IV. Self-Insurance Through International Reserves

Paolo Mauro, Torbjorn Becker, Jonathan Ostry, Romain Ranciere, and Olivier Jeanne
Published Date:
April 2007
  • ShareShare
Show Summary Details

On the asset side of countries’ external balance sheets, international reserves constitute the main form of self-insurance against damaging crises, partly because of the speed and ease with which they can be used in a period of balance of payments pressure and partly because of their role in underpinning the credibility of the monetary and exchange regime, such as to help ward off external pressures and currency crises. Reserves are costly to hold, however, because they yield a return that is generally lower than the interest rate that the authorities must offer on their debt. The authorities must therefore strike a balance between the cost of reserves in noncrisis times and their benefits in crisis times.

Trends in Reserve Accumulation, and Benefits and Costs of Reserve Holdings

Total international reserves in the world economy have more than tripled since 1990, partly reflecting the rapid deepening in trade and financial integration (IMF, 2003; and Flood and Marion, 2002). The increase has been particularly pronounced in Asian emerging economies, whose share of global reserves increased by more than 20 percentage points between 1990 and 2005 (Figures 4.1 and 4.2).17

Figure 4.1.International Reserves, by Country Group, 1990–2005

Source: IMF, International Financial Statistics database.

Note: Data for 2005 refer to stocks at the end of the second quarter of 2005.

Figure 4.2.International Reserve Ratios, by Country Group, 1990–2005

Sources: IMF, International Financial Statistics (IFS) and World Economic Outlook (WEO) databases; and World Bank, World Development Indicators database.

Notes: Data for 2005 refer to the end of the second quarter for the stocks of reserves and M2, and to WEO projections for GDP and imports. “Emerging others” includes emerging market economies in Eastern Europe, the Middle East, and Africa. For each country group, the data refer to unweighted cross-country averages. For the developing countries, reserves as a share of M2 and reserves as a share of short-term debt are medians to avoid undue influence of outliers.

The causes of rapid reserve accumulation have been the subject of fervent debate—for example, regarding whether accumulation in Asia reflects a desire to keep exchange rate stability in the face of significant current and financial account inflows, rather than crisis-prevention efforts (Dooley, Folkerts-Landau, and Garber, 2004; Goldstein, 2004; and Genberg and others, 2005). An analytical framework to gauge the benefits and costs of international reserves would thus seem helpful for shedding light on recent developments as well as giving policy guidance. This subsection reports results based on a calibrated model for emerging market countries (Jeanne and Rancière, 2006). The model focuses on an open economy that is hit by financial flow reversals, which, in turn, lead to falls in output. The optimal level of reserves in the model equates the opportunity cost of holding foreign reserves with the marginal benefit from being able to smooth domestic consumption in the event of a sudden stop.


From an insurance perspective, countries hold reserves to achieve a range of objectives, whose relative importance depends on several factors, such as the exchange rate regime and the degree of integration into international financial markets. Such objectives include limiting volatility in the exchange rate; providing liquidity to the foreign exchange market, thus making a floating exchange rate regime more efficient; buffering the domestic economy against shocks to the balance of payments; and providing liquidity to the domestic financial markets and the banking sector, especially if there is significant dollarization (Jeanne and Wyplosz, 2003).

With countries’ increasing international financial integration, considerations regarding reserve adequacy have shifted from an emphasis on trade (traditionally associated with the “three-months-of-imports” rule) to the financial account and balance-sheet fragilities (associated with the Greenspan-Guidotti rule, according to which reserves should cover short-term debt).18 Indeed, if one considers a sample of 33 middle-income countries during 1980–2003, 31 out of 40 episodes in which reserves decreased by more than 10 percentage points of GDP are associated with a sudden stop, a currency crisis, or a banking crisis.19

Reserves have played an important role in cushioning domestic absorption (that is, consumption and investment) during financial account reversals. The typical sudden stop—which sees average net inflows of 5 percent of GDP (in the year prior to the stop) turn to net outflows of 4 percent of GDP (in the year of the stop)—is accompanied by a large drop in reserves (by 4 percentage points of GDP). This drop helps to hold the observed fall in domestic absorption to 2½ percentage points of GDP, which is much smaller than the counterfactual fall in domestic absorption (6½ percentage points of GDP) that would have been observed if reserves had been held constant during each episode.

It has also been argued that reserves help reduce the likelihood of crises, including by discouraging speculation against the domestic currency; making it easier for the public and private sectors to roll over their foreign currency debt; and instilling confidence in the domestic financial sector. Empirical evidence on these benefits is mixed, owing in part to methodological difficulties. An increase in the ratio of reserves to short-term debt has been found to be associated with a lower probability of a currency crisis or sudden stop (Bussière and Mulder, 1999; and Garcia and Soto, 2006).20 However, the robustness of this finding has been questioned out of sample (Berg, Borensztein, and Pattillo, 2004) and because of reverse causality (Detragiache and Spilimbergo, 2001).


The cost of holding reserves is usually measured as the difference between the return on short-term foreign currency assets and the return on more profitable alternative investment opportunities. The simplicity of this approach, however, masks thorny questions regarding the appropriate definition of alternative investment opportunities, which have traditionally been interpreted as the repayment of foreign debt (Frenkel and Jovanovic, 1981; and Flood and Marion, 2002) or higher-yielding investment opportunities in the domestic business sector or public infrastructure (Ben-Bassat and Gottlieb, 1992; and Hauner, 2005).

Even following the traditional approach based on the opportunity cost of repaying long-term foreign debt, plausible arguments can be made both in favor of and against using only some subcomponents of the yield on bonds issued by an emerging market country. The full yield can be thought of as the sum of four components: (i) the short-term rate on U.S. treasury bonds, (ii) the term premium—the difference between long-term and short-term U.S. interest rates, (iii) the default premium, and (iv) the risk premium owing to a possible correlation between the tendency for defaults and the global cycle or global asset prices. Although in practice the default premium and the risk premium cannot be observed separately, they are conceptually distinct and may need to be treated differently. There is some evidence to suggest that the risk premium is small, because there is essentially no relationship between emerging market defaults and the “world market portfolio” held by international investors (Borensztein and Mauro, 2004). However, to the extent that investor classes are segmented, and investors specialize in emerging market bonds and other assets that comove strongly with them, the risk premium could be significant. Regarding the default premium, one could argue that it should be excluded on the grounds that, on average, it is a fair reflection of the probability of nonrepayment. At the same time, country authorities that have no intention of defaulting will likely include the default premium when computing the cost of holding reserves. Given the size and variation of the default premium across countries and over time, the choice of whether or not to include it has important implications for the results, as is discussed later on. The opportunity cost of holding reserves is thus estimated using two variants: (a) the U.S. term premium—that is, the difference between long and short rates excluding the default and risk premiums—which is the baseline in this paper; and (b) the full difference between emerging market yields and U.S. short-term treasury bonds.21

On average, the total cost of holding reserves was substantially lower in Latin America than in Asia in 2001–2005 (ranging between 0.2 and 0.4 percentage points of GDP in the former and 0.3–0.8 percentage points of GDP in the latter) using the term premium (Table 4.1). The cost was, however, relatively similar in the two regions using the term premium plus the spread (ranging between 1.0–1.6 percentage points of GDP in Latin America and 0.7–1.2 percentage points of GDP in Asia). This is explained by the fact that although, on average, the reserves/GDP ratio is twice as high for Asian countries as for Latin American countries, the sovereign spread is substantially higher in Latin America than in Asia. The cost of holding reserves has increased moderately for emerging market countries over the past few years, as the impact of reserve accumulation has been partly offset by a reduction in sovereign spreads.

Table 4.1.Cost of Reserves in Emerging Market Countries, 2001–2005
Reserves/GDP (in percent)
Latin America12.111.612.513.514.0
Cost of reserves (in percent of GDP)
based on term premium
Latin America0.
Cost of reserves (in percent of GDP)
based on term premium + sovereign spread
Latin America1.
Sources: IMF, International Financial Statistics and World Economic Outlook databases; and JPMorgan.Notes: The cost of reserves is computed as the stock of reserves times either the term premium (the differential between the yields on 10-year and 3-month U.S. treasuries) or the term premium plus the EMBIG spread. All regional averages are unweighted.
Sources: IMF, International Financial Statistics and World Economic Outlook databases; and JPMorgan.Notes: The cost of reserves is computed as the stock of reserves times either the term premium (the differential between the yields on 10-year and 3-month U.S. treasuries) or the term premium plus the EMBIG spread. All regional averages are unweighted.

Although the previous analysis assumes that reserves are held in liquid assets such as short-term, fixed-income instruments issued by the major advanced countries, the cost of holding reserves might be reduced by investing in longer-term, higher-yielding foreign assets (Genberg and others, 2005). For the specific case of economies experiencing large financial inflows and accumulating substantial reserve stocks, a related idea is to use these favorable external circumstances to make gradual progress toward financial account liberalization, with a portion of financial inflows being securitized through closed-end mutual funds that issue shares in domestic currency and use the proceeds to purchase foreign exchange from the central bank and then invest abroad (Prasad and Rajan, 2005). This would eliminate the fiscal costs of sterilizing inflows, give domestic investors opportunities for international diversification, stimulate the development of domestic financial markets, and allow central banks to control both the timing and quantity of outflows.

Another way of reducing the cost of self-insurance through reserves may be to hold assets that provide liquidity at the time when countries most need it, thus obviating the need to hoard large amounts of reserves in noncrisis times. For example, a commodity exporter might hold instruments whose return is inversely related to the price of its main exports; or an emerging market country could hold instruments that provide liquidity when international liquidity conditions are tight. One possibility would be for the central bank’s portfolio to include instruments whose payoff is positively related to the implied volatility index on the S&P 500—the so-called VIX, which has displayed an empirical association with recent sudden-stops episodes. Such a proposal—advanced by Caballero and Panageas (2004 and 2005)—is summarized in Box 4.1.

Framework for Assessing Optimal Level of Reserves

While it has been recognized that the benefits of reserves are multifaceted, reserve adequacy has traditionally been assessed using rules of thumb based on ratios of reserves to imports or, reflecting greater international financial integration, to short-term debt. Although these provide useful guidance, they lack fully developed analytical foundations. The remainder of this section seeks to fill this gap by developing an analytical framework that may be useful for making judgments about the level of reserves that is warranted by a country’s particular fundamentals. The analysis focuses on the role of reserves as self-insurance against vulnerabilities resulting from changes in the financial account balance and may thus be especially relevant for emerging markets facing a risk of sudden stops.22

Box 4.1.State-Contingent Reserves Based on Volatility Index

Given the cost of holding reserves, a strategy that made it possible to have reserves available only when they are needed would yield considerable savings. With that goal in mind, a hedging approach for the management of international reserves based on the use of contingent securities has been proposed by Caballero and Panageas (2004 and 2005). The authors suggest that a central bank would find it optimal to invest part of its reserves in contingent securities that provide a return when a sudden stop occurs. The authors argue that such a hedging strategy can be efficiently implemented by using options on the Standard and Poor’s (S&P) 500 implied Volatility Index (VIX).

The VIX is an index of near-term U.S. financial market volatility based on the prices of S&P 500 stock index options. It tends to increase when investors are more reluctant to participate in risky markets. More importantly, the authors show that spikes in the VIX are correlated with sudden stops in emerging markets. (The probability of observing a statistically significant jump in the VIX, conditional on a sudden stop, is about 70 percent.) The authors derive the optimal hedging portfolio of a central bank using risk-free liquid assets and call options on the VIX. This portfolio increases the expected reserves during sudden stops by as much as 40 percent, compared with a portfolio invested only in risk-free assets.

An advantage of VIX derivatives for central banks is that the VIX index is correlated with sudden stops but is independent of country policies, thus limiting moral hazard and measurement issues. The current size of the VIX market is relatively small, however. (The notional value of open futures contracts is about US$1.5 billion. Options are traded over the counter.) Two potential concerns would remain even if the market were to become much larger. First, it is not clear whether the correlation observed in recent years between occurrences of sudden stops and spikes in the VIX index will persist in the years ahead. Second, investors with short positions on call options on the VIX might already be exposed to emerging markets through their holdings of other assets; in the event of a sudden stop, the central banks might then face a serious counterparty risk. Investors might be unable to honor their obligations undeer the VIX contracts, should they sustain losses on other assets as well.

The main benefit of reserves in the context of the model is that they help to smooth domestic consumption in response to decreases in financial inflows and output. Based on a calibration of the model, the warranted level of reserves can be derived as an explicit function of factors that include the probability of a sudden stop, the size of a sudden stop, the output cost of a crisis, the opportunity cost of reserves, and the authorities’ risk aversion regarding crisis episodes. The probability of a sudden stop is based upon panel probit regressions that relate sudden stops to country fundamentals (including the exchange rate regime, financial openness, the level of public debt, de facto dollarization, output growth, and exchange rate overvaluation).23

Warranted Reserves as Function of Macroeconomic Fundamentals

The model can be used to compute the impact of a change in fundamentals (for example, a change in the debt/GDP ratio) on the warranted level of reserves.24 A change in fundamentals will affect the probability of a sudden stop (to an extent estimated by the probit model), and the calibration will determine the implications for the optimal level of reserves. Plausible changes in fundamentals have a substantial impact on the optimal level of reserves (Table 4.2). The (comparative statics) results may be summarized as follows:

  • Moving from no real exchange rate overvaluation to an overvaluation of 20 percent increases the estimated annual probability of a sudden stop by approximately 4.3 percentage points and generates an increase in the optimal level of reserves of 2.7 percentage points of GDP.

  • A rise in the ratio of public debt to GDP from 40 percent to 60 percent implies an increase in the optimal reserve ratio by 1.7 percentage points of GDP.

  • A large buildup of the ratio of foreign liabilities to money, calibrated on the experience of Thailand between 1990 (45 percent) and 1997 (262 percent), raises the optimal level of reserves by 3.7 percentage points of GDP.

  • An increase of one standard deviation in the degree of financial openness (measured by the absolute value of gross inflows, divided by GDP) leads optimal reserves to increase by 3.2 percentage points of GDP.

  • A change from a floating to a fixed exchange rate regime induces an increase in the optimal level of reserves of 3.4 percentage points of GDP.

Table 4.2.Changes in Fundamentals and Optimal Reserves: Simulations for Emerging Markets(In percentage points unless otherwise noted)
FundamentalsSample MeanParameter Change (in percent)Estimated Change in Sudden Stop ProbabilityChange in Optimal Ratio of Reserves to GDPChange in Optimal Level of Reserves (in months of imports)Change in Optimal Ratio of Reserves to Short Term
Exchange rate overvaluation00 →
Public debt/GDP4040 → 602.21.70.821.8
Foreign liabilities/money4645 (Thailand, 1990) → 262 (Thailand, 1997)
Financial openness as (|gross inflows|)/GDP5.55.5 →
Exchange rate regimeFloating → Fixed5.33.41.643.5
Source: See Appendix I.Notes: The initial optimal level of reserves is equivalent to 8.2 percent of GDP (about 3 months of imports, or 105 percent of short-term debt). The results are based on regression (1) in Table A3.2 (in Appendix III)—except for the exchange rate regime, which is based on regression (2). All fundamentals are averages of the first and second lags.
Source: See Appendix I.Notes: The initial optimal level of reserves is equivalent to 8.2 percent of GDP (about 3 months of imports, or 105 percent of short-term debt). The results are based on regression (1) in Table A3.2 (in Appendix III)—except for the exchange rate regime, which is based on regression (2). All fundamentals are averages of the first and second lags.

The changes in fundamentals analyzed previously are sizable but have certainly been observed in the sample considered. Interestingly, such assumed changes in one fundamental variable lead to large changes in optimal reserves, in some cases by one-third to one-half of the initial optimal level of reserves. The next subsection considers the impact on optimal reserves of the observed combination of changes in all fundamentals simultaneously.

Trends in Optimal Reserves for Country Groups

The optimal level of reserves—based on the calibrated model—is computed for each country and year for the 33 middle-income countries used in the probit estimation over 1980–2003. To trace the implications of the model for trends in optimal reserves, results are then summed up to obtain regional averages for Latin American and Asian emerging markets.25 For each country and year, the probability of a sudden stop is computed on the basis of the probit estimates. The size of the sudden stop is set to its realized mean value in each region and each decade. The only unobservable parameter, risk aversion, is set equal to six—a value selected to match the actual mean level of reserves to GDP in Asia in the middle of the sample period (1991).26 Using this approach, it is possible to compare changes over time in the optimal level of reserves, actual reserves, and the three-months-of-imports and Greenspan-Guidotti rules of thumb (Figure 4.3).

Figure 4.3.Asia and Latin America: Reserves as Shares of GDP, 1980–2003

(In percentage points of GDP)

Sources: IMF staff calculations using data from IMF, International Financial Statistics database; and World Bank, Global Development Finance database.

Note: Total reserves minus gold (excludes IMF financing). See text and Jeanne and Rancière (2006).

For the group of Asian emerging markets, the model suggests that reserves should have declined somewhat between the early and the late 1980s: in the aftermath of the debt crisis of the early 1980s, a slowdown in financial flows to emerging markets contributed to reducing the probability of sudden stops. Beginning in the early 1990s, the model envisages a rapid increase in optimal reserves, owing to rising international financial integration.27 The slight decline in optimal reserves following the Asian crisis is primarily accounted for by the reduction in public debt and financial flows. Although the model is intended to be normative, it is interesting to note that it outperforms the rules of thumb in predicting the actual level of reserves for most of the period under consideration. In particular, the upward trend in reserves in Asia during the 12 years prior to the Asian crisis (1985–96) is matched more closely by the model than by the alternative rules depicted in Figure 4.3.

For the Asian emerging market countries following 1997–98, however, the model suggests that the buildup in reserves may have been excessive—a finding consistent with previous analyses (IMF, 2003). A possible caveat is that the Asian crisis may have led to an upward revision of the size of sudden stops or of the associated output loss, though the revision would need to be very large for actual accumulation to be consistent with the increase in optimal reserves predicted by the model. For example, in order for the model to explain the increase in the average level of reserves held by emerging Asian countries between 1997 and 2003, the expected size of either the sudden stop or the output cost would have had to more than double relative to its average level observed in the 1990s.28 Another possible caveat is that the crisis may have led some countries to revise their views on the availability of other sources of insurance—such as financial support by the international financial institutions. A further note of caution on the result of excessive reserve accumulation is that some countries may envisage measures to increase their degree of integration in international financial markets and may thus be preparing for greater exposure to international financial flows.

In Latin America, warranted reserves appear to have been well in excess of actual reserves in the 1980s, a turbulent period for the region. Predicted reserves are lower in the early 1990s, partly on account of improved fundamentals (public debt and economic growth), but the rise over the course of the past decade seems to be in line with heightened international financial integration and increasing de facto dollarization. The close match between the model and the data for 1991–2003 is notable, considering that no individual-year level of reserves for Latin America has been used in the calibration.29 The model might be interpreted to suggest that the current level of reserves is, on average, adequate in Latin America. It is important to recall, however, that although the model includes de facto dollarization among the determinants of the likelihood of sudden stops, it does not take into account the possibility that output losses may be greater for dollarized economies in the event of a crisis.

Note: This section was prepared by Olivier Jeanne and Romain Rancière.

Considering the ratio of reserves to M2 (a measure of the country’s ability to withstand a sudden shift in demand from local currency to foreign currency), Latin America displays the highest ratio, as might be expected in light of widespread de facto dollarization in several countries, but the smallest increase. Of the 34 emerging economies in the sample, 7 are defined as highly dollarized by Honohan and Shi (2002); of these, 5 are located in Latin America.

The operational implications of this greater emphasis on balance-sheet considerations are further analyzed in Mulder (2000) and IMF (2000 and 2001).

Banking and currency crises are drawn from Caprio and Klingebiel (2003) and Ghosh, Gulde, and Wolf (2002), respectively.

A higher level of reserves has also been found to reduce the portion of exchange rate volatility that is unrelated to macroeconomic fundamentals (Hviding, Nowak, and Ricci, 2005).

The analysis considers the ex ante cost of holding foreign reserves and thus, in light of the unpredictability of exchange rates, abstracts from exchange rate gains or losses, even though these can turn out, ex post, to have been substantial (Hauner, 2005).

For further details, see Jeanne and Rancière (2006). Other related studies include Aizenman, Lee, and Rhee, 2004; Lee, 2004; Aizenman and Lee, 2005; and Caballero and Panageas, 2005.

More specifically, in the version of the model presented in Appendix III, the probability of a sudden stop is estimated for each country and each year by applying the regression coefficients to the observed country fundamentals. The size of the sudden stop and the output cost of a crisis are assumed constant (and identical across countries) and are estimated as the observed averages for the 33 countries in the sample over 1980–2003. Risk aversion is also assumed constant, across countries and over time, at a level that is well within the range of plausible estimates available from existing studies. The opportunity cost of holding reserves is the term premium and therefore varies over time but not across countries.

The exercise is based on a benchmark economy characterized by the calibration parameters presented in Appendix III (Table A3.1), with the cost of holding reserves based on the term premium only. The effect of a change in each fundamental on the optimal level of reserves is analyzed by comparison with the optimal level of reserves in the benchmark economy.

The remaining emerging markets in the sample form a limited and heterogeneous group—the related average is therefore not reported.

Ogaki, Ostry, and Reinhart (1996) provide estimates by region of this parameter.

A further factor contributing to the increase in optimal reserves in the mid-1990s and the post-crisis decline is the buildup and subsequent unwinding of foreign currency liabilities, particularly in Thailand.

Increasing the risk aversion parameter from 6 to 10 (the maximum value considered in existing studies on growth and business cycles) would lead the warranted level of reserves to increase by less than 3 percentage points of GDP, much less than is needed to explain the post-crisis buildup in emerging Asia’s reserves.

Extensions of the model might help explain a portion of the difference in reserve holdings between Asia and Latin America: this would be the case, for example, if one were to consider the opportunity cost based on the sum of the term and the individual-country sovereign spreads, because spreads are lower in Asia than in Latin America.

    Other Resources Citing This Publication