JPMorgan

Fixed Income Research

Diversifying into Emerging Markets

  • Emerging markets have offered high returns but considerable exposure to event risk
  • Emerging markets offer substantial diversification benefits, even if
  • they are not expected to outperform the principal government bond markets
  • they are expected to be as prone toevent risk as in the past
  • Shifting 510% of global bond investments into emerging fixed income markets produces lower downside risk than a 100% global bond portfolio

1. Introduction

Risks and returns in major government bond markets have been transformed by the prospect of European monetary unification. Currency volatility has fallen substantially, and spreads in traditionally highyielding European markets have narrowed by several hundred basis points. These developments affect both the strategic and tactical aspects of managing global bond portfolios. Investors are now forced to take more extreme tactical positions (deviations from benchmarks) in these markets in order to produce expectations of returns comparable to those of the past~ and the strategic diversification benefits offered by spreading portfolios across European markets have diminished.

Emerging fixed income markets are an obvious place to look for new sources of diversification to broaden this shrinking investment universe. Emerging market bond spreads have narrowed over the last year, but still offer roughly 500 basis points, relative to comparable US Treasuries. However, these markets in the past have moved dramatically in response to adverse news concerning the fiscal health of the debtor economies. Is exposure to these risks justified by the prospective returns and diversification offered by emerging markets? Put simply, the question is whether one should think of emerging markets as filling the "highyielder" role that Sweden, Italy and Spain occupied until recently among global bond markets. The principal conclusion we reach is that a strategic alloca- portfolio is warranted on the basis of the diversification benefits alone. This result is not just an unremarkable consequence of some "more the merrier" principle: including further markets in a portfolio does not automatically improve diversification. Of the thirteen countries in J.P. Morgan's Global Bond Index (GBI), only six pass the test to which we subject emerging markets.

Our analysis is intentionally tilted against finding a strategic role for emerging fixed income markets. While we assume that emerging markets will be as prone to event risks as in the past, we do not assume they offer any risk premium relative to global bonds to compensate for bearing the event risks. This creates a bias for two reasons. FM in the downside risk framework we employ, any risk premium attributed to emerging markets assets would lower the risk of portfolios involving them. Second, it may well be that emerging markets only improve the riskreturn tradeoff at higher levels of expected return than that associated with the least risky portfolio. In short, emerging markets may "shift out the efficient frontier", but only at higher levels of expected return. The advantage of proceeding this way is that we can be sure that our results do not hinge on forecasts of the emerging markets risk premium. And to the extent that any risk premium is to believed to be present, the case for diversifying into emerging markets only becomes stronger.

Our analysis follows the common procedure of using historical returns to construct a picture of potential future return movements. History provides a good estimate of return dynamics that derive from a market's institutional characteristics, and so are slow to change. For example, the liquidity of a market and the constraints that drive its central participants will affect how dramatically returns swing in response to news and to movements of other markets. However. there is no guarantee that the news affecting the market will have the same character as in the past, although using history amounts to assuming this as well. In the case of emerging markets, the implied assumption is that, every three years, on average, there will be a crisis whose effect is similar to that of the Mexican peso crisis of 199495. This assumption is arguably conservative, because the structural reforms introduced in Latin American countries in the wake of that crisis, as well as Eastern Europe, have gone a long wav towards correcting external imbalances, and so such a

We look at emerging markets as an asset class, represented by LP. Morgan's emerging market bond (EMBI+) and local currency (ELMI) indices, rather than as individual instruments, or as individual country exposures. The risks of these indices have been lower than those of individual emerging markets, substantially so in the case of the Elm countries, but to a much lesser extent for the EMBI+ countries. Historical return patterns suggest that a GBI investor can limit event risk exposure by weighting the component countries equally: this strategy diversifies best in the absence of specialised knowledge to assess individual countries' exposure to adverse events.

Section Two of this paper describes the sources of the event risks in emerging markets. in terms of the behaviour of the component markets and their interaction with each other Section 3 examines the scope for diversifying event risk within the group of emerging markets, while Section 4 looks at diversification between emerging markets as a group, and the GBL

2. Risks in Emerging Markets: the Historical Record

Table I outlines the composition of J.P. Morgan's emerging market bond (EMBI+) and local currency (ELMI) indices. The EMBI+ tracks primarily USSdenominated long-term traded debt instruments of the governments lof 14 countries. Brady and Eurobonds make up the vast majority of Latin American issuance, while Russia's debt is in the form of commercial bank loans. The ELMI tracks the returns of 3month money market instruments in each country's local currency, and so its returns are driven by foreign exchange movements.

As Figure 1 and Table 2 attest, the EMBI+ has fluctuated greatly over the last three years. These movements have been mirrored by the individual markets composing the EMBI+. Even though these countries are geographically dispersed, their individual return profiles are highly correlated. Over the last three years, the average correlation of monthly returns between pairs of EMBI+ countries was 56%, and no country was negatively correlated with any other. Anticipated or actual movements in the US Treasury market have occasioned a general movement across the EMBI+ markets. For example, the rise in Treasury yields in early 1994 initiated a sell-off across the EMBI+ markets, leading to a decline in value over two months of 22%. However, the historical performance of the EMBI+ is in large part a story of sharp movements in response to positive or negative news of economic developments, structural reform programs, and the fiscal health of governments in the individual countries. Although these economic developments were only related to domestic fundamentals in each country, individual country indices responded not only to their own news, but also to other countries’. For example, the assassination of a Mexican presidential candidate in early 1994 precipitated a marketwide decline in prices.

The high degree of correlation has not been evenly spread through good times and bad: adverse moves have been more contagious than positive ones. For example, in December 1994 and January 1995, at the height of the Mexican debt crisis, all EMBI+ markets moved very closely together from one day to the next. Figure 2 is based on the daily correlations during each month of each country index with the EUBI+ aggregate index. It shows the level, each month, of the average and minimum correlation (across the 14 countries). In January 1995, the average correlation rose to 83%, the minimum to 79%. Indeed, every time the EMBI+ return has moved into negative territory, there has been a sharp jump in the average and minimum correlation of the individual markets with the aggregate EMBI+ index. Apart from the Mexican debt crisis, the most notable events are early 1994, as mentioned above, and February 1996, when urtmet expectations of the Fed easing prompted selling of EUBI+ instruments. However, the contagion relationship persists in less dramatic episodes, causing the history of the average monthly correlation to move closely (inversely) with the EMBI+ return itself. the correlation between the two series is 54%.

These results suggest that the EMBI+ markets offer the least diversification just when it is need most, or, equivalently, that the event risk is difficult to diversify. However, this does not change the overall picture presented by the EUBI+ in Figures I and 2: the EUBI+ markets as a whole have paid a monthly return premium of nearly 4% for bearing the exposure to the event risk. The individual components of the ELMI paint a very different picture. The average correlation between monthly returns of pairs of the ten component markets is 10%, against the EMBI+ average of 56%, mentioned above. One third of the pairwise correlations among ELMI countries are negative. Similarly, there is no observable tendency in any month for the daily correlations of individual countries with the ELMI aggregate index to rise when the ELMI return falls. During the Mexican crisis starting at the end of 1994, the average correlation of countries with the EIM actually fell from about 40% to close to zero, reflecting the fact that most countries other than Mexico experienced positive returns during this period.

Figures 3 and 4 provide a direct comparison of the different ways in which the aggregate ELM and EMBI+ indices relate to their component markets. Each point in the Figures represents one of the 36 months from 199496, its coordinates being the return of the best and worst performing countries in that month. A straight line drawn through the EMBI+ points would have a positive slope (of roughly 1.5), which indicates that worst and bestperforming markets have tended to move closely together, as if a single "emerging markets factor" dominates returns across the EMBI+ countries. (If the best and worstperforming markets moved pointforpoint, all the points would he on a line with a slope of 1). In contrast, no such relationship is evident among the EUVH countries (Fig 4), where the return of the bestperforming market in any month is independent of the size of that month's worstperforming market's return.

What explains these differences in interaction between the components of the two emerging markets indices? Obviously, the individual components of the EN[BI+ should be mom sensitive to domestic economic developments that impinge upon the solvency of the debtor. However, these domestic developments have a tendency to spill over into other EMBI+ markets' performance. This contagion, plus the frequent linking of EMBI+ performance with the US Treasury market, suggest that the principal reason why EMBI+ markets move so closely together emanates from investors who view the bonds of the ENIBI+ countries as an asset class. Indeed, the emerging bond markets developed first as the "Brady market", a name which suggests primarily an asset class perspective as opposed to a country perspective. While the immediate holders of emerging markets debt may be specialist mutual fund managers, their endinvestor clients invested in emerging markets as a group: redemptions of emerging markets mutual funds in response to losses in one market would thus be spread over all markets covered by the funds.

In contrast, the assets in the ELMI do not appear to be held predominantly by the same investors. The bulk of TBills and deposits are held primarily by domestic residents, who have less of a tendency to value their holdings in hard currency. Thus, for example, them is little motivation for holders of Czech Konma deposits to liquidate if the Turkish lira falls suddenly. as happened in early 1994. This, of course, does not say anything about the likelihood of adverse circumstances befalling any individual ELM market.

Diversification should be the expected result of individual countries' central banks pursuing their own monetary policies. This contrasts with the debt comprising the EMBI+, which is not an instrument of monetary policy. As a result, the "demand" or investor side of the market should be expected to have greater influence on the EMBI+. It should also be mentioned that most of the ten component currencies operate to a greater or lesser extent under managed floating regimes. As a result, adjustments of exchange rates to fundamentals of ELMI markets; win tend to be occasional and large, coinciding with the authorities' inability to sustain further outflows or inflows of reserves rather than spread over time. So individual markets in the ELMI can be quite prone to event risk. The argument is merely that a run on one currency will not spill over into a run on another.

3. Diversifying among emerging markets

Index weighting rules

Both emerging markets indices are constructed with a view to representing the "size of the market", which is the customary way of providing exposure to an asset class. Each country in the EMBI+ is represented in proportion to the market value of its outstanding externalcurrency debt that meets certain liquidity criteria. To be included in the ELMI, a country must have no restrictions on currency convertibility or participation by nonlocals in its short term money market, and the money market must be of sufficient size and liquidity. Since there is no available direct measure of the size of each market, the weights are based on the combined value of exports and imports.

Equal weighting to diversify event risk

Satisfaction of the above criteria is certainly necessary for any country to be included in its respective index, and to be of interest to GBI investors. The weighting schemes that are used are also relevant for large holders of the assets, who could move the market in some of the smaller countries if they were not to pay attention to their relative size. However, it does not follow that the weights are the most appropriate to the GBI investor who is marginal to emerging markets, and puts a premium on limiting exposure to event risk. This investor would reason as follows: "Because (initially) I do not envisage investing in expertise specific to emerging markets, I am unable to distinguish one market's event risk from that of another. To diversify my event risk exposure, I should therefore weight each market equally?". Of course, an emerging markets specialist would have views on which markets are most likely to experience adverse events, and on their size, and could well advocate weighting markets differently, not only from equal weights, but from the index weights. The point is that equal weights are broadly consistent with the GBI investor's knowledge of the likelihood of events, and overriding desire to avoid exposure to them.

The ability of the equal weighting strategy to control event risk is limited by the degree of correlation among the individual markets. In view of the results presented so far, them is much mom scope to produce such diversification among the EIM countries than among those in the EMBI+. Nevertheless, because the GBI investor will not want to rely on specialist knowledge, equal weighting of the EMBI+ is still worth considering. In order to ensure that liquidity problems would not arise, we limit the equal weighted EMBI+ portfolio to the nine largest countries. This excludes five countries, whose combined capitalisation currently accounts for 5.7% of the total outstanding.

Table 3 compares the equally weighted (EW) indices' performance to those discussed so far. Equal weighting raises the average return of the EMBI+, but does little to affect the other statistics. In particular, the negative events remain, and the range of variation has actually increased. The effect of equal weighting on the volatility of the ELM is quite substantial, however. The negative event regime is diversified away, and as a result, the measured return volatility falls, and the average return rises. The lowest return over the 36 months is now less than three standard deviations from the mean, quite in keeping with a normal distribution. A drawback of the equally weighted ELMI is that it is more correlated with the GBI.

4. Diversification between emerging and GBI markets

We now examine the lowest risks that can be achieved by combining the GBI with the emerging markets aggregates discussed so far. While volatile, the returns of the emerging markets indices have a very low overall correlation with the GBI (Table 3). Taking these volatility and correlation figures at face value. a simple calculation reveals that diversifying into emerging markets would have lowered portfolio volatility for a GBI investor' . For example, switching 20% into the ELMI would have resulted in a monthly return volatility of 1.25%, compared with the GBI's 1.42%. However, these encouraging figures need to be treated with caution: measuring volatility in markets subject to event risks can conceal a potential for large losses, which, although infrequent~ may be unacceptable to investors who require stable portfolio values over short time horizons. so it does not necessarily follow that risks would have been lower (see Appendix).

To retain a hold on the potential for exposure to event risks, we look at the downside risk of a portfolio, which we measure here by its propensity to produce negative returns over a horizon of one month. Specifically, the goal is to identify the amount (if any) of emerging markets exposure that, in combination with GBI exposure, produces the lowest risk of negative returns in a month. To quantify a portfolio's risk, we measure its average shortfall: the average size of negative returns, multiplied by their probability (see Appendix, where the calculations are illustrated using a threshold of 4%, for ease of exposition). For purposes of comparison, we will also show shortfall probability, which is the probability of experiencing negative returns.