Benchmark-Driven Investments in Emerging Market Bond Markets: Taking Stock

Serkan Arslanalp; Dimitris Drakopoulos; Rohit Goel; Robin Koepke

doi:10.5089/9781513557588.001.A001

Author: Mr. Serkan Arslanalp, Dimitris Drakopoulos¹, Rohit Goel², and Mr. Robin Koepke https://orcid.org/0000-0002-9308-9619

¹ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
| ² 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: SArslanalp@imf.org; DDrakopoulos@imf.org; RGoel@imf.org; RKoepke@imf.org.

Publication Date:: 25 Sep 2020

eISBN:: 9781513557588

Language:: English

Keywords:: WP; investment; investor; JPMorgan; Capital flows; portfolio choice; push and pull; benchmarks; J.P. Morgan emerging market bond Index; benchmark index; benchmark-driven investment; J.P. Morgan government bond index-emerging markets; investor base; surprise Index; Emerging and frontier financial markets; Securities markets; Bonds; Mutual funds; Global; J.P. Morgan GBI-EM index; government debt markets; EM surprise Index; investment benchmark; EM surprise

Download PDF (2.9 MB)

This paper reviews the role of benchmark-driven investments in EM local bond markets. We provide an overview of how key EM bond benchmark indices are constructed, how they affect the behavior of investment funds, and what are the likely implications for capital flows and policy-making. Several methods are presented suggesting that the amount of assets benchmarked against widely followed EM local-currency bond indices have risen fivefold since the mid-2000s to around $300 billion. Our review suggests that the benefits of index membership may be tempered by portfolio outflow risks for some countries. This is because benchmark-driven investments may increase the importance of external factors at the expense of domestic factors, raising the risks of outflows unrelated to recipient country fundamentals. Some countries may be disproportionately exposed to these risks, reflecting the way the indices are constructed.

Abstract

I. Introduction

Benchmark driven investments (BDI) have received growing attention in the recent literature on portfolio flows to emerging markets. An investment is benchmark-driven to the extent that its portfolio allocation across countries is guided by the country weights in a benchmark index (IMF 2019, Arslanalp and Tsuda 2015). A number of influential studies have found evidence that investment benchmarks are increasingly shaping portfolio allocation dynamics in emerging market (EM) countries. For example, Raddatz and others (2017) find that 70 percent of country allocations of investment mutual funds are influenced by benchmark indices.

The growing role of benchmark-driven investments appears to reflect the confluence of several factors and may be contributing to more synchronized portfolio flows. These factors include rising assets under management of passive funds, increasing number of active funds “hugging” the index, and a growing number of countries represented in the benchmark indices. Together, these factors may have contributed to more correlated portfolio flows to countries included in benchmarks indices. For example, portfolio flows to EM local bond markets have traditionally been much more synchronized for countries in a benchmark index (the J.P Morgan GBI-EM Global) than those that were not in the index (Figure 1).

A burgeoning literature suggests that the growing role of benchmark-driven investments entails both benefits and risks for portfolio flows to EMs. On the upside, inclusion in major benchmark indices provides countries with access to a larger and more diverse pool of external financing (Sienaert 2012). On the downside, benchmark-driven investments may raise the sensitivity of portfolio flows to global factors and, more generally, to factors common to EMs included in benchmark indices (IMF 2019, Arslanalp and Tsuda 2015). This is because benchmark-driven investments are investment in EMs as an asset class, focusing mainly on factors that affect EMs as a group, rather than on country-specific developments. As a result, benchmark-driven portfolio flows may be more sensitive to common factors and therefore more correlated across countries.

The study by Raddatz and others (2017) empirically demonstrates that benchmark indices affect asset allocations, capital flows, and asset returns of international mutual funds in a statistically significant and economically important way. The study shows that the use of benchmarks generates pricing and flow distortions that represent deviations from the theoretical predictions of standard models (such as the CAPM). Related research shows that the use of benchmarks can give rise to correlated behavior on the part of even “actively” managed funds (Miyajima and Shim 2014) and that EMs with higher shares of financing from investment funds are more sensitive to external/push factors (Cerutti and others 2015).

Building on this literature, this paper takes stock of the rise of benchmark-driven investments in EM bond markets. We focus particularly on local currency government bonds—the single largest segment of EM bond markets. First, we provide an overview of how key EM bond benchmark indices are constructed, how they affect the behavior of investment funds, and what are the likely implications for capital flows dynamics. Second, we provide up-to-date estimates of the size of benchmark-driven investments in local EM bond markets, estimated at around $300 billion as of end 2019 before the COVID-19 shock in early 2020 (Figure 2). Third, we provide empirical results demonstrating the heightened sensitivity of benchmark-driven investments to external factors, which leads to an elevated correlation of such flows across countries.

During the COVID-19 pandemic, the combined effect of this heightened sensitivity and the growing assets under management appears to have contributed to the record outflows of foreign portfolio capital from emerging markets (IMF 2020) (Figure 3). In this episode, portfolio outflows from local currency bond markets were quite strongly correlated with the weight of the country in the GBI-EM benchmark index, suggesting that benchmark-driven investors may have played a significant role in driving the reversal of portfolio flows (Figure 4).

Overall, the picture that emerges from our review is that benchmark-driven investments are changing the landscape for EM bond markets, with important implications for capital flows dynamics. Different investor types exhibit fundamentally different behaviors in deciding their portfolio allocations, and different countries exhibit widely different investor bases. As a result, the same kinds of external and domestic shocks can prompt different capital flow responses in different countries, in turn, shaping the way such shocks are transmitted to the domestic economy.

The rest of this paper is organized as follows. Section 2 explains key concepts for benchmark-driven investments and its implications for capital flows. Section 3 provides country-level estimates of benchmark-driven investments through regression analysis and event studies. Section 4 discusses the effects of benchmark-driven investments on capital flows dynamics. Section 5 provides concluding remarks.

II. What are Benchmark-Driven Investments and What are the Implications for Capital Flows?

Benchmark-driven investments come about mainly in the context of investment funds, most of which use benchmark indices to guide their portfolio allocation. Investment funds vary in the degree to which they are benchmark-driven in that some follow the country weights of their benchmark more closely than others. Passive investment funds explicitly aim to replicate the performance of their benchmark, while active funds deviate from their benchmarks to varying degrees as fund managers try to outperform their respective benchmarks (Figure 5). By contrast, unconstrained funds choose portfolio allocations irrespective of benchmark indices. It is worth noting that the ultimate investors of benchmark-driven funds can be retail or institutional investors, though retail investors generally have only limited access to unconstrained funds.

A. What are Key EM Bond Benchmarks and How are They Constructed?

Among the multitude of investment funds focused on emerging market local bonds, most are benchmarked to a small number of indices. The most widely followed benchmark indices for emerging market bonds are the J.P. Morgan Emerging Market Bond Index (EMBIG; for dollar-denominated bonds) and the J.P. Morgan Government Bond Index–Emerging Markets (GBI-EM; for local currency bonds), both of which come in several variants. Benchmark-driven investments discussed in this paper primarily refer to investment funds benchmarked to these two indices (as discussed in greater details in Annex I).

The number of countries represented in benchmark indices has grown substantially over the years, especially among emerging markets. Since 2007, the number of countries in the EMBIG has doubled to more than 70 with the inclusion of many countries that have issued in international bond markets for the first time. In addition, the growing size and liquidity of local bond markets in many emerging markets have allowed the number of countries in the GBI EM to increase from 12 to 18. By contrast, although many EMs have earned investment grade status from rating agencies, the number of countries with local currency debt represented in global investment grade bond indices has been relatively stable, given the more demanding investability criteria of these indices (Annex I).

Due to the way indices are constructed, some countries are disproportionately exposed to benchmark-driven investors. This reflects both index inclusion rules and discretionary choices of index providers. Different inclusion criteria for each index lead to varying compositions of investor types, exposing issuers to diverse portfolio flow dynamics. Inclusion criteria help determine the universe of investors attracted to a given country (Annex II). In particular:

Global bond benchmarks: Investors tracking global bond benchmarks, are less likely to react to risks specific to EMs, given that these countries represent only a small fraction of the total index and hence make a small contribution to their overall returns. However, EMs that are part of global bond benchmarks have a larger share of rating-sensitive investors, given that these indices use ratings as criteria for index inclusion/exclusion. As a result, they can face large selling from benchmark-driven investors in the event of a loss of investment grade status (e.g., Pemex and South Africa in April 2020, Petrobras in September 2015).²
EM bond benchmarks: Contrary to global bond benchmarks, the most widely used EM bond benchmarks are not rating sensitive and bonds remain eligible as long as the liquidity criteria are satisfied. This may have a stabilizing effect on portfolio dynamics given the procyclical nature of credit ratings that can aggravate crisis episodes (Ferri, Liu, and Stiglitz 2003).

An important feature of the most commonly followed version of EM benchmarks is a weighting method that reduces the weight of larger issuers and redistributes the excess to smaller countries. This can create distortions for asset allocation portfolio flows. For local currency government bonds, for example, these “diversified” indices limit the maximum weight to 10 percent, which leads to more concentrated positions of benchmark driven investors in some smaller issuers. For example, Brazil’s weight is capped 8 percentage points lower than it would be under the market capitalization weights used in global benchmarks (Figure 5). On the other hand, smaller issuers such as Colombia, Hungary and Peru see an increase in their weights by 1 to 2 percent. As the index is tracked by an estimated $300 billion (see next section), a 2 percentage points higher weight would mean $6 billion additional benchmark driven investments due to index rules, which can be substantial for smaller countries.

Index inclusion decisions by index providers can also manifest at the security level depending on whether certain securities meet liquidity criteria. One recent example of a large liquidity related weight change is Malaysia’s weight reduction from in GBI-EM GD from 10 percent on February 2016 to 6 percent on August 2017. Part of this reduction was due to measures taken by the authorities in December 2016 that, in the view of J.P Morgan, impacted the liquidity of Malaysian government debt securities due to changes in the non-deliverable forward (NDF) market (J.P. Morgan, 2017).³ Such a weight reduction, unlike reductions driven by market value, would naturally lead to a reduction of benchmark driven investors who may need to mechanically rebalance their positions.

Benchmark effects can affect countries due to the inclusion of new countries in the benchmark index. Index inclusion decisions can lead to substantial rebalancing of portfolios and can alter the risk characteristics of the asset class. For example, J.P. Morgan recently included China in the GBI-EM Global index (Annex II). Over the transition period, in addition to boosting flows to the Chinese bond market, this will also lead to an index weight reduction for other countries and consequently to some rebalancing by benchmark-driven funds.

B. How Do Benchmark Indices Affect the Behavior of Investment Funds?

Previous work has shown that even supposedly active investment funds tend to follow their benchmark relatively closely. This is evident in standard metrics of fund behavior, such as the “active share” of a fund, which is defined as the sum of the absolute value of deviations of the fund’s country weights from those of the benchmark (Cremers and Petajisto, 2009). Miyajima and Shim (2014) show that the median “active share” is only 17 percent, among active EM local-currency bond funds tracked by the Emerging Portfolio Fund Research (EPFR) Global. This indicates a very high (83 percent) overlap in country weights between fund portfolios and the relevant benchmark. Accordingly, the authors label funds with an active share of 0–10 percent as “closet index” funds, and funds with an active share of 10–20 percent as “weakly active” funds. Based on this definition, they show that nearly 70 percent of actively managed EM bond funds tracked by EPFR Global are either a “closet index” or a “weakly active” funds—that is, benchmark driven.

We confirm that this trend has remained in place by calculating the active share for EM local currency bond funds. We find that the active share has been declining steadily since 2008 among EM local currency bond funds tracked by the EPFR (Figure 7, panel 2).This finding is corroborated using other metrics, such as the average tracking error—the difference between the return of a fund and its benchmark—of EM local bond funds (IMF, 2019). The trend that active funds are becoming more passive may partly be explained by underperformance of active funds over the last decade as well as the trend towards low-cost funds by retail investors.

C. What are Potential Implications for Capital Flows?

Previous work has shown that the role of benchmark-driven investments results in a higher degree of similarity in cross-country asset allocations and capital flows, especially when comparing emerging market investors. For example, Raddatz and others (2017) find that benchmarks can have significant effects on international investments and affect capital flows through both direct and indirect channels. In particular, the authors find that benchmarks explain, on average, around 70 percent of country allocations even after controlling for macroeconomic, industry, and country-specific effects. They define the “benchmark effect” as the channels through which “prominent international equity and bond market indices affect asset allocations and capital flows across countries,” differentiating it from the role domestic and external factors play in country allocations.

Similarly, Cerutti and others (2019) find that EMs relying more on global mutual funds are more sensitive to external factors in terms of their gross equity and bond inflows. Recipient market liquidity and inclusion in global benchmark indices also increase sensitivities. They also find little robust evidence that institutional and macroeconomic fundamentals dampen sensitivities. At the same time, in a multiple event study, Pandolfi and Williams (2020) show that index inclusion events lead to significant declines in government bond yields and to appreciation of the domestic currencies.

Relatedly, studies have shown that mutual funds can propagate shocks through several channels, including (i) directly via their holdings (Broner, Gelos, and Reinhart, 2006), (ii) indirectly through overlapping ownership of emerging and advanced economies (Jotikasthira and others, 2012); and (iii) via fire sales (Coval and Stafford, 2007). Miyajima and Shim (2014) specifically explore the use of benchmark indices in EMs. They find that the use of benchmarks can give rise to correlated behavior on the part of even “actively” managed funds. They argue that managers of those funds tend to be evaluated by whether the returns of their investments match or exceed those of a particular index. As a result, although active managers do not replicate the portfolio weights of the benchmark, the career risk of short-term underperformance against their peers can induce them to form similar portfolios or to “hug” their benchmarks, increasing correlation of asset managers’ portfolio choices.

For policymakers, two contrasting implications from these effects are worth highlighting. On the one hand, a rising share of benchmark-driven investments can be a source of vulnerability because it increases the country’s exposure to external shocks, as capital flows become more sensitive to external conditions (Miyajima and Shim 2014, Raddatz and others 2017). On the other hand, a rising share of benchmark-driven investments can be a source of resilience because it reduces the country’s exposure to domestic shocks, as capital flows become less sensitive to deteriorating country conditions/fundamentals.

Figure 8 illustrates these two effects based on a stylized financial accelerator framework (Bernanke and others 1994). The financial accelerator describes the self-reinforcing feedback loop between economic fundamentals and financial conditions. For example, deteriorating economic conditions result in higher credit spreads as lenders become more concerned about borrowers’ creditworthiness (feedback loop A below). This tightening in financial conditions further slows the economy (feedback loop B). In emerging markets, external factors play an additional important role in affecting domestic financial conditions. For example, a jump in global risk aversion that tightens external financial conditions will also result in deteriorating domestic financial conditions (feedback loop C). Severe financial distress typically arises through some combination of domestic and external shocks, which set off a vicious cycle of deteriorating economic and financial conditions (Kamin and others, 2001).

In this framework, a low share of benchmark-driven investments means that foreign portfolio flows are highly responsive to domestic factors (i.e., feedback loop A is relatively strong), but not so responsive to external developments (i.e., feedback loop C is relatively weak). An example would be Argentina, which has only a small weight in EM local bond indices and most of its local debt is owned by unconstrained institutional investors.

Conversely, a high share of benchmark-driven investors means that portfolio flows are less responsive to domestic factors (i.e., feedback loop A is relatively weak), but highly responsive to external developments (i.e., feedback loop C is strong). An example would be Colombia, which has a relatively high weight in the EM local bond indices and whose foreign investor base has been relatively stable despite several domestic shocks.

III. How Large are Benchmark-Driven Investments in EM Bond Markets?

The past decade has seen a remarkable rise in the importance of asset managers for portfolio flows to emerging markets and a consequent increase in the importance of various benchmarks. Over that period, foreign investors doubled their holdings of EM government debt to more than $1.5 trillion. More than 60 percent of this increase came from foreign asset managers (Figure 9, panel 1). In this section, we discuss the magnitude foreign bond holdings of benchmark driven investors, particularly in the local currency bond market.

A useful concept for gauging the size of benchmark driven investments is to think of them as a pool of money that is allocated purely according to the index weights. This concept has the advantage that it can be used to calculate the dollar amount of a portfolio shift corresponding to a change in the benchmark index, which is highly relevant from a policy perspective because it helps assess and predict shifts in the availability of external financing. Specifically, based on this concept, a change in the weights of the underlying benchmark would trigger a reallocation of funds equivalent to the percentage change in the index weight times the dollar amount of benchmark-driven investments. For example, if the weight of country A in the index were reduced from 10 to 9 percent (say, due to the inclusion of another country), and assuming a pool of benchmark-driven investments of $100 billion, one would expect a reallocation of $1 billion away from country A.

In practice, there is much uncertainty about the holdings of benchmark driven investors. One reason is that funds follow benchmark indices to varying degrees. Another reason is that data providers like EPFR can only report assets under management (AUM) of those mutual funds and ETFs that reveal their holdings. Moreover, these vehicles are mostly used by retail investors, whereas most AUMs for major asset managers are from institutional investors (e.g., pension funds, endowments). These institutional investors may be invested in similar strategies as the mutual funds but details on their portfolio holdings at the country and security level are very limited.

Some information about the magnitude of benchmark driven investments can be gained from surveys of index providers on the amount of funds tracking their index, but these surveys also have shortcomings. For example, J.P. Morgan provides a monthly investor survey on AUM tracking its EM bond indices (Figure 9, panel 2). However, the survey likely does not fully capture AUM that are invested with mixed mandates and use a combination of benchmarks (e.g., EM hard and EM local currency or EM and DM mandate). Additionally, several other less popular benchmarks exist by other providers that have similar index construction but are not captured by the survey. As of Q3:2019, JPM survey of local bond funds tracking their indices were close to $230 billion.

We use two alternative approaches to estimate the size of benchmark driven investments. The first is a regression-based analysis, while the second is based on event studies focused on key index inclusion events in recent years. The two approaches are complementary and result in broadly consistent estimates of the size of benchmark-driven investments, suggesting that these stood around $300 billion at end-2019, somewhat above survey-based estimates.⁴

A. Regression Analysis

We obtain regression-based estimates of benchmark-driven investments in EM local bond markets using the approach described in Arslanalp and Tsuda (2015). The approach is based on Ballston and Melin (2013). It assumes that the entire pool of capital invested in EM local markets can be divided into two pools: one benchmarked to the GBI-EM Global Diversified (including only the countries in the index) and the other benchmarked to all EMs (including those outside the index) based on the size/market capitalization of their government bond markets (unconstrained). By looking at the amount allocated to individual markets, one could then estimate the relative proportions of these two pools over time by solving for a_t and b_t in the following equations:

\begin{matrix} F_{i t} = a_{t} w_{i, t} + b_{t} W_{i, t} + ε_{i, t} & (1) \end{matrix}

Subject to:

\begin{matrix} a_{t} + b_{t} = \sum_{i = 1}^{N} F_{i t} & (2) \end{matrix}

Where a_t is the pool of benchmark-driven investments at time t. b_t is the pool of unconstrained investors at time t. w_i,t is the weight of country i in the J.P. Morgan GBI-EM Global Diversified index at time t. W_i,t is the weight of country i’s bond market at time t based on market capitalization. F_i,t is the nominal amount of foreign holdings of country i’s bonds at time t, in U.S. dollars. ε_i,t is the extent to which portfolio managers are over-/under-weight country i at time t.

The estimation is conducted using a constrained least squares (CLS) approach given that a_t and b_t should add up to total foreign holdings for each month. The sample covers the period from January 2010 to June 2020 and includes 17 countries (Table 1). Of these 17 countries, 13 are “benchmark” countries (i.e., part of the GBI-EM Global Diversified index for most of the sample period), while 4 are “off-benchmark” (the Czech Republic, India, Israel, and Korea).⁵ The mix of countries provides useful heterogeneity for the estimation of a_t and b_t.

^{Table 1.}

Sample of Countries ^1/

^{^1/}Countries that are not in the GBI-EM Global Diversified index are shown in italics. The Czech Republic re-entered the index in April 2017. EMEA = Europe, Middle East, and Africa.

^{Table 1.}

Sample of Countries ^1/

Asia	Latin America	EMEA-EU	EMEA-Non EU
India	Brazil	Czech Republic	Israel
Indonesia	Colombia	Hungary	Russia
Korea	Mexico	Poland	South Africa
Malaysia	Peru	Romania	Turkey
Thailand

^{^1/}Countries that are not in the GBI-EM Global Diversified index are shown in italics. The Czech Republic re-entered the index in April 2017. EMEA = Europe, Middle East, and Africa.

^{Table 1.}

Sample of Countries ^1/

Asia	Latin America	EMEA-EU	EMEA-Non EU
India	Brazil	Czech Republic	Israel
Indonesia	Colombia	Hungary	Russia
Korea	Mexico	Poland	South Africa
Malaysia	Peru	Romania	Turkey
Thailand

^{^1/}Countries that are not in the GBI-EM Global Diversified index are shown in italics. The Czech Republic re-entered the index in April 2017. EMEA = Europe, Middle East, and Africa.

The relevant data for the estimation (total outstanding and foreign holdings of local-currency government debt securities) come from national data sources, as discussed in Arslanalp and Tsuda (2014). Figure 10 provides a summary of the data as of end-2019. Figure 11 provides the country weights (w_i,t and W_i,t) as of end-2019.

The results of the regression analysis for a_t and b_t are summarized in Figure 12. The results suggest that the pool of benchmark-driven investments in EM local currency debt markets was $330 billion at end-2019. The detailed results for 2014–19 are presented in Table 2.

^{Table 2.}

Regression Results of the Constrained Least Squares Estimation

Source: IMF staff estimates.

^{Table 2.}

Regression Results of the Constrained Least Squares Estimation

	2014	2015	2016	2017	2018	2019
a_t	259.6	247.1	258.1	359.0	311.1	327.0
t-stat	4.32	4.98	4.67	8.64	8.62	7.23
P>\|t-stat\|	0.00	0.00	0.00	0.00	0.00	0.00
b_t	367.3	318.4	339.6	365.8	362.5	400.9
t-stat	6.11	6.41	6.14	8.80	10.05	8.86
P>\|t-stat\|	0.00	0.00	0.00	0.00	0.00	0.00
F-stat	37.36	41.13	37.68	77.43	100.97	78.49
Prob>F	0.00	0.00	0.00	0.00	0.00	0.00

Source: IMF staff estimates.

^{Table 2.}

Regression Results of the Constrained Least Squares Estimation

	2014	2015	2016	2017	2018	2019
a_t	259.6	247.1	258.1	359.0	311.1	327.0
t-stat	4.32	4.98	4.67	8.64	8.62	7.23
P>\|t-stat\|	0.00	0.00	0.00	0.00	0.00	0.00
b_t	367.3	318.4	339.6	365.8	362.5	400.9
t-stat	6.11	6.41	6.14	8.80	10.05	8.86
P>\|t-stat\|	0.00	0.00	0.00	0.00	0.00	0.00
F-stat	37.36	41.13	37.68	77.43	100.97	78.49
Prob>F	0.00	0.00	0.00	0.00	0.00	0.00

Source: IMF staff estimates.

Figure 13 shows how the composition of foreign investors (between benchmark-driven and unconstrained) have changed over time during 2010–2019 and where they stand across countries as of end-2019 (Annex II provides the country-specific time series). During the three recent episodes of significant capital flow reversals, portfolio flows driven by benchmark investors were less sticky than other types of flows.⁶ The prospect of faster-than-anticipated policy normalization by the Federal Reserve in mid-2013 triggered large portfolio outflows driven by emerging-market-dedicated investment funds. Similarly, there were sharp portfolio outflows by emerging market benchmark-driven funds during the 2018 emerging market sell-off (April-August 2018), after almost two years of a steady buildup of their positions in local currency bonds, and the recent COVID-19 related shock in March 2020.

Estimates of benchmark-driven investments also vary significantly across countries (Figure 13, panel 2, Figure 14). As a result, the same kinds of shocks can prompt different capital flow responses in different countries. For instance, Turkey has a high share of benchmark-driven investments as a proportion of the total local currency foreign holdings, but it is relatively small as a percent of GDP. This is in contrast to Hungary where the systemic importance of such investors is much higher.

B. Event Studies

In this section, we present the results of two event studies that provide supporting evidence on the size of benchmark-driven investments. The approach follows Arslanalp and Tsuda (2015). Overall, the event studies suggest that the pool of benchmark-driven investments was around $370 billion during the third quarter of 2017 and $340 billion during the second quarterly of 2018, broadly in line with the regression results presented earlier.

The event studies could be based on episodes, in which a country’s weight in the J.P. Morgan GBI-EM index rises either due to a country or security inclusion event. When J.P. Morgan adds to the index a new country (country inclusion) or a new set of debt securities for a country already in the index (security inclusion), the country’s weight in the index rises. In either case, to track the index, benchmark-driven investors would need to allocate more capital to the country. In contrast, this should not affect unconstrained investors as, by definition, they are not guided by index weights. As a result, such events can create portfolio inflows by benchmark-driven investors, but not unconstrained investors, due to the technical (not fundamental) nature of the event. Arslanalp and Tsuda (2015) examined three security inclusion events (Colombia, Peru, Romania). In this paper, we examine two country inclusion events (Dominican Republic and Uruguay) that took place since 2015.⁷

A country’s weight in the index can change due to valuation effects or inclusion events. For example, Russia’s weight in the GBI-EM Global Diversified fell rapidly from 10 percent to 4.7 percent between March 2014 and end-March 2015, mainly due to the sharp drop in the value of the ruble during this period. This highlights the importance of separating valuation effects from exogenous changes in country weights. For that, we use the approach of Raddatz and others (2017) to separate the change in a country’s index weight (w_it) into two components: (i) the buy-and-hold component, which captures the valuation effects; and (ii) the exogenous component, which captures inclusion events.

\begin{matrix} w_{i t + 1} - w_{i t} = \underset{E x o g e n o u s c o m p o n e n t B u y - a n d - h o l d c o m p o n e n t}{\underset{︸}{(w_{i t + 1} - w_{i t} * \frac{R_{i t}}{R_{b t}})}} + \underset{}{\underset{︸}{(w_{i t} * \frac{R_{i t}}{R_{b t}} - w_{i t})}} & (3) \end{matrix}

Where w_it is the weight of country i in the J.P. Morgan GBI-EM Global Diversified index at time t. R_it and R_bt are the total gross returns of country i’s bonds and the benchmark, respectively, from time t to t+1. Moreover, as benchmark-driven investments should only respond to the exogenous part of the country weight change, the size of the benchmark-driven investor base can be estimated as follows:

\begin{matrix} B_{t} = f_{i t} / (w_{i t + 1} - w_{i t} * \frac{R_{i t}}{R_{b t}}) & (4) \end{matrix}

Where B_t is the benchmark-driven investor base at time t in U.S. dollars. f_it is the net foreign purchases of country i’s government bonds between t and t+1 in U.S. dollars.

Below, we estimate the benchmark-driven investor base using two country inclusion events that took place during 2017–18. Specifically, we estimate (B_t) in Equation 4 when Uruguay and Dominican Republic were included in the J.P. Morgan GBI-EM Global Diversified index during the third quarter of 2017 and second quarter of 2018, respectively.

Uruguay

Uruguay joined the J.P. Morgan GBI-EM index during July-September 2017. In June 2017, J.P. Morgan announced that it would include Uruguay in the GBI-EM index over a three-month period (July-September 2017) in a phased manner. As a result, Uruguay’s weight in the GBI-EM Global Diversified index rose from zero at end-June 2017 to 0.29 percent at end-September 2017.

To track the index, a benchmark-driven investor would need to allocate capital to Uruguay during this period. Indeed, net foreign inflows into Uruguay’s government bonds were exceptionally large when the country was included in the index—$1.1. billion over July-September 2017—the highest quarterly figure recorded in official balance of payments statistics between 2000–17. Based on Equation 4, this would imply a benchmark-driven investor base of around $370 billion at the time of the event (Table 3), slightly above the regression results presented in the previous section.

^{Table 3.}

Uruguay: Event Study