In the 2000s, thanks to a steady debt deleveraging and de-dollarisation process, and to a lower dependence on external finance, emerging economies reduced their business cycle co-movement with advanced G7 economies (The Economist 2008). This decoupling with the G7 is partially replaced by a stronger Chinese influence (Levy Yeyati 2009). Did this newly gained macroeconomic resilience and autonomy enhance the relevance of local fundamentals as drivers of emerging market returns? Not at all, possibly due to the role of benchmarked institutional investors.1
In a recent paper (Levy Yeyati and Williams, forthcoming), we document the steadily growing influence of global factors and the persistently high emerging-market betas, even prior to the collapse of Lehman Brothers2. For example, the share of the time variability of asset prices in individual emerging markets that can be explained by their first principal component (a quick and dirty way to capture common factors) is remarkably high for all three main assets (stocks, bonds and currencies), and has been growing over time, even before the late-2008 to early-2009 selloff (Figure 1).
Figure 1. Asset co-movement in emerging markets across countries and over time
Notes: The figure shows median R-squares from estimations of assets monthly log returns in individual emerging markets against the first principal component of monthly log returns corresponding to the same asset in all emerging markets. Source: Bloomberg, MSCI, GEM World Bank.
Moreover, the influence of global assets’ performance on emerging markets has also been rising over time. This is readily illustrated by a back-of-the-envelope estimation of asset “betas” relative to standard global references, ie the S&P 500 stock market index for equity; the US high-yield corporate spreads for credit spreads; and the Broad US Dollar Index for exchange rates (Figure 2).3
Figure 2. Common factors in emerging markets and global assets
Note: This table presents correlation coefficients of the PC1 of EM/AM Equity/Spreads/FX against global assets for three periods during the 2000s. Source: Bloomberg, MSCI, GEM World Bank, US Treasury.
Why have market betas to global drivers remained so high? Why do a country’s fundamentals explain so little of short-run asset returns? In principle, this could be seen as the natural consequence of financial globalisation, to the extent that the latter increases the global nature of emerging markets’ investor base, thereby making it more homogeneous. In this context, one could think of common shocks to the supply of funds. For example, if the representative global investor faces a liquidity crunch, he would tend to liquidate assets across the board.4
Does foreign participation increase the market betas to global returns? A first glance at the data appears to contradict this hypothesis. Although in principle there seems to be a significant positive link between stocks holdings of foreign securities by US-domiciled investors and betas to US markets (Didier et al, 2010), a closer look reveals that this link is entirely accounted for by the group of frontier markets (Figure 3).5
Figure 3. US holdings and financial recoupling in equities: No pattern
Note: This figure shows the scatter plot of betas during the crisis (from country-by-country regressions of local equity returns on returns on the S&P500) against US holdings of foreign equity normalized by equity market capitalization at the end of 2007. ** denotes significance at the 5% level. Source: Didier et al,. 2010, US Treasury, Bloomberg.
The benchmarking effect
There are important differences, however, between the global investor and the professional portfolio manager. One of them is the fact that the latter is usually benchmarked (that is, his performance is evaluated relative to a predetermined index), which limits his margin to deviate from the benchmark weights, forcing him to buy and sell in all markets proportionally in the event of massive allocations or redemptions.
Do global mutual-fund flows increase the financial betas of emerging market assets? Indeed, they do. While they do not fully explain the high cross-market co-movement, they do display an amplifying effect. Larger fund flows are associated with higher betas (Figure 4). Interestingly, the effect is asymmetric. The volume of fund flows plays a significant role only during sharp sell-offs. A similar asymmetric pattern is found for credit spreads and currency returns.6,7
Figure 4. Benchmarking: Global funds amplify the “selloff” beta
Note: FG Flows are the absolute values of net inflows to a country by global and emerging mutual funds normalized by market capitalization. Selloff and Rally divide the sample between positive (or zero) and negative returns on the global asset. Robust standard errors in parentheses.
Source: MSCI, GEM World Bank, Bloomberg, US Treasury, EPFR, Barclays Capital, GEM World Bank, IMF IFS Statistics.
The underpinnings of benchmarking can be illustrated using individual fund data on flows and country weights. For a comprehensive sample of emerging market-dedicated funds, panel regressions of current against past country weights show a time correlation of weights that is close to one (Figure 5). This correlation reappears almost as strongly when we compute weights using the funds’ aggregate stock of assets under management.8
Figure 5. The stickiness of mutual funds’ portfolios
Note: Country returns are from MSCI country-specific stock market indexes. We exclude data when country weight is equal to zero. Fund Detail indicates that we use individual mutual fund data. Aggregated indicates we use aggregated flow data by country. Data is from 2005-November 2010. Robust standard errors in parentheses.
Source: MSCI, GEM World Bank, Bloomberg, US Treasury, EPFR, Barclays Capital, GEM World Bank, IMF IFS Statistics.
Thus, country weights in international funds tend to remain close to their benchmarks or, at any rate, adjust slowly (even in times of turmoil). In turn, flows in and out of individual countries replicate those stable weights and, in times of liquidation, induce correlated sales and price action.
It is easy to show how the influence of market weights on the cross-country correlation of fund flows increases in times of massive swings in fund allocations. After netting out monthly fund returns, we can decompose individual fund flows to a given country into those arising from the re-allocation flows within the fund due to portfolio rebalancing, and those reflecting new (net) flows to the fund, which are allocated according to current weights.9
It follows that, in times of little or no new flows to equity funds, the second term becomes irrelevant and the correlation between country flows and past weights due to rebalancing is negative. By contrast, in times of massive inflows or outflows, the second term dominates, and therefore hysteresis in portfolio composition (as noted, influenced by the composition of the benchmark) induces a strong cross-country correlation of flows in line with the results in Figure 4. It also suggests that, in a volatile global context dominated by systemic risks, the influence of country fundamentals on the behaviour of asset prices will continue to be dwarfed by the global risk factors.
Isn’t all this obvious? Financial globalisation, in the sense of exposure to a common pool of global investors, strengthens the impact of swings in risk appetite and liquidity preferences of those investors. Swings in global risk appetite would tend to hit risk assets in a globally diversified portfolio in a similar way. Indeed, the co-movement is not specific to emerging markets. The same pattern can be found in core markets in advanced economies.
True, but our finding points at a more specific, mechanical channel (benchmarking), with important practical implications. From the government standpoint, the increased co-movement translates into greater exposure to global financial shocks (for example, through their influence on credit spreads and exchange rates).
From the investor’s perspective, in turn, it casts doubts on the relevance of country fundamentals to explain short-run asset performance, and on the efficacy of relative value strategies that aim at profiting from these fundamental differences.
In fact, the benchmarking channel has a troubling circular nature, epitomised by the surge of Exchange Traded Funds, ie perfectly-passive mutual funds built to mimic a benchmark index. By increasing asset co-movement, such funds narrow the scope for diversification, reducing the efficacy of active management strategies, and feeding back into further “ETFication” – thereby partitioning the world into classes and regions that are likely to be fundamentally diverse.
Broner, F, G Gelos and C Reinhart (2006), “When in Perils, Retrench: Testing the Portfolio Channel of Contagion,” Journal of International Economics 69 (1), 203-230.
Didier, T, I Love and S Martinez Peria (2011), “What Explains Co-movement in Stock Market Returns during the 2007-2008 Crisis?,” International Journal of Finance and Economics, forthcoming.
The Economist (2008) “The decoupling debate”, 6 March
Forbes, K, and R Rigobon (2002), “No Contagion, Only Interdependence: Measuring Stock Market Co-movements,” Journal of Finance 57, 2223-2261.
Jotiskasthira, P, C Lunbland and T Ramadorai (forthcoming), “Asset Fire Sales and Purchases and the International Transmission of Shocks”, Journal of Finance.
Levy Yeyati, E (2009), “On Emerging Markets Decoupling and Growth Convergence,” VoxEU.org, 7 November.
Levy Yeyati, E and T Williams (2011), “Financial Globalization in Emerging Economies: Much ado about Nothing?,” Policy Research Working Paper Series 5624, The World Bank.
Levy Yeyati, E and T Williams (forthcoming), “Emerging Economies in the 2000s: Real Decoupling and Financial Recoupling”, Journal of International Money and Finance.
Raddatz, C and S Schmukler (forthcoming), “On the International Transmission of Shocks: Micro-Evidence from Mutual Fund Portfolios”, Journal of International Economics.
1 The role of institutional investors in transmitting shocks across markets (including through benchmarking) has been recently highlighted by Broner et al. (2006), Jotiskasthira et al. (forthcoming), and Raddatz and Schmukler (forthcoming).
2 Beta refers here to the coefficient of a regression of the asset returns on the global market return. The drawbacks of using standard correlations to estimate market interdependence have been repeatedly highlighted in the finance literature, most notably by Forbes and Rigobon (2002).
3 As before, the pattern cannot be attributed to the global financial crisis, as the pre-crisis period (2005-2008M02) exhibits betas for emerging market assets that are comparable to betas during the early 2000s. Interestingly, the pattern persists when we move to lower (quarterly or annual) frequencies.
4 A popular hypothesis in the 90s attributed this co-movement to imperfect information and herd behavior. However, the story is less appealing in the late 2000s given the increased sophistication of emerging market practitioners and the wealth of data now publicly available, and can hardly explain why co-movements should rise.
5 Tatiana Didier kindly provided the data for this table.
6 Because there is no data on currency flows to make comparable estimations, we use data from equity funds as a proxy. The assumption is that many professional investors go into equity markets unhedged, so that currency and equity cross-border flows are closely correlated.
7 The directional causation of the previous findings is not trivial. They could reflect the behavior of benchmarked professional investors facing contributions and redemptions while keeping a balanced portfolio. However, they could alternatively reflect money-chasing returns (or rushing to the exit) whereby asset managers increase their positions in countries with excess returns and cut their exposure in falling markets. However, similar results are found after controlling for endogeneity issues (Levy Yeyati and Williams, 2012).
8 The new, broader sample also includes country-dedicated funds and global funds. The persistence of country weights is not altered when crisis periods are singled out.
9 The decomposition is as follows
which, following the results of Table 6 and expressing wi,j,t as a function of wi,j,t-1, yields
with f ´ <0, g´ >0, where Flow i,j, t , AUM i,j, t and wi,j,t are flows, assets under management and weights in fund i to country j at time t. The first term corresponds to flows due to portfolio rebalancing, whereas the second represents the allocation of new flows.