Performance comparison between exchange-traded funds and closed-end country funds

Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Performance comparison between exchange-traded funds and closed-end country funds Joel T. Harper a , Jeff Madura b , Oliver Schnusenberg c,∗ a

Oklahoma State University, OK, USA Florida Atlantic University, FL, USA The University of North Florida, Coggin College of Business, Department of Accounting and Finance, 4567 St. Johns Bluff Road, Jacksonville, FL 32224, USA b

c

Received 3 June 2004; accepted 15 December 2004 Available online 10 August 2005

Abstract The objective of this study is to compare the risk and return performance of exchange-traded funds (ETFs) available for foreign markets and closed-end country funds. We utilize 29 closedend country funds (CEFs) for 14 countries over the sample period from April 1996 to December 2001. The performance proxies are mean returns and risk-adjusted returns. Results indicate that ETFs exhibit higher mean returns and higher Sharpe ratios than foreign closed-end funds, while CEFs exhibit negative alphas. This indicates that a passive investment strategy utilizing ETFs may be superior to an active investment strategy using CEFs. The findings reported here offer some insight on the relative advantages of each type of investment. Specifically, there may be some potential for additional types of ETFs that offer higher risk-adjusted returns than closed-end funds. Such ETFs may be able to offer higher risk-adjusted returns as part of an internationally diversified portfolio. © 2005 Elsevier B.V. All rights reserved. JEL classification: G12; G14; G15 Keywords: Exchange-traded fund; Closed-end country fund; Active investment strategy

∗

Corresponding author. Tel.: +1 904 620 1224; fax: +1 904 620 3861. E-mail address: [email protected] (O. Schnusenberg).

1042-4431/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.intfin.2004.12.006

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

105

1. Background and purpose Recent years have seen a rapid expansion of the creation and trading of exchange-traded funds (ETFs). ETFs are specialized investment trusts that are designed to mirror specified stock indexes ranging from broad country-specific indexes to specialized industry indexes, with newer developments in fixed income and other assets. While there are many open-end mutual funds that have mirrored many of these indexes, ETFs are unlike open-end mutual funds in that they are traded continuously on an exchange, can be purchased, sold, and even shorted any time the market is open. Furthermore, ETFs restrict the creation and redemption of shares to large in-kind transactions. These in-kind transactions prevent the price of the ETF from deviating from the net asset value through arbitrage and reduce the observed discount and premium found in closed-end funds. The objective of this paper is to assess the risk-adjusted return performance of ETFs based on a comparison to corresponding closed-end funds matched by country. Since both securities are exchange-traded, they are comparable in market trading structure. We first investigate (and test for significance) the tracking error of the ETFs (we use the iShares ETFs in the empirical analysis) with respect to the underlying index. Second, we assess the individual return and risk characteristics of the ETF and of a corresponding closed-end fund (CEF). Next, we compute the risk-adjusted return characteristics using the Sharpe ratio and Jensen’s alpha for the CEF relative to the ETF. Finally, we test the factors that explain the CEFs alpha relative to the ETF as a benchmark. We find that ETFs have very low and statistically insignificant tracking errors relative to their corresponding MSCI index. Furthermore, we find that ETFs exhibit higher mean returns and higher risk levels than their counterpart CEFs. Based on the Sharpe ratio and Jensen’s alpha, ETFs have, on average, higher risk-adjusted returns than their counterpart closed-end funds. When considering the abilities of investors to diversify among ETFs to reduce risk, ETFs serve as a viable substitute for closed-end funds when investing in an internationally diversified portfolio. This would indicate that a more passive investment style through ETFs provides better risk-adjusted returns than a more active management of assets in CEFs.

2. Related research One of the oldest exchange-traded funds are Standard & Poor’s Depository Receipts (SPDRs or “spiders”), which began trading on the American Stock Exchange (AMEX) in the Fall of 1992. The success of this ETF gave rise to several others including Diamonds, Cubes and several industry sector ETFs. ETFs are also available on an international basis. The MSCI indexes were initially traded under the name of World Equity Benchmark Shares (WEBS) for various countries, but now have several different names depending upon the sponsoring institution, including iShares, which are sponsored by Barclays. In the construction of an MSCI index ETF, 60% of the capitalization of each industry group is targeted for inclusion in each MSCI country index. Hence, MSCI’s stated objective is to provide investors with investable country indexes while minimizing the tracking error of the formed indexes.

106

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

ETFs can mimic an index while limiting the expense ratio, and may therefore be a suitable, more cost effective alternative to closed-end funds. Given their goal to mirror an index, they do not require aggressive portfolio management. Thus, the expenses associated with portfolio trading and management of exchange-traded funds are very low. For example, the S&P 500 ETF has an expense ratio of 0.11% and the France and Singapore ETFs have an expense ratio of 0.84%. Conversely, the France Growth Fund has an expense ratio of 3.4%, and the Singapore Fund has an expense ratio of 2.12%. Our objective is to evaluate the relative return and risk performance of ETFs currently available for foreign markets that have a longer time series for study. The underlying question is whether ETFs exhibit characteristics that are more favorable than their traditional closed-end fund counterparts, especially in light of the higher expenses charged by closed-end funds. The general characteristics of ETFs relative to mutual funds have been discussed extensively in the literature in recent years. For example, Gastineau (2001) discusses the low expense ratios of ETFs and how ETFs manage to avoid significant capital gains contributions. Fuhr (2001) points out that ETFs’ characteristics make them a viable alternative to trading futures for investors looking to increase or reduce their exposure to countries, sectors, industries, and styles. ETFs allow investors the flexibility to use them for numerous applications, which can be appealing to both individual and institutional investors. One important characteristic of exchange-traded funds that distinguish them from their mutual fund counterparts is their tax characteristics. ETFs are considered very tax efficient, as shares do not need to be sold to fund cash redemptions. This is because shares are only created and redeemed via in-kind share contributions and redemptions. These redemptions can be accomplished tax free for the ETF. Poterba and Shoven (2002) compare the pre-tax and after-tax returns on the SPDR trust and the Vanguard index 500 fund. Results suggest that between 1994 and 2000, the before- and after-tax returns on the SPDR trust and this mutual fund were very similar. Both the after-tax and the pre-tax returns on the fund were slightly greater than those on the ETF. These findings suggest that ETFs offer taxable investors a method of holding broad baskets of stocks that deliver returns comparable to those of low-cost index funds. Bernstein (2004) compares the tax efficiency of ETFs, openend mutual funds, and closed-end mutual funds, and concludes that it is difficult to make a generalization about the tax efficiency of the various types of funds. Dellva (2001) finds that transaction costs limit ETF attractiveness for small investors, but that in-kind creation and redemption processes provide the ETFs with significant tax efficiencies. Dellva also finds that tax deferred, long-term retirement investors have little or no advantage from using ETFs as opposed to traditional mutual funds. Gastineau (2004) directly compares the operational efficiency of ETFs with conventional mutual fund competitors. He finds that the ETFs for the Russell 2000 index and the Standard & Poor’s 500 index have underperformed their most comparable conventional mutual fund competitors. Gastineau argues that the reason for this underperformance is due to the inability of ETF managers to reduce transaction costs embedded in the index modification process (also see Blume and Edelen, 2002), since they have to wait until the end of the trading day to know what creations or redemptions will occur. Conversely, conventional mutual fund managers try to anticipate upcoming changes in indexes in order to perform modifications by trading at a better time.

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

107

Much of the research on closed-end funds has focused on the behavior of premiums or discounts. Thompson (1978), Anderson (1986), Cheng et al. (1994), and Pontiff (1995) document favorable performance from purchasing closed-end funds with relatively high discounts (or low premiums). Related research by Lee et al. (1990) and Zweig (1993) suggest that a closed-end fund’s discount or premium level can be used to predict future performance. Richards et al. (1980) show evidence of systematic trading behavior, and that filter rules can achieve abnormal returns. Internationally, DeLong and Schleifer (1992) find more pronounced premiums for international closed-end funds. Leonard and Shull (1996) suggest that international closed-end fund premiums (or discounts) can be used as an indicator of future returns. Chiang and Kim (2003) show that closed-end country fund prices are significantly correlated with the US stock index in the short run and with the NAV and foreign stock indexes in the long run. Khorana et al. (1998) examine changes in discounts and trading volume on closed-end funds around the introduction of WEBS and find that the WEBS’ performance is similar to that of closed-end funds in the 6-month period following the introduction of the WEBS. Furthermore, the authors find that closed-end country funds experience a decrease in trading volume and an increase in discounts from net asset value following the introduction of the WEBS. These studies suggest that ETFs may provide a more effective, low-cost strategy of diversifying internationally than CEFs. More closely related to the present study is the article by Patro (2001a), who investigates the announcement effect of listing seventeen WEBS (early name for international ETFs) on the returns of the corresponding market index returns and closed-end fund premiums. The author finds a positive market reaction for the market indexes and a decline in the premium for closed-end funds. Patro (2005), using a panel of 34 country funds from 18 countries during the 1981–1999 period, finds that closed-end fund premiums decrease in the months beginning with the announcement of new closed-end funds. Consequently, new funds are not timed when old funds are trading at large premiums. Patro concludes that existing premiums reflect barriers to international investments, which may be overcome by listing new financial instruments in foreign markets. Our study complements Patro’s studies by investigating the return and risk performance of ETFs subsequent to their listing compared to CEFs. Another study investigating ETFs is that by Olienyk et al. (1999), who utilize these exchange-traded funds to investigate the co-integration across 18 countries and find that substantial co-integration exists. They also utilize closed-end funds to investigate the existing co-integration. Pennathur et al. (2002) find that international iShares do replicate the foreign index but also have a high degree of US market exposure. Consequently, the potential for diversification is limited. The present study therefore extends this previous literature by comparing the risk and return performance of closed-end country funds, which represent another potential diversification method, and ETFs. Recently, Patro (2001b) conducts an empirical analysis of the performance of 45 international closed-end funds. The author uses alternate measures of performance to compare the sample of funds and 35 national market indices. The empirical evidence indicates that the risk-adjusted performance of the shares or the net asset values of the funds match the performance of their respective local market indices, as well as the world market indices and do not exhibit superior timing ability. The author’s findings are robust to conditioning on information. Our paper complements the findings of earlier papers by Patro in several

108

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

ways. First, we compare the performance of closed-end funds to that of corresponding ETFs in addition to the corresponding market indexes. This allows us to directly compare the performance of two tradable alternative investment vehicles. Second, we also investigate the CEFs’ alpha relative to the ETF, which enables us to focus on the factors that cause the difference in the risk-adjusted performance of ETFs versus closed-end funds. Patro shows that CEFs’ risk-adjusted performance is similar to the indexes themselves. Consequently, investing in CEFs is beneficial from a diversification perspective. By comparing the performance of ETFs to that of CEFs, we investigate whether investing in ETFs (which are based on the indexes) is more beneficial than investing in CEFs to diversify.

3. Hypotheses and research design Because traditional closed-end funds and their ETF counterparts have different characteristics, they may exhibit different risk and return characteristics, as discussed next. The question of superior performance of either investment vehicle has not been answered. We first compute the tracking error of the traded ETF relative to the untraded underlying index. We then compare the return and risk of the country ETFs to that of corresponding country funds. Finally, we compare the risk-adjusted performance of these investments and investigate what we call the “tradable Jensen’s alpha” of closed-end funds and try to explain any differences in performance. 3.1. Tracking error of ETFs If the ETF tracks an index perfectly, the return of an ETF is, on average, the dollar return on the index. The tracking error is the difference in the MSCI country index return and the corresponding ETF return. Formally, the tracking error, TE, is: TE = Ret − RMSCI ,

(1)

where Ret is the return on the iShare ETF and RMSCI is the return on the country’s MSCI index (or the S&P 500 index for the US). If the iShares mirror the underlying index perfectly, the tracking error should be equal to zero for each individual ETF, and the average tracking error across all ETFs should be zero as well. We therefore test the tracking errors for significance using a t-test for each country. Finally, we test the average differences across the countries’ tracking error. 3.2. Return Pennathur et al. (2002) find that ETFs closely mirror the underlying index and have low tracking errors to the index. However, since a closed-end fund is typically managed and not intended to mirror an index, the return of a closed-end fund may differ from the return of the index, and therefore the ETF, due to its portfolio composition. That is, closed-end funds could offer higher returns if they are capable of outperforming the corresponding index or ETF.

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

109

Since traditional closed-end funds are often comprised of assets not represented in the underlying index or of not all assets represented in the index, it is possible that these funds will either outperform or underperform the underlying index. In fact, financial research has found that closed-end funds generally do not outperform the market consistently and exhibit discounts (see Bal and Leger, 1996). Furthermore, as mentioned above, the expense ratio of traditional closed-end funds is substantially higher than the expense ratio of ETFs. Malhotra and McLeod (2000), for example, suggest that investors should include the expense ratio as a criterion for fund selection in addition to performance investment objectives and risk of the fund. Consequently, we hypothesize that traditional closed-end funds experience lower returns than their ETF counterparts. 3.3. Risk-adjusted returns While the traditional closed-end funds are expected to have returns that differ from exchange-traded funds (due to their higher expenses and discounts), they may also exhibit differing levels of risk (due to their investment decisions and exposure to premium volatility). Consequently, it is theoretically unclear whether traditional closed-end funds or ETFs will have higher risk-adjusted returns. To measure risk-adjusted returns, we utilize Sharpe ratios for both the traditional closed-end funds and the exchange-traded funds, as follows: SRce =

Rce − Rf σce

(2)

where SRce is the Sharpe ratio of the closed-end fund, Rce the monthly return of the closedend fund, Rf the risk-free rate, measured by the 1-month T-bill rate,1 and σ ce is the standard deviation of monthly closed-end fund returns. The Sharpe ratio for ETFs is measured as: SRet =

Ret − Rf σet

(3)

where SRet is the Sharpe ratio of the ETF, Ret the monthly return of the ETF, and σ et is the standard deviation of monthly ETF returns, and all other variables are defined as previously. We also compute the “tradable Jensen’s alpha” for each closed-end fund using the ETF as the benchmark index. We use the ETF instead of the actual index for two reasons. First, if the ETF has no tracking error, which we test above, then the ETF is essentially the same as the underlying index. Second, since both the ETF and the CEF are exchange-traded, they are viable investable alternatives for one another and should be directly compared to each other.2 To compute our tradable Jensen’s alpha, we regress the excess return (above the

1 We use the 1-month U.S. T-bill rate as a proxy for the risk-free interest rate for all countries investigated. Consequently, our focus is on a U.S. investor’s viewpoint who chooses between an investment in ETFs versus closed-end funds, both of which are traded in the US market. 2 We do compute the traditional Jensen’s alpha for the closed-end fund as well. We find that the matched pair difference for our tradable Jensen’s alpha and the traditional Jensen’s alpha are not statistically different for the sample of foreign closed-end funds and are marginally different when including US closed-end funds.

110

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

1-month T-bill rate) of the ETF against the excess return of the CEF, as follows: Rce − Rf = α + β × (Ret − Rf ) + εt ,

(4)

where Rce and Ret are defined above as the monthly return for the closed-end fund and the ETF, respectively, at time t, and Rf is the monthly risk-free rate proxied by the 1-month T-Bill rate. The intercept term, α, is the excess, risk-adjusted return above that of the benchmarked index. If closed-end fund managers have superior investment skill, especially in developing markets, then a positive alpha should be observed. However, market efficiency theory would indicate that the alpha should be zero or negative over an extended time period, such as in our sample.

4. Data and results 4.1. Data Our sample period includes monthly returns based on prices, not NAV, for international ETFs and 22 closed-end country funds from April 1996 to December 2001, resulting in 69 monthly observations. Furthermore, our sample includes monthly returns for SPDRs and seven US closed-end equity funds. While the US ETF began in 1992, we use a common time period beginning in April 1996. All comparisons between traditional closed-end funds and their counterpart ETFs require a research design that avoids distortions due to confounding factors such as economic conditions in various countries. For this reason, only those ETFs and traditional closed-end funds that can be matched up based on a specific country are included in the sample.3 Country index data were collected from Morgan Stanley Capital International (MSCI). ETF iShare data were collected from MSCI and the iShare returns (including dividends) from the Center for Research in Security Prices (CRSP), and closed-end funds data were obtained from CRSP for the period from April 1996 to December 2001. CEFs’ net asset values (NAVs) were collected from Compustat. This resulted in a sample of 29 closed-end funds for the 14 countries for which ETFs were available. Presently, more country ETFs have been created and are trading than are included in the study. However, the history of these funds is much shorter and not enough observations exist at this point to include in this study. The traditional closed-end funds that could be matched to counterpart ETFs are shown in Table 1. As shown in the table, data on closed-end funds for the 14 countries for which iShares were investigated were collected. When there was more than one closed-end fund representing a specific country, all traditional funds are included in the sample. Table 1 also shows the market value and premium (or discount) of the traditional closedend funds in March 1996, since ETFs were introduced in this month. The average market value of the closed-end funds was US$ 319 million in March of 1996, with a standard 3 We acknowledge that economic differences between countries render a comparison across countries difficult. However, a consistent superiority of ETFs versus closed-end funds for a majority of the countries investigated here illustrates that potential economic differences do not distort the results from an investor’s viewpoint.

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

111

Table 1 List of closed-end funds by countrya Country

Fund name

Ticker

Inception date

Market value

Premium (%)

Australia

First Australia Fund First Australia Prime Income

IAF FAX

12/19/1985 4/24/1986

145.033 1385.941

−15.12 −5.33

Austria France

Austria Fund France Growth Fund

OST FRF

9/28/1989 5/18/1990

103.864 155.368

−18.05 −15.84

Hong Kong

China Fund Greater China Fund Jardine Fleming China Fund

CHN GCH JFC

7/10/1992 7/15/1992 7/16/1992

144.758 149.75 102.386

3.12 1.13 −7.41

Germany

Emerging Germany Fund Germany Fund New Germany Fund

FRG GER GF

3/29/1990 7/18/1986 1/30/1990

105.06 158.554 398.137

−21.79 −18.63 −22.66

Italy

Italy Fund

ITA

2/28/1986

77.212

−13.01

Japan

Japan Equity Fund Japan OTC Equity Fund

JEQ JOF

7/24/1992 3/14/1990

138.792 99.61

15.89 16.20

Malaysia

Malaysia Fund

MF

5/4/1987

193.225

−6.82

Mexico

Emerging Mexico Fund Mexico Equity and Income Mexico Fund

MEF MXE MXF

10/10/1990 8/14/1990 6/11/1981

62.666 90.051 573.211

−27.86 −12.42 −9.45

Singapore

Singapore Fund

SGF

7/31/1990

98.023

5.39

Spain

Growth Fund Spain Spain Fund

GSP SNF

2/14/1990 6/28/1988

194.164 94.003

−13.46 −16.67

Switzerland UK

Swiss Helvetia Fund United Kingdom Fund

SWZ UKM

8/27/1987 8/14/1987

271.297 48.144

−14.51 −19.52

US

Gabelli Equity Trust General American Investors Liberty All-Star Equity Royce Value Trust Salomon Brothers Trust Source Capital Zweig Fund

GAB GAM USA RVT SBF SOR ZF

8/14/1986 1/30/1927 10/24/1986 11/26/1986 9/24/1929 10/24/1986 10/3/1986

850.434 503.668 872.94 283.146 1126.192 300.983 535.747

−3.66 −15.59 −1.17 −11.74 −14.47 −4.20 0.63

319.39 346.72

−9.21 10.88

Average Standard deviation

a Market value and premiums are measured as the month end market value and premium in March 1996 (month iShares (WEBS) were introduced).

deviation of US$ 347 million. Furthermore, the average discount for the closed-end funds was 9.21% in March 1996, with a standard deviation of 10.88%. 4.2. Tracking errors The average monthly tracking errors for each individual ETF are reported in Table 2. With the exception of Malaysia, the tracking errors are uniformly negative. However, none

112

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Table 2 Average monthly tracking error of returns Country

Average TE (%)

Standard error (%)

p-Value

Australia Austria Hong Kong France Germany Italy Japan Malaysia Mexico Singapore Spain Switzerland UK US

−0.025 −0.069 −0.163 −0.179 −0.012 −0.014 −0.069 0.812 −0.159 −0.035 −0.071 −0.190 −0.044 −0.0056

0.204 0.261 0.372 0.979 0.181 0.152 0.208 2.634 0.225 0.401 0.181 0.236 0.184 0.453

0.903 0.794 0.663 0.856 0.949 0.929 0.742 0.759 0.480 0.932 0.695 0.423 0.814

Statistical test of tracking errors International ETFs Average TE (%) Standard deviation (%) t-Statistic

−0.017 0.257 −0.235

US and international ETFs Average TE (%) Standard deviation (%) t-Statistics

−0.016 0.247 −0.242

Tracking error is measured by the difference between the monthly iShare return for each country and the corresponding dollar return on the MSCI index, TE = Ret − RMSCI . Tracking errors are measured over the April 1996–December 2001 period.

of the tracking errors are statistically significant different from zero. Thus, we conclude that the ETF tracks the underlying benchmark well and can be used as a tradable benchmark index. We also test the tracking errors across countries and find that while they are mostly negative, in aggregate, they are not significantly different from zero. The findings in this table support the claim that the ETFs closely mirror the underlying index, even though they may not be exact replication of the underlying index. 4.3. Comparison of returns The mean returns as well as the standard deviations of those returns for the traditional closed-end funds and for the ETFs in the sample are presented in Table 3 . The returns used are average monthly returns computed for CRSP data for the period April 1996–December 2001. Panel A of Table 3 shows the mean monthly returns for the ETFs and closed-end funds for 14 countries and 69 months in the sample. Interestingly, most of the closed-end funds have lower returns than the corresponding ETF. There are six international closed-end funds and two domestic funds with average returns greater than the ETF. Most of these positive differences in average returns are smaller with regard to the absolute differences. ETFs with

Table 3 Monthly average returns and standard deviations of ETFs and closed-end funds Mean return (%)

Mean return (%)

6.327

0.006

5.455

−0.100

EWA 0.308

Austria France Hong Kong Germany

5.880

0.206

10.135

0.346

Italy

6.659

1.049

Japan

7.103

0.584

6.912

−1.019

Malaysia Mexico Singapore

19.105

−1.047

10.138

0.184

Spain

10.799

−0.899

−0.018

JFC 6.922

−0.348

5.412

−0.273

GER

6.398 GF 5.850

5.601 JOF 5.756

−0.408

6.871

6.577

1.291

7.581 MXE 5.960

0.731

MXF 7.411

0.627

6.630

Mean return (%)

Standard deviation (%)

SGF

EWP 1.114

3.169

MEF

EWS −0.620

0.174

MF

EWW 1.250

GCH 6.274

JEQ

EWM 0.207

4.722

ITA

EWJ −0.728

3.559

7.169

FRG

EWI 0.982

Standard deviation (%)

FAX

CHN

EWG 0.641

−0.043

Mean return (%)

FRF

EWH 0.267

4.401

Standard deviation (%)

OST

EWQ 0.868

Mean return (%)

IAF

EWO −0.283

Standard deviation (%)

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Panel A Australia

Standard deviation (%)

5.641 GSP

SNF 2.906

0.469

7.785 113

114

Mean return (%) Switzerland

Standard deviation (%)

Mean return (%)

5.491

0.176

EWL 0.351

UK US

4.179

0.726

4.940

0.654

Mean return (%)

3.288

0.960

2.918

0.599

Standard deviation (%)

Mean return (%)

Standard deviation (%)

3.466 2.057 GAB

GAM 3.254

1.114

4.316

1.054

SBF 0.440

Panel B Average return (%) Test statistics

Standard deviation (%)

UKM

SPY 1.048

Mean return (%)

SWZ

EWU 0.690

Standard deviation (%)

RVT

SOR

4.380 USA

ZF 3.433

0.231

3.379

International funds

Domestic funds

All funds

−0.348 −3.677***

−0.326 −2.599***

−0.342 −4.457***

Average monthly mean returns and standard deviations for ETFs and the corresponding closed-end funds from April 1996–December 2002 (69 observations). FRG, MEF, GSP, and UKM had less than full observations. For US ETFs and CEFs, only data for the sample period above is reported even though the SPDR began in 1992. Statistical analysis for the US market does include the full sample period. *** Significant at the 1% level.

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Table 3 (Continued)

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

115

the highest average return are the Spanish and Mexican ETFs. The closed-end funds with the greatest difference from the ETF are a German fund and Japanese fund. In general, the results disclosed in Table 3, Panel A shows that the ETFs have a higher mean return than their corresponding closed-end funds. However, they also have lower standard deviations of monthly returns. Nevertheless, there are a few exceptions where the closed-end fund’s standard deviation is greater than that of the corresponding ETF. Panel B of Table 3 directly tests whether the combined differences in average monthly returns between CEFs and ETFs are statistically significant. The statistically negative results indicate that the ETFs outperform the closed-end funds for all funds combined, for foreign funds only, and for US funds only. For example, the average difference in monthly returns between closed-end funds and ETFs is −0.34% for both domestic and international funds, which is statistically significant at the 1% level. 4.4. Comparison of risk-adjusted returns Since the previous section indicates that ETFs have higher returns and higher standard deviations than their corresponding closed-end funds, a comparison of risk-adjusted returns is appropriate. As previously discussed, ETFs can be actively traded in the market throughout the day and have been shown to track the corresponding country index. Closed-end funds are also actively traded and their prices adjust throughout the day. ETFs are, by definition, a passive investment strategy. However, CEFs are typically actively managed and the NAV can differ from the market price, resulting in a premium or discount. Consequently, comparing the Sharpe ratios for ETFs and closed-end funds provides an indication whether an active trading strategy (closed-end funds) provides superior results to a passive trading strategy. Table 4 displays the Sharpe ratios for ETFs and closed-end funds. The first column displays the Sharpe ratios for the ETFs, while the remaining columns display the Sharpe ratios for the corresponding closed-end funds. In general, the Sharpe ratios for the ETFs are substantially higher than for the CEFs; only 7 of the 29 closed-end funds exhibit higher risk-adjusted returns than the corresponding ETF. The differences in the Sharpe ratios between the CEF and the ETF are tested using the Jobson and Korkie (1981) test statistic. While most CEF Sharpe ratios are lower than the corresponding ETF, there are four that are statistically significant. However, none of the CEFs investigated exhibit a statistically higher Sharpe ratio than the corresponding ETF. The results in Table 4 show that passive investing (i.e., investing in ETFs) provides better risk-adjusted returns in most country markets. 4.5. Closed-end funds and Jensen’s alphas The Sharpe ratio results displayed in Table 4 indicate that a passive investment strategy works well in foreign markets. To further investigate this possibility, we compute Jensen’s alpha for each of the 29 closed-end funds in the sample. Jensen’s alpha is a measurement of the portfolio manager’s added value. Since the results reported in other studies and verified in our sample indicate that ETFs track the underlying index closely and have higher Sharpe ratios than traditional closed-end funds (Table 4), we should expect the funds’ tradable

116

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

alphas to be negative.4 By utilizing the ETFs’ return to determine the market risk premium (to US investors), the resulting alpha gives a direct comparison between an investment in ETFs versus closed-end country funds. For example, a significantly negative tradable Jensen’s alpha for a country would indicate that the active management of a closed-end fund underperforms relative to a passive investment strategy utilizing ETFs. Results from estimating the tradable Jensen’s alpha using Eq. (4) for the 29 funds are displayed in Table 5. We report the alpha for each fund as well as the slope coefficient for the risk premium and the fit of the model (R-square).5 Panel A shows the results for the 22 international funds. As shown in the panel, six funds have statistically significantly negative alphas, indicating that the funds’ portfolio managers do not add value compared to ETFs. However, even though only six of the alphas are significant, 20 out of the 22 funds have negative alphas. Furthermore, the average alpha of −0.397% is highly significant across all international funds (t = −5.11). Panel B of Table 5 shows the results for the US funds, which are very similar to the findings for the foreign funds. Of the seven closed-end funds, three funds exhibit negative alphas, but none of the estimated alphas are statistically significant. Overall, the results reported in Table 5 support the notion that a passive investment strategy of investing in ETFs is superior to investing in closed-end country funds. To further investigate the negative tradable Jensen alphas reported in Table 5, a multiple regression model is utilized in Table 6. Specifically, alpha is regressed on several variables including the premium standard deviation (SDP), the percentage change in the premium over the sample period (PRE), and the change in net asset value (NAV) over the sample period. The broad model takes the form of: αi = bi,t + b1 SDPi + b2 PREi + b3 NAVi + b4 Sizei + b5 Expensei + b6 Turni + εi , (5) where αi , Jensen’s alpha reported in Table 5 for closed-end fund i; SDPi , the standard deviation of the closed-end fund’s premium for fund i over the sample period; PREi , the percentage change in the closed-end fund’s premium for fund i over the sample period; NAVi , the percentage change in net asset value for fund i over the sample period; Expensei , the average CEF expense ratio minus the average ETF expense ratio; it is the excess expense ratio of the CEF; Turni , the average CEF portfolio turnover ratio minus the average ETF portfolio turnover ratio; it is the excess turnover ratio for the CEF and measures CEF trading activity; Sizei , the relative logged size of the closed-end fund; average logged CEF NAV minus average logged ETF NAV. If active fund management is successful, we would expect higher alphas to be negatively related to the standard deviation of premiums but positively related to the fund’s net asset 4 We did compute Jensen’s alpha for the sample of ETFs separately. Because of the negative tracking error (discussed above) the ETF does have a negative alpha. However, the tradable Jensen’s alpha is not statistically different from the traditional Jensen’s alpha (see footnote 2). In addition, the ETFs alphas are higher than the CEF alphas. We continue to use the tradable Jensen’s alpha in empirical analysis since it represents alternative investment vehicles. 5 A simple t-test on the mean alpha was conducting by dividing the mean by the standard error of estimated alphas. The null hypothesis is that the average alpha is equal to zero.

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

117

Table 4 Sharpe ratios for iShares for each country and closed-end country fundsa Country

iShare

Closed-end funds

Australia

EWA −0.014

IAF −0.089

Austria

EWO −0.125

OST −0.070

France

EWQ 0.080

FRF −0.041*

Hong Kong

EWH −0.013

CHN −0.008

GCH −0.033

JFC −0.117

Germany

EWG 0.036

FRG 0.202

GER −0.077

GF −0.115

Italy

EWI 0.082

ITA 0.033

Japan

EWJ −0.163

JEQ −0.246

Malaysia

EWM −0.010

MF −0.191**

Mexico

EWW 0.084

MEF −0.038*

Singapore

EWS −0.094

SGF −0.230

Spain

EWP 0.109

GSP 0.302

Switzerland

EWL −0.009

SWZ −0.064

UK

EWU 0.070

UKM 0.155

US

SPY 0.131

GAB 0.078 SBF 0.009

FAX −0.124

JOF −0.118

MXE 0.045

MXF 0.034

SNF 0.012

GAM 0.217 SOR 0.224

RVT 0.128 USA 0.058

ZF −0.050*

a

Sharpe ratios computed over the sample period April 1996 (the first full month iShares were issued) to December, 2001 for international funds and for the entire period for SPDRs. Risk free rate is proxied by the 1-month T-Bill rate. * Significant at the 10% level. ** Significant at the 5% level.

value and premium. However, we expect alphas to be negatively related to expense ratios and turnover ratios as these increase the cost of closed-end fund operations and reduce potential tradable Jensen’s alpha. We also include Expense, Turn, and Size as control variables. Data were collected from the annual reports of each of the closed-end funds and ETFs over the sample period and then averaged.

118

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Table 5 Estimates of Jensen’s alpha for closed-end fundsa N

Jensen’s alpha (%)

Beta

R2

Panel A: Foreign funds IAF FAX OST FRF FRG GER GF CHN GCH JFC ITA JEQ JOF MF MEF MXE MXF SGF GSP SNF SWZ UKM

69 69 69 69 36 69 69 69 69 69 69 69 69 69 36 69 69 69 32 69 69 36

−0.344 −0.415 −0.184 −0.522* 0.219 −0.574 −0.825* 0.005 −0.115 −0.682 −0.184 −0.767 −0.068 −1.388** −0.769** −0.172* −0.258 −0.880** −0.076 −0.363 −0.199 −0.163

0.544*** 0.305*** 0.463*** 0.700*** 0.462*** 0.650*** 0.632*** 0.442*** 0.528*** 0.497*** 0.633*** 0.578*** 0.657*** 0.304*** 0.504*** 0.592*** 0.572*** 0.410*** 0.374*** 0.606*** 0.499*** 0.328***

0.605 0.283 0.111 0.757 0.710 0.634 0.511 0.501 0.592 0.613 0.639 0.474 0.427 0.577 0.872 0.746 0.759 0.610 0.890 0.465 0.622 0.404

Panel B: US funds GAB GAM RVT SBF SOR USA ZF

69 69 69 69 69 69 69

0.022 0.422 0.252 −0.367 0.480 −0.106 −0.470

0.359*** 0.421*** 0.477*** 0.629*** 0.270*** 0.471*** 0.465***

0.286 0.392 0.278 0.512 0.196 0.455 0.458

International funds only Tests of average alphas for funds Average (%) −0.397 S.D. (%) 0.371 t-Statistics −5.011***

International and domestic funds −0.293 0.410 −3.846***

a Jensen’s alpha computed as R − R = α + β × (R − R ) + ε , using country iShare returns (SPDRs for ce f et f t US) as the benchmark index. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

Table 6 displays the regression results for various alternative models and for all closedend funds. The control variables excess expense ratio (Expense) and excess portfolio turnover (Turn) do not significantly explain the estimated tradable alpha for the international sample or the sample including domestic funds. Hence, we drop these variables from subsequent models.

Table 6 Parameter estimates of the tradable Jensen’s alpha of closed-end funds Variable

Estimated coefficient

R-square Adjusted R-square N

0.256 −0.042 22

Panel B: Foreign and domestic closed-end funds Intercept −0.0197 (−4.59)*** SDP 0.0028 (0.25) PRE 0.0008 (1.22) NAV 0.0051 (1.21) Size −0.0021 (−2.19)** Expense −0.0039 (−0.86) Turn 0.0003 (0.58) R-square Adjusted N

R-square

0.472 0.327 29

Estimated coefficient

Estimated coefficient

−0.0208 (−9.02)*** −0.0420 (−1.93)* 0.0008 (1.41)

−0.0209 (−8.66)*** −0.0599 (−2.02)* −0.0011 (−0.28)

−0.0206 (−8.82)*** −0.0467 (−1.93)* 0.0009 (1.44) −0.0019 (−0.49)

0.264 0.186 22

0.190 0.105 22

0.274 0.153 22

−0.0203 (−9.46)*** −0.0008 (−0.07)

−0.0215 (−11.23)*** 0.0106 (0.23) 0.0012 (2.13)**

0.0058 (1.39) −0.0023 (−3.26)***

−0.0021 (−2.99)***

−0.0207 (−10.06)*** 0.0014 (0.13) 0.0011 (1.86)* 0.0042 (1.03) −0.0020 (−2.92)***

0.367 0.291 29

0.422 0.353 29

0.447 0.355 29

119

Estimates of regression model to explain the computed Jensen’s alpha for closed-end funds. SDP is the standard deviation of the closed-end fund premium, PRE the percentage change in the closed-end fund premium over the sample period and NAV is the change in net asset value (value added to portfolio through investment) over the sample period. The closed-end fund premium is calculated as (Price − NAV)/NAV. Size is the difference in the logged average NAV size of the CEF and ETF. Expense and Turn are the differences in the average expense ratio and portfolio turnover ratio for the CEF and the ETF. * Significance at the 10% level. ** Significance at the 5% level. *** Significance at the 1% level.

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Panel A: Foreign closed-end funds only Intercept −0.0167 (−3.06)*** SDP −0.0458 (−1.49) PRE 0.0007 (0.90) NAV −0.0004 (−0.08) Size −0.0009 (−0.69) Expense −0.0017 (−0.37) Turn −0.0003 (−0.45)

Estimated coefficient

120

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Panel A estimates the model for the sample of only international closed-end funds’ alphas. The first point to note is that the intercept is negative and statistically significant for all models, indicating that the tradable Jensen’s alpha for closed-end funds is negative, and confirming the results of the univariate tests in Table 5. For international closed-end funds, the standard deviation of the premium coefficient, SDP, is negative and significant at the 10% level in all three models, indicating that a changing or volatile premium will decrease performance. For example, for the model containing all three variables, a 1% increase in the premium standard deviation is associated with a reduction in alpha by over 4.5%. Irrespective of the model used, the change in the premium and the NAV are insignificant. Using the full sample, which includes domestic and international closed-end fund, fund size becomes the dominant determinant of the tradable alpha, with larger funds exhibiting statistically significant worse performance. These results are strongly influenced by US funds, which are much larger than foreign closed-end funds. As a result, it is difficult for these funds to consistently outperform the ETF benchmark, especially with larger fees in these closed-end funds. The premium standard deviation loses explanatory power in all models. However, the tradable Jensen’s alpha is positively related to the increase in the closed-end fund premium but not to changes in the net asset value of the fund. This relationship is statistically significant for all models in the full sample. A positive relationship is also observed for the international only sample, but is not statistically significant. The finding that the fund’s alpha is related to changes in the premium and not asset value supports the notion that investment managers do not make superior investment decisions.

5. Summary and conclusion Our objective is to compare the risk and return performance of ETFs to closed-end funds. Consequently, the present study contributes to the existing literature on exchange-traded funds by verifying the small tracking error of the ETF to the underlying index and then directly comparing the ETF performance to that of closed-end country funds. We utilize 29 closed-end country funds for 14 countries over the sample period from April 1996 to December 2001. The proxies for performance are mean returns and risk-adjusted returns over the sample period. Our results show that ETFs exhibit higher mean returns than foreign closed-end funds, which we attribute to lower expense ratios. Second, our results indicate that ETFs have higher Sharpe ratios, on average, than corresponding closed-end funds. This indicates that a passive investment strategy utilizing ETFs may yield superior to an active investment strategy using closed-end country funds. This result is further supported by the finding that a majority of the 29 closed-end funds utilized here exhibit negative alphas over the sample period. Moreover, these alphas are significantly negatively related to the standard deviation of the fund premium or discount for international funds and changes in the premium or discount for domestic funds, which suggests that active fund managers may wish to focus on the premium. The comparison of ETFs and closed-end funds offers some insight on the relative advantages of each type of investment. Given the results found here, there may be some potential for additional types of ETFs that are less exposed to risk. Specifically, ETFs could track

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

121

contrived indexes monitored by the market. Such ETFs may be able to offer lower risk while remaining focused on a particular country.

Acknowledgements We would like to thank Ike Mathur (editor), an anonymous reviewer of this journal, Anita Pennathur, Dilip Patro, Mario Reyes, session participants at the 2002 FMA meetings, seminar participants at Louisiana Tech University, The University of New Orleans, and The University of North Florida for helpful comments and suggestions.

References Anderson, S.C., 1986. Closed-end funds versus market efficiency. Journal of Portfolio Management 13, 63–65. Bal, Y., Leger, L.A., 1996. The performance of UK investment trusts. The Service Industries Journal 16, 67– 81. Bernstein, R.S., 2004. A comparison of tax efficiencies of ETFs, Vipers, and open and closed-end funds. Corporate Taxation 31, 34–38. Blume, M.E., Edelen, R.M., 2002. On replicating the S&P 500 index. Working Paper, The Wharton School University of Pennsylvania. Chiang, T.C., Kim, D., 2003. On country-fund price behavior—an empirical analysis of co-integrating factors. Advances in Financial Planning and Forecasting 11, 85–112. Cheng, A., Copeland, T., O’Hanlan, J., 1994. Investment trust discounts and abnormal returns: U.K. evidence. Journal of Business, Finance, and Accounting 21, 813–831. DeLong, J.B., Schleifer, A., 1992. Closed-end fund discounts. Journal of Portfolio Management 18, 46–53. Dellva, W.L., 2001. Exchange-traded funds not for everyone. Journal of Financial Planning 14, 110–124. Fuhr, D., 2001. Exchange-traded funds: a primer. Journal of Asset Management 2, 260–273. Gastineau, G.L., 2001. Exchange-traded funds: an introduction. Journal of Portfolio Management 27, 88–96. Gastineau, G.L., 2004. The benchmark index ETF performance problem. Journal of Portfolio Management 30, 96–103. Jobson, J.D., Korkie, B.M., 1981. Performance hypothesis testing with the Sharpe and Treynor measures. Journal of Finance 36, 889–908. Khorana, A., Nelling, E., Trester, J.J., 1998. The emergence of country index funds. Journal of Portfolio Management 24, 78–84. Lee, C.M.C., Schleifer, A., Thaler, R.H., 1990. Anomalies, closed-end mutual funds. Journal of Economic Perspectives 4, 153–164. Leonard, D., Shull, D.M., 1996. Investor sentiment and the closed-end fund evidence: impact of the January effect. Quarterly Journal of Economics and Finance 36, 117–126. Malhotra, D.K., McLeod, R.W., 2000. Closed-end fund expenses and investment selection. The Financial Review 35, 85–104. Olienyk, J.P., Schwebach, R.G., Zumwalt, J.K., 1999. WEBS, SPDRs, and country funds: an analysis of international co-integration. Journal of Multinational Financial Management 9, 217–232. Patro, D.K., 2001a. Market segmentation and international asset prices: evidence from the listing of World Equity Benchmark Shares. Journal of Financial Research 24, 83–98. Patro, D.K., 2001b. Measuring performance of international closed-end funds. Journal of Banking and Finance 25, 1741–1767. Patro, D.K., 2005. Stock market liberalization and emerging market country fund premiums. Forthcoming Journal of Business, 78. Pennathur, A.K., Delcoure, N., Anderson, D., 2002. Diversification benefits of iShares and closed-end country funds. Journal of Financial Research 25, 541–557.

122

J.T. Harper et al. / Int. Fin. Markets, Inst. and Money 16 (2006) 104–122

Pontiff, J., 1995. Closed-end fund premia and returns, implications for financial market equilibrium. Journal of Financial Economics 37, 341–370. Poterba, J.M., Shoven, J.B., 2002. Exchange-traded funds: a new investment option for taxable investors. The American Economic Review 92, 422–427. Richards, M., Fraser, D.R., Groth, J.C., 1980. Winning strategies for closed-end funds. Journal of Portfolio Management 7, 50–55. Thompson, R., 1978. The information content of discounts and premiums on closed-end fund shares. Journal of Financial Economics 6, 151–186. Zweig, M.E., 1993. An investor expectations stock price predictive model using closed-end fund premiums. Journal of Finance 28, 67–79.