1730*/−0.0605 0.1039/0.1063/0.1066Greater China −0.1520* 2.37 −0.2780**/−0.1075 −0.2347**/−0.0658 −0.0701/−0.0621 0.1105/0.1102/0.1109India −0.1043 1.95 −0.1392/−0.0934 −0.1072/−0.0705 −0.0544/−0.0687 0.0816/0.0824/0.0807India Growth −0.0501 1.23 0.0646/−0.0755 0.0590/0.0601 0.0036/−0.0589 0.0826/0.0827/0.0824Indonesia −0.1383* 5.30** −0.2955*/−0.0882 −0.1616/−0.0710 −0.2042*/−0.0668 0.0850/0.0842/0.0851J. F. China −0.1545** 3.89* −0.3058/−0.0907 −0.2185/−0.0815 −0.1040/−0.0793 0.0852/0.0849/0.0855J. F. India −0.0216 1.02 0.0479/−0.0956 0.0502/0.0822 0.0029/−0.0804 0.0809/0.0806/0.0808Korea equity −0.1457** 4.15** −0.3012*/−0.1084 −0.2149**/−0.0734 −0.3387**/−0.0712 0.1313/0.1317/0.1302Korea −0.1491** 3.22* −0.2834/−0.0795 −0.2283/−0.0626 −0.2322**/−0.0618 0.1278/0.1274/0.1277Malaysia −0.2199** 5.54** −0.3256**/−0.0770 −0.1984/−0.0645 −0.0721*/−0.0633 0.1081/0.1075/0.1068Morg Stan India −0.0085 0.98 −0.1393/−0.0729 −0.1280/−0.0658 −0.0021/−0.0632 0.0799/0.0802/0.0716Taiwan, ROC −0.1466* 1.93 −0.3179/−0.1131 −0.1922/−0.0846 0.0963/0.0817 0.1222/0.1215/0.1210Singapore −0.0928** 1.65 −0.2612*/−0.1023 −0.1678*/−0.0905 −0.0231/−0.0885 0.1329/0.1334/0.1312Taiwan −0.1538* 1.94 −0.1843/−0.0905 −01247/−0.0842 0.0708/−0.0809 0.1274/0.1266/0.1142Templeton China −0.2440* 5.08** −0.2730*/−0.1008 −0.1112**/−0.0957 −0.0237/−0.904 0.0747/0.0789/0.0715Thai Capital −0.1967** 5.17** −0.2993**/−0.1254 −0.2277/−0.1004 −0.0752**/−0.0774 0.1038/0.1046/0.1005Thai −0.1891** 5.01** −0.2927**/−0.1203 −0.2616*/−0.1023 −0.0835*/−0.0822 0.1073/0.1062/0.0988

Developed markets (Asian)Aberdeen Australia −0.1498 1.35 −0.2272/−0.1008 −0.2055/−0.0884 −0.1392/−0.0826 0.1309/0.1311/0.1193Japan equity −0.1590 1.88 −0.3200/−0.0937 −0.2401/−0.0868 −0.1174/−0.0793 0.1294/0.1283/0.1089Japan small cap −0.1606 1.89 −0.3006/−0.0954 −0.2367/−0.0872 −0.1250/−0.0784 0.1244/0.1257/0.1216

Developed markets (European)Austria −0.0308 0.71 −0.3862/−0.1021 −0.2678/−0.0992 −0.1573/−0.0835 0.1245/0.1236/0.1218France Growth −0.1842 1.60 −0.2055/−0.0989 −0.1440/−0.0847 −0.1862/−0.0795 0.1287/0.1279/0.1133Germany −0.1503 0.96 0.1828/0.1046 0.1636/−0.0855 0.0646/−0.0802 0.1310/0.1299/0.1258Italy −0.1410 2.05 −0.2447/−0.1105 −0.2177/−0.0906 0.0400/0.0774 0.1052/0.1054/0.0965New Germany −0.1418 1.14 −0.2783/−0.0910 −0.2233/−0.0861 0.0449/0.0750 0.1323/0.1329/0.1280New Ireland −0.1574 1.90 −0.2647/−0.0965 −0.1995/−0.0788 0.0088/0.0689 0.1026/0.1029/0.0974Portugal −0.1615 2.43 0.1833/−0.0989 0.0891/−0.0856 0.0079/0.0577 0.1079/0.1083/0.1005

882P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 3 (Continued )

Asian crisis period

NAV returns do not Granger-cause flows U.S market impacts/U.S. investor sentiment impacts R2

Eq. (6) Eq. (4) Eq. (5) Eq. (6) Eq. (4)/Eq. (5)/Eq. (6)Null hypothesis H0:

∑d32i = 0 H0: d311 = d321 H0:

∑d13i = 0/

∑d14i = 0 H0:

∑d23i = 0/

∑d24i = 0 H0:

∑d33i = 0/

∑d34i = 0

Fund name d321 F-Value d131/d141 d231/d241 d331/d341

Spain −0.1144 2.25 −0.2938/−0.0904 −0.1847/−0.0872 −0.0358/−0.0825 0.1066/0.1069/0.0987Swiss Helvetia −0.1612 1.40 −0.2046/−0.0878 −0.1410/−0.0806 −0.0072/−0.0933 0.0992/0.1006/0.0843

Test results for Granger causality incorporating Asian crisis effects are based on the following equations:

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i +L∑

i=1

a1i PDWt−i + d10D +L∑

i=1

d11iD × NAVRt−i +L∑

i=1

d12iD × Ft−i +L∑

i=1

d13iD × USRt−i

+L∑

i=1

d14iD × PDWt−i + e1t

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i +L∑

i=1

a2i PDWt−1 + d20D +L∑

i=1

d21iD × PRt−i +L∑

i=1

d22iD × Ft−i +L∑

i=1

d23iD × USRt−i

+L∑

i=1

d24iD × PDWt−i + e2t

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i +L∑

i=1

a3i PDWt−i + d30D +L∑

i=1

d31iD × PRt−i +L∑

i=1

d32iD × NAVRt−i +L∑

i=1

d33iD × USRt−i

+L∑

i=1

d34iD × PDWt−i + e3t

A dummy variable, D, is set equal to one during the Asian crisis period, otherwise its value is zero. The crisis period covers July 1997–December 1998. d10, d20 and d30 are the interceptdummy variables; d11i , d12i , d13i , d14i , d21i , d22i , d23i , d24i, d31i , d32i , d33i , and d34i are the interactive dummy variables. The test statistics are corrected for heteroskedasticity and serialcorrelation. Significance at the 5% level is denoted by **, and significance at the 10% level is denoted by *.

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 883

caused by the crisis may amplify the existence of asymmetric market information across local and U.S.investors.

The U.S. market influences on both price and NAV returns were generally negative and significantduring the crisis. Thus, the effects of the U.S. market on both fund price and NAV returns are diminishedduring the crisis period. On the other hand, the interactive U.S. market returns dummy variable forflows is significant and negative in a few emerging market Asian funds, but not significant for anydeveloped market funds, indicating a relatively strengthened role of the U.S. market on flows duringthe crisis period.

5.3. Mean and volatility linkages

Volatility measures are added to the models shown in Table 2 and the results are reported in Table 4.Addition of volatility measures leaves the original variables basically unchanged.

As set forth in Eqs. (7) and (8), there are no significant volatility effects in Latin American region. ForAsia, flow volatility effects are significant for 10 of 18 funds (10 cases on price returns and 3 cases on NAVreturns), where the signs are negative as hypothesized. The negative parameters of the flow volatilityeffects imply that fund returns are driven down by increasing uncertainty of flows. In addition, thereare significant flow volatility effects on fund returns for developed market funds: significant volatilitycoefficients on both price and NAV returns for seven funds, and significant volatility coefficients onlyon price returns for Aberdeen Australia, Austria, Italy, and Swiss Helvetia. That is, there exist strongcross-border information linkages between flow volatility and fund returns.

The effects of flow volatility on price returns are significantly different from the correspondingeffects on NAV returns for most emerging market funds and a few developed market funds. Equityinflows/outflows by foreign institutional investors may increase fluctuations of local markets. Thus,differences between emerging market and developed market funds may be reflecting the degrees ofindirect and/or direct investment restrictions on foreign investments.

Based on Eq. (9), volatility of fund NAV (price) returns significantly affects the flows for 20 (7)funds, where the effect is significantly negative as hypothesized. The evidence could be explainedthat increasing risk of local markets will drive the international capital flows away from local mar-kets.

Significant coefficient difference between price return volatility and NAV return volatility for flowsis for seven emerging market Asian funds and three developed market funds only. In general, volatilityof fund returns on flows has more significant impacts on Asian markets and European markets thanon Latin American markets.

U.S. market volatility does not significantly impact any of fund price returns, and significantlyimpacts NAV returns only for developed markets (except Portugal), where the signs are negative. Thisunexpected result for developed markets that price returns are not affected by U.S. volatility while NAVreturns are affected may be indicating the degree of integration of the U.S. and developed markets.U.S. volatility effects on price returns may be contemporaneous and not identifiable with monthlyobservation periods. All significant U.S. volatility coefficients for flows are negative (only 11 of 35 fundsare significant). It appears that uncertainty in the U.S. domestic market should cause U.S. investors toreduce international investments in other countries.

Tests results when the impacts of the Asian crisis are included are reported in Table 5. These coef-ficients are similar to those reports in Table 3 where the Asian crisis effects are incorporated into themodel but volatility measures are not. The most significant difference is that when volatility is added,the intercept dummy variable in the price return equation is negatively significant for all Asian fundsexcept Taiwan. There is no virtually intercept change in the NAV return and flow models.

As described in Eqs. (10) and (11), the interaction changes for the flow volatility show negativesignificance only for the Asian funds (emerging and developed). For the Asian markets, flow volatilityinteractive dummies are significant for both the price return and NAV return equations in roughlyhalf the funds. The negative signs of the flow volatility interactive dummy variables suggest that theimpact of the flow fluctuation on fund returns is amplified for the Asian funds during the crisis period.In addition, the magnitudes of significant flow volatility interactive dummies for price returns aredifferent from those of significant flow volatility interactive dummies for NAV returns, and significant

884P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 4Estimation results of Granger causality tests incorporating volatility effects.

Entire sample period

Long-run adjustments Flow volatility doesnot Granger-causeprice returns

Flow volatility doesnot Granger-causeNAV returns

Price returnvolatility does notGranger-causeflows

Eq. (7) Eq. (8) Eq. (9) Eq. (7) Eq. (8) Eq. (9)

Null hypothesis H0: ı1 = 0 H0: ı2 = 0 H0: ı3 = 0 H0:∑

ω1i = 0 H0:∑

ω2i = 0 H0: ω11 = �21 H0:∑

�3i = 0Fund name ı1 ı2 ı3 ω11 ω21 F-Value �31

Emerging marketsLatin American

Brazil −0.0068* 0.0405 −0.0166 −0.1732 −0.1448 1.98 −0.1594Brazilian equity −0.04003* 0.0472 −0.0520 −0.2071 −0.1417 4.22** −0.1340Chile −001107 0.0388 0.0045 0.0857 0.0665 1.43 −0.1517Mexico equity & income −0.0476** 0.0180 −0.0267 −0.1730 −0.1013 4.82** −0.2678Mexico −0.0256* 0.0778 −0.1652 −0.2109 −0.1336 5.46** −0.1452

AsianChina −0.0416** 0.0079 −0.0981 −0.2397* −0.1788 4.04* −0.1560

First Philippine −0.1058* 0.0223 −0.0517 −0.3851** −0.2715 9.37** −0.1738Greater China −0.0482* 0.0206 −0.0577 −0.2096* −0.1224 6.03** −0.1829India −0.0577** 0.1504 −0.0654 0.0042 0.0077 0.98 0.0034India Growth −0.0043* 0.0267 0.0765 0.0101 0.0226 1.30 −0.0047Indonesia −0.0565** 0.0185 −0.0746 −0.2560* −0.1428 9.35** −0.2023J. F. China −0.0289* 0.0218 −0.0349 −0.2178 −0.1586 3.96* −0.1548J. F. India na na na 0.0040 0.0089 1.01 0.0050Korea equity −0.0345** 0.0164 −0.2062 −0.3085** −0.2463* 4.08* −0.1245**Korea −0.0140** −0.0217 −0.1673 −0.3616** −0.2880* 5.04** −0.1871Malaysia −0.0569** 0.0046 −0.0058 −1.2644** −1.1672 7.25** −1.0654Morg Stan India na na na 0.0096 0.0035 1.22 0.0104ROC Taiwan −0.0276* 0.0404 −0.1533 −0.2124 −0.1615 3.45* −0.1332Singapore −0.0448** 0.0457** −0.0280 −0.2067* −0.1814** 1.74 −0.1047*Taiwan −0.0058* 0.0546 −0.1181 −0.2469 −0.1853 4.06* −0.1405Templeton China na na na −0.2365 −0.1747 4.07* −0.1221Thai Capital −0.0643** 0.0001 −0.0114 −1.2377** −1.0908 11.61** −0.2308Thai −0.0271** 0.0162 −0.0052 −1.2345** −1.1332 7.34** −0.2413

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894885

Developed markets (Asian)Aberdeen Australia −0.0066** 0.0563** −0.0116 −0.2174** −0.1810 2.29 −0.2353Japan equity −0.0285* 0.0318* −0.0314 −0.2033* −0.1622* 2.26 −0.2626**Japan small cap −0.0230* 0.0285** −0.0103 −0.1962* −0.1705* 1.76 −0.1605*

Developed markets (European)Austria −0.0221* 0.0477* 0.0556 −0.2413* −0.2001 2.27 −0.1315France Growth −0.0115** 0.0210* −0.0442 −0.2376* −0.1955* 2.32 −0.2009*Germany −0.0612** 0.0083* −0.0133 −0.2066** −0.1802** 1.80 −0.2272*Italy −0.0778** 0.0033 −0.0665 −0.2405** −0.1674 5.03** −0.1109New Germany −0.1243** 0.0169** −0.0237 −0.2381** −0.1979** 2.15 −0.1201*New Ireland −0.0814** 0.0273* −0.1280 −0.2103* −0.1725* 2.34 −0.0517Portugal −0.0091* 0.0166* −0.1024 −0.1824 −0.1230 3.97* −0.1368Spain −0.0517* 0.0428 −0.0609 −0.2407** −0.2021* 2.41* −0.1744Swiss Helvetia −0.1022** 0.0105 −0.1075 −0.1929* −0.1284 4.10* −0.1540

Entire sample period

NAV return volatilitydoes not Granger-causeflows

U.S market volatility impacts/U.S. investor sentiment volatility impacts R2

Eq. (9) Eq. (7) Eq. (8) Eq. (9) Eq. (7)/Eq. (8)/Eq. (9)

Null hypothesis H0:∑

ω3i = 0 H0: �31 = ω31 H0:∑

�1i = 0/∑

b1i = 0 H0:∑

�2i = 0/∑

b2i = 0 H0:∑

�3i = 0/∑

b3i = 0Fund name ω31 F-Value �11/b11 �21/b21 �31/b31

Emerging marketsLatin American

Brazil −0.1665 1.07 −0.0519/−0.0904 −0.0210/−0.0578 −0.0361/−0.0562 0.1130/0.1133/0.1135Brazilian equity −0.1769 2.25 −0.0720/−0.0926 −0.0126/−0.0663 −0.0064/−0.0504 0.1032/0.1042/0.1038Chile −0.1745 1.53 0.1722/−0.1005 −0.1655/−0.0709 −0.0033/−0.0626 0.1084/0.1089/0.1105Mexico equity & income −0.2732 0.95 −0.1978/−0.1237 −0.1546/−0.0801 −0.0386*/−0.0601 0.1247/0.1254/0.1166Mexico −0.1755 2.00 0.1726/−0.1229 0.1201/−0.0835 −0.0655*/−0.0653 0.1034/0.1027/0.1040

886P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 4 (Continued )

Entire sample period

NAV return volatilitydoes not Granger-causeflows

U.S market volatility impacts/U.S. investor sentiment volatility impacts R2

Eq. (9) Eq. (7) Eq. (8) Eq. (9) Eq. (7)/Eq. (8)/Eq. (9)

Null hypothesis H0:∑

ω3i = 0 H0: �31 = ω31 H0:∑

�1i = 0/∑

b1i = 0 H0:∑

�2i = 0/∑

b2i = 0 H0:∑

�3i = 0/∑

b3i = 0Fund name ω31 F-Value �11/b11 �21/b21 �31/b31

AsianChina −0.1972* 2.14 −0.1380/−0.0883 −0.1700/−0.0604 −0.0082*/−0.0599 0.1011/0.1018/0.0963

First Philippine −0.2241* 3.44* 0.2473/−0.0978 0.1444/−0.0773 −0.0043*/−0.0678 0.1048/0.1072/0.1078Greater China −0.2188 2.04 −0.1788/−0.1006 −0.0993/−0.0752 −0.0940/−0.0705 0.1116/0.1119/0.1122India 0.0086 0.94 0.0074/−0.0785 0.0034/−0.0679 0.0008/−0.0664 0.0837/0.0845/0.0818India Growth −0.0092 0.92 0.0145/−0.1177 0.0106/−0.0684 0.0035/−0.0601 0.0840/0.0849/0.0836Indonesia −0.2614** 3.86* −0.1648/−0.1346 −0.1472/−0.0902 −0.0071*/−0.0852 0.0865/0.0857/0.0863J. F. China −0.1916** 2.20 −0.1426/−0.1322 −0.1212/−0.1006 −0.0192/−0.0857 0.0868/0.0866/0.0871J. F. India 0.0125 1.08 0.0077/−0.0755 0.0046/−0.0598 0.0017/−0.0301 0.0823/0.0825/0.0820Korea equity −0.1821* 3.78* −0.1341/−0.1259 −0.1038/−0.0955 −0.0586*/−0.0775 0.1327/0.1329/0.1315Korea −0.2555* 4.59** −0.1095/−0.1260 −0.1210/−0.0947 −0.0254/−0.0837 0.1299/0.1286/0.1292Malaysia −1.1683* 8.78** −0.1506/−0.1365 −0.1138/−0.0965 −0.0523*/−0.0822 0.1104/0.1093/0.1080Morg Stan India 0.0096 0.71 0.0082/−0.0552 0.0011/−0.0502 0.0012/−0.0489 0.0811/0.0813/0.0724Taiwan, ROC −0.1509 1.35 −0.2676/−0.1037 −0.1792/−0.0957 −0.0193/−0.0782 0.1234/0.1228/0.1225Singapore −0.1320* 1.95 −0.1633/−0.1248 −0.1529/−0.0976 −0.0835/−0.0888 0.1339/0.1356/0.1326Taiwan −0.1707 1.99 −0.1300/−0.1303 −0.1054/−0.1015 −0.0331/−0.0939 0.1297/0.1288/0.1155Templeton China −0.1618 2.22 −0.1574/−0.1296 −0.1308/−0.1048 −0.0388/−0.0856 0.0772/0.0803/0.0728Thai Capital −0.2785** 3.41* 0.1945/−0.1365 0.1387/−0.1062 −0.0130*/−0.0845 0.1065/0.1069/0.1014Thai -0.2974** 3.76* 0.3002/−0.1378 −0.1713/−0.1055 −0.0126/−0.0861 0.1088/0.1086/0.1005

Developed markets (Asian)Aberdeen Australia −0.2842* 3.43* −0.1385/−0.1024 −0.1017*/−0.0944 −0.0549/−0.0907 0.1327/0.1329/0.1206Japan equity −0.2846* 1.43 0.1413/−0.1035 −0.0863*/−0.0879 0.0136/−0.0833 0.1308/0.1299/0.1114Japan small cap −0.1707* 1.21 0.2820/−0.1029 −0.1477**/−0.0905 0.0425/−0.0849 0.1283/0.1315/0.1277

Developed markets (European)Austria −0.1646 2.02 −0.1688/−0.1129 −0.1356**/−0.0994 −0.0714/−0.0890 0.1280/0.1268/0.1244France Growth −0.2300* 1.97 0.2594/−0.1008 −0.1971*/−0.0967 0.0882/−0.0851 0.1314/0.1302/0.1175Germany −0.2583** 2.01 −0.2860/−0.1037 −0.1283**/−0.0855 −0.0217/−0.0739 0.1333/0.1324/0.1302Italy −0.1478 2.21 −0.1445/−0.1056 −0.0615*/−0.0806 0.0065/−0.0760 0.1107/0.1123/0.1016New Germany −0.1250* 0.93 0.2211/−0.1080 −0.1552*/−0.0843 −0.0450/−0.0722 0.1349/0.1364/0.1328New Ireland −0.0674 1.33 0.2644/−0.1052 −0.2004*/−0.0969 −0.0029*/−0.0877 0.1093/0.1085/0.1008

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894887

Portugal −0.1771* 2.23 −0.1641/−0.0978 −0.1174/−0.0923 0.0460/−0.0882 0.1107/0.1122/0.1029Spain −0.2312** 3.77* −0.2587/−0.1167 −0.1922**/−0.1008 −0.0317*/−0.0845 0.1103/0.1115/0.1017Swiss Helvetia −0.2051* 3.46* 0.2267/−0.1072 −0.1105**/−0.1094 −0.1087**/−0.0936 0.1029/0.1039/0.0944

Test results for Granger causality incorporating volatility effects are based on the following equations:

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i +L∑

i=1

a1i PDWt−i +L∑

i=1

�1iV NAVRt−i +L∑

i=1

ω1iVFt−i +L∑

i=1

�1iV USRt−i

+L∑

i=1

b1iV PDWt−i + e1t

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i +L∑

i=1

a2i PDWt−i +L∑

i=1

�2iV PRt−i +L∑

i=1

ω2iVFt−i +L∑

i=1

�2iV USRt−i

+L∑

i=1

b2iV PDWt−i + e2t

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i +L∑

i=1

a3i PDWt−i +L∑

i=1

�3iV PRt−i +L∑

i=1

ω3iV NAVRt−i +L∑

i=1

�3iV USRt−i

+L∑

i=1

b3iV PDWt−i + e3t

where V PRt is the moving variance of fund price returns at period t, V NAVRt is the moving variance of fund NAV returns at period t; VFt is the moving variance of international net equityflows at period t; VUSRt is the moving variance of the U.S. market index returns at period t; V PDWt is the moving variance of the premiums/discounts of an equally weighted index of U.S.domestic closed-end funds at period t. The test statistics are corrected for heteroskedasticity and serial correlation. Significance at the 5% level is denoted by **, and significance at the 10%level is denoted by *.

888P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 5Estimation results of Granger causality tests incorporating volatility effects and Asian crisis effects.

Asian crisis period

Structural interceptchanges

Flow volatility does notGranger-cause pricereturns

Flow volatility does notGranger-cause NAVreturns

Price return volatilitydoes not Granger-causeflows

Eq. (10) Eq. (11) Eq. (12) Eq. (10) Eq. (11) Eq. (12)

Null hypothesis H0: d10 = 0 H0: d20 = 0 H0: d30 = 0 H0:∑

d16i = 0 H0:∑

d26i = 0 H0: d161 = d261 H0:∑

d35i = 0Fund name d10 d20 d30 d161 d261 F-Value d351

Emerging marketsLatin American

Brazil −0.0212 −0.0065 −0.0078 −1.1521 −1.1112 1.67 −2.6823Brazilian equity −0.0235 0.0408 0.0305 −1.0458 −1.0220 1.09 −1.9044Chile −0.0662 −0.0212 −0.0452 0.1226 0.0916 1.39 −0.7572Mexico equity & income −0.0218 −0.0116 −0.2177 −1.0677 −1.0245 1.75 −1.0001Mexico −0.0315 −0.0046 −0.1903 −1.0446 −1.0028 1.70 −2.4908

AsianChina −0.0452* −0.0311* −0.0242* −1.9533** −1.7592** 8.01** −2.0469First Philippine −0.0052* −0.0047* −0.0010** −1.9024** −1.6901* 9.07** −1.1010Greater China −0.0837** −0.0522 −0.0348* −3.1371** −3.0606** 4.13* −3.0205India −0.0101* 0.0064 −0.0574 0.0615 0.0468 0.53 0.1044India Growth −0.0095* 0.0021 −0.0092 0.0118 0.0040 0.25 −0.0883Indonesia −0.0506* −0.0388** −0.0143** −4.0989** −3.9102** 6.33** −3.2912J. F. China −0.0278* −0.0021 −0.0350** −3.3856** −3.2354* 5.26** −1.2806J. F. India −0.0114* 0.0058 −0.0012 0.0583 0.0122 1.99 −0.1101Korea equity −0.0147** −0.0062 0.1926 −0.9136** −0.6990* 9.15** −2.8038Korea −0.0149* −0.0028 0.2009 −1.7092** −1.5301* 6.07** −3.5586Malaysia −0.0918** −0.0507** −0.0666** −3.9355** −3.6167** 11.96** −4.2107Morg Stan India −0.0112* 0.0002 −0.0091 0.0704 0.0388 1.42 0.1698ROC Taiwan −0.0236* −0.0314 −0.0377 −1.4255** −1.3072 4.86** −2.5587Singapore −0.0217** −0.0125** 0.0208 −1.2486* −1.2331** 0.67 −2.8926**Taiwan −0.0242 0.0109 −0.0375 −1.0012** −0.9554* 1.87 −2.0590Templeton China −0.0449* −0.0216 −0.0231** −1.9035** −1.7644 5.04** −2.1577Thai Capital −0.0156** −0.0103** −0.0100** −3.2749** −3.0725 8.78** −3.3026Thai −0.0206** −0.0156** −0.0201** −4.5126** −4.2307 10.42** −3.2063

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894889

Developed markets (Asian)Aberdeen Australia −0.0318 −0.0153 0.0607 −1.6329 −1.6178 0.64 −1.6917Japan equity −0.0133* −0.0026 −0.0173 −1.4618** −1.4277* 1.55 −1.4155Japan small cap −0.0116* −0.0148 −0.0155 −1.0885* −1.0474* 1.68 −1.0216

Developed markets (European)Austria −0.0003 0.0036 0.0006 −1.7330 −1.6923 1.64 −1.7797France Growth −0.0204 0.0148 −0.0041 −1.7416 −1.6901 2.39 −1.6498Germany −0.0068 0.0005 −0.0128 −1.6855 −1.6364 2.23 −1.9021Italy −0.0288 0.0139 −0.0173 −2.6832 −2.6360 2.06 −1.4806New Germany −0.0014 0.0007 −0.0149 −1.3613 −1.3266 1.60 −1.5808New Ireland 0.0053 0.0061 −0.0414 −1.8717 −1.8293 1.72 −1.7525Portugal −0.0215 −0.0112 0.0348 −1.5895 −1.5563 1.51 −2.1017Spain 0.0279 0.0082 −0.0216 −1.1704 −1.1306 1.62 −2.3560Swiss Helvetia −0.0055 0.0047 −0.0532 −1.0903 −1.0499 1.63 −2.9979

Asian crisis period

NAV return volatilitydoes not Granger-causeflows

U.S market volatility impacts/U.S. investor sentiment volatility impacts R2

Eq. (12) Eq. (10) Eq. (11) Eq. (12) Eq. (10)/Eq. (11)/Eq. (12)

Null hypothesis H0:∑

d36i = 0 H0: d351 = d361 H0:∑

d17i = 0/∑

d18i = 0 H0:∑

d27i = 0/∑

d28i = 0 H0:∑

d37i = 0/∑

d38i = 0Fund name d361 F-Value d171/d181 d271/d281 d371/d381

Emerging marketsLatin American

Brazil −2.7179 1.55 1.1858/−0.7952 −0.6140**/−0.1022 0.1677/−0.0459 0.1152/0.1158/0.1161Brazilian equity −1.9540 2.31 −1.9806/−0.8033 −1.0934**/−0.0836 0.0417/−0.0336 0.1083/0.1085/0.1090Chile −0.7106 2.03 0.2526/−0.0572 −0.1505/−0.0344 0.0802/−0.0223 0.1095/0.1099/0.1117Mexico equity & income −1.0463 1.99 −2.3103/−1.0304 −1.5512/−0.9102 0.2215/−0.0775 0.1286/0.1298/0.1213Mexico −2.5375 2.10 −1.5045/−0.9627 −1.0985/−0.8555 −0.0728/−0.0724 0.1095/0.1086/0.1102

AsianChina −2.2873** 9.03** −1.9604/−1.0052 −1.8245**/−0.8842 −0.1277/−0.0495 0.1078/0.1081/0.1016First Philippine −1.3024* 8.14** 1.2275/−0.8645 −1.0028**/−0.5694 −0.1065/−0.0414 0.1072/0.1085/0.1090Greater China −3.1677* 5.17** −1.3703/−0.8550 −1.0101/−0.6359 −0.0812/−0.0382 0.1148/0.1154/0.1158India 0.1290 1.26 0.1385/−0.0624 0.0986/−0.0573 0.0053/−0.0039 0.0933/0.0954/0.0882India Growth −0.0996 0.49 −0.1801/−0.0681 −0.0126/−0.0520 −0.0012/−0.0025 0.0889/0.0905/0.0877Indonesia −3.5438** 9.26** −1.8079/−0.9096 −0.9455**/−0.7485 0.0347/−0.0106 0.0920/0.914/0.0919

890P.-J.Tsai/Int.Fin.M

arkets,Inst.andM

oney19

(2009)862–894

Table 5 (Continued )

Asian crisis period

NAV return volatilitydoes not Granger-causeflows

U.S market volatility impacts/U.S. investor sentiment volatility impacts R2

Eq. (12) Eq. (10) Eq. (11) Eq. (12) Eq. (10)/Eq. (11)/Eq. (12)

Null hypothesis H0:∑

d36i = 0 H0: d351 = d361 H0:∑

d17i = 0/∑

d18i = 0 H0:∑

d27i = 0/∑

d28i = 0 H0:∑

d37i = 0/∑

d38i = 0Fund name d361 F-Value d171/d181 d271/d281 d371/d381

J. F. China −1.3285 2.26 −1.3402/−0.9477 −0.9170/−0.7294 −0.0812/−0.0227 0.0927/0.0914/0.0923J. F. India −0.1294 1.08 −0.1246/−0.0589 −0.1074/−0.0471 −0.0648/−0.0358 0.0892/0.0884/0.0880Korea equity −3.0132** 8.66** 1.6385/−0.9332 1.3671*/−0.8926 0.1444/−0.0679 0.1389/0.1386/0.1355Korea −3.7644** 8.29** 0.6477/−0.7520 0.2413/−0.6930 0.1549/−0.0708 0.1337/0.1326/0.1318Malaysia −4.4728** 9.35** −1.1704/−0.9736 −0.6762/−0.7901 0.0126*/−0.0095 0.1175/0.1162/0.1155Morg Stan India 0.2062 1.70 −0.1033/−0.0589 −0.0764/−0.0233 −0.0582/−0.0128 0.0890/0.0898/0.0776Taiwan, ROC −2.6093 2.38 −1.6404/−0.9280 −1.4958/−0.6782 −0.0957/−0.0203 0.1292/0.1285/0.1267Singapore −2.9255* 1.43 1.1927/−0.6825 −1.0836/−0.5479 0.0163*/−0.0082 0.1387/0.1403/0.1375Taiwan −2.1048 1.95 −0.6501/−0.7811 −0.2073/−0.3844 −0.0310/−0.0094 0.1308/0.1307/0.1216Templeton China −2.2003 1.81 −1.5129/−0.8285 −1.4029**/−0.7026 −1.0506/−0.7395 0.0813/0.0826/0.0869Thai Capital −3.5705** 9.37** −2.9567/−1.2306 −1.6901/−0.8238 0.7738/−0.5114 0.1126/0.1122/0.1094Thai −3.4509* 9.10** −1.2302/−0.8720 0.9162/−0.5799 0.0845/−0.0289 0.1134/0.1139/0.1105

Developed markets (Asian)Aberdeen Australia −1.7275 1.64 1.7234/−0.7889 1.8843*/−0.9242 1.0596/−0.8831 0.1384/0.1389/0.1294Japan equity −1.4290 0.88 1.9855/−0.8722 1.2024/−0.9545 0.7916/−0.6628 0.1377/0.1336/0.1180Japan small cap −1.0528 1.41 1.5016/−0.9456 1.0975/−0.9200 −0.3769/−0.4447 0.1312/0.1365/0.1302

Developed markets (European)Austria −1.8102 1.32 −1.3405/−0.8551 −1.3173**/−0.8037 1.0026/−0.8379 0.1352/0.1333/0.1314France Growth −1.6925 1.85 −1.1856/−0.7632 −1.0322/−0.7229 0.0864/−0.0293 0.1377/0.1358/0.1285Germany −1.9399 1.72 −1.1338/−0.7907 −1.1259/−0.7047 −1.0165/−0.6281 0.1399/0.1387/0.1356Italy −1.5220 1.76 1.1202/−0.7810 −1.1024/−0.6102 0.0448/−0.0045 0.1187/0.1216/0.1062New Germany −1.5914 0.42 −0.8437/−0.8042 −0.7383/−0.5290 −0.0972/−0.0018 0.1417/0.1434/0.1405New Ireland −1.7963 1.88 1.0635/−0.9582 −0.4937*/−0.3004 0.0337/−0.1047 0.1144/0.1130/0.1078Portugal −2.1315 1.30 1.7448/−0.6792 −1.2002/−0.2890 −0.3028/−0.1011 0.1193/0.1210/0.1105

P.-J.Tsai/Int.Fin.Markets,Inst.and

Money

19(2009)

862–894891

Spain −2.4012 1.92 −1.8626/−0.7021 −1.0923/−0.5289 0.4030/−0.2956 0.1146/0.1198/0.1091Swiss Helvetia −3.0346 1.68 −1.2573/−0.5807 −1.0764/−0.2955 −0.8589/−0.1045 0.1103/0.1116/0.1017

Test results for Granger causality incorporating volatility effects and the Asian crisis effects are based on the following equations:

PRt = ˛1 + ı1 ECTt−1 +L∑

i=1

ˇ1i PRt−i +L∑

i=1

�1i NAVRt−i +L∑

i=1

�1iFt−i +L∑

i=1

�1i USRt−i +L∑

i=1

a1i PDWt−i +L∑

i=1

�1iV NAVRt−i +L∑

i=1

ω1iVFt−i

+L∑

i=1

�1iVUSRt−i +L∑

i=1

b1iV PDWt−i + d10D +L∑

i=1

d11iD × NAVRt−i +L∑

i=1

d12iD × Ft−i +L∑

i=1

d13iD × USRt−i +L∑

i=1

d14iD × PDWt−i +L∑

i=1

d15iD × V NAVRt−i

+L∑

i=1

d16iD × VFt−i +L∑

i=1

d17iD × VUSRt−i +L∑

i=1

d18iD × V PDWt−i + e1t

NAVRt = ˛2 + ı2 ECTt−1 +L∑

i=1

ˇ2i PRt−i +L∑

i=1

�2i NAVRt−i +L∑

i=1

�2iFt−i +L∑

i=1

�2i USRt−i +L∑

i=1

a2i PDWt−i +L∑

i=1

�2iV PRt−i +L∑

i=1

ω2iVFt−i +L∑

i=1

�2iVUSRt−i

+L∑

i=1

b2iV PDWt−i + d20D +L∑

i=1

d21iD × PRt−i +L∑

i=1

d22iD × Ft−i +L∑

i=1

d23iD × USRt−i +L∑

i=1

d24iD × PDWt−i +L∑

i=1

d25iD × VPRt−i +L∑

i=1

d26iD × VFt−i

+L∑

i=1

d27iD × VUSRt−i +L∑

i=1

d28iD × V PDWt−i + e2t

Ft = ˛3 + ı3 ECTt−1 +L∑

i=1

ˇ3i PRt−i +L∑

i=1

�3i NAVRt−i +L∑

i=1

�3iFt−i +L∑

i=1

�3i USRt−i +L∑

i=1

a3i PDWt−i +L∑

i=1

�3iV PRt−i +L∑

i=1

ω3iV NAVRt−i +L∑

i=1

�3iVUSRt−i

+L∑

i=1

b3iV PDWt−i + d30D +L∑

i=1

d31iD × PRt−i +L∑

i=1

d32iD × NAVRt−i +L∑

i=1

d33iD × USRt−i +L∑

i=1

d34iD × PDWt−i +L∑

i=1

d35iD × V PRt−i +L∑

i=1

d36iD × V NAVRt−i

+L∑

i=1

d37iD × VUSRt−i +L∑

i=1

d38iD × V PDWt−i + e3t

where D is as defined earlier. d10, d20 and d30 are the intercept dummy variables; d16i , d26i , d35i and d36i are interactive dummy variables. The test statistics are corrected for heteroskedasticityand serial correlation. Significance at the 5% level is denoted by **, and significance at the 10% level is denoted by *.

892 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

effects are found only in twelve emerging market Asian funds. This finding suggests that crisis effectsamplified asymmetric market information across local and U.S. investors.

Based on Eq. (12), all coefficients of the interactive dummy variables are insignificant except foremerging market Asian funds. The Asian crisis affected nine Asian funds through interactive NAVreturn volatility dummy variable and affected Singapore fund through both interactive dummy vari-ables (price and NAV return volatility). And, these effects are negative, indicating that the crisis effectstrengthened the impacts of fund return volatility on flows. Furthermore, significant return volatilitycoefficient difference between price returns and NAV returns for flows is only for nine emerging mar-ket Asian funds. Thus, the effects of the Asian crisis were predominantly local for Asian funds only anddid not appreciably affect relationships between flows and fund returns for funds from other parts ofthe world.

6. Summary and conclusions

This study investigates the relationships between U.S. equity flows in foreign countries and returnsof closed-end country funds for emerging Latin American markets, emerging Asian markets and devel-oped markets based on two basic models (information contribution and feedback trading). The dataset includes 35 closed-end single country funds, representing 22 countries. All flows are scaled flows,flows which are adjusted for market capitalization. The effects of volatility and of the 1997 Asian crisison these relationships are studied. The data cover the period from January 1994 to December 2007.

Throughout the tests, the evidence of Granger causality from flows to fund returns supports theinformation contribution argument; the result of Granger causality from fund returns to flows favorsthe feedback trading argument. All the significant effects are positive, in keeping with the findingsof Froot and Ramadorai (2008). Furthermore, the analysis of differential effects between price-flowand NAV-flow relationships provides strong evidence of market segmentation effects over and aboveinvestor sentiment effects. However, premiums/discounts of an index of U.S. domestic closed-endfunds show a positive correlation with fund price returns, but not with NAV returns. This providessome evidence consistent with the presence of investor sentiment in country fund prices.

Impulse response functions, used to determine whether information contribution effects are long-term or short-term, show that fund return responses to flow innovations decay slowly. The impact offlows on fund returns can be explained by information revelation.

The Asian crisis effects are reflected almost solely in the Asian fund relationships. Surprisingly,adding the crisis dummy variables appreciably affects the significant levels of fund return (flow) effectson flows (fund returns). The information contribution results with the Asian crisis effects were evenstronger; the crisis intensified the influence of flows for emerging Asian markets. The uncertainties thecrisis created also induced stronger asymmetric information between U.S. investors and local investorsacross markets. However, including Asian crisis effects weakened feedback trading results.

The findings of volatility effects in the initial information contribution and feedback trading testsshow negative relationships between flow volatility and fund returns and between return volatility andflows. Thus, fund returns react significantly when international capital flows become more uncertain;and, there is a symmetric effect on capital flows when local returns become increasing volatile. Theresults support Nelson (1991) and Glosten et al. (1993) who find a negative relation between condi-tional expected monthly returns and conditional variance of monthly returns. When the crisis is added,the causal linkages between flow (fund return) volatility and fund returns (flows) are strengthened forAsian funds only.

The differential relationships between price returns and flows relative to NAV returns and flowsprovide better insight of international investors’ effects on local and U.S. markets. This is especiallyimportant since the crisis amplified the influence of foreign equity flows on local markets. The poten-tially dangerous capital outflows during the crisis may require emerging markets restrict foreign equityflows. These results, plus the strong volatility effects on the relationship between flows and fundreturns, suggest needed close monitoring of equity markets. Such monitoring of international capitalflows should help make equity markets less vulnerable to foreign speculative attacks. These find-ings also aid policymakers of emerging markets in reducing market volatility and recognizing foreigninfluences in their markets, thereby making more effective policy adjustments.

P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894 893

References

Barberis, N., Shleifer, A., Vishny, R., 1998. A model of investor sentiment. Journal of Financial Economics 49, 307–343.Bekaert, G., 1995. Market integration and investment barriers in emerging equity markets. The World Bank Economic Review

9, 75–107.Bodurtha, J., Kim, D.-S., Lee, C., 1995. Closed-end country funds and U.S. market sentiment. Review of Financial Studies 8,

879–918.Bowe, M., Domuta, D., 2001. Foreign investor behaviour and the Asian financial crisis. Journal of International Financial Markets,

Institutions and Money 11, 395–422.Chordia, T., Swaminathan, B., 2000. Trading volume and cross-autocorrelations in stock returns. Journal of Finance 55,

913–935.Dann, L.Y., Mayers, D., Raab, R.J., 1977. Trading rules, large block trades and the speed of price adjustment. Journal of Financial

Economics 4, 3–22.Delcoure, N., Zhong, M., 2007. On the premiums of iShares. Journal of Empirical Finance 14, 168–195.Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive time series with a unit root. Journal of American

Statistical Association 74, 427–431.Engle, R.F., Granger, C., 1987. Cointegration and error correction: representation, estimation, and testing. Econometrica 55,

251–276.Engle, R., Sarkar, D., 2002. Are exchange traded funds fairly priced? New York University. Working paper.Errunza, V., Senbet, L.W., Hogan, K., 1998. The pricing of country funds from emerging markets: theory and evidence. International

Journal of Theoretical and Applied Finance 1, 111–144.Eun, C.S., Janakiramanan, S., 1986. A model of international asset pricing with a constraint on the foreign equity ownership.

Journal of Finance 41, 897–914.Frankel, J.A., Schmukler, S.L., 1996. Country fund discounts, asymmetric information and the Mexican crisis of 1994: did local

residents turn pessimistic before international investors? NBER Working Paper No. W5714.Froot, K., O’Connell, P., Seasholes, M.S., 2001. The portfolio flows of international investors. Journal of Financial Economics 59,

151–193.Froot, K., Ramadorai, T., 2001. The information content of international portfolio flows. NBER Working Paper No. W8472.Froot, K., Ramadorai, T., 2008. Institutional portfolio flows and international investments. Review of Financial Studies 21,

937–972.Gemmill, G., Thomas, D., 2002. Noise-trading, costly arbitrage, and asset prices: evidence from closed-end funds. Journal of

Finance 57, 2571–2594.Glosten, L., Jagannathan, R., Runkle, D., 1993. On the relation between the expected value and the volatility of the nominal excess

return on stocks. Journal of Finance 48, 1779–1801.Goetzmann, W.N., Ivkovic, Z., Rouwenhorst, K.G., 2001. Day trading international mutual funds: evidence and policy solutions.

Journal of Financial and Quantitative Analysis 36, 287–309.Granger, C.W.J., 1988. Some recent developments in a concept of causality. Journal of Econometrics 39, 199–212.Griffin, J.M., Nardari, F., Stulz, R.M., 2004. Are daily cross-border equity flows pushed or pulled? Review of Economics and

Statistics 86, 641–657.Hardouvelis, G., La Porta, R., Wizman, T.A., 1994. What moves the discount on country equity funds? In: Frankel, J.A. (Ed.), The

Internationalization of Equity Markets. University of Chicago Press, Chicago, pp. 345–397.Harris, L., Gurel, E., 1986. Price and volume effects associated with change in the S&P 500 list: new evidence for the existence of

price pressure. Journal of Finance 41, 815–829.Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efficiency. Journal

of Finance 48, 65–91.Jegadeesh, N., Titman, S., 2001. Profitability of momentum strategies: an evaluation of alternative explanations. Journal of Finance

56, 699–720.Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on cointegration—with applications to the demand

for money. Oxford Bulletin of Economics and Statistics 52, 169–210.Kim, E.H., Singal, V., 1997. Are open markets good for foreign investors and emerging nations? Journal of Applied Corporate

Finance 10, 18–33.Kraus, A., Stoll, H.R., 1972. Price impacts on block trading on the New York Stock Exchange. Journal of Finance 27,

569–588.Lee, B.-S., Hong, G., 2002. On the dual characteristics of closed-end country funds. Journal of International Money and Finance

21, 589–618.Merton, R.C., 1980. On estimating the expected return on the market: an exploratory investigation. Journal of Financial Economics

8, 323–361.Nagayasu, J., 2001. Currency crisis and contagion: evidence from exchange rates and sectoral stock indices of the Philippines

and Thailand. Journal of Asian Economics 12, 529–546.Nelson, D.B., 1991. Conditional heteroskedasticity in asset returns: a new approach. Econometrica 59, 347–370.Nishiotis, G., 2004. Do indirect investment barriers contribute to capital market segmentation? Journal of Financial and Quan-

titative Analysis 39, 613–630.Nishiotis, G., 2006. Further evidence on closed-end country fund prices and international capital flows. Journal of Business 79,

1727–1754.Officer, R.R., 1973. The variability of the market factor of the New York stock exchange. Journal of Business 46, 434–453.Pan, M.-S., Chan, K., Wright, D., 2001. Divergent expectations and the Asian financial crisis of 1997. Journal of Financial Research

24, 219–238.Patro, D.K., 2005. Stock market liberalization and emerging market country fund premiums. Journal of Business 78, 135–168.Phillips, P.C.B., Perron, P., 1988. Testing for a unit root in time series regression. Biometrica 75, 335–346.Pontiff, J., 1997. Excess volatility and closed-end funds. American Economic Review 87, 155–169.

894 P.-J. Tsai / Int. Fin. Markets, Inst. and Money 19 (2009) 862–894

Richards, A.J., 1996. Volatility and predictability in national stock markets: how do emerging and mature markets differ?International Monetary Fund, Staff Papers 43, 461–501.

Scholes, M., 1972. The market for securities: substitution vs. price pressure and the effect of information on share price. Journalof Business 45, 179–211.

Shleifer, A., 2000. Professional Arbitrage in Inefficient Markets: An Introduction to Behavioral Finance. Oxford University Press,pp. 89–111.

Swanson, P.E., 2004. A dual approach to measuring equity market interrelationships. Global Business and Finance Review 9,15–26.