Chapter 16. International Reserves in Emerging Market Countries: Too Much of a Good Thing?

Christopher Crowe, Simon Johnson, Jonathan Ostry, and Jeronimo Zettelmeyer
Published Date:
August 2010
  • ShareShare
Show Summary Details

Olivier Jeanne1


With international reserves four times as large, in terms of their GDP, as in the early 1990s, emerging market countries seem more protected than ever against shocks to their current and capital accounts. Some have argued that this buildup in reserves might be warranted as insurance against the increased volatility of capital flows associated with financial globalization.2 Others view this development as the unintended consequence of large current account surpluses and suggest that the level of international reserves has become excessive in many of these countries (see, for example, Summers, 2006). Do emerging market countries hold too much international reserves, and are there better ways to use those funds?

Answering these questions requires a normative benchmark for the optimal level of reserves. I present in this chapter a simple welfare-based model of the optimal level of reserves to deal with the risk of capital account crises or of “sudden stops” in capital flows. On the basis of this model, I derive some formulas for the optimal level of reserves and compare them with conventional rules of thumb, such as the Greenspan-Guidotti rule of full coverage of short-term debt. I then calibrate the model for emerging market countries and compare its predictions with the actual data.

One lesson from this exercise is that the optimal level of reserves is subject to considerable uncertainty, because it is sensitive to certain parameters that are difficult to measure. The model nevertheless produces ranges of plausible estimates against which the data can be compared. I find that it is not difficult for the model to explain a reserves-to-GDP ratio on the order of 10 percent for the typical emerging market country (close to the long-run historical average), and that even higher ratios can be justified if one assumes that reserves have a significant role in crisis prevention. The levels of reserves observed in many countries in the recent period, in particular in Latin America, are within the range of the model’s predictions.

Ultimately, however, the insurance model fails to account for the recent pattern of reserves accumulation in emerging market countries. The reason is that most of the reserves accumulation has taken place in Asian emerging market countries, where the risk of a capital account crisis seems much too small to justify such levels of self-insurance. The insurance model can account for the reserves accumulation observed in the Asian emerging market countries only if one assumes that the expected cost of a capital account crisis is unrealistically large (more than 60 percent of GDP for one of the two major types of crisis examined).

The conclusion that most of the current buildup of reserves is not justified by precautionary reasons has some implications for reserves management. There is little reason for countries to invest these funds in the liquid but low-yielding foreign assets in which central banks tend to invest. Rather, reserves should be viewed as a component of domestic external wealth that is managed by the public sector on behalf of the domestic citizenry, taking full advantage of the portfolio diversification opportunities available abroad. Indeed, an increasing number of emerging market countries are transferring a fraction of their reserves to “sovereign wealth funds,” mandated to invest in a more diversified way and at a longer horizon than central banks normally do. This is a trend that might take on considerable importance looking forward.

The last part of the chapter discusses some policy challenges and opportunities implied by the buildup in emerging market countries’ “sovereign wealth.” I discuss, first, the impact of sovereign wealth diversification on global financial markets, and second, some ways in which this wealth could be used in collective international arrangements—to insure against future crises or to promote financial development.


The growth in the international reserves of emerging market countries is striking when compared with the contemporaneous trends in reserves in industrial countries (Figure 16.1).3 Whereas reserves in a group of industrial countries have remained stable below 5 percent of GDP, reserves in the emerging market countries have grown more than fourfold in terms of GDP since 1990. Much of this accumulation cc;more than half of the dollar amount “has taken place in Asia since the 1997–98 Asian crisis. China now has the largest stock of international reserves in the world, having overtaken Japan at the end of 2005, and it accounts for an important share of the buildup in emerging market reserves. However, China is not very different from the other Asian emerging market countries in terms of its ratio of reserves to GDP.

Figure 16.1.International reserves in emerging-market and industrial countries, 1980–2005.

Total reserves minus gold. Countries included in each group are listed in Table 16.10. The lower panel shows unweighted averages of reserves as a percentage of GDP in the countries in each group.

(Sources: IMF, International Financial Statistics; World Bank, World Development Indicators.)

This development is an important dimension of what Lawrence Summers calls the “capital flows paradox” in the current world financial system (see Summers, 2006), namely, that capital is flowing upstream from developing and emerging market countries toward the industrialized world and principally the United States. The reserves accumulated in my sample of emerging market countries between 2000 and 2005 are equal to a significant fraction (about 40 percent) of the U.S. current account deficit in the same period and may thus have contributed to keeping global interest rates low.

Table 16.1. provides some insights on whether the reserves buildup has tended to be financed by current account surpluses or through capital inflows. The first line of the table reports cumulative net capital inflows as a percent of the increase in reserves over 2000–05 for the sample of emerging market countries, with a breakdown for Asia and Latin America. About 40 percent of the reserves buildup has been financed by capital inflows on average. Whereas Asia has relied more than the average on net exports to accumulate reserves, Latin America has run current account deficits, so that its (relatively smaller) increase in reserves has had to be financed more than one for one by capital inflows.

Table 16.1Reserves Accumulation and the Financial Account in Emerging Market Countries, 2000–05 *
ItemAll Emerging-Market CountriesAsiaLatin America
Net capital inflows as percent of change in reserves40.636.6137.0
Composition of the Increase in Gross Foreign Assets (Percent)
Direct investment8.85.620.9
Portfolio investment8.78.713.0
Other investment22.311.736.0
Reserve assets60.273.930.0
Composition of the Increase in Gross Foreign Liabilities (Percent)
Direct investment67.963.3104.0
Portfolio investment20.928.26.8
Other investment11.28.5−10.8
Source: IMF, Balance of Payments Statistics.

Another way to look at reserves is in the broader context of the country’s external balance sheet. The bottom two panels of Table 16.1 show the composition of the increase in both external assets and external liabilities that were traded in the financial accounts of emerging market countries between 2000 and 2005. More than 60 percent of their foreign asset accumulation consisted of reserves (more than 70 percent in Asia). By contrast, foreign direct investment (FDI) accounted for almost 70 percent of the new liabilities accumulated by these countries.

That emerging market countries tend to have external assets that are more liquid than their external liabilities is confirmed by looking at stocks rather than flows. Figure 16.2 compares the external balance sheets of emerging market and industrial countries (taking the average over 2000–05), using the IMF data on international investment positions. The share of reserves in gross foreign assets is almost nine times as large in the emerging market countries as in the industrial countries, whereas the share of FDI in their liabilities is almost twice as large.

Figure 16.2Composition of stock and foreign assets and liabilities in emerging market and industrial countries, 2000–05 averages.

Source: IMF, Balance of Payments Statistics

The level of reserves in emerging market countries has thus increased since the early 1990s, but so has their trade and financial integration—and with it the associated risks. How much of the increase in reserves can be explained as selfinsurance in response to an increase in the hazards of globalization?

As numerous studies have pointed out, the recent accumulation of reserves by emerging market countries seems difficult to explain using the conventional rules of thumb for reserves adequacy. Figure 16.3 tracks three conventional reserves adequacy ratios in emerging market countries since 1980: the ratios of reserves to imports, to short-term external debt, and to broad money (M2).4 Although imports and M2 have increased over time in these countries, international reserves have increased by much more. All three reserves adequacy ratios have increased markedly and are now much higher than any of the conventional rules of thumb would prescribe. In 2005, reserves in emerging market countries were close to seven months of imports and five times the level of short-term debt. That reserves deviate even more from the Greenspan-Guidotti rule than from the three months- of-imports rule is surprising, since the latter was developed to better capture the risks stemming from the capital account after the crises of the 1990s.

Figure 16.3Reserve adequacy in emerging market countries, 1980–2005.

The conventional range of the reserves-M2 ratio is 5 to 20 percent.

(Sources: IMF, International Financial Statistics; World Bank, Global Development Finance.)

The reserves buildup is also difficult to explain using regression-based empirical models for precautionary reserves. A large empirical literature explains the cross-country and time variation in reserves by a few key variables: economic size of the country, current and capital account vulnerability, and exchange rate flexibility. Recent studies find that although such regressions do a good job of predicting reserve holdings over a long period, they significantly underpredict the reserves accumulation of emerging market countries after the Asian crisis, especially in Asia (see IMF, 2003; Aizenman and Marion, 2003; and Aizenman, Lee, and Rhee, 2007).

It could be, however, that such regressions fail to capture the impact that the severe capital account crises of the 1990s had on how these countries perceived the risks associated with their international financial integration. It has been argued that the Asian crisis marked a watershed, in that emerging market countries became painfully aware that even sound macroeconomic policies did not insulate them from contagion and sharp reversals in capital flows. The buildup in reserves could be a rational adaptation to this new, more volatile world.

The concept that came to epitomize the capital account instability of the 1990s is that of “sudden stop” in capital inflows. Figure 16.4 shows that although sudden stops were not a total novelty for emerging market countries as a whole, they were a relatively new phenomenon in Asia. For the five Asian countries most affected by the 1997–98 crisis, furthermore, the size of the shock to the capital account and the loss of reserves were unprecedented, in recent decades at least, as Figure 16.5 shows. It may not be a coincidence, from this point of view, that most of the recent buildup in international reserves has taken place in Asia.

Figure 16.4Sudden stops in emerging market countries, 1980–2000, using the SS2 definition of a sudden stop.

(See Table 16.11 in Appendix I for crisis definitions.)

Sources: Frankel and Cavallo (2004); IMF, International Financial Statistics; and World Bank, World Development Indicators.

Figure 16.5Yearly changes in reserves-to-GDP ratios in five Asian countries (Indonesia, Korea, Malaysia, the Philippines, and Thailand), 1980–2000. Data include crisis loans received from the IMF.

Source: IMF, International Financial Statistics; and World Bank, World Development Indicators.

In sum, the recent buildup in emerging market countries’ international reserves cannot be explained by conventional adequacy ratios or by simple linear regressions. But it may be that neither approach fully captures how the instability of the 1990s changed the perception of risks and the desire for insurance on the part of the countries most affected. For this reason, looking at the implications of a costbenefit analysis of the optimal level of reserves might be more informative than historical regressions. This is the approach that I take in the rest of the chapter.


I present in this section a simple framework for a cost-benefit analysis of the optimal level of reserves to deal with capital account crises. The model features a small, open economy that is subject to being hit by a capital account crisis. Reserves are useful both in terms of crisis prevention (reducing the probability of a crisis) and in terms of crisis mitigation (reducing the welfare cost of a crisis, once it has occurred). I start with a brief review of the literature on cost-benefit analyses of international reserves, before presenting the model.

16.3.1. Cost-Benefit Analyses of the Optimal Level of Reserves

The idea of a cost-benefit approach to the optimal level of reserves has inspired a long line of literature that goes back to a seminal contribution published by Robert Heller in 1966.5 In Heller’s analysis the optimal level of reserves was determined in the context of a trade-off between their opportunity cost and the risk of an external disequilibrium leading to a costly adjustment—a contraction in domestic absorption. Heller simply posited that the optimal level of reserves should minimize the sum of the expected cost of adjustment plus the opportunity cost of reserves.

One problem with traditional models of optimal reserves is that the objective function maximized by the authorities is only loosely related to domestic welfare. This leaves room for ambiguity in the definition and in the measurement of key variables of the model. First, it is not very clear how the cost of an external disequilibrium should be measured.6 Second, the lack of a rigorous welfare criterion also leads to some ambiguity in the definition of the opportunity cost of reserves, as I will show later.

I will therefore rely on a model of the optimal level of reserves that is welfarebased but preserves some of the simplicity of the earlier literature. This section concludes with a brief summary of the main features of my analytical framework. After reading this summary, those primarily interested in my predictions on the optimal level of reserves can skip the remainder of this section, which presents the model in more detail, and proceed directly to the discussion of the numerical findings.

The model features a small, open economy that is vulnerable to crisis, defined as a loss of access to external credit associated with a fall in output. The economy is populated by a representative consumer who holds a certain amount of foreign assets, or “sovereign wealth.”7 This wealth can be invested in liquid international reserves or an illiquid asset. Reserves yield benefits in terms of crisis prevention and crisis mitigation but entail an opportunity cost relative to the more profitable illiquid investment. The optimal level of reserves will depend on the following parameters of the model:

  • L and ΔY, the size of the capital flight and of output loss in a crisis, respectively, expressed in terms of potential output;

  • δ, the opportunity cost of accumulating reserves;

  • σ, the relative risk aversion of the domestic consumer; and

  • π, the probability of a crisis (which is endogenous to the level of reserves if there is crisis prevention).

16.3.2. Assumptions

The model assumes a small open economy and three periods of time t = 0, 1, 2. The last period (period 2) represents the long term. The intermediate period (period 1) is the time during which a crisis could occur. During the initial period (period 0) the country adjusts its reserves to the risk of a crisis in period 1. This simple time structure makes it possible to preserve the simplicity of Heller’s original approach but does not preclude a more dynamic interpretation of the model, as I will show shortly.8

At the end of period 0, a representative consumer in the small open economy structures his or her external assets and liabilities to deal with the risk of a crisis in period 1. To keep the problem simple, I assume that the consumer allocates wealth W0 between two assets: liquid bonds (or reserves, R0) and an illiquid asset I. This asset can be defined as a negative variable, in which case the consumer issues a long-term external liability D =–I. The welfare of the representative consumer is given by

where u(·) is an increasing and concave function of consumption, and W2 is the consumer’s net foreign wealth at the beginning of period 2. Foreign wealth can be traded between periods at interest rate r. The consumer thus desires a level of consumption C* in period 1 that satisfies the first-order condition,

The reserves are more liquid than the asset in the sense that they are the only form of wealth that can be sold in period 1. The illiquid asset cannot be sold in period 1 but brings a higher return in the long run (period 2). The difference between the return on the illiquid asset and the return on reserves is the opportunity cost of reserves, the price that the consumer must pay in order to keep wealth in liquid form.

The sequence of events and actions is as follows.

  • Period 0. The consumer allocates wealth between reserves and the illiquid asset,

  • Period 1. An external liability L comes due. The consumer repays L and consumes C1 under the budget constraint,

where Y1 is domestic output, L′ is new debt issued in period 1, R = (1 + r) R0 is the stock of reserves at the beginning of the period, and R is the stock of reserves at the end of the period.

  • Period 2. The consumer’s net foreign wealth is equal to output in period 2 plus the net return on net foreign assets,

where r is the interest rate between period 1 and period 2, and δ is the excess return on the illiquid asset (or “illiquidity premium”).

In period 1, the economy can be in either of two states that differ by the level of output and the consumer’s access to external credit:

  • The no-crisis state: output is at its potential, Y1 = Y, and the representative consumer has complete access to external credit (there is no restriction on L′), or

  • The crisis state: output is below potential, Y1 = Y –ΔY, and the representative consumer has no access to external credit in period 1 (L′ is equal to zero).

The crisis state thus consists of both an output drop and a sudden stop in capital flows. As equation (4) shows, the negative impact of the fall in output and capital inflows on domestic consumption can be mitigated by running down reserves (R′= 0).9 I shall assume, as a matter of normalization, that Y = 1, so that the output cost of a crisis ΔY and the size of the sudden stop L are expressed in terms of potential output. I also assume that the desired level of consumption is equal to potential output (C* = Y) so that there is no predictable trade deficit in period 1.

The ex ante probability of a crisis is denoted by π. To capture the idea that reserves might provide a benefit in terms of prevention, I assume that the probability of crisis is a decreasing function of the ratio of reserves to short-term debt,

where F(·) is an increasing function, and v is a measure of vulnerability to a crisis, summarizing the fundamentals other than reserves. (I will at times refer to the coefficient a as the prevention benefit parameter.) In calibrating the model I will use a probit specification, implying that F(·) is the cumulative distribution of a normal function.

The interesting question is how the optimal level of reserves R depends on the relevant determinants: the country’s vulnerability to a crisis, measured by v; the magnitude of the crisis, measured by the size of the shock to the capital account L and of the output loss ΔY, and the opportunity cost of reserves, δ.

16.3.3. The Optimal Level of Reserves

As shown in Appendix II, the optimal level of reserves minimizes a loss function that equals the opportunity cost of reserves plus the expected welfare cost of a crisis:

where f(R), the welfare cost of a crisis, is increasing with the size of the crisis (L and ΔY) and decreasing with the level of reserves R.

Equation (7) is reminiscent of the loss function postulated in some earlier costbenefit analyses of optimal reserves (see, for example, Heller, 1966; Ben-Bassat and Gottlieb, 1992; and Garcia and Soto, 2004). It captures in a simple way the tradeoff between the opportunity cost of reserves δR and their benefits in terms of crisis prevention π(R) and crisis mitigation f(R). It can be interpreted, in a more dynamic context, as the average intertemporal loss of a country maintaining a constant level of reserves R. The consumer bears the opportunity cost δR in every period but pays the welfare cost of a crisis with a frequency π(R). Equation (7) thus sums up the average cost of crises and the average cost of insurance against those crises. As shown in Figure 16.6, for low levels of reserves the gains from increasing reserves, in terms of crisis prevention and crisis mitigation, dominate the opportunity cost, whereas the opposite holds for high levels of reserves.

Figure 16.6Total loss and the optimal level of reserves.

Closed-form expressions for the optimal level of reserves can be obtained if one assumes that reserves have no benefits in terms of prevention—that is, if π is exogenous. The first-order condition for the minimization of the loss function in equation (7) can then be written as

where R¯=L+ΔY is the “full insurance” level of reserves (i.e., the minimum level of reserves sufficient to maintain consumption at the desired level in a crisis). This condition implies that the optimal level of reserves is increasing with the probability of a crisis and decreasing with the opportunity cost of holding reserves—as one would expect.

If the consumer has constant relative risk aversion σ, then the optimal level of reserves is given by the formula

In words, the optimal level of reserves is equal to short-term external debt plus the output cost of a crisis minus a term reflecting the opportunity cost of holding reserves.

Note that in this model the optimal level of reserves could be higher than under the Greenspan-Guidotti rule (R = L), because reserves smooth the impact on consumption of the fall in output, and not only the impact of the debt rollover crisis. The optimal level of reserves could also be lower than short-term debt because of the opportunity cost of holding reserves, which the Greenspan-Guidotti rule ignores.

The optimal level of reserves does not have a closed-form expression in the general case where the probability of a crisis is endogenous to the level of reserves. Then the optimal level of reserves minimizes

Taking into account the benefits of crisis prevention leads to an increase in the optimal level of reserves, other things equal. In fact—and this is an important difference from the case where the probability of a crisis is exogenous—the optimal level of reserves may now exceed the “full insurance” level R¯=L+ΔY Crisis prevention could make it optimal for a country to hold more reserves than it is willing to spend in a crisis.


I now turn to the calibration of the model, starting with the benefits of reserves. In my model reserves yield benefits in terms of crisis prevention π(R) and crisis mitigation f(R). To calibrate the model I thus try to identify each type of benefit in the data.10

16.4.1. Crisis Prevention

The international financial crises of the 1990s triggered a search for reserves adequacy ratios that would capture the vulnerability of emerging market countries’ balance sheets and capital accounts in a world with highly mobile capital flows. The staff of the IMF concluded that the ratio of reserves to short-term external debt was the “single most important indicator of reserves adequacy in countries with significant but uncertain access to capital markets” (IMF, 2000, p. 6), although this ratio should be taken as only a starting point for an analysis that should also look at other reserves adequacy ratios in light of each country’s specific conditions.11

This view was supported by a vast body of empirical research showing that the ratio of reserves to short-term external debt tended to perform well as an early indicator of currency crises. By contrast, the (relatively smaller) empirical literature on sudden stops in capital flows has been less conclusive, generally failing to detect a significant preventive role for reserves.12

In order to take a broad view of the preventive role of reserves with respect to both currency crises and sudden stops, I ran a number of univariate probit regressions using various crisis definitions and reserves adequacy ratios. The regression results are based on four different definitions of a currency crisis (denoted by CC1 to CC4) and four different definitions of a sudden stop (denoted by SS1 to SS4). Table 16.11 gives these definitions, and Table 16.12 lists the years when each type of crisis occurred in each country. For the first of the currency crisis definitions (CC1), I use Frankel and Rose’s criterion of a nominal depreciation of the currency of at least 25 percent relative to the previous year that is also at least a 10-percentage-point increase in the rate of depreciation (Frankel and Rose, 1996). The other three definitions (CC2 to CC4) are based on a crisis pressure index that adds the percentage nominal depreciation of the currency to the percentage loss in foreign reserves (Frankel and Wei, 2005).

I first identify sudden stops as those years in which net capital inflows fell by more than 5 percent of GDP (SS1). This simple criterion has been criticized for various reasons, in particular because it captures some episodes in which capital net inflows slowed down but remained positive (such as Malaysia in 1994, following the imposition of controls on capital inflows). For robustness, I also consider three sudden-stop measures that are more stringent (SS2 to SS4).13

Table 16.2 summarizes the results of 160 univariate regressions using various reserves adequacy ratios, crisis definitions, and probit specifications. For each crisis definition and reserves adequacy ratio, I ran four probit regressions of the crisis dummy variable on the lagged reserves ratio and a constant: without fixed effects, with country fixed effects, with time fixed effects, and with both country and time fixed effects. Since currency crises and sudden stops each have four different definitions, each cell in the table is based on 16 probit regressions. The table reports the number of regressions in which the coefficient on reserves was both negative and significant at the 10 percent level or better.

Table 16.2Regressions of Crisis Variables on Alternative Measures of Reserves, 1980–2000
Dependent Variable (Type of Crisis)
Measure of Reserves AdequacyCurrency CrisisSudden Stop
Ratio of reserves to imports91
Ratio of reserves to short-term debt (World Bank measure)160
Ratio of reserves to short-term debt (BIS measure)41
Ratio of reserves to M204
Ratio of reserves to GDP121
Source: Author’s calculations.

Several facts stand out. First, the denominator of the reserves adequacy ratio that “works” best to predict a currency crisis is short-term debt.14 The benefit of increasing reserves in terms of crisis prevention, furthermore, is economically significant. To illustrate, Figure 16.7 shows how the probability of a crisis varies with the Greenspan-Guidotti ratio R / L for values of a in the range of estimation of the probit. As the figure shows, if a = 0.3, doubling the ratio of reserves to short-term debt from 1 to 2 reduces the probability of a crisis by almost 4 percent. However, there are diminishing returns to further increasing reserves: increasing R / L from 5 to 6 reduces the probability of crisis by less than 1 percent.

Figure 16.7Reserves and crisis prevention.

The variable a is the prevention benefit parameter, the coefficient on the Greenspan- Guidotti ratio in the crisis probability (equation 6 in the text). It is assumed that the probability of crisis is 10 percent for R = L.

Source: Author’s calculations.

Second, the reserves adequacy ratios do not perform as well at predicting sudden stops as they do at predicting currency crises. The ratio that works best is that based on M2, but even this ratio is significant in only one-fourth of the regressions. This result also seems consistent with the empirical literature, which is ambiguous with regard to the benefits of reserves in preventing sudden stops rather than currency crises.

One important caveat is necessary before one accepts the conclusion that reserves help to prevent crises. The existing empirical studies do not really distinguish between two possibilities: whether high levels of reserves allow countries to prevent crises, or whether spending the reserves merely postpones the crises.15 This identification problem does not affect the rationale for using reserves as an early warning indicator of crisis, but it may lead to an exaggeration of the benefits of reserves in terms of crisis prevention. In many cases, countries might actually have hastened the crisis, and not reduced its probability, by trying to maintain a high level of reserves in the face of a loss of confidence in domestic policies.

16.4.2. Crisis Mitigation

There are two ways in which reserves can help to mitigate the impact of a balance of payments crisis on domestic welfare. First, the reserves can be used to mitigate the fall in domestic output. Second, the reserves can be used to buffer the impact of the balance of payments shock on domestic absorption.

The authorities can reduce the output cost of a crisis by using international reserves through various channels. Foreign exchange interventions can mitigate the depreciation of the domestic currency, and thus the disruption induced by currency mismatches in balance sheets. Reserves help the monetary authorities in providing liquidity to the domestic financial markets, the banking sector, and even exporters; this is especially valuable if there is significant dollarization of bank deposits and other domestic liabilities.16

As for the second benefit, I present a simple accounting exercise that shows the extent to which international reserves help smooth domestic absorption in the face of balance of payments shocks. In a small, open economy, domestic absorption can be written as the sum of domestic output, capital inflows, and reserves decumulation (net income from abroad is omitted because it typically varies little in a crisis):17

There is an exact correspondence between this decomposition and equation (4) of the model, which can be written

Thus information about the behavior of the components of equation (11) can help in calibrating the model. I now look at how the components of equation (11) behave in observed sudden stop episodes. Sudden stops will be identified, in my sample of emerging market countries, as a year in which net capital inflows fall by more than 5 percent of GDP (definition SS1).

Figure 16.8 shows the average behavior of domestic absorption and the contribution of the various components on the right-hand side of equation (11) in a five-year event window centered around a sudden stop. Real output is normalized to 100 in the year before the sudden stop. All the variables are converted from current dollars into constant local currency units so that the changes in output and domestic absorption can be tracked in volume terms.18

Figure 16.8Domestic absorption and output, net capital inflows, and reserves in sudden stops

(percent of GDP in year of sudden stop). A sudden stop is defined as a fall in the financial account of more than 5 percent of GDP (SS1). Events that occurred before 1980 or within the five-year window of a previous sudden stop are excluded from the calculation. The solid line is the sample mean; the dotted lines indicate the mean plus and minus one standard deviation.

(Source: Author’s calculations using data from the IMF, International Financial Statistics and the World Bank, World Development Indicators.)

A large fall in net capital inflows is observed in the year of the sudden stop, amounting to almost 10 percent of the previous year’s output on average. This is not surprising, since a large fall in those inflows is the criterion used to identify sudden stops. More interestingly, most of the negative impact of the capital account reversal on domestic absorption is offset by a fall in reserves accumulation. Thus domestic absorption falls by only 3 percent of GDP on average in the year of the sudden stop—much less than the capital inflows. Figure 16.8 also shows that the contribution of output is relatively small: real growth merely falls to zero at the time of the sudden stop.

This evidence is consistent with the view that emerging market countries accumulate reserves in good times so as to be able to decumulate them, thereby smoothing domestic absorption, in response to sudden stops. This smoothing effect is potentially large. To illustrate, if reserves accumulation were equal to zero in the year of the sudden stop, domestic absorption would fall by 9 percent of output on average instead of 3 percent, other things equal. This counterfactual experiment should be interpreted with caution, because the magnitude of capital flight could in part be endogenous to the fall in reserves. It does suggest, however, that foreign exchange reserves may well make a sizable contribution to the smoothing of domestic absorption in response to sudden stops.

The case of Uruguay in 2002 provides a striking illustration of the role of reserves in a very severe sudden stop episode. Following the Argentine crisis, net capital inflows to Uruguay fell by 26 percentage points of precrisis GDP. The Uruguayan government used a large amount of foreign exchange reserves (a significant part of which was made available in the context of an IMF arrangement) to cover the withdrawal of dollar-denominated deposits from the domestic banking system. As a result, the decline in domestic absorption, although quite substantial (14 percent of GDP), was much smaller than the shock to the capital account.


The cost of holding reserves is measured in the literature“as in the model” as the difference between the return on the reserves and the return on more profitable alternative investment opportunities.19 One term of the comparison, the return on the reserves, is generally proxied as the return on short-term foreign currency assets. The appropriate definition of alternative investment opportunities, on the other hand, raises several thorny questions.

One approach is to consider higher-yielding investment opportunities in the domestic business sector or in the building of public infrastructure. However, the marginal product of capital is difficult to measure in a way that is comparable across a large number of countries. Caselli and Feyrer’s recent estimates can be used to compute an average annual real return to capital of 7.8 percent in 17 emerging market countries in my sample.20 This, together with an estimate for the short-term real interest rate of 2 percent a year–roughly the average U.S. real short-term rate over 1980–2005—would lead to an opportunity cost of around 6 percent a year.

Given the difficulties involved in measuring the returns to physical investment, most measures in the literature assume that the alternative to holding international reserves is to invest in other financial assets or to repay existing financial liabilities. One approach defines the opportunity cost of reserves as the quasi-fiscal cost of sterilization by the central bank, that is, the difference between the return on the central bank’s domestic currency assets and the return on international reserves (see, for example, Frenkel and Jovanovic, 1981; Flood and Marion, 2002; and Mohanty and Turner, 2006). This differential is generally positive, but in countries where domestic interest rates are very low—such as China recently—this approach leads to a negative opportunity cost of reserves.

There are two serious issues with measuring the opportunity cost of reserves in this way. First, this measure is not adjusted for the expected appreciation or depreciation of the domestic currency. For example, the fiscal cost of reserves could be found to be negative because the domestic currency is expected to appreciate relative to the dollar—and interest rate parity applies—but this measure fails to take into account the expected valuation loss on the reserves. Second, the central bank’s profit is not a measure of domestic welfare. Selling high-yielding domestic bonds for reserves may reduce the central bank’s flow of profit but increases the income of the domestic investors who purchase the bonds. The opportunity cost of reserves should therefore be measured by looking at the budget of the country as a whole rather than that of the central bank. This might be a reason to measure the opportunity cost of reserves by reference to external—rather than domestic—assets and liabilities.

Reserves can be accumulated by issuing—or can be used to repay—external debt. Given this observation, some authors measure the opportunity cost as the spread between the interest rate on external debt and the return on reserves.21 By this measure the opportunity cost of reserves was 8.4 percent a year in emerging market countries on average in 2000–05, but this figure masks important disparities between Asia, where the spreads were low, and Latin America, where they were much higher (Figure 16.9).

Figure 16.9Alternative measures of the opportunity cost of reserves, 2000–05. The term premium is the difference between the return on long-term dollar assets and liquid dollar assets. The spread is the difference between the interest rate on external debt and the return on reserves.

Sources: Author’s calculations using data from Bloomberg; and IMF, International Financial Statistics.

One might argue that these spreads overstate the true opportunity cost of holding reserves, because they include the default risk premium on foreign debt. As shown more formally in Appendix II, the welfare-based approach suggests that the default risk premium should not be included, because it is, on average, a fair reflection of the probability of less than full repayment. Pushed to its logical extreme, this approach suggests that the true opportunity cost of reserves is the U.S. term premium, that is, the opportunity cost of financing a stock of liquid dollar assets with default-free long-term dollar debt. This would lead to a much lower measure of the opportunity cost of reserves of at most 2 percent.22

Table 16.3 presents some measures of the average opportunity cost of reserves in terms of domestic GDP in my sample of emerging market countries over the period 2000–05. The measures are based on a uniform opportunity cost of 6 percent as well as the term premium, with and without a spread. With an opportunity cost of 6 percent a year, the average cost of reserves amounts to 1 percent of GDP.23 The estimated cost of reserves is significantly lower if one considers the term premium, but larger if one includes the emerging market spread. On average, the total cost of holding reserves was substantially lower in Latin America than in Asia if one uses the same opportunity cost per unit of reserves for both regions, but it was relatively similar in the two regions when one uses the term premium plus the spread. This is explained by the fact that, whereas on average the reserves-to-GDP ratio is more than twice as high for Asian countries as for Latin American countries, the sovereign spread is substantially higher in Latin America than in Asia.

Table 16.3Opportunity Cost of Reserves, 2000–05*
AssumptionAll Emerging MarketsAsiaLatin America
Opportunity cost of reserves is 6 percent a year0.931.290.65
Opportunity cost of reserves is the term premium (2 percent a year)0.320.450.22
Opportunity cost of reserves is the term premium plus the spread on external debt1.060.991.00
Sources: Author’s calculations using data from Bloomberg; World Bank, World Development Indicators; and IMF, International Financial Statistics.


The model presented above is used here to predict the optimal level of reserves in emerging market countries. This is done in two steps. First, I calibrate the model by reference to an average emerging market country, as a way of getting a broad sense of the quantitative implications of the model and their sensitivity to the parameters chosen. Second, I calibrate the model by reference to country-specific data, to study how far the model can go in explaining the reserves buildup in emerging market countries.

16.6.1. Benchmark Calibration and Sensitivity Analysis

The benchmark calibration is based on the parameter values given in Table 16.4. I assume that reserves provide no benefits in terms of prevention, so that the formula in equation (9) applies. The probability of crisis was set to the unconditional frequency of sudden stops (SS1) in my sample of emerging market countries, which is close to 10 percent a year. The value for the opportunity cost of reserves, δ= 3 percent, is close to the middle of the range of estimates discussed earlier. The chosen values for risk aversion and its range of variation are standard in the growth and real business cycle literature.

Table 16.4Benchmark Calibration Parameters
Size of sudden stopL = 0.10[0, 0.3]
Probability of sudden stopπ = 0.10[0, 0.25]
Output lossΔY = 0.10[0, 0.2]
Opportunity costδ = 0.03[0.01, 0.06]
Risk aversionσ = 2[1, 10]
Prevention benefit parametera = 0[0, 0.3]

Capital flight (L) and the output loss (ΔY) are both set to 10 percent of GDP. These figures are in line with the behavior of capital flows and of output during the sudden stops documented in Figure 16.8.24 The output cost figure was obtained by cumulating the average output gap in the year of a sudden stop and the following year, under the assumption that output would have grown at the same rate as before the crisis in the absence of a sudden stop. An output loss of 10 percent of GDP is in the ballpark of the estimates reported in the literature on currency crises and sudden stops.25

The benchmark calibration implies an optimal level of reserves of 7.7 percent of GDP, or 77 percent of short-term external debt. This is close to the ratio of reserves to GDP observed in the data on average over 1980–2000, but significantly below the level observed in the most recent period, especially in Asia. It would be interesting to know what changes in the parameters are required to increase the optimal level of reserves to something approaching the recently observed level.

Figure 16.10 shows the sensitivity of the optimal level of reserves to the probability of crisis, the opportunity cost of reserves, the degree of risk aversion, and the elasticity of the crisis probability to the level of reserves. In each case the level of reserves computed using the sudden stop model is contrasted with that implied by the Greenspan–Guidotti rule. Several interesting results emerge.

Figure 16.10Sensitivity analysis of the optimal level of reserves

(percent of GDP). Dashed line indicates the optimal level of reserves using the Greenspan-Guidotti rule, assuming that short-term debt is 10 percent of GDP.

(Source: Author’s calculations.)

The optimal level of reserves is quite sensitive to the probability of crisis, the opportunity cost of reserves, and the risk aversion parameter. This offers an interesting contrast with the Greenspan–Guidotti rule, which does not depend at all on these parameters. The optimal level of reserves is zero if the probability of crisis falls below 5 percent, but it almost doubles, from 7.7 percent to 13.3 percent of GDP, if the probability of crisis increases from 10 percent to 20 percent. Risk aversion also has a first-order impact on the optimal level of reserves. A shift in the risk aversion parameter from 2 to 8 increases the optimal level of reserves from 7.7 percent to 16.8 percent of GDP.

Figure 16.10 Sensitivity analysis of the optimal level of reserves (percent of GDP). Dashed line indicates the optimal level of reserves using the Greenspan-Guidotti rule, assuming that short-term debt is 10 percent of GDP. (Source: Author’s calculations.)

Figure 16.10 also shows that the optimal level of reserves can be significantly larger if one assumes that reserves have benefits in terms of crisis prevention (parameter a). If, in line with my univariate probit results for currency crises, a is set between 0.2 and 0.3, then the optimal level of reserves can reach 23 percent of GDP, about three times the optimal level if reserves have no effectiveness at crisis prevention.

To summarize, there are two ways in which the model can potentially explain a level of reserves of the order of magnitude currently observed in Asia. The first is to assume very large numbers for capital flight or for the output cost of a crisis. To illustrate, if the size of the sudden stop or the output cost amounted to 40 percent of GDP, instead of 10 percent in the benchmark calibration, the model would predict an optimal level of reserves in excess of 35 percent of GDP. Such an assumption, however, seems out of line with the historical record on currency crises and sudden stops. The second and perhaps more plausible way in which the model can predict a higher level of reserves is if reserves offer substantial benefits in terms of crisis prevention.

16.6.2. Country Estimates

I now bring the model closer to the data by estimating the optimal level of reserves for each emerging market country in my sample in 2000. For each country I estimate the level of reserves that minimizes the loss function in equation (10), that is, the sum of the opportunity cost of reserves and of the expected welfare cost of a crisis,

where i is the country index. This loss function is calibrated based on a probit estimation of the crisis probability for each country. The model indicates excess or insufficient reserves, depending on how the optimal level of reserves, Ri*, compares with the observed level, Ri.

The first step is to estimate the probability of a crisis for each country. This is done by running a probit regression of the probability of crisis on the countries’ economic fundamentals in my sample of emerging market countries over 1980– 2000. The preferred specifications are reported in the top panel of Table 16.5 for sudden stops (defined as SS1) and in the bottom panel for currency crises (CC1). The explanatory variables have been selected using a general-to-specific approach, starting from a set of 18 potential regressors, which are listed in Table 16.13 in Appendix I. All explanatory variables are lagged at least one year and are thus predetermined with respect to the crisis. The results are robust to the inclusion of time and country fixed effects.

Table 16.5Probit Regressions of the Probability of Crisis on Macroeconomic Fundamentals, 1980–2000*
Regression Specification
Type of Crisis and Independent VariableNoneCountry Effects OnlyYear Effects OnlyCountry and Year Effects
Crisis is Sudden Stop SS1
Real exchange rate deviation from Hodrick-Prescott trend−1.240***−1.295**−1.102**−1.192**
GDP growth−2.047***−2.028**−2.511***−2.856***
Ratio of foreign liabilities to money††0.025***0.031**0.028***0.037***
Ratio of current account to GDP††−0.045**−0.053***−0.044**−0.056***
Ratio of total public debt to GDP††0.544**0.3330.578**0.324
Number of observations511394511394
Crisis is Currency Crisis CC1
Ratio of reserves to short-term debt††−0.162**−0.261**−0.130**−0.201*
Real exchange rate deviation from Hodrick-Prescott trend−1.441***−1.332***−1.598***−1.547***
Consumer price inflation††0.331**0.1340.392***0.161
No. of observations560483560483
Source: Author’s regressions.

I find that the main explanatory variable is the real exchange rate (or, more precisely, its deviation from a trend), which appears with the expected sign in both probit regressions. Consistent with the univariate evidence presented earlier, the ratio of reserves to short-term debt is significant for currency crises but not for sudden stops. The GDP growth rate, the ratio of foreign liabilities to money (a measure of dollarization in the banking sector), the current account, and total public debt are also significant in the regressions for sudden stops. Finally, the probit estimation for currency crises finds a role for inflation.

Figure 16.11 tracks the estimated probability of crisis over time in my sample of emerging market countries (the averages are GDP-weighted and based on the regressions without fixed effects). The probability of crisis is significantly lower in Asia than in the other emerging market countries, especially at the end of the 1990s because of the weak real exchange rates, large current account surpluses, and strong economic growth that prevailed in that region. To illustrate, the probability of a sudden stop is estimated at 2.7 percent in China in 2000, and that of a currency crisis is less than 0.2 percent.

Figure 16.11Probabilities of currency crises and sudden stops, 1980–2006. Results are GDPweighted country averages (excludes Russia and Ukraine).

(Source: Author’s calculations.)

In the second step, I compute the optimal level of reserves Ri* for each country in 2000. Parameter vi, which captures the country’s intrinsic vulnerability to a crisis, is calibrated using the probit regression for sudden stops reported in Table 16.5 Capital flight, the output cost of the crisis, and the values for the opportunity cost of reserves and for the risk aversion parameter remain the same as in the benchmark calibration (see Table 16.4).

The results of this exercise are reported in Table 16.6. At $234 billion, the total predicted level of reserves for all countries in the sample is significantly below the actual level observed in 2000 (just over $650 billion). However, the discrepancy comes mainly from the Asian countries, where the predicted level of reserves is extremely low. The estimated optimal level of reserves is zero in several important Asian countries (China, Korea, and Malaysia), because the probability of a sudden stop was below the 5 percent threshold (see Figure 16.10.). By contrast, the model works well for Latin America, where the observed level of reserves is actually slightly below the model prediction.

Table 16.6Actual and Model-Predicted Levels of Reserves
Country GroupActual, 2000 (billions of dollars)Predicted Benchmark* (billions of dollars)Implied Risk Aversion (σ)Implied Output Cost of a Crisis (ΔY), percent of GDP
All emerging market countries6512345.220.8
Latin America1451531.99.6
Source: Author’s calculations.

The last two columns of Table 16.6 give the “implied” values for the risk aversion parameter σ and the expected output loss in a crisis ΔY, that is, the values that one must assign to these parameters for the model to explain the observed level of reserves. In Latin America the implied values are very close to those in the benchmark calibration (reflecting the fact that the model fits the observations well in that region). By contrast, in Asia the implied values are implausibly high—almost 12 percent of GDP for risk aversion and more than 30 percent of GDP for the output cost of a crisis.

The results in Table 16.6 assume that reserves have no benefits in terms of crisis prevention. As mentioned before, the optimal level of reserves may be significantly higher if reserves have preventive benefits. Might this explain the reserves buildup in Asia? I look into this question by estimating the benefits of the reserves accumulation between 2000 and 2005 in terms of crisis prevention. For simplicity, I assume that the welfare cost of a crisis is equal to the output cost. Then increasing the level of reserves from R to Rσis optimal if

that is, if the decrease in the expected output cost of a crisis exceeds the opportunity cost of increasing reserves. To calibrate this condition, I compute for each country in my sample the decrease in the crisis probability induced by the reserves accumulation observed between 2000 and 2005, Δπ = πi2000 – πi2005. The probabilities are estimated using the probit regression for currency crises reported in the bottom panel of Table 16.5 The benefits and costs of the observed reserve accumulation are computed under the assumption that a crisis costs 10 percent of potential output and that the opportunity cost of reserves is 3 percent.

Table 16.7 reports the results of this exercise for emerging market countries as a whole as well as for Asia and Latin America separately (country averages weighted by GDP). It appears that, on average, the cost of reserves accumulation exceeded the benefits in terms of crisis prevention by a factor of about 3. But again the average masks an important difference between Asia, where the cost was more than five times larger than the benefit, and Latin America, where the benefit of reserves accumulation in terms of crisis prevention actually exceeded the cost.

Table 16.7Benefit and Cost of Reserves Accumulation, 2000–05*
ItemAll Emerging MarketsAsiaLatin America
Change in ratio of reserves to GDP (percentage points)8.713.82.8
Reduction in crisis probability (percentage points)
Benefit of reserves accumulation (percent of 2000 GDP)
Cost of reserves accumulation (percent of 2000 GDP)**0.631.040.14
Implied output cost of crisis (percent of GDP)††28.562.75.5
Source: Author’s calculations using IFS and WDI data.

The reason for this difference is that the probability of a currency crisis was much lower in Asia than in Latin America in 2000 (see Figure 16.11), implying that the marginal returns to reserves accumulation in terms of crisis prevention were much higher in Latin America than in Asia. To illustrate, in 2000 Mexico could have reduced its estimated crisis probability from 9.6 percent to 5.6 percent by doubling its reserves. By contrast, in China the estimated probability of crisis was 0.2 percent in 2000 and so could not have been reduced much further. It is nevertheless in emerging market Asia that most of the recent reserves accumulation has taken place.

Finally, the last line of Table 16.7 reports the “implied” output loss for each country, that is, the minimum output cost that one must assume for the observed accumulation of reserves between 2000 and 2005 to be worth the cost. To rationalize the reserves buildup in Asian emerging market countries, one needs to assume that the output cost of a crisis amounts to more than 60 percent of GDP; the implied output cost is one-tenth that size in Latin America. implied output cost is one-tenth that size in Latin America.

The conclusion is that the model cannot reasonably account for the increase in reserves in Asian emerging market countries as self-insurance against capital account crises. It can only do so by assuming that a capital account crisis costs more than 60 percent of one year’s output, which is out of line with the historical experience.26


To summarize, one justification for emerging market countries holding liquid international reserves is as a means of dealing with capital flow volatility and the risk of capital account crises, but the evidence suggests that most countries (especially in Asia) hold more international reserves than can be justified by this objective. This raises several questions. Why have Asian emerging market countries accumulated such large reserves? How should those reserves be managed? And looking forward, what are the implications of this buildup in emerging market countries’ foreign assets for the international financial system?

16.7.1. Trade Surpluses and Sovereign Wealth

Having rejected the view that the recent reserves accumulation can be justified on a precautionary basis, one has to consider as the main alternative explanation that these reserves are the unintended consequence of large current account surpluses.27 The “mercantilist” variant of this view holds that the central banks of these countries are accumulating reserves in order to resist the appreciation of the domestic currency.28 For this effort not to be defeated by domestic inflation, it must be augmented by policies that repress domestic demand—for example, capital controls or domestic financial repression.

Table 16.8. shows, for the same sample of emerging market countries, the cross-country correlations between the increase in the reserves-to-GDP ratio between 2000 and 2005 and some key macroeconomic variables. It appears that reserves accumulation is strongly correlated with the current account surplus and not correlated at all with the change in gross external liabilities. This suggests that, to a first approximation, the accumulation of reserves reflects net export flows rather than balance sheet operations.

Table 16.8Correlation Between the Change in Reserves-to-GDP Ratio and Selected Macroeconomic Variables, 2000–05
Change in Reserves to GDPAverage Current Account to GDPChange in Gross External Liabilities to GDPCapital Account Restrictions IndexAverage Real GDP Growth Rate
Change in Reserves to GDP1
Average Current Account to GDP0.585***1
Change in Gross External Liabilities to GDP0.0720.0141
Capital Account Restrictions Index0.337*0.184−0.2811
Average Real GDP Growth Rate0.460***0.2230.422**0.2881
Sources: IMF, Balance of Payments Statistics; World Bank, World Development Indicators; and Chinn and Ito, 2005.Note: The table presents the simple cross-country correlations for the emerging market countries reported in Table 16.10. The change in reserves is the difference in the reserves-to-GDP ratio between 2000 and 2005. The average of the current-account-to-GDP ratio is over the period 2001–05. The change in gross external liabilities is the difference in the ratio of gross external liabilities to GDP between 2000 and 2005. The capital account restrictions index is defined as one minus the average Chinn-Ito measure of capital account openness over the period 2000–04 (Chinn and Ito, 2005). The average real growth rate is over the period 2000–05. Asterisks denote statistical significance at the ***1 percent, **5 percent, and *10 percent levels.

The change in the reserves-to-GDP ratio is also positively correlated with capital account restrictions and with the real GDP growth rate.29 The correlation with capital account restrictions is the opposite of what one would expect based on the precautionary view of reserves accumulation, which predicts that countries with a more open capital account should hold more precautionary reserves because they are more vulnerable to the volatility of capital flows. The positive correlation with the growth rate is also puzzling if one thinks that high-growth developing countries should be importing foreign capital to finance their development.30

One could develop a cost-benefit welfare analysis of a mercantilist development strategy in the same way as I have done for the precautionary view, but the trade-offs involved would be very different. On the cost side, one would have to count the various distortions that are necessary to repress domestic demand, as well as the valuation loss on the foreign assets accumulated by the authorities when the inevitable real appreciation eventually takes place. The benefit side would include the gains in terms of productivity and growth from stimulating the export sector.

It is important to understand that what such a cost-benefit analysis would endogenize is not the level of reserves R, but rather the level of total publicly held foreign assets, which was denoted by W0 and taken as exogenous in my model of reserves. Endogenizing W0 would not affect my conclusion that most emerging market countries in Asia have excess reserves from the point of view of crisis insurance. Those excess reserves are costly, first in terms of forgone returns and portfolio diversification, and second because they generate difficulties for domestic monetary control that can be mitigated only by introducing or maintaining costly distortions in the domestic banking and financial system.

The governments of emerging market countries have started to mitigate these costs by transferring a fraction of foreign exchange reserves from the central bank to “sovereign wealth funds.”31 These funds are mandated to invest in a more diversified portfolio and at a longer horizon than central banks—not unlike the natural resource-based stabilization funds set up by a number of commodity exporters. For example, since July 2005 a fraction of Korea’s reserves have been managed by an independent entity, the Korean Investment Corporation, with the aim of seeking higher yields. China recently established the State Foreign Exchange Investment Corporation to manage reserves outside of the central bank.

According to some estimates, the holdings of sovereign wealth funds already amount to more than $2 trillion, mostly consisting of funds derived from oil and gas exports, but their size could increase to $12 trillion by 2015, surpassing official reserves within five years (see Jen, 2007). If those estimates are correct, sovereign wealth funds are set to become a major force in the international financial system.

16.7.2. Portfolio Diversification

Although, as just described, central banks in emerging market countries have recently been diversifying their allocation of reserves, this trend has been slow, and central banks continue to allocate their portfolios in a significantly different manner than private investors do.32 To illustrate, Figure 16.12 compares the allocation of U.S. assets held by the foreign official sector with that of foreign private investors. The foreign official sector invests much more in U.S. government debt and much less in equity or corporate debt than do private investors. Clearly there remains significant scope for diversification, a trend that should be facilitated by the transfer of emerging market countries’ reserves to sovereign wealth funds.

Figure 16.12Composition of foreign official and nonofficial holdings of U.S. assets, 2005.

Source: U.S. Treasury, Treasury International Capital database.

Some have expressed concern that the diversification of emerging market countries’ reserves could lead to disruptions in exchange rates and the relative prices of financial assets. To shed light on this question, consider, for the sake of argument, the following experiment. The total stock of foreign exchange reserves in my sample of emerging market countries amounted to approximately $2 trillion dollars in 2005. Assume that $1.2 trillion of this (60 percent of the total) was invested in dollar assets, of which $900 billion was invested in the asset classes represented in Figure 16.11.33 Assume further that the emerging market countries in my sample reinvest half of the assets currently invested in the official sector’s portfolio shown in Figure 16.11 ($450 billion) in the global financial portfolio. What would be the impact on the net supply of financial assets for the rest of the global investor community?

Table 16.9 details the current structure of the global portfolio of financial assets. The table was constructed by aggregating World Bank cross-country data on stock and bond market capitalizations in the industrial countries. The table also shows, for each asset class, the net demand from emerging market central banks that would be induced by the assumed portfolio reallocation, as a percentage of the outstanding stock. For example, the demand for U.S. bonds would decrease by 1.34 percent of the outstanding stock, though the demand for Japanese equity would increase by 0.66 percent of the outstanding stock.

Table 16.9Impact of Emerging Market Reserves Diversification on Financial Market Capitalization
ItemUnited StatesEuro AreaJapanUnited Kingdom
Current stock (billions of dollars)16,800600042003000
Expected change in demand* (percent)(+0.40)(+0.66)(+0.66)(+0.66)
Current stock (billions of dollars)19,800840087001000
Expected change in demand (percent)(–1.34)(+0.66)(+0.66)(+0.66)
Of Which: U.S. Treasury Marketable Debt
Current stock (billions of dollars)4000
Expected change in demand (percent)(–7.1)
Source: Beck, Demirgüç-Kunt, and Levine (1999) and author’s calculations.

As one would expect, the selling pressure would play against the dollar, especially fixed-income dollar assets (net demand for U.S. equity would actually increase with the diversification). Net demand for U.S. assets would decrease by 0.5 percent of the outstanding stock, while that for non-U.S. assets would increase by 0.66 percent.

Overall, this back-of-the-envelope calculation shows that the changes in net demand would amount to relatively small fractions of the outstanding stocks. This suggests that moderate price and exchange rate changes would suffice to restore equilibrium. This conclusion, however, comes with several caveats. First, the net supply exceeds 7 percent of the outstanding stock if one restricts one’s attention to marketable U.S. Treasury debt. This results from the fact that the foreign official sector holds a significant fraction—about one-third—of outstanding U.S. government debt (see Parisi-Capone and Setser, 2006). The impact on the interest rate that the U.S. government pays on its debt might thus be nonnegligible, depending on its substitutability with other forms of dollar debt in the portfolios of global investors.34

Second, the short-run price effects of portfolio diversification will depend on the pace of the diversification and on the reaction of private investors. Whereas the literature on sterilized foreign exchange intervention suggests that such interventions have moderate and transitory effects on exchange rates, the microstructure literature shows that their impact might be large (at least in the short run), especially in markets that lack depth and in which information is fragmented. Furthermore, private speculation may not be stabilizing—private investors might want to get out in front of any government moves rather than offset them as they occur. So, although it is unlikely that large price and exchange rate adjustments must result, in the long run, from increased diversification of emerging market countries’ foreign assets, there certainly is a need for the international community to assess and monitor the risks in the transition.

16.7.3. Collective Arrangements

The abundance of reserves held by emerging market countries reduces the need for collective insurance—such as that provided by the IMF at the global level, or by the Chiang Mai Initiative or the Latin American Reserve Fund at the regional level. Indeed, the resources of collective insurance arrangements have become relatively small compared with the reserves that emerging market countries have recently accumulated. For example, the increase in reserves in the Asian emerging market countries over 2000–05 amounts to more than 4 times the IMF’s usable resources at the end of 2005, and more than 20 times the bilateral swap agreements under the Chiang Mai Initiative signed over 2001–05. The buildup in reserves explains in part the recent decline in IMF credit outstanding, which is likely to persist for some time.35

Looking forward, one question is whether the large accumulated stocks of sovereign wealth could be used to collectively insure risks other than capital account crises. Emerging market countries face other risks that are now largely uninsured, such as natural disasters, epidemics, terms of trade shocks, and severe output drops (see Becker and others, 2010). Although some of these risks may be uninsurable because of the potential for moral hazard, there might be scope for expanding insurance through appropriate collective intervention at the regional or global level.

Finally, sovereign wealth can be used to induce the development of regional financial markets. An example of this is the Asian Bond Fund, created in 2003 to diversify the investment of Asian central banks’ reserves away from U.S. and European securities into Asian bonds. Since 2005 the Asian Bond Fund has also invested in domestic currency bonds issued by regional sovereign issuers, as a catalyst for private investment in Asian issues.36 Such initiatives might enable emerging market countries to develop debt instruments (with long maturities and domestic currency denomination) that are safer for borrowers.

16.8. ConclUSION

This chapter has argued that reserves accumulation in Asian emerging market countries is difficult to justify—at least since 2000—in terms of self-insurance against capital flow volatility and capital account crises. The main piece of evidence behind this claim is the failure of a simple cost-benefit model of optimal reserves to account for the reserve buildup in these countries since 2000: their vulnerability to a capital account crisis was too low in that year to justify the cost of the accumulated reserves. That reserves were excessive from the point of view of crisis insurance is also suggested by recent moves to reallocate reserves from central banks to sovereign wealth funds investing in less liquid, higheryielding assets.

Even if the rate of accumulation of reserves were to abate—and notwithstanding the good reasons that it should—the public sectors of a number of emerging market countries, especially in Asia, will have to manage stocks of foreign financial assets of unprecedented size for some time to come. This generates both policy challenges and opportunities for the international community. One challenge is to ensure that the diversification of those assets is conducted in an orderly manner, to avoid large or abrupt changes in the relative prices of financial assets or in exchange rates. An opportunity lies in the fact that this increase in sovereign wealth could provide the basis for cross-country insurance arrangements against risks other than capital account crises, or could catalyze regional financial development.

Appendix I

Data and Definitions

Table 16.10Countries in the Sample
Emerging Market Countries*Industrial Countries
ColombiaPeruFinlandUnited Kingdom
Cotê d’IvoirePhilippinesFranceUnited States
Dominican RepublicPolandGermany
EgyptSouth AfricaIceland
El SalvadorThailandIreland
MalaysiaVenezuelaNew Zealand
Table 16.11Crisis Definitions
Crisis TypeSourceDefinition
Currency Crises
CC1Frankel and Rose (1996)A nominal depreciation of the currency of at least 25 percent relative to the previous year that is also at least a 10 percent acceleration, year over year, in the rate of depreciation.
CC2Frankel and Wei (2005)A year is identified as a crisis year if, in at least one month, the sum of the monthly percentage nominal deprecia tion and the percentage loss in foreign reserves exceeds 15. The index of the nominal depreciation and loss in reserves must also accelerate by 10 percent over the previous month. In cases where successive years may satisfy the crisis criterion, only the first year of crisis is counted within any three-year window.
CC3Frankel and Wei (2005)Same as CC2 except that the sum of the index of nominal depreciation and the loss of foreign reserves must exceed 25 percent.
CC4Frankel and Wei (2005)Same as CC2 except that the sum of the index of nominal depreciation and the loss of foreign reserves must exceed 35 percent.
Sudden Stops
SS1Jeanne and Ranciére (2006)The ratio of net capital inflows* to GDP falls by more than 5 percent relative to the previous year.
SS2Frankel and Cavallo (2004), “sudden stop 1”A reduction in the financial account from a surplus position with respect to the previous year that is two standard deviations above the mean standard deviation (the average of standard deviations of the financial account over the entire sample). A fall in GDP per capita and in the current account deficit must accompany the financial account reduction, either during the same year or the next year.
SS3Frankel and Cavallo (2004), “sudden stop 2”Same as SS2 except that the mean standard deviation of the financial account is that over the corresponding decade only.
SS4Frankel and Cavallo (2004), “sudden stop 3”Same as SS2 except that the mean standard deviation is omputed for the year-to-year change in the financial account rather than the level.
(source: IMF Balance of Payments Statistics).
Table 16.12Currency Crises and Sudden Stops in Emerging Market Countries, 1980–2000
Currency CrisesSudden Stops
Argentina1981 1984 19871981 1984 19871981 1984 19871981 1984198919891989
1990 19951990 19951987 1990
Brazil1981 1983 19871982 1985 19881982 1985 19881982 198719831983
1992 19991991 1995 19981991 19981990 1998
Bulgaria1990 1993 19961990 1994
1996 1998
Chile1982 19851982 1985198519851982 1983 19851982 1983 199819981998
1981 1995 1998
China1984 19941980 1989 19921980 19921992
Colombia1985 19971984 199519851998 1999
Côte d’Ivoire19941982 1985 19881980 1983 19861980 1983 19861983 1984 19961984
1991 1994 19981989 19921989 1992
1996 2000
Dominican Rep.1985 1987 19901981 1984 19871980 1983 19861982 19851981 1993
1990 1994 19981989 19941988 1994
Ecuador1982 1985 19881982 1985 19881982 19861982 19861983 1986 19881983 199919831983
1995 19981991 19981990 19981990 19991992 1999 2000
Egypt19891981 1984 19891981 1986 199019901990 1993199019901990
El Salvador1986 19901982 1985 19881981 1985 19901980 1986 1990
1991 1995
Hungary1994 1996
India19911988 199119911991
Indonesia1983 19861982 1986 19971983 1986 19971983 1986 19981997 199819971997
1997 2000
Korea19971980 1983199719971986 1997199719971997
1986 1997
Malaysia19971992 19971984 19871997
1994 1999
Mexico1982 1985 19941981 1985 19881981 19851982 19851982 19951982 1994 199519821982 1995
1993 19981988 19941990 1994
Morocco1981 1984 19871982 1985 19881980 19831995199519951995
1986 1989
Nigeria1986 19891981 1984 19871982 1986 19921982 19861984 1985 19871999
1992 19991992 1995 19991995 19991992 19991992 1996 1999
Pakistan19821980 1984 19871985 1988 19911988 1991 1996
1990 19931995 1998
1996 1999
Panama1981 1984 19871982 1985 19881980 19831982 1983 1985200020002000
1991 19971992 19971986 19971987 1988 2000
Peru1982 1987 19901982 1987 19901987 19901987 19901983 1984 1998199819981998
Philippines1983 19971982 1985 19881982 19851982 1985 19901983 19971997 199819831983
1991 19971988 1997
Poland1982 19861988 1990
1989 1992
Russia19981998 1999
South Africa19841982 1985 19881982 1985 19881981 1984 19941985
1991 1994 19981992 1995 1998
Thailand1997199719971982 1997 1998199719971997
Tunisia1980 1984 19871984 1987 19911986 1991
1991 1994 2000
Turkey1988 1991 19941982 1985 19881980 19831980 1983 199419941991 1994 19981988
1996 19991991 1994 19981989 1994
Uruguay1982 1984 19891981 1984 19911982 1985 19911982 1985198219831983
Venezuela1984 19861980 19841984 1989 19941984 1989 19941980 19891994
1989 19941989 19941990 1994
Frequency (%)*
Sources: Author’s computations, Frankel and Cavallo (2004), and Frankel and Wei (2005).
Table 16.13Variables Considered in the Probit Analysis
Annual growth in GDPWDI
Current account balanceIFS
Ratio of lagged real public debt to real GDPGDF, WDI
Ratio of lagged short-term debt to real GDPGDF, WDI
Exchange Rate
Second lag of exchange rate regime dummiesReinhart and Rogoff (2004)
Lagged real effective exchange rate deviation from Hodrick-Prescott trendIFS
Ratio of lagged sum of exports and imports to GDPWDI
Lagged growtd in terms of trade (percent)IFS
Index of current account opennessQuinn (2000)
U.S. Interest Rates
Interest rate on Treasury bills (percent a year)IFS
Change in the interest rate on Treasury bills (basis points)IFS
Business Cycle Indicators
Average of first and second lags of real GDP growthWDI
Financial Account Openness
Ratio of lagged absolute gross inflows to GDPIFS
Ratio of lagged sum of absolute gross inflows and absolute gross outflows to GDPIFS
Stocks of Foreign Assets and Foreign Liabilities
Ratio of lagged net foreign assets to GDPLane and Milesi-Ferretti (2010)
Ratio of lagged stock of foreign liabilities to GDPLane and Milesi-Ferretti (2010)
Ratio of stock of debt liabilities to stock of total liabilitiesLane and Milesi-Ferretti (2010)
Ratio of lagged stock of FDI to stock of total liabilitiesLane and Milesi-Ferretti (2010)
Ratio of foreign liabilities to money in tde financial sectorIFS
Consumer price inflation (percent a year)IFS
GDF, Global Development Finance (World Bank); IFS, International Financial Statistics (IMF); WDI, World Development Indicators (World Bank).
Appendix II

Solving for the Optimal Level of Reserves

Reserves Management in a Crisis. How does the country in the model use its reserves in period 1? If there is no crisis, the consumer achieves the desired level of consumption C* and saves any residual wealth (which could be positive or negative) as net reserves. But if there is a crisis, the consumer may be unable to consume C*. Then, using equation (4), Y1 = Y—ΔY, and L′ = 0, period-1 consumption is given by

The question is whether the consumer can achieve the desired level of consumption C1 = C* = Y by running down reserves (R′ = 0). This is the case if reserves R exceed the following threshold:

R is the “full insurance” level of reserves, that is, the amount that allows the consumer to maintain consumption as if there were no crisis. It is also the maximum amount of reserves that the consumer is ready to spend in a crisis. The “full insurance” level of reserves is equal to the sum of the capital outflow and output fall in a crisis, plus the period-1 trade deficit. The last term is equal to zero because of the assumption that C* = Y.

The Loss Function [Equation (7)]. Period-0 welfare is given by

where U1n is the welfare level conditional on no crisis, U1c is that conditional on a crisis, and f(R)=U1n-U1c is the welfare cost of a crisis. Using equations (3) and (4) to substitute out I and R′ - L′ from equation (5), we can write period-1 welfare as

(I set Y2 = 0 to reduce the amount of algebra.) If there is no crisis, C1 = Y1 = C* = Y, so that welfare is given by

where Ū is the consumer’s ex ante welfare if there is no crisis risk and W0 is invested in the illiquid asset. By contrast, if there is a crisis, Y1 = Y—ΔY and R′ = 0 (assuming RR), so that C1 = Y—ΔYL + R = C*—(R-R), implying a welfare level of

Taking the difference, one obtains U1nU1c=f(R), with

(if RR). The welfare cost of a crisis is the sum of two components: the output cost of the crisis and the cost of distorting the path of domestic consumption away from the unconstrained equilibrium. The second component, which is decreasing with R and equal to zero if RR, captures the benefit of reserves in terms of crisis mitigation in my model. Finally, period-0 welfare can be written

An Extension with Debt and Default

Assume that reserves are financed by long-term debt D = –I, and assume W0 = 0, so that R = D. There is no need to assume that debt is illiquid in the same sense as the physical investment—the debt could be traded in a liquid market. What is important for my results is that I cannot be decreased (or, equivalently, D cannot be increased) in a crisis. For debt this is an implication of the credit constraint to which the consumer is subject during a crisis.

Let us assume that the consumer fails to repay long-term debt D with probability μ and therefore pays a risk premium [1/(1 –μ)]—1. Then the expression for final wealth [equation (5)] is

where η takes the value of 1 if the consumer repays the debt and 0 if not. The expression for the expected wealth E1 (W2), and thus the expressions for U1n and U1c, are the same as before. Hence the default probability µ has no impact on the optimal level of reserves; only δ should be counted in the opportunity cost of reserves.

However, the risk of default could be relevant if there are default costs. Assume that Y2 is stochastic and default occurs only if the debt repayment exceeds the cost of default γY2. Then, given D = R, the probability of default μ is endogenous and is solved by

This equation implicitly defines a default threshold Y(R) that is increasing with R. The ex ante loss becomes Loss

The opportunity cost of reserves includes a term for the deadweight cost of default, which is increasing with reserves. This term is not the same as the default risk premium, 1/(1—μ)—1.


    Aizenman,Joshua, and JaewooLee, 2005, “International Reserves: Precautionary vs. Mercantilist Views, Theory and Evidence,”IMF Working Paper 05/198(Washington: International Monetary Fund).

    Aizenman,Joshua, YeonhoLee, and YeongseopRhee,2005, “International Reserves Management and Capital Mobility in a Volatile World: Policy Considerations and a Case Study of Korea,”Journal of the Japanese and International Economies, Vol. 21,No. 1, pp. 1–15.

    Aizenman,Joshua, and NancyMarion,2003,“The High Demand for International Reserves in the Far East: What Is Going On?”Journal of the Japanese and International Economies, Vol. 17, No. 3, pp. 370–400.

    Beck,Thorsten,Asli Demirgüç-Kunt, and Ross E. Levine, 1999, “A New Database on Financial Development and Structure,”Policy Research Working Paper 2146 (Washington: World Bank).

    Becker,Törbjörn, and PaoloMauro,2006, “Output Drops and the Shocks That Matter,”IMF Working Paper 06/172 (Washington: International Monetary Fund).

    Becker,Törbjörn,OlivierJeanne,PaoloMauro,Jonathan D.Ostry, and RomainRancière,2010,“Country Insurance: The Role of Domestic Policies,”Chapter 15 in this volume.

    Ben-Bassat,Avraham, and DanielGottlieb,1992,“Optimal International Reserves and Sovereign Risk,”Journal of International Economics, Vol. 33 (November), No. 3, pp. 345–62.

    Berg,Andrew,EduardoBorensztein, and Catherine Pattillo, 2005, “Assessing Early Warning Systems: How Have They Worked in Practice?”IMF Staff Papers, Vol. 52, No. 3, pp. 462–502.

    Bird,Graham, and RamkishenRajan,2003, “Too Much of a Good Thing? The Adequacy of International Reserves in the Aftermath of Crises,”The World Economy, Vol. 26, No. 6, pp. 873–91.

    Bussière,Matthieu, and ChristianMulder,1999,“External Vulnerability in Emerging Market Economies: How High Liquidity Can Offset Weak Fundamentals and the Effects of Contagion,”Working Paper 99/88(Washington:International Monetary Fund).

    Caballero, and RicardoJ.,2006, “On the Macroeconomics of Assets Shortages,”Working Paper 12753 (Cambridge, MA: National Bureau of Economic Research).

    CaballeroRicardoJ., and Arvind Krishnamurthy,2004, “Smoothing Sudden Stops,”Journal of Economics, Vol. 119, No. 1, pp. 104–27.

    Caballero,Ricardo J., and StavrosPanageas,2005,“A Quantitative Model of Sudden Stops and External Liquidity Management,”Working Paper 11293 (Cambridge,MA: National Bureau of Economic Research).

    Calvo, and GuillermoA.,1996, “Capital Flows and Macroeconomic Management: Tequila Lessons,”International Journal of Finance and Economics, Vol. 1, No. 3, pp. 207–23.

    Calvo, and GuillermoA.,2006,“Monetary Policy Challenges in Emerging Markets: Sudden Stop, Liability Dollarization, and Lender of Last Resort,”Working Paper 12788 (Cambridge, MA: National Bureau of Economic Research).

    Calvo,GuillermoA.,AlejandroIzquierdo, and Luis F. Mejía, 2004, “On the Empirics of Sudden Stops: The Relevance of Balance-Sheet Effects,”Working Paper 10520 (Cambridge, MA: National Bureau of Economic Research).

    Caselli,Francesco, and James Feyrer,2007, “The Marginal Product of Capital,”Quarterly Journal of Economics, Vol. 122, No. 2, pp. 535–68.

    Chang,Roberto, and AndresVelasco,2000,“Liquidity Crises in Emerging Markets: Theory and Policy,”in NBER Macroeconomics Annual 1999, ed. by Ben S. Bernanke and Julio Rotemberg(Cambridge, MA: MIT Press).

    Chinn,MenzieD., and HiroIto,2006,“What Matters for Financial Development? Capital Controls, Institutions, and Interactions,”Journal of Development Economics, Vol. 81, No. 1, pp. 163–92.

    DeBeaufort-Wijnholds,J.Onno, and Arend Kapteyn,2001,“Reserve Adequacy in Emerging Market Economies,”Working Paper 01/143(Washington: International Monetary Fund).

    Detragiache,Enrica, and AntonioSpilimbergo,2001, “Crises and Liquidity: Evidence and Interpretation,”Working Paper 01/2(Washington: International Monetary Fund).

    Dooley,MichaelP.,DavidFolkerts-Landau, and Peter Garber,2004, “The Revived Bretton Woods System: The Effects of Periphery Intervention and Reserve Management on Interest Rates and Exchange Rates in Center Countries,”NBER Working Paper 10332(Cambridge, MA: National Bureau of Economic Research).

    Duffie,Darrell,P.,Lasse HejePedersen, and Kenneth J.Singleton,2003, “Modeling Sovereign Yield Spreads: A Case Study of Russian Debt,”Journal of Finance, Vol. 58, No. 1, pp. 119–59.

    Durdu,CeyhunBora,Enrique G.Mendoza, and Marco E.Terrones,2007,“Precautionary Demand of Foreign Assets in Sudden Stop Economies: An Assessment of the New Mercantilism,”NBER Working Paper 13123 Journal of Finance, Vol. 58, No. 1, pp. 119–59.(Cambridge, MA: National Bureau of Economic Research).

    Edwards,Sebastian,1985, “On the Interest-Rate Elasticity of the Demand for International Reserves: Some Evidence from Developing Countries,”Journal of International Money and Finance, Vol. 4, No. 2, pp. 287–95.

    Edwards,Sebastian,2001, “Capital Mobility and Economic Performance: Are Emerging Economies Different?”NBER Working Paper 8076 (Cambridge, MA: National Bureau of Economic Research).

    Eichengreen, and Barry,2006,“Insurance Underwriter or Financial Development Fund: What Role for Reserve Pooling in Latin America?”NBER Working Paper 12451 (Cambridge, MA: National Bureau of Economic Research).

    Eichengreen,Barry, and Ashoka Mody,2000,“What Explains Spreads on Emerging Market Debt?”In Capital Flows and the Emerging Economies: Theory, Evidence and Controversies,ed.bySebastianEdwards(Chicago:University of Chicago Press).

    Flood,RobertP., and PeterM.1984,“Collapsing Exchange-Rate Regimes: Some Linear Examples,”Journal of International Economics, Vol. 17 (August), pp. 1–13.

    Flood,Robert, and NancyMarion,2002, “olding International Reserves in an Era of High Capital Mobility,”Brookings Trade Forum 2001 (pp. 1–47), ed.bySusanM. Collins andDaniRodrik(Washington:Brookings Institution).

    Frankel,Jeffrey A., and EduardoA. Cavallo,2004,“Does Openness to Trade Make Countries More Vulnerable to Sudden Stops, or Less? Using Gravity to Establish Causality,”Brookings Trade Forum 2001 (pp. 1–47),ed.bySusanM. Collins andDaniRodrik(Cambridge, MA:National Bureau of Economic Research).

    Frankel,JeffreyA., and EduardoA. Cavallo,2008,“Does Openness to Trade Make Countries More Vulnerable to Sudden Stops, or Less? Using Gravity to Establish Causality,”Journal of International Money and Finance, Vol. 27, No. 8, pp. 1430–52.

    Frankel,JeffreyA., and AndrewK. Rose,1996,“Currency Crashes in Emerging Markets: An Empirical Treatment,”Journal of International Economics,Vol. 41 (November),pp. 351–66.

    Frankel,Jeffrey A., and Shang-JinWei,2005,“Managing Macroeconomic Crises: Policy Lessons” in Managing Economic Volatility and Crises: A Practitioner’s Guide,ed. byJoshuaAizenman and BrianPinto(Cambridge, MA:Cambridge University Press).

    Frankel,Jacob A., and BoyanJovanovic,1981,“Optimal International Reserves: A Stochastic Framework,”Economic JournalVol. 91No. 362,pp. 507–14.

    Garcia,Pablo S., and ClaudioSoto,2004,“Large Hoarding of International Reserves: Are They Worth It?”Working Paper 299(Santiago:Central Bank of Chile).

    Ghosh,Atish,ManuelaGoretti,BikasJoshi,AlunThomas, and JuanZalduendo,2007,“Modeling Aggregate Use of Fund Resources—Analytical Approaches and Medium-Term Projections,”IMF Working Paper 07/70(Washington:International Monetary Fund).

    Gourinchas,Pierre-Olivier, and OlivierJeanne,2006,“Capital Flows to Developing Countries: The Allocation Puzzle,”NBER Working Paper No. 13602(Cambridge, MA:National Bureau of Economic Research).

    Hamada,Koichi, and KazuoUeda,1977,“Random Walks and the Theory of Optimal International Reserves,”Economic Journal, Vol. 87,No. 848,pp. 722–42.

    Hauner,David,2005,“A Fiscal Price Tag for International Reserves,”IMF Working Paper 05/81(Washington: International Monetary Fund).

    Heller,H. Robert,1966,“Optimal International Reserves,”Economic Journal,Vol. 76,No. 302,pp. 296–311.

    Hutchison,Michael M., and IlanNoy,2006,“Sudden Stops and the Mexican Wave: Currency Crises, Capital Flow Reversals and Output Loss in Emerging Markets,”Journal of Development Economics,Vol. 79,No. 1,pp. 225–48.

    Hviding,Ketil,MichaelNowak, and LucaAntonio Ricci,2004,“Can Higher Reserves Help Reduce Exchange Rate Volatility?”IMF Working Paper 04/189(Washington:International Monetary Fund).

    International Monetary Fund,2000,“Debt-and Reserve-Related Indicators of External Vulnerability,”IMF Board Paper.Available via the Internet:

    International Monetary Fund,2001,“Issues in Reserves Adequacy and Management,”IMF Board Paper.Available via the Internet:

    International Monetary Fund,2003,“Are Foreign Exchange Reserves in Asia Too High?,” in World Economic Outlook,pp. 78–92(September).(Washington:International Monetary Fund).

    Jeanne,Olivier, and RomainRancière,2006,“The Optimal Level of International Reserves for Emerging Market Countries: Formulas and Applications,”IMF Working Paper 06/229(Washington:International Monetary Fund).

    Jeanne,Olivier, and CharlesWyplosz,2003,“The International Lender of Last Resort: How Large Is Large Enough?” in Managing Currency Crises in Emerging Markets, ed.byMichaelP. Dooley and JeffreyA. Frankel(Chicago:University of Chicago Press).

    Jen,Stephen,2007,Currencies: How Big Could Sovereign Wealth Funds Be by 2015?(New York:Morgan Stanley Research Global).

    Johnson-Calari,Jennifer, and MalanRietveld, eds.,2007,Sovereign Wealth Management(London:Central Banking Publications).

    Kaminsky, Graciela,SaulLizondo, and Carmen M.Reinhart,1998,“Leading Indicators of Currency Crises,”IMF Staff Papers,Vol. 45,No. 1,pp. 1–48

    Knight,Malcolm D.,2006,“International Reserve Diversification and Disclosure,” Speech at the Swiss National Bank/Institute for International Economics Conference,Zurich,September 8.Available via the Internet:

    Krugman,Paul,1979,“A Model of Balance-of-Payments Crises,”Journal of Money, Credit and Banking,Vol. 11,No. 3,pp. 311–25

    Lane,Philip R., and Gian MariaMilesi-Ferretti,2010,“The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970–2004,”Chapter 8 in this volume

    Levy-Yeyati,Eduardo,2006,“The Cost of Reserves,”Working Paper 10/2006(Buenos Aires:Centro de Investigación en Finanzas, Universidad Torcuato Di Tella).

    Miller,Marcus, and LeiZhang,2006,“Fear and Market Failure: Global Imbalances and ‘Self-Insurance’,”Discussion Paper 6000(London:Centre for Economic Policy Research).

    Mohanty,M. S., and PhilipTurner,2006,“Foreign Exchange Reserve Accumulation in Emerging Markets: What Are the Domestic Implications?”BIS Quarterly Review, September, pp. 39–52.

    Morris,Stephen, and HyunSong Shin,1998,“Unique Equilibrium in a Model of Self-Fulfilling Currency Attacks,”American Economic Review,Vol. 88,No. 3,pp. 587–97.

    Mulder,Christian,2000,“The Adequacy of International Reserve Levels: A New Approach” in Risk Management for Central Bankers,ed. byStevenF. Frowen,RobertPringle, and BenedictWeller(London:Central Bank Publications).

    Parisi-Capone,Elisa, and BradSetser,2006,“Central Banks Active Participants in the Market for Treasuries with a Maturity of up to 10 Years,”RGE Monitor(New York:Roubini Global Economics). Available via the Internet:;tgid=10000&;cli d=4042&;cid=152505.

    Prasad,Eswar S., and Raghuram G.Rajan,2005,“Controlled Capital Account Liberalization:A Proposal,”IMF Policy Discussion Paper 05/7(Washington:International Monetary Fund).

    Pringle,Robert, and NickCarver, eds.,2005,RBS Reserve Management Trends 2005(London:Central Banking Publications).

    Quinn,Dennis P.,2000,“Democracy and International Financial Liberalization,”IMF Working Paper 08/250(Washington:McDonough School of Business, Georgetown University).

    Reinhart,Carmen M., and Kenneth S.Rogoff,2004,“The Modern History of Exchange Rate Arrangements: A Reinterpretation,”Quarterly Journal of Economics,Vol. 119,No. 1,pp. 1–48.

    Rodrik,Dani,2006,“The Social Cost of Foreign Exchange Reserves,”International Economic Journal,Vol. 20,No. 3,pp. 253–66.

    Rozanov,Andrew,2005,“Who Holds the Wealth of Nations?”Central Banking Journal,Vol. 15,No. 4,pp. 52–7.

    Rudebusch,Glenn D.,Brian P.Sack, and Eric T.Swanson,2007,“Macroeconomic Implications of Changes in the Term Premium,”Federal Reserve Bank of St. Louis Review,Vol. 89,No. 4,pp. 241–69.

    Stiglitz,Joseph E.,2006,Making Globalization Work(New York:W.W. Norton).

    Summers,Lawrence H.,2006,“Reflections on Global Account Imbalances and Emerging Markets Reserves Accumulation,”L. K. Jha Memorial Lecture, Reserve Bank of India, Mumbai, March 24.Available via the Internet:

    Truman,Edwin M., and AnnaWong,2006,“The Case for an International Reserve Diversification Standard,”Working Paper 06-2(Washington:Institute for International Economics).

    Warnock,Francis E., and VeronicaCacdac Warnock,2009,“International Capital Flows and U.S. Interest Rates,”Journal of International Money and Finance, Vol. 28, pp. 903–19.

    Wooldridge,Philip D.,2006,“The Changing Composition of Official Reserves,”BIS Quarterly Review, September,pp. 25–38.

This chapter is a slightly revised version of an article previously published in Brookings Papers on Economic Activity 1:2007, William C. Brainard and George L. Perry, eds., pp. 1–55. © The Brookings Institution. Reprinted with permission.

I thank Ioannis Tokatlidas for superb research assistance. I also thank Joshua Aizenman and Lawrence Summers, as well as Eduardo Borensztein, Stijn Claessens, Fernando Goncalves, Pierre-Oliver Gourinchas, Nancy Marion, Jonathan D. Ostry, Brad Setser, and Shang-Jin Wei for comments on earlier drafts. This chapter benefited from discussions with Romain Rancière (who also generously shared data) and Christian Mulder.

See, for example, Aizenman and Marion (2003) and Stiglitz (2006). According to a survey of central bankers of developing and emerging market countries, the main reason for the recent buildup in reserves was to “secure protection from volatile capital flows” (Pringle and Carver, 2005). In the words of Stiglitz (2006, p. 248) “The East Asian countries that constitute the class of—97’the countries that learned the lessons of instability the hard way in the crises that began in that year: have boosted their reserves in part because they want to make sure that they won—t need to borrow from the IMF again. Others, who saw their neighbors suffer, came to the same conclusion’it is imperative to have enough reserves to withstand the worst of the world’s economic vicissitudes.”

My sample of emerging market countries is based on JP Morgan’s Emerging Markets Bond Index Global (EMBIG); my sample of industrial countries includes all countries that were members of the Organisation for Economic Co-operation and Development (OECD) in 1990. Appendix I lists the countries in both samples. Neither sample includes three large reserves holders in Asia: Hong Kong SAR, Singapore, and Taiwan Province of China.

The ratio of reserves to imports should equal 0.25 according to the three-months-of-imports rule. The ratio of reserves to short-term external debt should equal 1 according to the Greenspan- Guidotti rule, the idea being that reserves should allow a country to live without foreign borrowing for up to one year. A conventional range for the ratio of reserves to broad money is 5 to 20 percent. The rationale for this ratio is that broad money reflects a country’s exposure to the withdrawal of assets (Calvo, 1996; De Beaufort-Wijnholds and Kapteyn, 2001).

See Heller (1966). The dynamic aspect of the authorities’ optimization problem was treated more rigorously in the buffer stock models of international reserves of Hamada and Ueda (1977) and Frenkel and Jovanovic (1981).

Whereas Heller (1966) interpreted the adjustment cost as a transitory fall in domestic absorption, Ben-Bassat and Gottlieb (1992) and Garcia and Soto (2004) define it as a fall in domestic output. The two are not equivalent for domestic welfare.

The representative-consumer assumption implies that one must look at the optimal level of reserves from the point of view of the country as a whole, without distinguishing between the private sector and the public sector. See, for example, Caballero and Krishnamurthy (2004) for a model of international reserves that includes a meaningful distinction between the private sector and the government.

Aizenman and Marion (2003) and Miller and Zhang (2006) present two-period precautionary savings models of reserves. Caballero and Panageas (2005) and Durdu, Mendoza, and Terrones (2007) present more dynamic precautionary savings models of international reserves. These models do not yield closed-form solutions for the optimal level of reserves but can be solved numerically.

Note that the consumer always repays the short-term debt that is not rolled over; that is, default is ruled out by assumption as a way of smoothing domestic consumption.

In line with the model, my discussion will focus on crisis management and will not deal with some benefits that reserves may have in noncrisis times, such as limiting exchange rate volatility (Hviding, Nowak, and Ricci, 2004) or providing liquidity to the foreign exchange market. Reserves can also yield benefits if the government is able to invest them more wisely than the average citizen, or if they promote capital market integration and domestic financial development.

Those conclusions were presented in two documents: “Debt- and Reserve-Related Indicators of External Vulnerability” (IMF, 2000) and “Issues in Reserves Adequacy and Management” (IMF, 2001). One study that contributed to crystallizing the official sector’s conventional wisdom about the importance of this ratio was Bussière and Mulder (1999). See also Mulder (2000).

The literature on early warning signals and the empirical determinants of crisis in probit/logit regressions is too large to be reviewed here’ the reader is referred to the reviews by Kaminsky, Lizondo, and Reinhart (1998); Berg, Borensztein, and Patillo (2005); and Frankel and Wei (2005). Another way in which reserves might stabilize the domestic economy is by lowering the interest rate on foreign debt (Levy-Yeyati, 2006). Evidence that larger reserves decrease the sovereign spread is provided in Hauner (2005); Duffie, Pedersen, and Singleton (2003); and Eichengreen and Mody (2000). By contrast with currency crises, Calvo, Izquierdo, and Mejía (2004) and Frankel and Cavallo (2008) did not find that reserves had a statistically significant effect of reducing the probability of sudden stops.

The precise definitions are given in Table 16.11. in Appendix I. The crisis dates for SS2 to SS4 are taken from Frankel and Cavallo (2008), who apply the criteria of Calvo, Izquierdo, and Mejía (2004) to a larger sample of countries and a longer time period.

More precisely, the measure of short-term debt that works best is that from the World Bank Global Development Finance database rather than that in the Bank for International Settlements (BIS) data. This result is surprising because the BIS data should be a better measure of the denominator in the Greenspan-Guidotti ratio (the BIS reports debt maturing in the following year, whereas the World Bank data are based on maturity at issuance). However, the BIS debt measure might be less significant because it is available for fewer of the countries in the regressions.

This ambiguity is certainly present in the theoretical literature on crises and reserves. In some models, a large volume of reserves effectively reduces the probability of crisis by making the economy more resilient to adverse shocks (Chang and Velasco, 2000; Aizenman and Lee, 2005) or to selffulfilling changes in market sentiment (Morris and Shin, 1998). By contrast, in the Krugman- Flood-Garber framework, a speculative attack made unavoidable by excessive money growth is merely delayed by a larger stock of reserves (Krugman, 1979; Flood and Garber, 1984). In addition, countries often shorten the maturity of their debt before a crisis, further reducing the Greenspan- Guidotti ratio (Detragiache and Spilimbergo, 2001).

Jeanne and Wyplosz (2003) and Calvo (2006) emphasize that lending the reserves to domestic agents is a more effective tool than foreign exchange intervention in preventing and mitigating crises. Calvo (2006) points to an interesting example of a nonstandard way of disposing of international reserves: in August 2002 the central bank of Brazil employed some of its international reserves to make loans to the export sector through commercial banks.

See Jeanne and Rancière (2006). This decomposition of domestic absorption results from two national accounting identities. First, domestic absorption (the sum of domestic private and public consumption and investment) is the difference between real output and the trade balance,At = Yt—TBt. Second, the balance of payments equation CA+KAt=ΔRtwhereCAt=TBt+ITt is the current account balance (the sum of the trade balance and income and transfer from abroad), can be used to substitute out the trade balance from the first identity.

The dollar value of output and domestic absorption falls by a larger amount than indicated in Figure 16.8 because of the real depreciation of the domestic currency. The variables are converted from current dollars to constant local currency units using the nominal dollar exchange rate and the local GDP deflator. IMF loans are counted as reserves rather than capital inflows.

My discussion focuses on the opportunity cost of carrying the reserves and does not deal with the challenges to monetary and financial stability posed by large-scale sterilization (see Mohanty and Turner, 2006, and European Central Bank, 2006, for a discussion of those costs). Another cost that I do not discuss is the false sense of confidence that reserves may instill in foreign investors, allowing the domestic authorities to postpone necessary adjustments. Finally, large-scale purchases and sales of reserves could induce exchange rate changes that cause valuation losses on the reserves.

Caselli and Feyrer (2007) compute the return to capital using production functions calibrated as in the development accounting literature. They find that the return to capital is not higher in developing countries than in industrial countries once one adjusts for nonreproducible capital (land).

This measure was initially proposed by Edwards (1985). It is used by Garcia and Soto (2004) and Rodrik (2006).

The differential between 10-year U.S. Treasury bonds and three-month U.S. Treasury bills was almost 2.5 percentage points on average over 2000–05. Expectation-adjusted measures lead to even lower estimates of less than 1 percentage point (Rudebusch, Sack, and Swanson, 2007).

This is consistent with the estimates obtained by Rodrik (2006) and Bird and Rajan (2003).

Using instead the ratio of short-term external debt to GDP would give similar values for L. For my sample this ratio is 8.2 percent on average over the period 1980–2000 according to the World Bank’s Global Development Finance (GDF) data set, and 11.7 percent according to the BIS database.

Hutchison and Noy (2006) find that the cumulative output loss of a sudden stop is around 13 to 15 percent of GDP over a three-year period. Becker and Mauro (2006) find an expected output cost of 10.2 percent of GDP for currency crises and 16.5 percent of GDP for sudden stops. On the one hand, the estimated output cost of a crisis can be significantly larger if the output gap is cumulated until output has returned to potential, which typically takes longer than two or three years. On the other hand, using the precrisis growth rate to estimate postcrisis potential output may exaggerate the size of the output gap if the crisis was preceded by an unsustainable economic boom.

However, this may not be an implausible order of magnitude for the cost of a severe banking crisis or of social unrest.

Another alternative is the view that the high-growth developing countries are exporting their savings abroad because of a shortage of domestic assets for their residents to invest in (Caballero, 2006). These capital outflows must take the form of reserves accumulation if residents’ holdings of foreign assets are restricted by capital controls.

The nonmercantilist variant would hold that these countries’ competitiveness results from natural factors (for example, that wages are kept low in the export sector by a reserve army of labor migrating from the traditional sectors) rather than policy-induced distortions. Mercantilism is at the core of the “Bretton Woods II” view (Dooley, Folkerts-Landau, and Garber, 2004) of the international financial system. Although many commentators find this view quite plausible, it is not obvious how to confirm or reject it empirically. For example, Aizenman and Lee (2005) find that variables associated with the mercantilist motive (lagged export growth and deviations from predicted purchasing power parity) explain very little of the cross-country difference in reserves accumulation.

The correlation is less significant for capital account restrictions than for the current account balance or the growth rate, and it seems less robust’it is no longer significant if one uses Edwards’ (2001) measure of capital mobility rather than Chinn and Ito’s (2006).

As Gourinchas and Jeanne (2006) have shown, high-growth developing countries tend to export capital, a puzzle that is explained in part by reserves accumulation.

See Rozanov (2005) and Johnson-Calari and Rietveld (2007). Another approach would be to give the private sector more direct control over the allocation of the country’s foreign assets, as in Prasad and Rajan’s (2005) proposal to set up closed-end mutual funds that purchase reserves from the central bank and invest the proceeds abroad.

On recent trends in reserves diversification see Knight (2006), Woolridge (2006), and Truman and Wong (2006).

Table 16.11, which is based on data from the Treasury International Capital (TIC) database, does not report foreign official investment in onshore or offshore dollar deposits and repurchase agreements, which amount to about one-fourth of the total (Knight, 2006, Table 2).

Warnock and Warnock (2009) find that foreign demand for Treasury securities has a significant impact on Treasury yields. A study by the European Central Bank (2006) finds that the interventions conducted by Asian central banks cannot be shown to be responsible for the low yields in the United States, although they have certainly played a role.

Using various models of the demand for IMF loans, Ghosh and others (2007) project that IMF credit outstanding will decline from an average of SDR 50 billion over 2000–05 to SDR 8 billion over 2006–10, in part because of the increase in the reserves-to-short-term-debt ratio in emerging market countries.

Eichengreen (2006) recommends that the Latin American Reserve Fund follow a similar course of action.

    Other Resources Citing This Publication