Chapter

II Implications of Dollarization for Financial Stability

Author(s):
Anne Gulde, David Hoelscher, Alain Ize, Dewitt Marston, and Gianni De Nicolo
Published Date:
June 2004
Share
  • ShareShare
Show Summary Details

Empirical evidence suggests that financial dollarization may increase the vulnerability of financial systems to solvency and liquidity risks. Simple cross-country estimates of the impact of dollarization on some key financial soundness indicators, while controlling for changes in underlying macro-volatility, are consistent with the hypothesis that increased dollarization may increase financial vulnerability. In particular, the variance of deposit growth is positively and significantly correlated with dollarization, suggesting that dollarized financial systems may be more exposed to credit cycles and liquidity risk (Table 3). Based on estimates of non-performing loans (NPLs) or a composite systemic risk measure, the Z-index, which measures the probability of insolvency of a firm, dollarized economies also may be more exposed to solvency risk (Box 3 and Table 4).

Table 3.Deposit Volatility and Dollarization
Dependent

Variable
AFCDINFVRERDVCOVAdjusted

R2
Number of

Countries
DEPGV1.56**0.30**0.013.76**0.9458
(0.52)(0.01)(0.01)(0.13)
Sources: National authorities; IMF, International Financial Statistics database; and IMF staff estimates.Note: DEPGV is the standard deviation of total deposit growth, AFCD is the 1995–2001 average ratio of foreign deposits to total deposits, INFV is the variance of inflation, RERDV is the variance of real exchange rate depreciation, and COV is the covariance between inflation and real exchange rate depreciation. DEPGV is computed with annual data for the 1990–2001 period. All other variables are computed with quarterly data for the 1995–2001 period. White heteroskedastic-consistent standard errors are shown in parentheses.

denotes results at the 99 percent significance level.

Sources: National authorities; IMF, International Financial Statistics database; and IMF staff estimates.Note: DEPGV is the standard deviation of total deposit growth, AFCD is the 1995–2001 average ratio of foreign deposits to total deposits, INFV is the variance of inflation, RERDV is the variance of real exchange rate depreciation, and COV is the covariance between inflation and real exchange rate depreciation. DEPGV is computed with annual data for the 1990–2001 period. All other variables are computed with quarterly data for the 1995–2001 period. White heteroskedastic-consistent standard errors are shown in parentheses.

denotes results at the 99 percent significance level.

Table 4.Impact of Dollarization on Financial Soundness Indicators
Dependent

Variable
FCDINFRERINFVRERVCOVAdjusted

R2
Number of

Countries
ZT−0.06**0.550.80−0.013−0.028**−0.0730.2735
(0.01)(0.54)(0.51)(0.117)(0.007)(0.15)
NPL0.065**1.16−0.07−0.1240.039−0.1490.0842
(0.003)(1.21)(1.09)(0.22)(0.033)(0.29)
Sources: National authorities; Bureau van Dijk, Bankscope database; IMF, International Financial Statistics database; and IMF staff estimates.Note: We find a negative and significant sign of FCD for the ZT regression, implying that systemic risk potential is positively correlated with financial dollarization across countries. When the 2001 FCD ratio is replaced with the 1995 FCD ratio, the sign of FCD in the first regression remains negative and significant, consistent with causality from dollarization to systemic risk potential. Furthermore, NPL appears larger in more highly dollarized countries because the sign of the coefficient associated with FCD is positive and significant. In sum, the evidence is suggestive of a positive relationship between financial dollarization and systemic risk potential. All of these results are confirmed by the augmented versions of these regressions reported in De Nicoló, Honohan, and Ize (2003). White heteroskedastic consistent standard errors are in parentheses.

denotes results at the 99 percent significance level.

Sources: National authorities; Bureau van Dijk, Bankscope database; IMF, International Financial Statistics database; and IMF staff estimates.Note: We find a negative and significant sign of FCD for the ZT regression, implying that systemic risk potential is positively correlated with financial dollarization across countries. When the 2001 FCD ratio is replaced with the 1995 FCD ratio, the sign of FCD in the first regression remains negative and significant, consistent with causality from dollarization to systemic risk potential. Furthermore, NPL appears larger in more highly dollarized countries because the sign of the coefficient associated with FCD is positive and significant. In sum, the evidence is suggestive of a positive relationship between financial dollarization and systemic risk potential. All of these results are confirmed by the augmented versions of these regressions reported in De Nicoló, Honohan, and Ize (2003). White heteroskedastic consistent standard errors are in parentheses.

denotes results at the 99 percent significance level.

Are Dollarized Financial Systems Inherently Vulnerable?

In comparing the financial vulnerability of dollarized and nondollarized systems, it is important to account for the fact that dollarization itself largely arises as protection against risk. Financial dollarization is, in large part, a response to currency instability. Domestic borrowers and lenders prefer to denominate contracts in foreign currency when that currency is expected to provide a more stable, and thus less risky, medium for intermediation.1 Hence, for a given underlying macrovolatility and monetary regime, local currency lending could well be equally risky, if not more so, than dollar lending, because of more volatile, and often higher, real lending rates. Instead, to the extent that borrowers can hedge against risk by borrowing in a mix of the two currencies, a case can be made that bicurrency—for example, partially dollarized—financial systems should be less risky than single currency systems. Explaining the apparent higher financial vulnerability of dollarized economies thus requires identifying systemic risks that are unique to dollarized systems or that are linked to abrupt regime changes. When these risks are not adequately internalized by financial market participants, excessive dollarization and exposure to dollarization risks may ensue.

Box 3.Systemic Risk Potential and Dollarization

Aggregate measures of bank risk taking for the entire banking system can be viewed as proxies of systemic risk potential. We consider two such proxy measures: the Z-index (Z) and the ratio of the nonperforming loans to total loans (NPL). The Z-index is a proxy of the probability of insolvency of a firm. It combines in a single indicator: profitability, given by a period average return on assets (ROA); leverage, given by the period average equity capital-to-asset ratio (K); and return volatility, given by the period standard deviation of returns on asset (S). It is measured by the ratio (ROA + K)/S. Thus, Z increases with profitability (higher ROA), and decreases with leverage (lower K) and return volatility (higher S). A larger value of Z indicates a smaller risk profile, which can be attained by improving efficiency (increasing ROA), greater diversification (decreasing S), or larger capital (increasing K), or through a combination of these. When measured for an aggregate of firms, it can be viewed as the probability of insolvency of a firm whose ROA, K, and S are weighted according to the size of its components. Thus, it is a proxy of systemic risk potential when measured for the aggregate banking system. Z-measures constructed on the basis of 1995–2000 data taken from national authorities and the Fitch-IBCA banking database, are taken from De Nicoló and others (2003).

Table 4 presents the results of two regressions. The dependent variables are a Z measure (ZT) and the 2001 NPLs, both for the entire banking system. The independent variable of interest is financial dollarization, as measured by the 2001 ratio of foreign deposits to total deposits (FCD). Cross-country differences in the macroeconomic environment are controlled for by the average and variance of inflation (INF and INFV, respectively), average and variance of real exchange rate depreciation (RERD and RERDV, respectively), and the covariance between inflation and real exchange rate depreciation (COV). Inflation and real exchange rate depreciation are computed with quarterly data for the 1995–2001 period.

What Are the Specific Financial Risks of Partial Dollarization?

Liquidity Risk

The dollar deposits held by non–U.S. banks are only partially covered by liquid U.S. dollar assets. Since these deposits can potentially be withdrawn in full, dollarization subjects the financial system to a very specific type of liquidity risk. Systemic liquidity risk in dollarized economies arises when the demand for local assets falls, because of a perceived increase in country risk or banking risk, prompting depositors to convert their deposits into cash dollars or transfer them abroad, or foreign banks to recall short-term lines of credit. Dollar liabilities need to be paid at par against foreign currency, inhibiting equilibrating adjustments in the relative price of the two assets (the exchange rate of local dollars against U.S. dollars is fixed).2 Thus, unless liquid dollar liabilities are backed by sufficient liquid dollar assets abroad, banks may run out of dollar liquid reserves and fail to pay off dollar deposits or other dollar liabilities on demand, or as they fall due. Similarly, central banks may run out of international reserves to provide dollar lender-of-last-resort support to distressed banks.3 When this happens, deposit (or loan) contracts may need to be broken and disruptive or confiscatory measures taken, thereby validating creditors’ fears and justifying the run (see Section III). The recent Argentine experience provides a fresh and vivid illustration of a liquidity-induced banking crisis but there have been many other similar crises or near-crisis episodes in dollarized countries in recent history, including Mexico and Bolivia in 1982; Turkey in 1994; Bolivia and Argentina in 1995; Bulgaria in 1996; Peru and Russia in 1998; and Uruguay, Bolivia, and Paraguay in 2002 (Table 5).

Table 5.Bank Runs in Partially Dollarized Economies
Country

Case
Dollarization

of Deposit
Trigger for

the Run
Extent of

the Run
Treatment of

Dollar Deposits
Treatment of Local

Currency Deposits
Outcome
Turkey (1994)46 percentSovereign downgrade and foreign exchange depreciation.17 percent (January–March 1994)Full guarantee of all household deposits, and liquidity of up to 200 percent of capital available to banks.Same as for dollar deposits.Outflows from sound banks were halted by guarantee, allowing insolvent banks to be closed, but lack of supervisory follow-up left banking system vulnerable.
Argentina (1995)50 percentMexican crisis.18 percent (January–March 1995)Heavy, but ad hoc, liquidity assistance provided in dollars.Liquidity assistance in domestic currency.Outflows diminished in due course, with 10 smaller banks closed and ad hoc liquidity provided to others. There was deposit flight to quality.
Bulgaria (1996)40 percentDeposit rationing by two banks.Closure of some banks, choice of full guarantee of foreign exchange deposits in domestic currency or gradual payout in dollars over two years.Full guarantee and liquidity in local currency.Outflows halted once a second wave of withdrawals eliminated all bankrupt banks from the system and full guarantee and earmarked funds were in place for remaining banks. Banking system remained healthy without liquidity needs once currency board was in place.
Russia (1998)18 percentUnsustainable fiscal expansion leading to government debt market collapse.7 percent (January–August 1998)Household deposits at six major banks transferred to a state bank, but repaid in rubles at a fixed exchange rate. Liquidity provided only in rubles to the state bank and to other banks.Same as for dollar deposits.The transfer of deposits to the state banks was an effective guarantee of almost all household deposits, but data do not show whether it stopped the run. The standstill, although unpopular, did not prevent capital reflows when the economy improved.
Ecuador (1999)76 percentFailures of major domestic banks.29 percent (January 1998–March 1999)The first attempt at resolution was a bank holiday followed by a deposit freeze and temporary securitization. When the run reemerged as the freeze was lifted, formal dollarization took place.Same as for dollar deposits.The deposit freeze halted the outflows initially, but social pressures caused the freeze to be gradually relaxed ahead of schedule. Further runs occurred as the easing took place, leading to renewed currency crisis and government default. Formal dollarization eventually was successful in easing deposit outflows.
Argentina (2001)70 percentExchange rate and government debt concerns.22 percent during 2001Partial deposit freeze, followed by forced dedollarization and maturity extension. Voluntary securitization of 12 percent of deposits.Similar partial deposit freeze and heavy liquidity assistance after dedollarization.Social unrest and sharpened recession. Leakages occurred under the deposit freeze, partly from legal challenges.
Uruguay (2002)85 percentArgentine crisis.10 percent during 2002Domestic banks: dollar liquidity to sight deposits, maturity extension of time deposits. Foreign banks: headquarters liquidity in case of need.Domestic banks: liquidity to all deposits. Foreign banks: headquarters liquidity in case of need.Outflows halted once 100 percent of unrestricted deposits were covered by earmarked funds, but government debt sustainability is now in question.
Sources: National authorities; and IMF staff estimates.
Sources: National authorities; and IMF staff estimates.

The triggers for runs on dollar liabilities can be of a diverse nature. Often, the runs were triggered by rapidly deteriorating macroeconomic conditions. For example, the Mexican 1982 crisis was triggered by an apparent loss of macroeconomic control, with a rapidly expanding fiscal deficit and public debt, and weak and confusing monetary and exchange rate management. The large claims of commercial banks on the government introduced a direct channel of transmission from fiscal insolvency to bank insolvency. A very similar sequence of events took place in Argentina during the more recent crisis. In other cases, as in Bolivia, the outflows were triggered by unfounded rumors of deposit confiscations or political turmoil. In still other cases, as in the recent Uruguayan, Paraguayan, and, to some extent, Bolivian crises, contagion was a major determinant, as illustrated by the correlation of deposit outflows (Figure 1). Contagion was in part motivated by the risk of a freeze or a forced conversion of dollar deposits into local currency spreading to other countries.4 Interestingly, however, not all countries were affected equally. Bolivia’s deposit outflows seemed to have responded primarily to the uncertainty resulting from the presidential elections. In Peru, while there was a reduction of foreign credit lines, deposits were not affected at all. In any event, it should be noted that liquidity crises in dollarized economies, like all banking crises, are, by their nature, hard to predict.

Figure 1.Foreign Currency Deposits

(December 2000 = 100)

Sources: IMF, International Financial Statistics database; and IMF staff estimates.

Dollar deposits are often more vulnerable to runs than local currency deposits, even in the absence of exchange rate adjustments. Except for Argentina, where a shift from local currency deposits to dollar deposits took place in the early stages of the run, reflecting fear of a currency depreciation, this was the case for the recent runs in the other Southern Cone countries (Figure 2).5 In highly dollarized countries, the relative stability of local currency deposits also reflects the fact that they are mostly held for transaction purposes. Thus, they are less affected by expected yield differentials than dollar deposits, which are predominantly held as stores of value and are close substitutes for deposits abroad or cash dollars.6 Moreover, even when the demand for local currency deposits is affected, the small size of these deposits in the most highly dollarized countries limits the threat they represent for banks’ liquidity.

Figure 2.Foreign and Local Currency Deposits

(In millions of U.S. dollars)

Sources: IMF, International Fincial Statistics database; and IMF staff estimates.

Note: Local currency deposits are converted at a constant exchange rate at the beginning of the period.

The lack of dollar monetary instruments can further inhibit the scope for interest rate defenses against deposit withdrawals. As in the case of fixed pegs, an interest rate defense that validates changes in risk premiums may be ineffective once a run has started. What is peculiar to dollarized systems, however, is that in the absence of a monetary policy in dollars, the lack of monetary signals from the central bank requires that banks take the initiative to raise their dollar deposit rate. Banks are often reluctant to do so, reflecting concerns that increases by individual banks may be interpreted as a sign of weakness, further exacerbating deposits withdrawals. Indeed, there was nearly no dollar deposit rate response to the recent deposit withdrawals in Uruguay and Paraguay, where central banks do not have dollar-denominated monetary instruments (Figure 3). Instead, in Argentina and Bolivia, where dollar-denominated monetary instruments exist, the interest rate response was more significant.

Figure 3.Interest Rates on Foreign Currency Deposits

(In percent)

Source: IMF, International Financial Statistics database.

Solvency Risk

The main solvency risk faced by dollarized financial systems results from currency mismatches in the event of large depreciations. In turn, currency mismatches can affect banks’ balance sheets directly or affect them indirectly by undermining the quality of their dollar loan portfolio. Banks’ direct exposure to currency risk is generally limited by tight regulatory limits on open foreign exchange positions. However, controlling banks’ positions in derivatives, which are often misreported, have been a frequent source of difficulties.7

Currency-induced credit risk is a key, mostly unregulated, source of vulnerability. The main source of currency risk for banks in highly dollarized economies generally is the exposed position of their borrowers, which makes them susceptible to default in the event of a large depreciation. The availability of local currency loans is not sufficient to meet the needs of clients in the nontradable sector whose cash flow is in local currency. At the same time, banks with large domestic dollar liabilities must balance their foreign exchange positions by either extending dollar lending to local currency earners or holding dollar assets abroad. To maintain their profitability—in view of the generally much lower rates of return on foreign assets than on local dollar assets—and satisfy the pent-up demand for loans, banks generally end up on-lending domestically a large share of their dollar deposits, effectively transferring the currency risk to their unhedged clients and retaining the resulting credit risk.8 The share of total dollar loans granted to borrowers in the nontradable sector thus reached, in mid–2002, more than 65 percent in Bolivia, about 50 percent in Costa Rica, 60 percent in Peru, and 80 percent in Paraguay. The scope for borrowers’ currency mismatch is enhanced by the fact that prices and wages may continue to be set in local currency (e.g., real dollarization remains limited) even when financial dollarization is widespread.9 Counterparty exposure is also amplified if the value of the collateral backing the loan obligation is denominated in domestic currency and declines relative to the loan, consequent on the exchange rate movement.

In the event of large depreciations, widespread currency mismatches can have macrosystemic ripple effects that compound the deterioration of banks’ financial situation. As a result of balance sheet effects, large devaluations in highly dollarized economies are more likely to be contractionary, thereby further undermining borrowers’ capacity to service their debts. By undermining the solvency of both borrowers and banks, the credit risk deriving from a large devaluation also increases the scope for a credit crunch and heightens the risk of deposit withdrawals by concerned depositors, whether in anticipation or as a reaction to the devaluation. Thus, solvency and liquidity risks are closely interrelated.

Figure 4.Deposit and Loan Dollarization

(Logarithms of ratios in percent)

Sources: National authorities; IMF, International Financial Statistics database; and IMF staff estimates.

Figure 5.Real and Financial Dollarization

(In percent)

Sources: National authorities; IMF, International Financial Statistics database; and IMF staff estimates.

The interaction between prudential risks and the monetary regime, which instills a “fear of floating,” subjects the financial system to risks similar to those incurred under a rigid exchange rate system. The more financially dollarized an economy, the more vulnerable it is to large exchange rate fluctuations, and, hence, the less disposed the monetary authorities are to let the exchange rate float. Concerns about the potential inflationary impact of large exchange rate adjustments can exacerbate the monetary authorities’ “survival” instinct to cling to the exchange rate as a lifeline. In turn, the less exchange rate volatility there is, the more dollarized an economy becomes. Dollarization can thus become a “trap” in which monetary policy ends up hostage to the need to protect the soundness of the financial system and keep inflationary expectations “in the bottle.” Indeed, there is good empirical evidence that both nominal and real bilateral exchange rates are less volatile in more dollarized economies.10 Instead, interest rates must bear the brunt of the adjustment to shocks, raising interest rate risk for both local currency and dollar intermediation. Heightened credit cycles, often a prelude to banking crises, constitute another key source of risk. Credit booms are accentuated by the fact that incoming dollar flows—capital flight repatriation, portfolio investment by foreigners, or foreign borrowing by banks—feed domestic lending and, through the banking multiplier, boost dollar intermediation. The tendency of real exchange rates to appreciate accentuates the cycle by reducing the burden of dollar debt and enhancing the value of collateral in the nontradable sector, thereby relaxing credit constraints.11 The scope for intervention by the monetary authorities is extremely limited.12

Figure 6.Dollarization and Volatility of the Bilateral Real Exchange Rate to the U.S. Dollar, 1990–2001

Sources: National authorities; IMF, International Financial Statistics database; and IMF staff estimates.

Once an exchange rate collapse occurs, the impact on banks’ solvency can be devastating. As long as the prevailing monetary regime remains unaltered, borrowers should choose the currency composition of their debt in a way that minimizes their exposure to risk. Thus, the risk from dollar lending mainly originates from a sudden, large change in the monetary and exchange rate regime, typically a large depreciation taking place after years of controlled and stable exchange rates. The stronger and more pronounced the monetary authorities’ commitment to a peg, as in the case of a currency board, the more damaging it is to break this commitment. At this juncture, time inconsistency and moral hazard arguments clearly take center stage. The maintenance of a stable exchange rate is perceived as an implicit commitment of the monetary authorities, and reneging on this commitment is a “catastrophic” event that warrants government intervention to limit the resulting economic disruption.13 Thus, expectations of a widespread bailout in the event of an abrupt exchange rate depreciation pervade the portfolio choices of depositors, borrowers, and banks, encouraging a level of financial dollarization above what would be socially desirable.14

Against this background, pegs or quasi-pegs that are not perceived to be fully sustainable can encourage dollarization by enhancing the value of the government guarantee. In particular, in strongly anchored regimes, economic agents must, by design, be assured that the fix will hold indefinitely. This may preclude the development of prudential regulations to properly manage exchange-rate-induced credit risks and maximizes incentives to borrow in foreign currency. Even if banks and borrowers remain aware that the current regime could collapse, they expect, and feel they deserve, to be bailed out if it does.

The dollarization of public debt can be an important collateral source of financial fragility when banks have large holdings of public securities.15 Sharp exchange rate depreciations can undermine the sustainability of the public debt, in turn undermining the solvency of banks when they hold large volumes of public securities. This can affect financial systems that, because of administrative restrictions, are not themselves dollarized.

    Other Resources Citing This Publication