Emerging Markets Review 8 (2007) 167 – 180 www.elsevier.com/locate/emr
What do bond holdings reveal about international funds' preferences? ☆ Yingbin Xiao ⁎ International Monetary Fund, 700, 19th Street NW, Washington, DC 20431, USA Received 12 December 2006; received in revised form 13 February 2007; accepted 4 March 2007 Available online 13 March 2007
Abstract International bond funds are important institutional investors in emerging markets and their asset allocation decisions have significant implications for bond market developments and debt management policies in developing countries. This paper studies emerging market bond holdings of international funds and analyzes economic and financial factors affecting their bond preferences. It shows that mutual funds prefer to invest in countries with sound fundamentals and more openness to trade. In addition, they favor bonds with high past returns and yields while averting bonds with high transaction costs and idiosyncratic risks. © 2007 Elsevier B.V. All rights reserved. JEL classification: G10; G11; G25; G23 Keywords: Bond funds; Portfolio holdings; Emerging markets; Asset allocation
Many studies have highlighted the importance of investigating institutional investors' preferences for assets. Institutional investors are different from individual investors. Their demand for assets is not solely driven by conventional wisdom of return, risk, and cash-flow considerations, as predicted by portfolio theory. Several studies (Falkenstein, 1996; Kang and Stulz, 1997; Dahlquist and Robertsson, 2001) document that institutional investors exhibit significant preferences for certain asset characteristics. Their preferences play a significant role in ☆
I am grateful to Jonathan Batten (the editor), Campbell Harvey, and Y. Julia Xiao for comments and suggestions. I would like to thank Robert Adler from AMG Data Services and Gloria Kim from JP Morgan for their help with data questions. The usual disclaimer applies. ⁎ Tel.: +1 202 623 8679. E-mail address:
[email protected]. 1566-0141/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ememar.2007.03.001
168
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
shaping asset demand and thus have implications for asset pricing, herding, and investor recognition (Gompers and Metrick, 2001). All these intellectual inquiries focus on one asset class-stocks. This begs the question of whether findings from stocks hold for another equally important asset class-bonds. Do institutional investors plow money into bonds the same way as into stocks? Despite the extensive literature on institutional investors' preferences for both domestic and international stocks,1 very little work has been done on their preferences for domestic bonds, much less international bonds. This paper tries to fill the gap by focusing on one class of international bonds— those of emerging market issuers. With emerging market economies maturing and growing in global importance, emerging market bond markets have become one of the fastest growing markets. In line with the broadening and deepening of emerging market bond markets, emerging market bonds have emerged as an increasingly important asset class in institutional investors' holdings, offering investors the potential for diversification and exposure to emerging market credits. Substantial research has been conducted to understand yield spreads (Bekaert and Harvey, 2003; Duffie et al., 2003; Batten et al., 2006). Knowing what emerging market bonds appeal to institutional investors would enhance our understanding of their behavior in allocating investments across major asset classes. In this paper, I document the revealed preferences of U.S. dedicated emerging market mutual funds for emerging market sovereign bonds. Emerging market bonds are held by a diverse group of investors, such as dedicated emerging market mutual funds, crossover international mutual funds, hedge funds, high-grade retail investors, and non-U.S. institutions, among which dedicated emerging market mutual funds are the dominant investors of emerging market bonds. Using the data from AMG Mutual Fund Holdings Database, an independent data provider of mutual fund holdings, I identify a set of common economic and financial variables important to mutual fund investment decisions over emerging market bonds. In particular, I ask the following questions: (i) Do a country's economic fundamentals affect emerging market bond fund demand? and (ii) Do a bond's financial characteristics affect emerging market bond fund demand? Empirical results show that mutual funds demonstrate strong preferences towards emerging market bonds with certain characteristics. Both country economic and bond financial characteristics have a bearing on emerging market bond fund demand. Specifically, bond funds prefer to invest in bonds issued by countries with sound fundamentals and more openness to trade. Sound fundamentals such as fiscal discipline, high reserves, and favorable current account (CA) position reduce the countries' balance sheet risk, strengthen their repayment capacity, and lower their leverage. More trade openness boosts the countries' visibility and increases investors' familiarity with the countries' bonds. Analogous to the findings on preferences for stocks, mutual funds favor bonds with high past returns and yields while averting bonds with high transaction costs and idiosyncratic risks. The rest of the paper proceeds as follows. Section 1 describes the data. Section 2 discusses determinants of the bond fund allocation. Section 3 details the methodology, results, and robustness checks. Section 4 concludes. 1. Data The dataset for analysis was created by merging data on bond holdings with data on country and bond attributes. Portfolio holdings data come from AMG Mutual Fund Holdings Database. It 1
Among others, see Falkenstein (1996), Kang and Stulz (1997), Dahlquist and Robertsson (2001), Aggarwal et al. (2003), Gelos and Wei (2005).
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
169
provides bond holdings, including zero holdings, reported by 44 bond funds in 2003. These funds have a stated objective of investing primarily in emerging market bonds. For each fund, the database gives the value of assets under management (AUM) and report dates, which range from May 2003 to December 2003. For each bond held by each fund, the database lists some characteristics of bonds including coupon type and rate, issuing date, maturity date, issuer, currency denomination, market value, and par value. Data on country and bond attributes come from International Financial Statistics (IFS), Bloomberg, Datastream, and JP Morgan. To make sure country and bond characteristics are available to bond funds before their investments and to steer clear of the endogeneity problems, I consider the country data at the end of 2002 and the financial data before the report date of each fund. The bond funds in the AMG database invest predominately in dollar-denominated sovereign bonds. The dominance can be explained by both supply and demand factors. Emerging market debt markets took off with the debut of Brady bonds as a solution to the debt crises in the 80s. In 1994, emerging market debt markets saw the Mexican financial crisis and in 1995 the subsequent collapse of the Latin American markets. Starting from the late 90s and early 2000, emerging market debt markets went through another structural change. Improved economic fundamentals through EM reforms, coupled with ample global liquidity, led many emerging markets to retire Brady debt by either debt buybacks or exchanges. Eurodollar bonds, Yankee bonds and global bonds have risen in importance. These bonds tend to be dollar-denominated sovereign bonds, which reflect “original sin”2 and comprise the largest share of emerging market debt. For U.S. emerging market bond funds, these bonds are liquid, do not incur exchange-rate risk,3 and enjoy the highest credit supported by the sovereign countries. The sample bond funds manage over $16 billion of assets. The AUM of each fund ranges from $30 million to $1.3 billion. Their investments span 31 emerging market countries: Argentina, Brazil, Bulgaria, Chile, China, Colombia, Cote d'Ivoire, Croatia, Dominican Republic, Egypt, Ecuador, El Salvador, Hungary, Lebanon, Malaysia, Mexico, Morocco, Nigeria, Panama, Peru, Philippines, Poland, Russia, South Africa, South Korea, Thailand, Tunisia, Turkey, Ukraine, Uruguay, and Venezuela. Each fund invests in at least 14 emerging market countries and 20 countries on average. Funds with higher asset value tend to diversify geographically by investing in greater number of countries. Each fund on average puts 2.5% of its portfolio holdings in any single country. The bonds included in the sample are restricted to fixed-rate bonds with more than one year remaining to maturity. They are non-callable, non-puttable, non-convertible, and nonsinking fund. The number of bonds each fund holds varies from 50 to 120. On average, each fund allocates 0.34% of its holdings in any single bond. 2. Determinants of emerging market bond fund allocation Fund portfolio allocation is measured by the percentage amount of each bond fund invested in each emerging market bond. It is obtained by dividing the market value of the bond by that of the fund portfolio. There are no theories on factors important to bond funds' allocation except for the risk-return argument applicable for all kinds of securities, including emerging market bonds. Finance theory 2
See Eichengreen and Hausmann (1999). Although investing in emerging market dollar-denominated bonds eliminates currency risk to U.S. bond funds, the elimination may come at a cost to investors. Burger and Warnock (2006) show that returns of emerging market dollardenominated bonds tend to display more negative skewness than those of local currency bonds. 3
170
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
predicts that investors prefer securities with high returns given risk, which is typically measured by second moments of returns.4 It is generally believed that emerging market bonds are analogous to high yield corporate bonds, whose yields appeal to investors. As emerging market bonds under study are issued by governments, it seems reasonable to look for determinants of allocation from both country and bond attributes. Presumably, country attributes that could affect governments' repayment capacity and bond attributes that could affect investors' demand should be critical to mutual funds' decision. Governments' repayment capacity is driven by fiscal and balance of payment (BoP) positions. Balance sheet items of the government's fiscal and external accounts could reflect the fundamental health of the sovereign. Drawing from the literature on demand for stocks, investors' demand for bonds could also be driven by return, risk, and transaction cost considerations. I consider the following select variables as independent variables: 2.1. Fiscal balance This is defined as the difference between fiscal revenue and fiscal expenditure as a percentage of GDP. A high fiscal balance reduces the borrowing requirements of a sovereign, thus decreasing debt accumulation and lessening debt-service burden. As the present value of fiscal balance comprises assets of a country's sovereign balance sheet, a high fiscal balance indicates strong financial health and suggests a lower vulnerability. Since sovereign bonds are generally issued by countries to finance their fiscal deficit or fulfill their borrowing requirements, the government's fiscal tightening or loosening would affect its capacity to repay. Fiscal strength (deterioration) would indicate a lower (higher) probability of default and lead investors to be more (less) willing to hold bonds issued by the government. 2.2. Foreign exchange reserves Like the present value of the fiscal balance, foreign exchange reserves also make up assets of a country's sovereign balance sheet. In addition, they represent resources available to finance imports and sovereign external debt. They can be used to pay off foreign currency denominated debt and result in savings in debt-service obligations. High reserves imply high debt payment capacity. The more foreign exchange reserves a country holds, the more foreign exchange is available for interest payments and principal repayments. The role of foreign reserves to sovereign bonds is similar to that of earnings before interest, taxes, depreciation, and amortization (EBITDA) to high yield corporate bonds. 2.3. Current account balance This is defined as the difference between current account receipts and payments as a percentage of GDP. It is an important component of a sovereign's external balance sheet and is related to the sustainability of a sovereign's balance of payment position. It can also been seen as an indicator of economic leverage as countries run current account deficits not only to smooth consumption, but also to capitalize on investment opportunities. In a corporate setting, Kang and Stulz (1997) find that foreign investors prefer to invest in low-leverage Japanese firms. 4
The importance of taking into account higher moments of returns such as skewness can be found in Kraus and Litzenberger (1976), Harvey et al. (2004), and Burger and Warnock (2006).
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
171
2.4. Bilateral trade with the United States This is defined as exports and imports between the United States and a country as a percentage of GDP. Not only can trade generate more foreign exchange cash-flows, but also the more trade a country has with the United States, the more likely the country's bonds are known to mutual funds in the States. In this sense, this variable also serves as a proxy for information asymmetry. Merton (1987) argues that investors are likely to invest in familiar firms because of investor recognition. Falkenstein (1996) shows that mutual funds overweight stocks with high visibility. Kang and Stulz (1997) find that exporting Japanese firms have a greater foreign ownership and they interpret this as an information advantage enjoyed by the exporting firms. Dahlquist and Robertsson (2001) note that foreign ownership is greater in Swedish firms with foreign sales. However, Edison and Warnock (2004) find some evidence that U.S. investors favor stocks of non-tradeable industries such as transportation and communications. 2.5. Return The average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Returns factor in coupon payments, accrued interest, and changes in clean prices. Specifically, a bond's return is calculated as the difference of dirty prices between two payment dates plus coupon divided by the dirty price of the previous payment date. There is large literature on both domestic and international stock funds chasing past returns. For example, Grinblatt et al. (1995) find that a majority of U.S. domestic stock funds are momentum investors who buy stocks with higher past returns. Froot, O'Connell and Seasholes (2001) examine daily international portfolio flows of international stock funds. They find strong evidence that international mutual funds engage in positive feedback trading and chase past stock winners. However, Gompers and Metrick (2001) observe that institutional investors are not trend followers. 2.6. Total variance This is calculated as the variance of the monthly returns over the past 24–60 months, depending on data availability, prior to the report date of each fund. Falkenstein (1996) shows that mutual funds overweight stocks with low variance.5 Including total variance will allow us to examine the relation between bond fund holdings and volatility. 2.7. Residual variance This variable measures unsystematic risks. It is derived by regressing each bond's monthly excess returns on those of JP Morgan's EMBI Global (EMBIG) country index, which is a benchmark for U.S. dollar-denominated emerging market debt instruments in a country. Monthly excess returns are obtained by subtracting from average monthly returns 1-month dollar risk-free rates from Ibboson Associates. Campbell and Taksler (2003) illustrate the distinction between total variance and residual variance. Kang and Stulz (1997) and Dahlquist and Robertsson (2001) use residual variance to proxy idiosyncratic risks and report some evidence on how idiosyncratic risks affect Japanese and Swedish stock holdings by international investors. Edison and Warnock (2004) Falkenstein (1996) calls it “idiosyncratic risk,” but he does not differentiate the total variance and idiosyncratic variance. 5
172
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
find that U.S. portfolios in EM are tilted towards stocks with high beta and low residual volatility. Including both total variance and residual variance measures will allow us to examine their respective roles and identify relative importance on emerging market bond fund holdings. 2.8. Yield to maturity This is the yield that equates the discounted cash-flows of the bond to the bond price prior to the report date of the fund. One of the often cited motives of investors holding emerging market debt is the search for yield. Emerging market bonds are regarded as an asset class with a high yield just like high yield corporate debt in the United States. In addition, the yield to maturity of a bond is analogous to the book-to-market ratio of a stock. Falkenstein (1996), Kang and Stulz (1997), and Dahlquist and Robertsson (2001) report some evidence on the impact of book-to-market ratio on domestic and international stock holdings. 2.9. Bid-ask spread This is the average bid-ask spread over the past one year prior to the report date of each fund. Many studies show that transaction costs are important barriers to international investments. Falkenstein (1996), Kang and Stulz (1997), and Dahlquist and Robertsson (2001) observe that mutual funds tend to hold more liquid stocks, which have low transaction costs. In addition, ease in liquidating in crunch time is important to asset holders. As such, bid-ask spreads are used to proxy transaction costs of bonds. The higher the spreads, the higher the costs, the tighter the liquidity. Table 1 provides summary statistics of all independent variables. As can be seen, these variables vary substantially among different countries and bonds. The fiscal balance ranges from a deficit of 15% to a surplus of 2% of GDP. The current account balance ranges from a deficit of 14% to a surplus of 9% of GDP. On average, the foreign reserves stand at over $26 billion, varying from $0.5 billion to $290 billion. The trade value with the United States ranges from 1.5% to 40% of GDP. The monthly return is on average about 1.5%. As expected, the residual variance is lower than the total variance. The yield is relatively high, on average over 6%. The bid-ask spread varies from 1 to 25 basis points. Table 1 Summary statistics of independent variables
Fiscal balance (%) Reserves ($ billion) CA balance (%) Trade (%) Return (%) Total variance (%) Residual variance (%) Yield (%) Bid-ask spread (%)
Mean
Maximum
Minimum
− 3.96 26.52 − 1.21 12.90 1.49 0.41 0.07 6.58 0.11
2.33 292.05 8.95 39.96 3.75 3.38 0.56 13.80 0.25
−15.07 0.47 −13.78 1.52 −0.21 0.01 0.005 1.33 0.01
This table summarizes independent variables used in all the regressions. Fiscal balance, reserves, CA balance, and trade variables take the end-2002 values from IFS. Return variable is the average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Monthly returns are calculated based on price and coupon information from Bloomberg and Datastream. Total variance variable is the variance of the past 24–60 monthly returns. Residual variance variable is the variance of the error term obtained by regressing excess returns on EMBIG country index returns. Yield and bid-ask spread variables are calculated based on information from Bloomberg.
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
173
3. Estimation results To uncover the relation between fund demand and various economic and financial variables, I run univariate and multivariate regressions as well as robustness checks. As bond funds are not allowed to short bonds, bond holdings are non-negative. To take into account the impact of no short-selling imposed on mutual fund investments, I use Tobit regressions for base results and OLS regressions for supplements and robustness. 3.1. Univariate regressions The univariate regressions regress shares of bonds in each fund portfolio on each of the explanatory variables identified above while controlling for fund effects. As the explanatory variables do not vary with the fund, the standard errors are adjusted to allow for the clustering effect and avoid the inflated significance level. Results from Table 2 indicate that bond fund investments are greater in countries with sound economic fundamentals such as the high fiscal balance, foreign reserves, and current account balance and in countries having more trade with the United States. In addition, fund ownership is positively related to the return and yield, but negatively related to the residual variance and bid-ask spread, suggesting that bonds with better cash-flows, low idiosyncratic risks, and transaction costs appeal to institutional investors. Although the coefficient of the total variance variable is positive, it becomes negative after controlling for other variables. Table 2 Univariate regressions of bond fund demand
Fiscal balance Reserves CA balance Trade Return Total variance Residual variance Yield Bid-ask spread # of Observation
Tobit
OLS
0.171 (0.00) 0.001 (0.05) 0.119 (0.00) 0.023 (0.01) 1.705 (0.00) 1.043 (0.00) − 2.308 (0.03) 0.382 (0.00) − 18.01 (0.00) 3872
0.030 (0.00) 0.001 (0.00) 0.016 (0.00) 0.006 (0.01) 0.399 (0.00) 0.269 (0.00) − 0.356 (0.02) 0.064 (0.00) − 2.559 (0.00) 3562
This table presents univariate Tobit and OLS regression results. The dependent variable is a bond's weight in a bond fund. Fiscal balance, reserves, CA balance, and trade variables take the end-2002 values from IFS. Return variable is the average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Monthly returns are calculated based on price and coupon information from Bloomberg and Datastream. Total variance variable is the variance of the past 24–60 monthly returns. Residual variance variable is the variance of the error term obtained by regressing excess returns on EMBIG country index returns. Yield and bid-ask spread variables are calculated based on information from Bloomberg. Fund dummies are included, but not reported. The figures in parentheses are p-values calculated based on robust standard errors allowing for clustering.
174
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
Since all these variables are statistically significant at the 5% level, they are included in the following multivariate analyses. Table 2 also reports the OLS results, which are quite similar to the Tobit results in the signs and statistical significance of the coefficients. 3.2. Multivariate regressions Model (i) of Table 3 includes only the fiscal balance and bond returns. Including returns allows us to explore whether the return-chasing pattern exists in emerging market bond funds. Both variables are positively and significantly related to fund demand, suggesting that a strong fiscal position and bond performance attract investments. They remain positive and significant after controlling for other variables in all other specifications. As we will see later, the fact that their statistical significance does not diminish is one of the most robust results we observe over many specifications. Models (ii) to (v) of Table 3 include all the bond characteristics in addition to some country characteristics. All coefficients of bond characteristics are statistically significant at the 5% level. Funds bias their portfolio towards bonds with low idiosyncratic risks and high yields. Bid-ask spreads are negatively related to fund ownership, suggesting that high transaction costs discourage investments. Moreover, the positive and significant relation between trade and fund ownership indicate that trade openness promotes investments. In multivariate regressions, reserves and the current account balance remain to be positive, but no longer significant. Model (vi) includes all variables concerned and confirms results in other regressions. Table 3 Multivariate Tobit regressions of bond fund demand
Fiscal balance
(i)
(ii)
(iii)
(iv)
(v)
(vi)
0.062 (0.00)
0.145 (0.00)
0.125 (0.00)
0.126 (0.00)
0.130 (0.00) 0.004 (0.38)
0.018 (0.04) 1.424 (0.00) − 0.606 (0.04) − 5.234 (0.00) 0.473 (0.00) − 16.338 (0.00) 3872
0.005 (0.82) 0.019 (0.04) 1.417 (0.00) − 0.601 (0.05) − 5.186 (0.00) 0.474 (0.00) − 16.497 (0.00) 3872
0.131 (0.00) 0.001 (0.37) 0.010 (0.80) 0.015 (0.05) 1.424 (0.00) − 0.681 (0.03) − 5.171 (0.00) 0.481 (0.00) − 16.514 (0.00) 3872
Reserves CA balance Trade Return
1.540 (0.00)
Total variance Residual variance Yield Bid-ask spread # of Observation
3872
1.300 (0.00) − 0.602 (0.04) − 5.283 (0.00) 0.470 (0.00) − 17.957 (0.00) 3872
0.016 (0.05) 1.433 (0.00) −0.685 (0.03) −5.228 (0.00) 0.483 (0.00) −16.327 (0.00) 3872
This table reports multivariate Tobit regression results. The dependent variable is a bond's weight in a bond fund. Fiscal balance, reserves, CA balance, and trade variables take the end-2002 values from IFS. Return variable is the average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Monthly returns are calculated based on price and coupon information from Bloomberg and Datastream. Total variance variable is the variance of the past 24–60 monthly returns. Residual variance variable is the variance of the error term obtained by regressing excess returns on EMBIG country index returns. Yield and bid-ask spread variables are calculated based on information from Bloomberg. Fund dummies are included, but not reported. The figures in parentheses are p-values calculated based on robust standard errors allowing for clustering.
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
175
The statistical significance of the fiscal balance and trade after controlling for bond variables demonstrates that country economic attributes matter. Not surprisingly, funds exhibit strong preferences for countries with solid economic fundamentals such as fiscal strength, high foreign exchange reserves and low macro leverage. The positive sign of trade openness after controlling for other variables indicates that more bilateral trade with the United States raises the visibility of these countries' bonds. It supports the theory of Merton (1987) and the empirical evidence found in Falkenstein (1996), Kang and Stulz (1997), and Dahlquist and Robertsson (2001) regarding American, Japanese, and Swedish stock markets. The statistical significance of all the coefficients of the bond variables after controlling for all other country variables show that bond financial attributes matter. Consistent with the results in univariate regressions, funds are tuned into bonds with high past returns, low idiosyncratic risks, and high yields. Analogous to the findings of stock by Grinblatt, Titman, and Wermers (1995) and Froot, O'Connell, and Seasholes (2001), EM bond funds employ the similar momentum strategy as stock funds do in chasing stocks. Their preferences for bonds with low idiosyncratic risks dovetail with observations of stocks by Kang and Stulz (1997) and Dahlquist and Robertsson (2001). The total variance variable remains significant but reverses its sign to negative in the presence of the residual variance variable. The negative relation between fund ownership and two variance measures underscores mutual funds' strong aversion to risky bonds.
Table 4 Multivariate OLS regressions of bond fund demand
Fiscal balance
(i)
(ii)
(iii)
(iv)
(v)
(vi)
0.012 (0.04)
0.026 (0.00)
0.020 (0.01)
0.022 (0.01)
0.023 (0.01) 0.002 (0.05)
0.006 (0.01) 0.360 (0.00) − 0.084 (0.50) − 1.029 (0.00) 0.084 (0.00) − 2.336 (0.00) 0.10 3562
0.009 (0.20) 0.007 (0.01) 0.356 (0.00) − 0.081 (0.52) − 1.024 (0.00) 0.088 (0.00) − 2.606 (0.00) 0.10 3562
0.025 (0.00) 0.001 (0.08) 0.010 (0.16) 0.005 (0.05) 0.355 (0.00) − 0.125 (0.35) − 0.997 (0.00) 0.092 (0.00) − 2.591 (0.00) 0.10 3562
Reserves CA balance Trade Return
0.344 (0.00)
Total variance Residual variance Yield Bid-ask spread Adjusted R2 # of Observation
0.04 3562
0.321 (0.00) − 0.090 (0.47) − 1.034 (0.00) 0.081 (0.00) − 2.715 (0.00) 0.09 3562
0.008 (0.05) 0.359 (0.00) − 0.126 (0.36) − 1.006 (0.00) 0.089 (0.00) − 2.300 (0.00) 0.10 3562
This table reports multivariate OLS regression results. The dependent variable is a bond's non-zero weight in a bond fund. Fiscal balance, reserves, CA balance, and trade variables take the end-2002 values from IFS. Return variable is the average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Monthly returns are calculated based on price and coupon information from Bloomberg and Datastream. Total variance variable is the variance of the past 24–60 monthly returns. Residual variance variable is the variance of the error term obtained by regressing excess returns on EMBIG country index returns. Yield and bid-ask spread variables are calculated based on information from Bloomberg. Fund dummies are included, but not reported. The figures in parentheses are p-values calculated based on robust standard errors allowing for clustering.
176
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
The economic significance of the coefficients from the Tobit regressions is hard to assess because they reflect both the probability and magnitude. However, as can be seen in the following robustness checks, the coefficients of the explanatory variables are not only statistically, but also economically significant in the OLS regressions. 3.3. Robustness checks To make sure these results are not driven by zero holdings, I also run OLS regressions of positive holdings as robustness checks. Models (i)–(vi) in Table 4 include variables in the same way as in the Tobit regressions, producing broadly similar results. All the coefficients display the same signs as in Tobit and they are statistically significant except for current account balance and total variance. As can be seen from Table 4, these statistically significant coefficients are also economically significant. For example, based on model (vi), increasing a country's fiscal balance from its 25th percentile (− 6.36%) to its 75th percentile (1.36%) raises a mutual fund's bond ownership by 12.5 percentage points. Similarly, increasing the bond return from its 25th percentile (1.11%) to its 75th percentile (1.69%) boosts a mutual fund's bond ownership by 20.6 percentage points. Reducing the residual variance from its 75th percentile (5.58 basis points) to its 25th percentile (1.37 basis points) leads a mutual fund to increase its investment by 4.2 percentage points. Narrowing the bid-ask spread from its 75th percentile (15.2 basis points) to its 25th percentile (5.75 basis points) translates into an increase of 24.5 percentage points in bond demand. The overall fit of the regression is about 0.1. Table 5 Multivariate Tobit regressions of aggregated demand
Fiscal balance
(i)
(ii)
(iii)
(iv)
(v)
(vi)
0.261 (0.00)
0.487 (0.00)
0.392 (0.00)
0.387 (0.00)
0.430 (0.00) 0.011 (0.00)
0.181 (0.00) 4.631 (0.00) 2.592 (0.01) − 4.194 (0.00) 0.300 (0.00) 1364
0.0431 (0.38) 0.179 (0.00) 4.556 (0.00) 2.664 (0.00) − 4.278 (0.00) 0.319 (0.00) 1364
0.442 (0.00) 0.012 (0.00) 0.072 (0.18) 0.178 (0.00) 4.132 (0.00) 3.156 (0.00) − 3.916 (0.00) 0.335 (0.00) 1364
Reserves CA balance Trade Return
3.068 (0.00)
Total variance Residual variance Yield # of Observation
1364
4.110 (0.00) 2.212 (0.02) −4.500 (0.00) 0.323 (0.00) 1364
0.176 (0.00) 4.058 (0.00) 3.228 (0.00) − 4.071 (0.00) 0.317 (0.00) 1364
This table reports multivariate Tobit regression results. The dependent variable is the aggregated weight of bonds from the same country in a bond fund. Fiscal balance, reserves, CA balance, and trade variables take the end-2002 values from IFS. Return variable is the average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Monthly returns are calculated based on EMBIG country return index from Datastream. Total variance variable is the variance of the past 24–60 monthly returns. Residual variance variable is the variance of the error term obtained by regressing excess returns on EMBIG index returns. Yield data are obtained from JP Morgan. Fund dummies are included, but not reported. The figures in parentheses are p-values calculated based on robust standard errors allowing for clustering.
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
177
The second robustness checks change the dependent variable. Instead of analyzing the percentage amount invested in each EM bond and associating it with economic and financial variables, I analyze the percentage amount of each bond fund invested in each emerging market country and associate it with those variables. The new dependent variable is obtained by aggregating the market value of bonds issued by each country and dividing it by the fund's market value. In line with the foregoing regressions, independent economic variables are the fiscal balance, reserves, current account balance, and trade. Independent financial variables are the return, risk, and yield of a country's bond market as proxied by the EMBIG country index. Monthly total return level series of the EMBIG index and EMBIG country indices are from Datastream. Total returns include both interest and price changes. Depending on data availability, for each country, I calculate the average monthly returns over the 24–60 months ending in May 2003. Average monthly returns of the EMBIG country index proxy a country's bond market returns. Each country's bond market idiosyncratic risk is measured by the residual variance from regressing monthly excess returns of the EMBIG country index on those of the EMBIG index. The stripped yield of a country's EMBIG index, obtained from JP Morgan, represents its bond market's yield. Information on bid-ask spreads of the indices is not included because of unavailability. Table 5 presents the results. All the coefficients are significant except for the current account balance. Consistent with baseline regression results, the fiscal balance, foreign reserves, and trade are positively related while CA balance is negatively related to mutual funds' bond investments in Table 6 Multivariate OLS regressions of aggregated demand
Fiscal balance
(i)
(ii)
(iii)
(iv)
(v)
(vi)
0.097 (0.09)
0.244 (0.00)
0.217 (0.00)
0.221 (0.00)
0.354 (0.00) 0.018 (0.00)
0.106 (0.00) 3.494 (0.00) 3.240 (0.00) − 3.115 (0.00) 0.168 (0.00) 0.32 1282
0.026 (0.62) 0.105 (0.00) 3.437 (0.00) 3.285 (0.00) − 3.175 (0.00) 0.162 (0.00) 0.33 1282
0.348 (0.00) 0.019 (0.00) 0.114 (0.02) 0.094 (0.00) 2.399 (0.00) 3.995 (0.00) − 2.529 (0.00) 0.165 (0.00) 0.50 1282
Reserves CA balance Trade Return
2.628 (0.00)
Total variance Residual variance Yield Adjusted R2 # of Observation
0.21 1282
3.065 (0.00) 2.768 (0.00) − 3.156 (0.00) 0.145 (0.00) 0.28 1282
0.091 (0.00) 2.220 (0.00) 4.142 (0.00) −2.812 (0.00) 0.141 (0.00) 0.49 1282
This table reports multivariate OLS regressions. The dependent variable is the aggregated non-zero weight of bonds from the same country in a bond fund. Fiscal balance, reserves, CA balance, and trade variables take the end-2002 values from IFS. Return variable is the average monthly return over the past 24–60 months, depending on data availability, prior to the report date of each fund. Monthly returns are calculated based on EMBIG country return index from Datastream. Total variance variable is the variance of the past 24–60 monthly returns. Residual variance variable is the variance of the error term obtained by regressing excess returns on EMBIG index returns. Yield data are obtained from JP Morgan. Fund dummies are included, but not reported. The figures in parentheses are p-values calculated based on robust standard errors allowing for clustering.
178
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
a country. In addition, funds invest significantly more in bond markets that offer high returns, low risks, and high yields. To ensure that these results are not biased by zero holdings, I also run OLS regressions of nonzero holdings. As shown in Table 6, OLS results match those from Tobit and have an adjusted R squared of 0.5. The magnitude of these coefficients confirms the economic significance of these variables in affecting bond fund demand. For instance, based on model (vi), a one percentage point rise in a country's fiscal surplus, or conversely, a one percentage point fall in its fiscal deficit, raises a mutual fund's bond investments in a country by 34.8 percentage points. By the same token, a one percentage point increase in annualized bond returns leads a mutual fund to increase bond ownership by 20 percentage points. A one percentage point drop in annualized residual variance boosts a mutual fund's investments by 21 percentage points. In short, the various regressions and robustness checks yield broadly similar results. They clearly show that economic fundamentals, trade openness, and bond financial characteristics are not only statistically significant, but also economically significant in explaining bond fund demand. 4. Conclusion This paper is the first study to identify economic and financial variables important to mutual fund investment decision over emerging market bonds. Despite a history punctuated by periodic crises, emerging market bond markets have become one of the fastest growing markets. By analyzing mutual funds' revealed preferences towards emerging market bonds, the paper complements existing studies and advances our understanding of the behavior of institutional investors. It shows that both country economic and bond financial variables affect investors' willingness to hold emerging market sovereign bonds. Mutual funds favor bonds issued by countries with sound fundamentals and openness to trade. They overweight bonds with high returns and yields, but avert bonds with high risks and transaction costs, which mirrors previous findings that mutual funds prefer stocks with certain attributes. Bond funds' aversion of high risks may reflect their dislike for volatility or negative skewness. Burger and Warnock (2006) find that U.S. investors are not in favor of volatile or negatively skewed local currency bonds. However, the variance and skewness of bond returns in their sample are highly correlated, which makes it difficult to disentangle the effects of the variance and skewness. The clear bias of bond funds toward solid economic fundamentals may have implications for preventing financial crises. The ability to differentiate fundamentals could imply that dedicated emerging market bond funds are less likely to withdraw money because of contagion concerns. As they are the steadfast market forces, they provide greater stability and reduce volatility of emerging market bond markets. Sovereigns should take measures to support these institutions since they can spot the difference instead of following the herd. Their preferences towards certain bond characteristics such as low risks and transaction costs highlight the need for these markets to reduce volatility and barriers. Bond funds' preferences may also have important implications for debt management policies in emerging markets. Since a stable investor base could help reduce the vulnerabilities and provide steady financing, policy makers need to take concrete steps to broaden and diversify the institutional investor bases in emerging markets. As shown in the paper, less information asymmetry is conducive to fund investments. To enhance information flows between emerging markets and institutional investors, policy makers could put in place transparency policies, improve disclosure about economic developments and reform measures, establish investor relations programs, and facilitate the development of market infrastructure.
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
179
References Aggarwal, Reena, Klapper, Leora, Wysocki, Peter, 2003. Portfolio preferences of foreign institutional investors. Working Paper. Batten, Jonathan, Fetherston, Thomas, Hoontrakul, Pongsak, 2006. Factors affected the yields of emerging market issuers: evidence from the Asia-Pacific Region. Journal of International Financial Markets, Institutions & Money 16 (1), 57–60. Bekaert, Geert, Harvey, Campbell, 2003. Emerging markets finance. Journal of Empirical Finance 10, 3–55. Burger, John, Warnock, Francis, 2006. Foreign participation in local currency bond markets. NBER Working Paper, vol. 12548. Campbell, John, Taksler, Glen, 2003. Equity volatility and corporate bond yields. The Journal of Finance 58, 2321–2349. Dahlquist, Magnus, Robertsson, G., 2001. Direct foreign ownership, institutional investors, and firm characteristics. Journal of Financial Economics 59, 413–440. Duffie, Darrell, Pedersen, Lasse Heje, Singleton, Kenneth J, 2003. Modeling sovereign yield spreads: a case study of Russian debt. The Journal of Finance 58, 119–159. Eichengreen, B., Hausmann, R., 1999. Exchange rates and financial fragility. NBER Working Paper, vol. 7418. Falkenstein, Eric, 1996. Preferences for stock characteristics as revealed by mutual fund portfolio holdings. The Journal of Finance 51, 111–135. Froot, Kenneth, O'Connell, Paul, Seasholes, Mark, 2001. The portfolio flows of international investors. Journal of Financial Economics 59, 151–193. Gelos, Gaston, Wei, Shang-Jin, 2005. Transparency and international portfolio holdings. The Journal of Finance 60, 2987–3020. Gompers, Paul, Metrick, Andrew, 2001. Institutional investors and equity prices. The Quarterly Journal of Economics 116, 229–259. Grinblatt, Mark, Titman, Sheridan, Wermers, Russ, 1995. Momentum investment strategies, portfolio performance, and herding: a study of mutual fund behavior. American Economic Review 85, 1088–1105. Harvey, Campbell R., Liechty, John, Liechty, Merrill W., Mueller, Peter, 2004. Portfolio Selection With Higher Moments. Available at SSRN: http://ssrn.com/abstract=634141. Kang, Jun-Koo, Stulz, Rene, 1997. Why is there a home bias? An analysis of foreign portfolio ownership in Japan. Journal of Financial Economics 46, 3–28. Kraus, A., Litzenberger, R., 1976. Skewness preferences and the valuation of risky assets. The Journal of Finance 31, 1085–1100. Merton, Robert C., 1987. Presidential address: a simple model of capital market equilibrium with incomplete information. Journal of Finance 42, 483–510. Warnock, Hali, Warnock, Francis, 2004. U.S. investor's emerging market equity portfolios: a security level analysis. Review of Economics and Statistics 86, 691–704.
Further reading Blake, Christopher, Elton, Edwin, Gruber, Martin, 1993. The performance of bond mutual funds. Journal of Business 66, 371–403. Borensztein, Eduardo, Gelos, Gaston, 2000. A panic-prone pack? The behavior of emerging market mutual funds. IMF Working Paper, vol. 198. IMF. Burger, John, Warnock, Francis, 2003. Diversification, original sin and international bond portfolios. International Finance Discussion Papers #755, Federal Reserve. Chan, K., Covrig, V., Ng, L., 2005. What determines the domestic bias and foreign bias? Evidence from mutual fund equity allocations worldwide. The Journal of Finance 60, 1495–1534. Chen, Hsiu-Lang, Jegadeesh, Narasimhan, Wermers, Russ, 2000. An examination of the stockholdings and trades of fund managers. Journal of Financial and Quantitative Analysis 35, 343–368. Claessens, Stijn, Klingebiel, Daniela, Schmukler, Sergio, 2003. Government bonds in domestic and foreign currency: the role of macroeconomic and institutional factors. Centre for Economic Policy Research Discussion Paper, vol. 3789. Coval, Joshua, Moskowitz, Tobias, 1999. Home bias at home: local equity preference in domestic portfolios. The Journal of Finance 54, 2045–2073. Coval, Joshua, Moskowitz, Tobias, 2001. The geography of investment: informed trading and asset prices. Journal of Political Economy 109, 811–841. Detzler, Miranda, 1999. The performance of global bond mutual funds. Journal of Banking and Finance 1999, 1195–1217.
180
Y. Xiao / Emerging Markets Review 8 (2007) 167–180
Elton, Edwin, Gruber, Martin, Blake, Christopher, 1995. Fundamental economic variables, expected returns, and bond fund performance. The Journal of Finance 50, 1229–1526. Gande, A., Parsley, D., 2005. News spillovers in the sovereign debt market. Journal of Financial Economics 75, 691–734. Gebhardt, William, Hvidkjaer, S., Swaminathan, B., 2005. The cross-section of expected corporate bond returns betas or characteristics? Journal of Financial Economics 75, 85–114. Kaminsky, G., Lyons, R., Schmukler, S., 2001. Mutual fund investment in emerging markets: an overview. The World Bank Economic Review 15, 315–340. Khorana, A., Servaes, H., Tufano, P., 2005. Explaining the size of the mutual fund industry around the world. Journal of Financial Economics 75, 145–185. Klapper, Leora, Sulla, Víctor, Vittas, Dimitri, 2004. The development of mutual funds around the world. Emerging Markets Review 5, 1–38. Sirri, Erik, Tufano, Peter, 1998. Costly search and mutual fund flows. The Journal of Finance 53, 1589–1622.