Taming original sin redux: The role of benchmark-driven flows

Emerging market capital flows remain highly sensitive to global shocks. This column analyses foreign investors in emerging market local-currency sovereign bonds depending on whether they are benchmarked to J.P. Morgan’s main local-currency government bond index or not. Benchmarked investors have increased their share of foreign holdings more than twofold over 2010–2025. Global push factors are strong determinants of benchmarked flows, whereas domestic pull factors determine non-benchmarked flows. Index design choices, such as country caps, can therefore materially affect flows to constituent countries.

Drivers of capital flows have regained research and policy interest as the world experiences many shocks and goes through far-reaching geoeconomic and political realignments (Bruno et al. 2026). While emerging markets (EMs) have been relatively resilient recently, the sensitivity of their portfolio flows to global shocks remains high, making them often innocent bystanders to developments in advanced economies (Bergant et al. 2020). Notably, monetary and financial conditions in countries such as the US, known as ‘push factors’, still greatly affect debt and equity portfolio capital flows to and from economies, leading to volatility in emerging markets’ external financing and complicating their macroeconomic and financial management. This volatility can arise in spite of countries’ fundamentals, ‘pull factors’, which have been improving for many emerging markets. Relatedly, the original sin redux literature (Carstens and Shin 2019) shows that borrowing in domestic currency does not insulate emerging markets from fluctuations in global financial conditions.

An important but at times overlooked aspect of these spillovers is the composition of investors. Not all investors are equally sensitive to push and pull factors (IMF 2026). In a recent paper (Arslanalp et al. 2026), we document the importance of foreign investors benchmarked to the J.P. Morgan GBI-EM Global Diversified index (B) and non-benchmarked investors (NB) in emerging market local-currency sovereign bond markets and analyse their drivers.

For a broad sample of emerging markets, benchmarked and non-benchmarked holdings amounted to $464 billion and $620 billion, respectively, at end-2025, up from $2 billion and $45 billion in 2004, but their growth has slowed in recent years. While benchmarked holdings initially increased, also reflecting the emergence of local-currency bond indexes, their share of total bonds outstanding has been broadly stable – up only from 2.8% in 2012 to 3.7% in 2025. Non-benchmarked holdings, the larger pool, have grown less than local markets and as a share declined from 11.1% to 5%. Benchmarked investors have become more important, increasing their share of foreign holdings from 16.9% in 2010 to 42.8% in 2025, as benchmark-driven assets under management (AUM) grew sevenfold, boosted by the inclusion of China (2020) and India (2024) in the GBI-EM index.

Importantly, the two investor groups allocate their funds very differently. Benchmarked investors commit funds to an index, after which country allocations are largely determined by the index. Non-benchmarked investors, who are broader and more heterogeneous, place greater weight on the fundamentals, risks, and expected returns of individual markets. When we analyse the drivers of changes in the respective holdings, we find that global push factors – proxied by the VIX and the US dollar index – are the only statistically significant drivers of benchmarked flows. This suggests that once a country enters the index, domestic fundamentals no longer determine the amount of benchmarked investment it attracts. In contrast, domestic pull factors – proxied by the bilateral exchange rate and sovereign credit rating – are the only significant determinants of non-benchmarked flows.

The global and domestic factors are comparable drivers of their respective investor bases. The variance decomposition, Figure 1, illustrates this. For benchmarked flows, push factors – the VIX and US dollar index – account for more than 90% of the explained variation. For non-benchmarked flows, pull factors – the bilateral exchange rate and the sovereign credit rating – account for nearly 75%. Flows of one investor type do not offset the other. Despite holding about one-third more assets than benchmarked investors, non-benchmarked investors do not systematically offset benchmarked outflows. And among domestic investors, only central and commercial banks tend to function as contrarian investors during global risk-off episodes, whereas nonbank investors do not respond systematically.

Figure 1 Determinants of benchmark-driven and non-benchmark-driven bond flows to emerging markets

Note: R-squared decomposition following Huettner and Sunder (2012).

Our finding that push factors are as important economically for benchmarked flows as pull factors are for non-benchmarked flows provides new perspectives. While benchmarked investors respond primarily to global factors, they do account for fundamentals indirectly by tracking an index whose country selection criteria reflect fundamentals. In contrast, domestic fundamental factors matter directly for non-benchmarked investors – countries with stronger macroeconomic fundamentals and institutions attract greater foreign participation in their local-currency bond markets and thereby overcome original sin. Non-benchmarked investors, however, pull back from countries when their exchange rates depreciate, possibly generating the perverse feedback loops emphasised in the original sin redux literature. Although benchmarked investors do not exhibit the same feedback loop, their simultaneous retrenchment in response to adverse global shocks can also generate an adverse dynamic through correlated outflows. This distinction between global shocks, captured by the US dollar index, and local shocks, captured by the bilateral exchange rate, is not always made clearly in the original sin redux literature (e.g. Hofmann et al. 2020 focus on bilateral exchange rates, while Bertaut et al. 2025 focus on the broad US dollar).

Our analysis has implications for index design. With benchmarked investors becoming more important, the allocation rules embedded in sovereign bond index benchmarks are increasingly consequential for constituent countries. Index weight alone, however, is an incomplete measure of a benchmark’s importance to the issuer. Benchmarked ownership also depends on how much of a country’s bond market qualifies for the index. The proportion of local-currency government bonds meeting requirements for bond structure, size, maturity, liquidity, and investor accessibility varies from about 11% in Türkiye to 93% in Peru (Figure 2). This means that a relatively small index weight can still translate into substantial benchmarked ownership relative to the overall market. Before China and India entered the index, benchmarked investors held, for example, 31.2% of Peru’s domestic bond market but only 3.2% of Brazil’s, despite the latter having a much larger index weight, as fewer of its bonds are eligible.

Figure 2 Benchmark-driven ownership and index-eligible market share

Note: Pro forma using projected post-India index weights, September 2023 benchmarked assets under management, and foreign holdings. Benchmarked holdings are shown as a share of the domestic bond market; eligible share is the proportion of that market qualifying for the index. Sources: IMF Sovereign Debt Investor Base; J.P. Morgan; SBI Research (2023); EPFR; national debt offices of the Czech Republic, Serbia, and the Dominican Republic; authors’ calculations.

Changes in the country universe also affect issuers differently than index weights alone would suggest. Holding benchmarked assets under management constant, adding China and India reduced Brazil’s index weight by 3.28 percentage points but lowered its benchmarked ownership by only one percentage point of its domestic market. Peru lost just 0.91 percentage points of index weight, yet its benchmarked ownership fell by 7.2 percentage points. Corresponding declines are 7.1 percentage points for Romania and 6.7 for Czechia. Without China and India, benchmarked investors would hold more than one-fifth of the market in these countries; with both included, that share falls to roughly one-seventh.

The country cap increases exposure to global shocks. In addition, a country cap (currently 10%, reducing to 9% by January 2027) limits the weights of the largest markets in the index. But this redistributes the excess to uncapped, smaller countries, which creates a ‘multiplier’ on their eligible-market shares: about 2.07 with a 10% cap and 2.32 with a 9% cap. The lower the cap, the less concentrated the index is in the largest markets, but the higher the index weight – and hence benchmarked ownership, and thus exposure to global shocks – for uncapped, smaller markets. Because benchmarked flows are highly sensitive to global push factors, higher benchmarked ownership relative to domestic market size means greater exposure to synchronised outflows when global financial conditions deteriorate. Adding large markets to the index when the cap is reduced allows them to absorb some of the weight that would otherwise flow to smaller markets.

Together, our results imply that investor diversification, issuer-side exposure, benchmarked assets under management, and transition costs need to be considered jointly when evaluating benchmark methodologies. Changes to the index can materially affect flows into and out of domestic bond markets and, in turn, bond pricing. For debt managers, index eligibility should be part of investor-base management rather than a goal in itself. Inputs to proposed changes in index methodology and the underlying pro-forma index weights need to be sufficiently transparent, and methodology reviews should address both steady-state allocations and transition effects.

Authors’ note: The views expressed are those of the authors and do not necessarily represent the views of the IMF, its Executive Board, or IMF management.

Source : VOXeu