What drives reserve accumulation (and at what cost)?

Eduardo Levy Yeyati 30 September 2010



The argument against reserve accumulation by emerging economies is often built on three premises:

  • reserves introduce negative externalities: they perpetuate global imbalances and depress interest rates, stimulating asset bubbles (Aizenman 2009);
  • reserves are costly: most indebted emerging economies need to pay a carrying cost roughly proportional to the sum of their sovereign credit risk premium and the term premium in the reserve currency of choice; and
  • reserves are precautionary: they are purchased to build self-insurance “liquidity war chests” to prevent or defend against a sudden capital-account reversal.1

Of these three premises, the first one reflects a complex coordination problem that exceeds cost-efficiency considerations by individual countries. The other two, in turn, merit some important qualifications.

The cost of reserves

The cost of reserves is often estimated as the gap between the yield of hard-currency public debt and the return on reserves. Because reserves are held in short risk-free assets, this gap is, in turn, a function of the sovereign risk spread and the hard-currency interest rate premium (see for example Jeanne and Ranciere 2006 and Rodrik 2006). In reality, the cost of reserves tends to differ from this simple formula.

  • First, to the extent that liquid reserves reduce credit risk (and the interest rate) paid on the total (public and private) debt stock, the marginal cost of carrying reserves for indebted economies may be significantly lower than the sovereign spread.2
  • Second, because the fact that reserves are held in short-dated instruments is related less to liquidity than to central banks’ reserve management practices (including, possibly, fear and of mark-to-market losses), the term premium is in most cases an unnecessary cost.
  • Third, and more importantly, reserves are typically purchased by central banks through interventions sterilised with the sale of local currency-denominated debt.

This may result in either:

  • the central bank bearing quasi-fiscal losses associated with steep interest rate differentials, or,
  • if intervention delays appreciation (and appreciation expectations depress local currency rates), changes in the local-currency value of international reserves as the exchange rate moves towards its new, more appreciated equilibrium.

It follows that “leaning-against-the-wind” reserve accumulation would sustain important valuation losses only if appreciation pressures are permanent. By contrast, if they are due, for example, to cyclical inflows or short-lived terms of trade shocks, the reversion of the exchange rate to its earlier, more depreciated level would eliminate much of the valuation losses.

A quick look at the 2005-2010 rollercoaster gives a first glance at the diversity of realised costs of reserves.

Figure 1 shows back-of-the-envelope estimates cumulative valuation and carry losses for a few central banks known to intervene actively in foreign exchange markets (see Levy Yeyati and Sturzenegger 2010b for details). Predictably, valuation losses accumulate during the appreciation phase, and decline during a sell-off, as the central bank sells at a high price reserves that it had bought on the cheap, and reserves stocks benefit from the revaluation of the dollar. Indeed, many central banks experienced valuation profits during the period, as the early appreciation reverted, and reserves were sold at higher parities to contain the currency run. On the other hand, carrying costs are larger for “carry currencies” (e.g., Brazil) with wide interest rate differentials often unrelated with exchange rate expectations.

In sum, while realised intervention costs tend to vary considerably, the conventional view that reserves are costly due to wide sovereign spreads or heavy quasi fiscal losses appears to be somewhat overstated.

Figure 1. Profits and losses from reserve purchases (USD billions unless otherwise indicated)

Source: Central bank bulletins and IMF.

Precautionary liquidity or exchange rate policy?

Regarding the motives for reserve accumulation, the evidence does not seem to support the precautionary story in its conventional form.3

On the one hand, the evolution of global international reserve stocks are largely explained by a few countries (China, oil exporters, Japan) with a limited need for hard-currency liquidity insurance (see Figure 2).

Figure 2. Distribution of official holdings of international reserves

Source: Levy Yeyati and Sturzenegger (2010b) based on BIS data

On the other hand however, even in those emerging economies where capital flow reversals may represent a clear and present danger, a casual look at the path of reserve accumulation seems to suggest that the latter is better explained by an exchange rate-smoothing policy than by a pure precautionary motive.4

Figure 1 already provides a hint at this leaning-against-the-wind pattern whereby central banks purchase reserves to contain appreciation, and sell them in the event of a run (see Kiguel and Levy Yeyati 2009 for a discussion5). A simple regression of the change in the reserve-to-GDP ratio on these the two motives (proxied, respectively, by the M2-to-GDP ratio, and by the financial account balance over GDP) points in the same direction. That is, in most cases, leaning-against-the wind appears to account for reserve accumulation better than precaution (see Levy Yeyati and Sturzenegger 2010b6 for details)

Two motives or one? Common grounds and distinct implications

A key aspect of the reserves debate that is usually overlooked is the fact that the two motives highlighted above may not be at odds with each other. One could see leaning-against-the-appreciation-wind during expansions as the countercyclical prudential response to procyclical capital flows and real exchange rates, the goal of which is to avoid current-account deficits in good years and prevent a dollar squeeze in the downturn.

That said, such a strategy would differ from the simple hoarding of liquid assets typically associated with the precautionary story. By the same token, if intervention is geared towards limiting a temporary (and possibly excessive) appreciation of the local currency, it is unrealistic to expect, as is often claimed, that an effective global financial safety net should lead to a significant decline in reserve accumulation (see Mateos et al. 2010).

As a final note, exchange rate smoothing does not require that reserves be held in short, low-yielding liquid assets. Even precautionary reserves – unlikely to be used in full – can afford to be partially invested in higher yielding long-run saving instruments as in the case of sovereign wealth funds.

Perhaps in the realisation of this inconsistency between goals (precautionary exchange rate smoothing) and instruments (liquid reserve assets) lies the hope to reduce the excessive demand for the latter that triggered the quest against reserve accumulation in the first place.


Aizenman, Joshua (2009), “Alternatives to sizeable hoarding of international reserves: Lessons from the global liquidity crisis”, VoxEU.org, 30 November.

Aizenman, J and J Lee (2007), “International Reserves: Precautionary Versus Mercantilist Views, Theory and Evidence”, Open Economies Review, 18(2).

Cordella, T and E Levy Yeyati (2010), “Global safety nets: The IMF as a swap clearing house”, VoxEU.org, 18 April.

Jeanne, O and R Ranciere (2006), “The Optimal Level of International Reserves for Emerging Market Countries: Formulas and Applications”, IMF Working Paper 06/229.

Kiguel, A and E Levy Yeyati (2009), “Back to 2007: Fear of appreciation in emerging economies”, VoxEU.org, 29 August.

Lane, Philip R. (2009), “International Financial Integration and Japanese Economic Performance”, Working Paper Series, Centre on Japanese Economy and Business.

Levy Yeyati, Eduardo (2008a). "The Cost of Reserves", Economic Letters, 100(1):39-42

Levy Yeyati, E (2008b), "Liquidity Insurance in a Financially Dollarized Economy" in Financial Markets Volatility and Performance in Emerging Markets, 185-218, NBER.

Levy Yeyati, E and F Sturzenegger (2010a), “Monetary and Exchange Rate Policies and Economic Development”, in M Rosenzweig and D Rodrik (eds.), Handbook of Development Economics 5, Chapter 64, Elsevier.

Levy Yeyati, E and Sturzenegger, F (2010b), “Leaning against the wind: Exchange rate policy in emerging markets in the 2000s”, mimeo, Universidad Torcuato Di Tella.

Mateos, Isabelle, Rupa Duttagupta, and Rishi Goyal (2009), “The Debate on the International Monetary System,” IMF Staff Position Note, SPN/09/26.

Obstfeld, M, J Shanbaugh, and A Taylor (2009), “Financial Instability, Reserves, and Central Bank Swap Lines in the Panic of 2008”, American Economic Review, 99(2):480-86, May.

Rodrik, D (2006), “The social cost of foreign exchange reserves”, NBER Working Paper No. 11952.

1 As the argument goes, individual reserves could be optimally substituted by centralised financial safety nets. Centralised holdings by an international agency would require a smaller stock than individual self-protection thanks to risk pooling and diversification gains –a claim qualified by the high correlation displayed during episode of global distress (Cordella and Levy Yeyati 2010).

2 If, for a given net debt stock, a larger stock of liquid foreign currency assets tightens the sovereign spread, the resulting gain in rollover costs should be net out from the spread (Levy Yeyati 2008a).

3 According to this view, the rapid accumulation of precautionary reserves in emerging economies in the early 2000s reflects the buildup of a hard-currency liquidity cushion, in the aftermath of the dollar shortages of the late 90s (see Aizenmann and Lee 2007 and Levy Yeyati 2008b).

4 For a discussion of leaning-against-the-wind intervention policies, see Levy Yeyati and Sturzenegger (2010a), and references therein.

5 The case of Brazil is illuminating, because they complemented intervention by introducing a Tobin tax on inflows by late 2009, thereby increasing the costs of dollar liquidity (and contradicting the liquidity hoarding argument).

6 The exercise replicates Lane's (2009) tests for Japan. The use of M2-to-GDP as a proxy for the self-insurance target was originally proposed by Obstfeld et al. (2009).



Topics:  International finance International trade

Tags:  exchange rates, global imbalances, exchange-rate policy

Dean, School of Government, Universidad Torcuato Di Tella (UTDT); Professor (on leave), School of Economics, Universidad de Buenos Aires