Information about Western Hemisphere Hemisferio Occidental
Journal Issue
Share
Article

Mexico: Selected Issues

Author(s):
International Monetary Fund. Western Hemisphere Dept.
Published Date:
November 2014
Share
  • ShareShare
Information about Western Hemisphere Hemisferio Occidental
Show Summary Details

Capital Flow Volatility and Investor Behaviour in Mexico1

Summary

The size and volatility of gross capital inflows (particularly portfolio inflows) to Mexico have increased significantly in recent years. This paper investigates how the volatility of gross capital flows could be affected by the behavior of foreign investors, especially during the periods of market stress, and whether domestic investors behaved differently from foreign investors. Our study of some 1000 foreign and domestic mutual funds in Mexico found strong evidence that foreign mutual fund investors exhibited potentially destabilizing trading behaviors that could contribute to market volatility. The evidence on domestic mutual funds’ destabilizing behavior was weaker. Our study of the Mexican sovereign bond markets also shows that foreign participation tended to amplify the impact of global financial shocks on these markets, notably in periods of market stress. Domestic institutional investors played some mitigating role, but the evidence is mixed and depends on the nature of the external shocks. These findings underline the importance of country insurance against global shocks and the potential role that a deep and diverse domestic investor base could play to mitigate such shocks.

A. Introduction

1. Gross capital inflows (particularly portfolio inflows) to Mexico have increased significantly in recent years (chart). Mexico has particularly strong links to the United States, its largest trading partner and the main source of portfolio and foreign direct investment. Mexico’s reputation as a prudently managed economy, with strong fundamentals, an open capital account, and relatively deep and liquid financial markets, has attracted significant portfolio inflows recently. In 2010, Mexico also became the first Latin American country to be included in the Citigroup’s World Government Bond Index (WGBI), attracting new groups of foreign investors. The share of non-resident holdings of domestic sovereign debt has reached 36 percent in April 2014, among the highest in emerging markets.

Net Capital Flows

(USD, billions; adjusted for errors and omissions)

Sources: Haver Analytics.

2. Portfolio capital inflows have also become more volatile. As observed in major emerging economies, capital inflows to Mexico surged prior to the global financial crisis (GFC), contracted sharply during the crisis (2008–09), rebounded to an higher level in 2010, fell again during European crisis in 2011, surged again in 2012 and contracted sharply after the U.S. Federal Reserve made its announcement about tapering in May 2013. Excessive volatility in capital flows could amplify economic cycles, increase financial system vulnerabilities, and aggravate overall macroeconomic instability. While our paper does not assess causation, past episodes of sharp contraction in portfolio inflows suggest that global factors could lead to a rapid rebalancing of investor portfolios away from assets in Mexico (as well as in other key emerging market countries) and Mexico remains vulnerable to the risks that the process of normalization of U.S. monetary policy may not proceed smoothly and geopolitical events could deteriorate further.2

3. Domestic investors’ flows have become more important. Mexico has a steadily expanding and diverse domestic investor base. Pension, insurance, and mutual funds now account for about half of the financial system (more than 40 percent of GDP). For example, over the last 10 years pension funds’ assets have increased by about 18 percent annually, and gradual changes in government regulations have allowed them to diversify their portfolios and invest abroad.3 While foreign investors rapidly increased their holdings of Mexican government debt in all currencies, domestic investors have increased their holdings at a much slower pace, and instead built up their holdings of foreign assets. When portfolio inflows stopped during the GFC, domestic residents retrenched, selling their foreign assets and bringing the money home.

4. Our study investigates how the volatility of gross capital flows in Mexico could be affected by the behavior of foreign and domestic investors, especially during the periods of market stress. Specifically, we try to reconcile the aggregated macro data with the high frequency market and fund-level data. In doing so, we aim to address the following questions:

  • Have foreign investors behaved differently from domestic investors, particularly during the periods of market stress?
  • Have they been more destabilizing?
  • Do domestic investors have a mitigating role to play?

5. Our study contributes to the literature and policy debate in two ways. First, we discuss capital flow volatility from the perspective of the behavior of different classes of investors; and second, we conduct empirical investigation using three unique datasets on Mexico, which to the best of our knowledge have not been used previously for studying the role of foreign investors in Mexico. For example, our analysis of the mutual funds investing in Mexico relies on an extensive fund flows dataset for about 400 international mutual funds (including global, emerging market-dedicated, Latin America-dedicated, and Mexico-dedicated funds) and 540 Mexican mutual funds active in the local markets in Mexico at a monthly frequency, with the latest data observation as recent as in April 2014. This enables us to gain valuable insight into their behavior during volatile periods, such as around the Fed tapering announcement in May 2013.

6. The rest of the paper is organized as follows: Section B discusses recent extreme capital flow episodes in Mexico following the methodology developed by Forbes and Warnock (2012); in section C, we study the behavior of domestic and foreign mutual funds active in Mexico during these episodes to assess their contribution to capital flow volatility.4 A time-series analysis in Section D investigates whether the participation of foreign investors in Mexico’s sovereign bond market has amplified the impact of external shocks during these episodes, and whether domestic investors (banks, pension and insurance funds, mutual funds, and other investors) have played a mitigating role; Section E concludes with policy implications.

B. Recent Episodes of Extreme Capital Movements in Mexico

7. Recent studies on capital flow volatility emphasized the importance of analyzing gross capital flows instead of net flows (Forbes and Warnock, 2012; Milesi-Ferretti and Tille, 2010). The literature’s earlier focus on net flows was largely based on the developments in the early and mid-1990s when net capital inflows roughly mirrored gross inflows, since the capital outflows of domestic investors tended to be small and changes in net inflows could thus be interpreted as being driven by changes in foreign flows. The new focus on gross flows in recent studies arises from the recognition that analyses based solely on net flows will overlook the significant changes in gross flows that have occurred recently—due to global financial integration and the development of a domestic investor base in many EM countries—and ignore important information contained in these flows, especially since foreign and domestic investors may well be motivated by different factors and respond differently to policies and shocks. In the case of Mexico, the size and volatility of gross flows have increased rapidly in recent years while net capital flows have been more stable, highlighting the importance of differentiating between gross inflows and gross outflows.

8. By focusing on gross capital inflows and outflows, we differentiate capital movements viewed as being initiated by foreigners from those initiated by domestic residents.5 To do so, we follow the approach in Forbes and Warnock (2012) to identify the following four types of extreme portfolio capital flow movements in Mexico over the period 1995 through 2013 at a quarterly frequency. Thus, a “surge” event corresponds to a sharp increase in nonresidents’ gross capital inflows, a sharp decrease in these gross inflows is referred to as a “stop” event; a sharp increase in residents’ gross capital outflows is a “flight”; and a “retrenchment” event refers to a sharp decrease in residents’ gross capital outflows.6 This allows us to differentiate the types of capital flow volatility episodes driven by foreigners (surges and stops) from those driven by domestic investors (flights and retrenchments). The reason for this approach is that we are interested in exploring whether domestic residents have mitigated the impact of a capital flow surge by foreigners and the subsequent sudden stop during the periods of markets stress.

9. Figure 1 suggests that domestic residents did act differently from their foreign peers during the GFC and also during the Fed tapering incident in May 2013 though to a lesser extent. Prior to the Lehman incident, foreign portfolio inflows increased sharply (a surge) and they were largely offset by an increase in capital outflows by residents (a flight). During the height of the global financial crisis, gross portfolio inflows from foreigners fell sharply (a stop) as they sold Mexican assets, while residents sold their foreign assets and brought the money home (a retrenchment). During the tapering talk in May 2013, foreign portfolio investors reduced their holding of Mexican assets as sharply as they did during the GFC. What was different, however, is that the recent sharp decline was offset only partially by opposite actions taken by the domestic residents.

Figure 1.Mexico: Extreme Capital Flow Episodes

C. Behavior of Foreign and Domestic Mutual Funds in Mexico

In this section, we study the behavior of mutual funds active in Mexico during the periods of market stress, especially shortly after the Fed tapering announcement in May 2013. We explore whether foreign mutual fund investors are more inclined than domestic investors to sell as others are selling (‘herding behavior’), and to buy when prices have risen (‘positive-feedback trading behavior’), relying on two extensive fund flows datasets, one covers about 400 foreign mutual funds and another covers 540 Mexican mutual funds, at monthly frequency and with the latest data observation as recent as April 2014. By doing so, we hope provide one plausible explanation for the large portfolio outflows during the episodes of market stress, from the perspective of individual investor behaviors.

10. Herding in financial markets emerges when investors mimic other investors. Such behavior can potentially destabilize financial markets, aggravate shocks, and lead to mispricing or asset price bubbles. While herding can be the result of cognitive biases or of “heuristic”-based decision making, it can also be a result of several other factors. For instance, herding may emerge if there is asymmetric information sharing, which induces less-informed asset managers to follow their possibly better-informed peers instead of relying on their own assessments, and in this context, improving transparency may help reduce herding behavior (Kim and Wei, 2002; Bikhchandani, Hirshleifer, and Welch, 1992). Herding may also occur if asset managers are evaluated against each other (Scharfstein and Stein, 1990) or vis-à-vis similar benchmarks (Maug and Naik, 2011).

11. We use fund-level data in our analysis of herding and momentum trading behaviors. To this end, we estimate the flow from each fund to Mexico using the country allocation data set from the EPFR Global (see Box 2 for details). As of April 2014, there are a total of 375 foreign mutual funds actively investing in Mexico, among which 112 are bond funds, and 263 are equity funds. Our full sample is from January 2007 to April 2014, covering two “stress” episodes, namely the Global Financial Crisis and the tapering announcement in May 2013, during which Mexico experienced significant declines in gross portfolio inflows. In order to facilitate comparison between domestic and foreign mutual funds active in Mexico, we use a shorter sample period (January 2011 to April 2014) for the analysis on momentum trading, as data on domestic mutual funds is only available from January 2011.

12. We use two (related) measures to quantify co-movements in trading patterns for funds—foreign or domestic—investing in Mexico:

  • A simple measure, defined by the proportion of all funds active in Mexico (in a particular month) that are net sellers. This gives intuitive and indicative evidence of whether “correlated” selling occurs at times of market stress.
  • A commonly used herding index originally introduced by Lakonshiok, Shleifer, and Vishy (1992). This index assesses whether funds move in the same direction more often than one would expect if they traded independently and randomly, and is computed as follows:

where pmex,t is the proportion of funds active in Mexico that are net buyers in month t, pt is its expected value, and AFmex,t is an adjustment factor so that HMmex,t is zero when there is no herding. pt is approximated by the share of funds that are net buyers across all emerging markets,7 and is allowed to be time-varying to control for common trends across countries, such as swings in aggregate inflows to emerging markets due to market-wide developments. The adjustment factor is equal to the expected value of the first term under the null hypothesis that there is no herding8

13. Both measures point to the evidence on “herding” behavior among the mutual funds, notably the foreign mutual funds (Figures 2 and 3). More specifically, we found:

Figure 2.Evidence of Herding

(Net sellers as a percent of total funds)

Figure 3.Evidence of Herding

(based on the herding index)
  • Foreign mutual funds (both bond and equity) exhibited a strong tendency to sell Mexican assets during the periods of heightened global uncertainty. For instance, after Lehman Brothers collapsed in September 2008, around 75 and 95 percent of equity and bond funds active in Mexico were net sellers of Mexican assets, respectively.
  • Foreign mutual fund investors are more inclined to exhibit “herding” behaviors than domestic investors. During the tapering announcement in 2013, around 50 percent of domestic mutual funds were selling Mexican assets, while at the same time the number of net sellers among foreign mutual funds rose to above 70 percent. This observation is consistent with the result based on the herding index (Figure 3), which shows herding among foreign mutual funds (bond and equity) increased significantly around the tapering announcement in May 2013, in comparison to the period beforehand.

14. Next, we examine the evidence of positive-feedback trading behavior among the mutual funds, differentiating foreign funds from domestic funds. Since foreign investors exhibit a stronger tendency to sell assets simultaneously at times of market stress, we also examine whether foreign investors show a stronger tendency to acquire (sell) more of an asset during periods of rising (falling) returns than domestic investors. We focus particularly on comparing the behaviors of bond and equity funds during stress and non-stress periods.

15. To do so we estimate the following equation that links the change in a fund’s current asset position to the past return,9 for a panel dataset consisting of 546 domestic mutual funds and 375 foreign mutual funds that includes global funds, EM dedicated funds, Latin America regional funds and funds dedicated Mexico only.

where Flowi,t is the flow of fund i to Mexico in month t. Returnt−1 is either the (peso-denominated) return on the Mexico’s 3-month government bonds (for bond funds) or the return in the Mexico’s stock market (for equity funds).10 We take a one-month lag to mitigate concerns about endogeneity. Stresst is a dummy variable that is equal to one for the periods of Global Financial Crisis and tapering announcement, and zero otherwise.11 The model also includes fund fixed effects αi.

16. We found strong evidence of “positive feedback trading” behavior among foreign bond funds during the periods of market stress (Tables 1a and 1b). The evidence that domestic bond funds exhibit “positive feedback trading” behaviors is not significant. For the foreign bond funds, the “positive feedback trading” behavior tended to be stronger during the episodes of market stress than during normal times. There is evidence that domestic equity funds appear to follow a “negative-feedback” trading strategy by which they would buy during periods of falling returns and sell during periods of rising returns. On the other hand, foreign equity funds do not exhibit statistically significant “positive feedback trading” behaviors during normal times, but do so during crisis times. The coefficient on the stress dummy is negative as expected, as it captures the overall tendency of funds to reduce exposures to Mexico at times of market stress. To check the robustness of our results, we used different sample periods (especially for the foreign funds) and U.S. dollar-denominated returns for the estimations (Table 2a and 2b).

Table 1a.Bond Funds: Evidence of Positive Feedback Trading Behavior
Domestic FundsForeign Funds
Explanatory Variables(1)(2)(1)(2)
Fund FlowFund FlowFund FlowFund Flow
Lagged Mex 3-month0.007260.006280.00725***0.00660***
bond yield(0.00541)(0.00543)(0.000765)(0.000770)
Lagged bond yield*0.00481**0.0171***
Stress(0.00225)(0.00567)
Stress−0.0177*−0.0677***
(0.00945)(0.0229)
Constant−0.0205−0.0166−0.0288***−0.0260***
(0.0220)(0.0221)(0.00319)(0.00321)
Fund Fixed EffectsYESYESYESYES
Number of Observations202022020234223422
R-squared0.0190.0190.1720.178
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2011 January – 2014 March (stress episode: tapering announcement)
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2011 January – 2014 March (stress episode: tapering announcement)
Table 1b.Equity Funds: Evidence of Positive Feedback Trading Behavior
Explanatory VariablesDomestic FundsForeign Funds
(1)(2)(1)(2)
Fund FlowFund FlowFund FlowFund Flow
Lagged stock−0.0211*−0.0269**0.00228**0.000615
market return(0.0115)(0.0123)(0.00113)(0.00105)
Lagged stock−0.0986*0.0138**
market return *
Stress(0.0517)(0.00657)
Stress−0.00417***−0.0000949
(0.00128)(0.000118)
Constant0.00236***0.00251***−0.000143***−0.000102**
(0.000351)(0.000377)(0.0000375)(0.0000398)
Fund Fixed
EffectsYESYESYESYES
Number of Observations18018180181859518595
R-squared0.0830.0840.1400.140
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2011 January – 2014 March (stress episode: tapering announcement)
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2011 January – 2014 March (stress episode: tapering announcement)
Table 2a.Robustness Check (1)—Using a Longer Sample for Foreign Mutual Funds
Foreign Bond FundsForeign Equity Funds
Explanatory Variables(1)(2)(1)(2)
Fund FlowFund FlowFund FlowFund Flow
Lagged return0.000557***0.000557***0.0172***0.00127*
(0.000204)(0.000204)(0.00604)(0.000673)
Lagged return * Stress0.0263***0.00235*
(0.00625)(0.00129)
Stress−0.101***−0.000202
(0.0237)(0.000137)
Constant−0.00346**−0.00346**0.000445**0.0000105
(0.00157)(0.00157)(0.000213)(0.0000303)
Month Fixed EffectsYESYESYESYES
Fund Fixed EffectsYESYESYESYES
Number of Observations728772873872438724
R-squared0.2200.2200.0750.070
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2005 January – 2014 March(Stress episodes: global financial crisis & tapering announcement)
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2005 January – 2014 March(Stress episodes: global financial crisis & tapering announcement)
Table 2b.Robustness Check (2) – Using dollar-denominated return on the 3-month Government Bonds
Domestic FundsForeign Funds
Explanatory Variables(1)(2)(1)(2)
Fund FlowFund FlowFund FlowFund Flow
Lagged dollar return on0.03550.02060.0705***0.0668***
Mex 3-month govt
bonds(0.0579)(0.0581)(0.00803)(0.00806)
Lagged bond yield *0.0592**0.261***
Stress(0.0280)(0.0701)
Stress−0.0192**−0.0852***
(0.00947)(0.0228)
Constant−0.002150.00267−0.0218***−0.0204***
(0.0184)(0.0185)(0.00261)(0.00261)
Fund Fixed EffectsYESYESYESYES
Number of
Observations202022020234223422
R-squared0.0190.0190.1680.176
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2011 January – 2014 March (stress episode: tapering announcement)
Standard errors (robust) in parentheses *p<0.10 **p<0.50 ***p<0.01Sample size: 2011 January – 2014 March (stress episode: tapering announcement)

17. These micro-level findings provide one plausible explanation for what we observe from macro-level data—the large portfolio outflows during the episodes of market stress—from the perspective of individual investor behaviors. For example, between Q2 and Q1 2013, the capital inflows (by non-residents) fell by US$24.5 billion, of which US$14 billion was due to a sudden stop in portfolio inflows. Much of this large decline seems to reflect a sharp reduction of foreign mutual funds’ investment in Mexico, especially by the small retail funds.

Gross Portfolio Inflows (by Foreigners)

(USD, billions)

Sources: Haver Analytics

Cumulative Flows of All Bond Funds to Mexico

(USD, Millions)

D. Does Foreign Participation Amplify External Shock? A Time-Series Analysis of Mexican Sovereign Bond Market

In this section, we estimate two empirical models, an OLS model and a GARCH model, to investigate whether higher foreign participation has amplified the impact of global financial shocks on the Mexican sovereign debt market and whether domestic investors played a mitigating role. The empirical investigation relies on an informative database of aggregate bond holdings by foreigners and residents (banking sector, insurance funds, pension funds, mutual funds, other investors) at the daily frequency.

18. The share of non-resident holdings of domestic sovereign debt in Mexico has risen rapidly, especially relative to residents’ holdings (charts below). Traditionally, Mexico has been a popular market for the U.S. investors, who represent a large share of the total foreign investors in Mexico. Since 2007, investors from Europe and Japan have also boosted their holdings of the Mexican assets. Moreover, the inclusion of Mexico in the Citibank’s World Government Bond Index (WGIB) in October 2010 paved the way for more participation by foreign investors. Moreover, it appears that many foreign investors have been able to hedge their currency risk exposures as the Mexican peso is the most traded EM currencies globally.

Tenure of All Government Securities by Sector (billion of MXN)

Government Securities Held by Residents Abroad*

19. Whether foreign participation increases market volatility is subject to ongoing debate. Foreign participation, to the extent that it increases market liquidity and exerts pressure for strong corporate governance and institutional reform, can be a stabilizing force in the long run (Prasad and Rajan, 2008). However, sudden withdrawals by foreign investors from domestic bond markets, as happened in EMs in 2008/2009 and more recently after the tapering announcement, could introduce greater bond yield volatility. Foreign investors in the EMs could be more sensitive to global and EM shocks than EM domestic investors, due to home bias and asset allocation decisions that could be influenced by information asymmetries and hedging costs of currency risks. Empirical evidence on the impact of foreign participation on market volatility is sparse and mixed. Several recent IMF working papers, based on cross-country evidence, found that while high foreign participation in the local markets helped reduce borrowing costs it was associated with higher yield volatilities (Ebeke and Lu, 2014; Andritzky, 2012).12 However, another IMF paper found that greater foreign participation does not necessarily result in increased volatility in bond yields in EMs and it could dampen volatility in some cases (Peiris, 2010).

20. In the case of Mexico, the effect of foreign holdings of sovereign bonds on their yield volatility is not evident upon first glance. The chart on the right plots the volatility of Mexican 10-year sovereign bond yields against the level of foreign participation using data at daily frequency. It shows no clear correlation between the two for the whole sample period of 2000–14. However, for the periods of market stress (for example, during the GFC and the Fed’s tapering announcement in May 2013), higher foreign participation seems to be associated with higher yield volatility. Thus, in our empirical analysis, we divided our sample period into a stress period and a non-stress period. Moreover, the chart reveals that Mexican 10-year sovereign bond yields exhibit the volatility clustering property, i.e., large changes in yields tend to be followed by large changes in volatility during the stress periods. Such a property has often been observed among time series of financial asset returns and linked to the behavior of market participants (Cont, 2005). We therefore also used GARCH models to address the presence of volatility clustering.

The Volatility of Mexican 10-Year Bond Yields and Foreign Participation

21. We estimate two time series models, an OLS model and a GARCH model, to investigate whether higher foreign participation has amplified the impact of global financial shocks on the local debt market and whether domestic investors acted differently. The empirical work focuses on the volatility of Mexican long-term (LT) local-currency sovereign bond yields. The estimation distinguishes normal time with low market volatility from the periods of market stress. The details of model specifications are discussed in Box 1. High frequency samples (at both weekly and daily frequencies) covering 2000–14 are constructed to conduct the estimations. Variables included in the OLS regressions are:

  • Volatility of the Mexican 10-year local-currency sovereign bond yields (dependent variable). For robustness, two measures of the yield volatility are used in the weekly and daily OLS regressions, respectively: (i) Within-week volatility, measured by the standard deviation of daily yields within each week normalized by the weekly mean; and (ii) 5-day rolling volatility, measured by the standard deviation of a 5-day rolling window of daily yields, normalized by the mean.
  • Foreign and domestic participation, defined as the daily holdings of Mexican LT local-currency sovereign bonds by (i) foreign investors; and (ii) five different types of domestic investors (pension funds, insurance companies, mutual funds, banks, and other domestic investors) as shares of the total values of outstanding LT sovereign bonds.
  • Global financial shocks, measured by (i) the VIX to capture global uncertainty; and (ii) the volatility in U.S. 10-year Treasury yields to capture the U.S. monetary policy shocks.
  • Domestic variables: official international reserves (weekly, excluding gold), peso/US dollar exchange rate (daily), and interbank interest rate (daily).13
  • Variables included in the (daily) multivariate GARCH models are the changes in the Mexican 10-year local-currency sovereign bond yields, the changes in the Mexican stock market returns (dependent variables), and the explanatory variables included in the daily OLS regressions.

22. The empirical results of the OLS regressions and multivariate GARCH models are presented in Tables 3a, 3b, 4a, and 4b. Several interesting results emerge:

Table 3a.OLS Regression Results (VIX Shock)
Explanatory variablesOLS Regressions
With-week volatility5-day rolling volatility
VIX*Foreign share*Stress0.0059***0.0053***
(6.56)(7.13)
VIX*Pension share*Stress−0.0073*−0.0071*
(1.89)(1.83)
VIX*Mutual share*Stress0.00450.0037
(1.25)(1.47)
VIX*Insurance share*Stress−0.0014−0.0016
(0.72)(0.73)
VIX*Bank share*Stress−0.0047**−0.0037**
(2.18)(2.73)
VIX*Other share*Stress0.0020*0.0022*
(1.69)(1.67)
VIX*Foreign share*Non-Stress0.0021***0.0023***
(5.33)(7.13)
VIX*Pension share*Non-Stress0.000520.00024
(0.76)(0.43)
VIX*Mutual share*Non-Stress−0.00049−0.00063
(1.02)(1.56)
VIX*Insurance share*Non-Stress−0.0059***−0.0063***
(3.80)(5.04)
VIX*Bank share*Non-Stress0.0014*0.0014**
(1.91)(2.44)
VIX*Other share*Non-Stress−0.000.00011
(0.094)(0.37)
Change in short-term interest rate0.180.22
(1.16)(0.90)
Change in reserves−0.00*
(1.89)
Exchange rate depreciation−0.078***
(2.79)
Constant0.140.14*
(1.45)(1.71)
Adjusted R20.330.32
Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.
Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.
Table 3b.OLS Regression Results (U.S. Tapering Shock)
Explanatory variablesOLS Regressions
With-week volatility5-day rolling volatility
Volatility of U.S. 10-year bond yields* Foreign share*Stress0.043***0.040*
(4.78)(5.57)
Volatility of U.S. 10-year bond yields* Pension share*Stress−0.023−0.059
(0.71)(1.46)
Volatility of U.S. 10-year bond yields* Mutual share*Stress−0.026−0.012
(0.83)(0.53)
Volatility of U.S. 10-year bond yields* Insurance share*Stress−0.00650.0076
(0.16)(0.18)
Volatility of U.S. 10-year bond yields* Bank share*Stress−0.074−0.054**
(1.61)(2.26)
Volatility of U.S. 10-year bond yields* Other share*Stress0.0210.034*
(1.43)(1.81)
Volatility of U.S. 10-year bond yields* Foreign share*Non-Stress0.00580.0095***
(1.43)(2.96)
Volatility of U.S. 10-year bond yields* Pension share*Non-Stress0.0076−0.00
(0.74)(0.0069)
Volatility of U.S. 10-year bond yields* Mutual share*Non-Stress−0.011−0.012**
(1.61)(2.04)
Volatility of U.S. 10-year bond yields* Insurance share*Non-−0.0096−0.018
Stress(0.46)(1.16)
Volatility of U.S. 10-year bond yields* Bank share*Non-Stress0.015**0.014**
(1.96)(2.14)
Volatility of U.S. 10-year bond yields* Other share*Non-Stress−0.00440.00019
(0.79)(0.046)
Change in short-term interest rate0.190.22
(1.11)(0.84)
Change in reserves−0.00*
(1.76)
Exchange rate depreciation−0.052*
(1.67)
Constant0.61***0.60***
(8.45)(16.08)
Adjusted R20.250.22
Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.
Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.
Table 4a.Multivariate GARCH Results (VIX Shock)
Explanatory variablesIII-AIV-A
ΔYtMexσY,t2
VIX*Foreign share*Stress−0.00130.0079***
(0.39)(9.13)
VIX*Pension share*Stress0.025*0.0036
(1.67)(0.70)
VIX*Mutual share*Stress−0.0090−0.0024
(0.62)(0.55)
VIX*Insurance share*Stress−0.015*−0.011***
(1.70)(3.54)
VIX*Bank share*Stress−0.022***−0.0099***
(2.86)(3.25)
VIX*Other share*Stress−0.0079**−0.00044
(2.04)(0.35)
VIX*Foreign share*Non-Stress−0.0037**0.0053***
(2.13)(8.18)
VIX*Pension share*Non-Stress−0.0038−0.00045
(1.07)(0.49)
VIX*Mutual share*Non-Stress−0.0059**−0.0042***
(2.31)(4.77)
VIX*Insurance share*Non-Stress0.020***−0.018***
(2.63)(6.66)
VIX*Bank share*Non-Stress0.00100.0078***
(0.32)(7.72)
VIX*Other share*Non-Stress0.000640.00068
(0.34)(1.32)
Change in short-term interest rate2.420.20
(1.39)(0.34)
Exchange rate depreciation2 87***0.0030
(16.84)(0.04)
ARCH term0.25***
(8.34)
GARCH term0.30***
(6.05)
Constant1.88***1 39***
(4.08)(7.87)
Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively
Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively
Table 4b.Multivariate GARCH Results (U.S. Tapering Shock)
Explanatory variablesIII-BIV-B
ΔYtMexσY,t2
U.S. 10-year bond yields*Foreign share*Stress 1/0.016*3.73***
(1.63)(5.58)
U.S. 10-year bond yields*Pension share*Stress0.076*2.04
(1.83)(0.65)
U.S. 10-year bond yields *Mutual share*Stress−0.020−7 55**
(0.48)(2.31)
U.S. 10-year bond yields *Insurance share*Stress−0.020−0.75
(0.71)(0.38)
U.S. 10-year bond yields *Bank share*Stress−0.053**−3.54
(2.11)(1.33)
U.S. 10-year bond yields *Other share*Stress−0.027**2.57**
(2.48)(2.34)
U.S. 10-year bond yields *Foreign share*Non-Stress0.015***1 49***
(2.98)(2.65)
U.S. 10-year bond yields *Pension share*Non-Stress0.0036−0.49
(0.33)(0.86)
U.S. 10-year bond yields *Mutual share*Non-Stress−0.012−0.77
(1.53)(1.14)
U.S. 10-year bond yields *Insurance share*Non-Stress0.0051−1.70*
(0.27)(1.76)
U.S. 10-year bond yields *Bank share*Non-Stress0.023**244***
(2.14)(4.04)
U.S. 10-year bond yields *Other share*Non-Stress−0.0049141**
(0.85)(2.43)
Change in short-term interest rate2.90*2.25**
(1.71)(2.54)
Exchange rate depreciation344***0.50***
(18.95)(3.44)
ARCH term0.26***
(7.83)
GARCH term0.56***
(10.95)
Constant−0.18*−2.84
(1.85)(1.07)

The “U.S. 10-year bond yields” denotes the change in U.S. 10-year bond yields and the volatility of U.S. 10-year bond yields in columns III-B and IV-B, respectively.

Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.

The “U.S. 10-year bond yields” denotes the change in U.S. 10-year bond yields and the volatility of U.S. 10-year bond yields in columns III-B and IV-B, respectively.

Note: t-statistics in parentheses. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.
  • First, foreign participation tends to amplify the impact of global financial shocks on the volatility of Mexican sovereign bond yields, notably during the periods of market stress. Results from both the OLS regressions and the multivariate GARCH models are consistent and robust across all specifications. This implies that the foreign participation in the local-currency bond market can make the market more susceptible to adverse shifts in the market sentiment caused by global financial shocks. From finance theory, due to information asymmetry or currency risks, local-currency assets tend to have larger risks for foreign investors than for domestic investors when global risk aversion rises or global liquidity tightens. As a result, foreign investors’ holdings of local-currency assets could be more sensitive to global financial conditions than local investors’ holdings, generating “extra” volatility in local markets. The results of the GARCH regressions show that higher participation by foreign investors in the Mbono market heightens the sensivity of the volatility of the 10-year Mbono yield, without necessarily affecting the level of the Mbono yield, to a shock to the VIX. At the same time, increased foreign participation does make both the level and volatility of the Mbono yield more sensitive shocks to the yield on 10-year US Treasuries. This discrepancy could reflect the fact that—over this sample period—surges in the VIX were often associated decline in the level of the yield on the 10-year US Treasuries, since this asset was viewed as the safe asset.
  • Second, domestic investors in the sovereign bond market played some mitigating role, but the empirical evidence was mixed, depending on the type of investors and the nature of the global shocks. In particular, banks seem to amplify the impact of global financial shocks on domestic yield volatility during normal times across all specifications, but their role during stress periods seems to be unclear. In addition, the results based on the OLS regressions find that domestic pension funds and banks seem to dampen the impact of VIX shock during stress periods, but there is no significant and robust evidence that the insurance sector or the mutual funds play any mitigating or amplifying role during stress periods. Moreover, the mitigating role of the pension funds no longer holds if the shock is to the U.S. 10-year Treasury yields. These mixed results tend to suggest that domestic investors are not significantly different from each other, and their behaviors would depend on the nature of the global financial shocks.14

23. These results are robust to the choices of dependent and independent variables. For the dependent variable, another volatility measure using the exchange rates rather than the sovereign bond yields is also used in the OLS regressions, and the main (qualitative) result that foreign participation tends to amplify global financial shocks does not change. For the independent variables, we replace the shares of respective investors in the interaction terms by their first-order differences, and the main results still hold.

E. Concluding Remarks

24. The size and volatility of gross capital inflows (particularly portfolio inflows) to Mexico have increased in recent years. Mexico remains vulnerable to global financial shocks, given its globally integrated financial markets. A resurgence of investor uncertainty triggered by an earlier or sharper-than-expected rise in U.S. interest rates (for example due to an unexpected rise in inflation or a decompression of U.S. term premia) presents a distinct risk in the near term. As our analysis of the Fed tapering incident in May 2013 has shown, such a shock could lead to a significant portfolio capital outflows from Mexico, resulted in high volatility in bond prices. A protracted period of financial market volatility could affect the confidence of long-term investors and threaten the ongoing progress on the implementation of structural reforms.

25. This paper investigates the behaviors of foreign and domestic investors, especially during the periods of market stress. More specifically, relying on three approaches using both macro- and micro-level data, our empirical investigations suggest that foreign and domestic investors do seem to behave differently. We found that foreign mutual fund investors exhibited potentially destabilizing trading behaviors that could contribute to market volatility. For examples, they tended to sell as others are selling (‘herding behavior’) and follow a “positive-feedback trading” strategy (i.e. buy high and sell low). We did not find significant evidence indicating that domestic mutual funds behaved similarly. Moreover, our study of the Mexican sovereign bond markets suggests that foreign participation tended to amplify the impact of global financial shocks, notably in periods of market stress. Domestic institutional investors played some mitigating role, but the evidence is mixed and depends on the nature of the external shocks.

26. These results imply that policy credibility will continue to be key to maintain market confidence at times of stress. Mexico has a strong policy framework,15 with sound public finances, low and stable inflation, a sustainable external sector position, and a healthy banking system. It also has one of the highest credit ratings among emerging markets. These strong fundamentals have helped attract foreigner investors and maintain a strong demand from local investors, especially during the periods of market stress. For example, during the GFC local investors reduced their holdings of foreign assets and brought local assets when foreigners left Mexican market. The situation was very different during the Mexican peso crisis in 2004, when both foreign and local investors deserted the country amid the great domestic policy uncertainty.

27. Country insurance against global risks may be necessary. While the exchange rate flexibility acted as an effective shock absorber, the high level of reserves has provided a useful buffer against temporary stress in foreign exchange markets.16 The FCL arrangement, which the authorities continue to treat as precautionary, was an important complement to reserve buffers, providing additional protection against tail risks. For instance, in the aftermath of the announcements by Federal Reserve Board members about imminent tapering, capital inflows by non-residents dropped by US$24.5 billion from Q1to Q2 2013. The exchange rate depreciated by 8 percent. Both of these developments were quickly reversed, as investors began differentiating between emerging markets, and Mexico stood out for its strong macroeconomic fundamentals, but also its large foreign exchange buffer, notably when including the FCL.

Gross Portfolio Inflows by Foreigners

(USD, billions)

Source: Haver Analytics.

28. Finally, a deep and diverse domestic investor base is important as well. Mexico has benefited from an expanding and diverse local investor base, thanks to pension reforms and the establishment of insurance and mutual fund industries. These investors could play a more important stabilizing role to the Mexican financial markets, for example, by improving liquidity in the secondary markets for government securities.

Box 1.OLS and Multivariate GARCH Models

Weekly OLS regression with within-week volatility:

where: VolMex is the within-week volatility of Mexican 10-year sovereign bond yields; Z denotes the global financial shocks measured by VIX and the within-week volatility of U.S. 10-year Treasury yields; Si represents the holdings of long-term (LT) Mexican sovereign bonds by each type of investors as a share of total outstanding bond values; is a dummy variable which takes the value of 1 if it’s the period with extreme capital flows (referred to as extreme periods hence after) and 0 if it’s not; X is the change in Mexico’s international reserves, as a domestic variable to control for partly the supply of sovereign bonds. We use the first lag of investors’ shares of holdings to avoid the endogeneity problem due to the potential impact of volatility on investors’ holdings.

Daily OLS regression with 5-day rolling volatility:

Where: Z denotes the global financial shocks measured by VIX and the 5-day rolling volatility of U.S. 10-year Treasury yields; X is the first-order log difference of the exchange rate. The 4th lag of investors’ shares of holdings is used to avoid the endogeneity problem.

Multivariate GARCH Model (MGARCH): To jointly and systematically model both bond market and equity market returns, as well as the levels and (conditional) volatilities of these returns.

  • The mean equation:

where: ΔYMex denotes the (daily) change in the Mexican 10-year sovereign bond yields, aed ΔEMex denotes the (daily) first-order log difference of the Mexican stock prices; Z denotes the global financial shocks to domestic asset returns measured by VIX and the (daily) change in U.S. 10-year Treasury yields; X is the first-order log difference of the exchange rate.

  • The volatility equation: We assume a diagonal conditional variance matrix where each diagonal element follows a GARCH(1, 1) process with exogenous regressors, and also assume a constant conditional correlation between the two asset returns for simplicity.

where: σY,t2 and σE,t2 are the conditional variances of ΔYtMex and ΔEtMex, respectively; Z denotes the global financial shocks to these conditional variances measured by VIX and the conditional volatility of U.S. 10-year Treasury yields.

Box 2.Data on Foreign Mutual Funds

  • Our data source for foreign mutual funds is EPFR Global. It covers in total about 11,000 equity funds and about 4,500 bond funds, all of which have $22 trillion in total assets as of the end of 2013. According to EPFR Global, its data track more than 95 percent of EM-focused bond and equity funds. EPFR data have several advantages over Balance of Payments data. First, EPFR Global provides high-frequency (weekly or monthly), detailed information at the fund-level. Second, it records data on a nationality basis, while the Balance of Payments data report in a residency basis.
  • A drawback of EPFR Global is that it generally tracks only mutual funds. However, this is not necessarily critical since mutual funds have been playing an important role in capital flows to Mexico. Moreover, the behavior of mutual funds itself is an important research agenda, since IMF (2014) reports that they are more sensitive to global financial conditions and are more likely to engage in return chasing than other types of investors.
  • EPFR Global provides various fund-level and country-level data. We use two different fund-level data sets: the fund flow data, and the country allocation data. The fund flow data set reports dollar-denominated flows, returns, assets under management (AUM), in addition to various fund characteristics, such as the domicile and geographic focus. However, the flows, returns, and AUM are not disaggregated by destination economy. On the other hand, the country allocation data set reports country allocation weights over more than 130 developed and emerging economies on a monthly basis.
  • EPFR Global also provides country-level data, which are estimated using the two fund-level data sets. It also enables us to obtain the country-level data decomposed by fund characteristics. Roughly speaking, EPFR Global estimates country-level flows by multiplying the country allocation weight at the end of month by the aggregate flow into funds with specific characteristics. Although the country-level data are useful, a potential drawback is that the estimation method generally can only capture changes in flows from ultimate investors but not changes in allocation weights. Hence, if asset managers shift their allocations from EM economies to advanced economies or cash in response to some deterioration in global financial conditions, the estimated outflows from EM economies are underestimated.
References

    Andritzky, Jochen R.(2012), “Government Bonds and Their Investors: What Are the Facts and Do They Matter?” IMF Working Papers, WP/12/158, June.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Bikhchandani, Sushil, DavidHirshleifer, and IvoWelch, 1992, “A Theory of Fads, Fashion, Custom and Cultural Change as Information Cascades,” Journal of Political Economy, Vol. 100, No. 5, pp. 992–1026

    • Crossref
    • Search Google Scholar
    • Export Citation

    Cont, Rama, (2005), “Volatility Clustering in Financial Markets: Empirical Facts and Agent-Based Models”,

    Ebeke, Christian and YinqiuLu, (2014), “Emerging Market Local Currency Bond Yields and Foreign Holdings in the Post-Lehman Period—a Fortune or Misfortune”, IMF Working Papers, WP/14/29, February.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Faruqee, H., Li, S., and I.K.Yan, (2004), “The Determinants of International Portfolio Holdings and Home Bias”, IMF Working Papers, WP/04/34, February.

    • Search Google Scholar
    • Export Citation

    Forbes, Kristin J., and Francis E.Warnock, (2012), “Capital Flow Waves: Surges, stops, flight, and retrenchments”, Journal of International Economics88: 235–251.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Frankel, J., (1982), “A Search of the Exchange Risk Premium: A Six-Currency Test Assuming Mean-Variance Optimization”, Journal of International Money and Finance, Vol. 1, pp. 255–274.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Gelos, Gaston, (2011), “International Mutual Funds, Capital Flow Volatility, and Contagion—A Survey”, IMF Working Papers, WP/11/92, April.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Hsieh, Meng-Fen, Tzu-YiYang, Yu-TaiYang, and Jen-SinLee, 2011, “Evidence of Herding and Positive Feedback Trading for Mutual Funds in Emerging Asian Countries,” Quantitative Finance, Vol. 11, No. 3, pp. 423–435.

    • Crossref
    • Search Google Scholar
    • Export Citation

    International Monetary Fund, (2014), “How Do Changes in the Investor Base and Financial Deepening Affect Emerging Market Economies”, Chapter 2 in Global Financial Stability Report, April.

    • Search Google Scholar
    • Export Citation

    Kim, Woo-Chan, and Shang-JinWei(2002), “Foreign portfolio investors before and during a crisis”, Journal of International Economics56: 77–96.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Maug, Ernst, and NarayanNaik, 2011, “Herding and Delegated Portfolio Management: The Impact of Relative Performance Evaluation on Asset Allocation,” Quarterly Journal of Finance, Vol. 1, No. 2, pp. 265–92.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Milesi-Ferretti, Gian-Maria and CedricTille, 2011, “The great retrenchment: international capital flows during the global financial crisis”, Economic Policy, Volume 26, Issue 66, pages 289–346, April.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Peiris, Shanaka J.(2010), “Foreign Participation in Emerging Markets’ Local Currency Bond Markets,” IMF Working Papers, WP/10/88, April.

    Prasad, Eswar S., and Raghuram G.Rajan(2008). “A Pragmatic Approach to Capital Account Liberalization.” Journal of Economic Perspectives, 22(3): 149–72

    • Crossref
    • Search Google Scholar
    • Export Citation

    Scharfstein, David, and JeremyStein, 1990, “Herd Behavior and Investment,” American Economic Review, Vol. 90, No. 3, pp. 465–79.

    Sidaoui, José, JulioSantaella and JavierPérez, 2012, “Banco de México and recent developments in domestic public debt markets”, BIS background paper.

    • Search Google Scholar
    • Export Citation
1

Prepared by Jianping Zhou, Han Fei, and Jasmine Xiao. The authors would llike to thank Dora Iakova, Hibiki Ichiue, Herman Kamil, Robert Rennhack, Alejandro Werner, and seminar participants at the IMF, Secretaría de Hacienda y Crédito Público, and Banco de México for comments and discussions.

2

For a discussion on U.S. monetary policy uncertainly, see “Fed chiefs debate monetary normalization while Yellen passes off financial stability”, CITI Research, July 11, 2014.

3

Reforms to the Mexican pension system have strengthened the demand for government securities. The transformation in 1997 of a pay-as-you-go system into an individual contributory pension system for private workers resulted in a surge of large pension funds. Later on in 2007, the pension system of public employees went through a similar reform which further increased assets managed by pension funds, hence stimulating additional demand for securities (see Sidaoui, José, Julio Santaella and Javier Pérez 2012).

4

An extension of this paper will estimate the impact of mutual funds’ trading behaviors on price volatility.

5

We use data from the International Monetary Fund’s International Financial Statistics, in which international capital flows are based on the residency criterion of the balance of payments, and cover transactions where one of the counterparties is a resident of the country (say Mexico) and the other a resident of the rest of the world. Capital outflows denote net purchases by domestic residents of financial instruments issued by non-residents, while capital inflows denote net purchases by foreign residents of domestic financial instruments. The difference between capital inflows and outflows (the financial account balance) corresponds to the current account balance (up to a statistical discrepancy).

6

Forbes and Warnock (2012) defines extreme capital flow episodes using three criteria: (1) current year-over-year changes in four-quarter gross capital inflows or outflows is more than two standard deviations above or below the historic average during at least one quarter of the episode; (2) the episode lasts for all consecutive quarters for which the year-over-year change in annual gross capital flows is more than one standard deviation above or below the historical average; and (3) the length of the episode is greater than one quarter.

7

Our sample of emerging markets include: Argentina, Bangladesh, Brazil, Bulgaria, Chile, China, Colombia, Croatia, Czech Republic, Egypt, Hungary, India, Indonesia, Israel, Jordan, Korea, Latvia, Lebanon, Lithuania, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Peru, Philippines, Poland, Romania, Russia, Serbia, South Africa, Sri Lanka, Taiwan Province of China, Thailand, Turkey, Ukraine, Uruguay, and Vietnam.

8

This is needed since the distribution of the first term is not centered on zero.

9

We followed the approach adopted in Hsieh et al (2011) and IMF (2014).

10

For robustness checks, we used (i) the return on the Mexico’s 10-year government bonds (instead of 3-month government bonds), and (ii) the dollar-denominated return on the Mexico’s government bonds.

11

We use these two episodes as “stress” episodes, because: (i) the height of the Global Financial Crisis (2008Q3–2009Q3) qualifies as an episode of sudden stop, according to the Forbes and Warnock (2012) methodology; and (ii) during the tapering announcement in May 2013, portfolio inflows to Mexico fell, drastically and significantly, two standard deviations below its five year rolling mean.

12

These papers have not advocated for capital control measures as their desirability and effectiveness are subject to debate, but they supported building up foreign exchange reserve buffers and developing a deep domestic investor base.

13

The 28 day TIIE interbank equilibrium rate is used as the interbank interest rate to capture domestic monetary policy. The first-order differences of the reserves and interest rates, as well as the first-order differences of log exchange rates are used in the regressions.

14

Our result is consistent with the finding in the IMF’s Regional Economic Outlook: Western Hemisphere (April 2014) that a shock in VIX (which may reflect global uncertainty) can have diffenrt impact from a shock in US interest rate.

15

Monetary policy is guided by an inflation targeting framework under a flexible exchange rate, and fiscal policy is anchored by a fiscal responsibility law.

16

International reserves stand at about US$190bn as of June 2014.

Other Resources Citing This Publication