Comments on Chapter 4

Adrián Armas, Eduardo Levy Yeyati, and Alain Ize
Published Date:
July 2006
  • ShareShare
Show Summary Details
Kevin Cowan

I want to focus my comments on the Nozaki and Rennhack paper. In particular, I would like to do three things. First, offer a very brief summary and discussion of their main results. Second, complement the facts presented in the paper with some additional empirical characteristics of dollarization in Latin America. Finally, discuss the policy implications of the paper’s main findings both for the optimal choice of monetary and exchange rate policies and for the specific regulatory changes in the financial sectors of Latin American economies.

Summary and discussion

The first very valuable contribution of the paper is to update existing data on the level of bank dollarization into 2004.1 By doing so the authors are able to document an important new development in dollarization in Latin America. Unlike the 1990s, which saw rising dollarization levels in most countries, the share of foreign currency deposits (loans) in total loans (deposits) has stabilized or has fallen in all but two countries in the region in the current decade. Granted, the level of dollarization is still high in many countries (most notably Bolivia, Nicaragua, Peru and Uruguay) but at least there is a reversal in the direction of change.

The second contribution is to extend existing cross-country empirical work on the determinants of dollarization. The paper does so in two directions. The first is to incorporate additional macroeconomic determinants of dollarization; the second to exploit time variation in dollarization levels (and their determinants) using a panel approach.2 Their main results confirm previous findings: in countries where inflation has historically been volatile vis-à-vis the real exchange rate, dollar debt is high, suggesting that it is used as partial insurance against future inflationary shocks. Next, the paper finds a high degree of persistence in the levels of financial dollarization. The paper also provides evidence that recent fiscal policy (as measured by the central government deficit), exchange rate asymmetry (measured as a bias towards depreciation) and currency mismatches all have significant positive effects on dollarization. Finally, the authors find no effect of the exchange rate behaviour (as measured by the Calvo and Reinhart fear of floating measure) on dollarization.

The model behind the empirical specification is a portfolio model, in which agents choose the optimal share of dollar assets (liabilities) based on the relative variances, and covariance of inflation and the real exchange rate. Since financial contracts are forward-looking, portfolios are determined by agents’ beliefs regarding the movement of these two variables. These beliefs are not observable, which leads to a choice of variables with which to proxy (or infer) the beliefs.

The initial approach followed in the paper (and indeed the rest of this literature) is to infer the population variances and covariances from a sample spanning the last ten years and construct a MVP based on this statistics. This approach gives the same weight to all observations in the sample (what happened ten years ago is the same as what happened last month), and no weight to observations outside the window. In addition to the constructed MVP, the authors add additional contemporaneous policy outcomes to the regression, which in principle allows for recent events to have a differential impact on dollarization decisions. It is in this vein that current exchange rate flexibility and inflation variables are incorporated. In turn, the simultaneous inclusion of government spending variable would imply that a government deficit will have an effect on future relative price uncertainty not captured by the history of inflation and real exchange rate movements nor by current inflation levels.

My first (and main) concern regarding this approach is in the interpretation of the empirical results. Does a non-significant coefficient on the current exchange rate regime imply that the choice of exchange rate regime does not matter for dollarization? The answer is clearly no. The only implication is that it has no additional effect beyond that which may be already captured in the MVP. For example, if a country has followed a similar exchange rate regime in the last ten years, then is should come as no surprise that the current exchange rate regime has no significant effect. The same applies to the level of inflation and government deficits.

An alternative approach would be to factor the effects of regime changes rather than regime levels into the analysis. One would expect a smaller effect of the historical MVP on dollarization in a country that has recently undergone important institutional reforms – be it fiscal reforms or the establishment of an independent central bank, since in this case past outcomes are poorer predictors of future outcomes.

This leads me to my second comment. Why not explore the policy determinants of the MVPs directly? In particular, it would be extremely interesting from a policy perspective to know whether MVPs are higher in countries with high average inflation. How exchange rate regimes (de jure or de facto) impact MVPs is another key question that could be explored in this framework.

Third, what is the correct measure of variances and covariances in the CAPM model? The variance of total inflation (or real depreciation) or the variances of the forecast errors? If uncovered interest parity holds (as in the Ize and Levy Yeyati [2003] model), then the model suggests that all expected changes will be factored in to the interest rate differential, leaving only the unexpected component as a determinant of the optimal portfolio.

Next, I have some specific concerns regarding the construction of the three exchange rate measures. First, following the MVP framework, it is not clear why the relative variance of the interest rate and reserves vs. the exchange rate should be a determinant of dollarization. If anything, one could argue more clearly for reverse causality, with dollarization determining fear of floating. This is in fact the specification that Calvo and Reinhart (2002) estimate in their paper. Second, I would argue that the measures of skewdness and asymmetry of the nominal exchange rate should be consistent with the measures of variance and covariance used in the MVPs, and hence should be based on the same ten-year sample. Third, it is not clear why the measure of exchange rate asymmetry should affect dollarization levels. The authors argue that an asymmetrical exchange rate, i.e., one with many depreciations and few appreciations is an incentive for dollarization. However, this should be factored in to the expected depreciation, and hence to the interest rate parity condition. For asymmetry to matter we would need to depart from the simple CAPM setting to a model in which depreciations have a different effect than appreciation. This is an interesting (and unexplored) avenue for additional research.

Finally, as this is a policy volume, I would like to flag some empirical results obtained elsewhere in the literature that are particularly relevant for the policy debate on dollarization in Latin America. The first is that widespread use of domestic price indexation has a significant negative effect on the level of financial dollarization (Ize and Levy Yeyati, 2003). Hence, the establishment of credible price indexation mechanisms appears as a plausible way of reducing financial dollarization without generating capital flight. This is an issue discussed extensively elsewhere in this volume. The second is that banking regulation; specifically currency-blind deposit insurance has significant positive effects on dollarization. Controlling for the existence of some form of deposit insurance, dollarization is higher in those countries with currency-blind insurance (Inter-American Development Bank [IADB] 2005).

Additional data

Financial vulnerabilities do not arise from financial dollarization, but from a mismatch between the currency composition of assets (or income) and liabilities. Therefore, when looking at currency mismatches across countries it is crucial to look both at the average levels of dollarized debt and at how this debt is distributed within the economy. With this in mind I would like to complement the data on the level and distribution dollarization provided in this paper with some additional data put together recently by the IADB (IADB, 2005).

Figure 4C.1 summarizes information on the share of total loans (in domestic and foreign currency) going to the tradable sector (agriculture and industry), and the fraction of total loans denominated in foreign currency in the banking system for countries in the Latin America and Caribbean region. A minimum estimate of the level of currency mismatches can be obtained by assuming that all loans absorbed by the tradable sector are denominated in dollars, with the remaining dollar loans being picked up by the non-tradable sector. The presumption is that the tradable sector is better prepared to deal with dollar loans either because it directly exports part of its output or because prices move in step with the exchange rate. The figure shows that in many economies in the region, loans denominated in dollars are considerably larger than the total loans to the tradable sector, suggesting that currency mismatches in these countries are substantial.

Figure 4C.1Mismatch in bank lending in Latin America and the Caribbean (percentage of total loans, 2001)

Sources: Bank superintendencies; de Nicoló, Honohan and Ize (2005); Arteta (2005).

Note: ‘The share of dollar loans over total loans was replaced by the share of dollar deposits over total deposits.

A more detailed picture of currency mismatches can be obtained for a smaller sample of countries for which currency composition data are available at the firm level. Figure 4C.2 shows the median shares of dollar-denominated liabilities in total liabilities for firms operating in both the tradable and non-tradable sectors in 2001 in selected countries in Latin America. The figure shows that firms in the non-tradable sector were highly leveraged in foreign currency debt in countries with high levels of financial dollarization: Argentina, Costa Rica, Peru and Uruguay. The figure also shows that firms in the tradable sector are more highly dollarized than firms in the non-tradable sector in countries with low overall financial dollarization: Brazil, Chile and Mexico. In highly dollarized economies, this is not the case; the gap between tradable and non-tradable dollarization disappears or reverts, as in the case of Costa Rica.

Figure 4C.2Liability dollarization in the tradable and non-tradable sectors (median value, per cent, 2001)

Source: Cowan and Kamil (2004).

Note: Figures in parentheses denote the level of deposit dollarization in each country in 2001.

This last finding has two possible explanations. First, if the share of dollar loans in total loans is sufficiently high, then it will be inevitable that some dollar loans spill over into the non-tradable sector. In other words, in countries where financial dollarization is high, no matter how much banks try to reduce mismatches, debtors from the non-tradable sector end up with debts denominated in tradables, increasing their exchange rate exposure. Second, differences across countries in the degree to which firms match their liabilities to the exchange rate elasticity of their revenues might reflect other important differences in the economic and institutional structure that affect the incentives for firms or banks to hedge. This is an issue that merits additional research.

Policy implications

I would like to finish these comments by discussing the implications of the paper’s empirical findings for policy design in Latin America.

Should dollarization be prohibited? It is common for policy discussion to address financial dollarization with a strong prior: dollarization poses a significant threat for economic stability. The empirical results presented in this paper (and its predecessors) suggest that this view is at best partly correct. While there is growing evidence that financial dollarization does indeed render countries vulnerable to exchange rate shocks, it ignores other shocks – in particular price shocks – and their effects on economic stability. Hence, the effects of the MVP on dollarization documented in this paper suggest that part of dollarization is an optimal decision, in which price risks are traded off against real exchange rate risks. Restricting dollarization fully by regulation is therefore likely to lead to capital flight (de Nicoló, Honohan and Ize, 2005; IADB, 2005) or excessive use of nominal financial contracts. Unless a market failure that leads to excess dollarization is identified, full restrictions will be welfare reducing.

Should economies alter their monetary and exchange rate policy regime to reduce dollarization? In as much as the MVP is a function of the relative variances of inflation and depreciation, then the results presented in this paper indicate that a regime that credibly reduces inflation volatility vis-à-vis exchange rate volatility regime should reduce dollarization. For example, countries that adopt some form of inflation-targeting (IT) regime in which inflation expectations are targeted while the exchange rate is allowed to fluctuate should see their levels of financial dollarization drop. Various additional pieces of evidence corroborate this result. There is evidence that this is what happened in Chile after fully fledged IT was adopted in 1999 (Cowan, Herrera and Hansen, 2005). There is additional evidence that IT regimes contribute to fixing inflation expectations (Mishkin and Schmidt-Hebbel, 2005). However, this paper does not provide direct evidence that changes in policy regimes affect dollarization levels. Additional work on the determinants of MVPs or on the differential effect of calculated MVPs across regimes are needed to complement existing evidence.

An important issue – closely related to the previous discussion on policy regimes – is the speed at which dollarization ratios adjust to changes in monetary/exchange rate policy. One possible interpretation of the high level of persistence of dollarization the authors find in the panel estimations is that adjustment towards the new long-run level of dollarization is very slow. In the interim, a regime in which the exchange rate is allowed to fluctuate will be left vulnerable to exchange rate shocks. This suggests that temporary restrictions may be optimal in the transition period. An alternative interpretation is that adjustment is fast, but the constructed MVP is a poor measure of agents’ beliefs.3 If regime changes alter these beliefs persistently (in ways not captured by the historical MVP), then the estimated coefficient on the lagged dollarization levels is simply capturing a slow-moving omitted variable. This being the case, dollarization is not persistent (in a partial adjustment sense) and will fall if credible regime changes are implemented. Unfortunately, the empirical results presented in this paper cannot be used to separate the two hypotheses. Additional work, which looks at credibility and the formation of beliefs in a more systematic manner, is therefore needed.

Finally, what regulatory changes are needed to reduce dollarization? The paper discusses the role of financial sector regulation in determining the level of financial dollarization. In particular, the paper argues that most bank regulations in the region are blind to currency denomination. Galindo and Leiderman (2005) arrive at similar conclusions in an earlier study. The problem is regulation that appears ‘neutral’ on paper may actually increase dollarization levels. Broda and Levy Yeyati (2003) show that if there is no discrimination against dollar deposits and a relatively high coverage under the deposit insurance scheme, the banking system will endogenously generate an inefficiently high level of deposit dollarization. Existing empirical evidence corroborates this argument by showing that currency-blind deposit insurance has significant effects on dollarization levels (IADB, 2005). This suggests that there is room for reform in Latin America in this area.


Most previous empirical work relied on samples that extended only until 2001.

The model behind their analysis is a CAPM portfolio model, from Ize and Levy Yeyati (2003).

As discussed above, measured MVPs ignore regime changes and their effects on credibility.


    Arteta, C. (2005) ‘Exchange Rate Regimes and Financial Dollarization: Does Flexibility Reduce Currency Mismatches in Bank Intermediation?’Board of Governors of the Federal Reserve System International Finance Discussion Papers No. 738.

    Broda, C. and E.Levy Yeyati (2003) ‘Endogenous Deposit Dollarization’. Staff Report No. 160 (New York: Federal Reserve Bank of New York).

    Calvo, G. and C.Reinhart (2002) ‘Fear of Floating’, Quarterly Journal of Economics, Vol. 117, No. 2, pp. 379408.

    Cowan, K. and H.Kamil (2004) ‘A New Database of Firm Level Dollarization in Latin America’, unpublished, Inter-American Development Bank.

    Cowan, K., L.O.Herrera and E.Hansen (2006) ‘Currency Mismatches, in Chilean Non Financial Corporations’, in External Vulnerability and Preventive Policies (edited byR.Cuballero, C.Calderón, and L.Céspedes) (Santiago: Central Bank of Chile).

    de Nicoló, G., P.Honohan and A.Ize (2005) ‘Dollarization of the Bank Deposits: Causes and Consequences’, Journal of Banking and Finance, Vol. 29, No. 7, pp. 1697727.

    Galindo, A. and L.Leiderman (2005) ‘Living with Dollarization and the Route to Dedollarization’, Inter-American Development Bank Working Paper No. 526 (Washington, DC).

    Inter-American Development Bank (2005) Unlocking Credit The Quest for Deep and Stable Bank Lending (Washington, DC: John Hopkins University Press) Chapter 4.

    Ize, A. and E.Levy Yeyati (2003) ‘Financial Dollarization’, Journal of International Economics, Vol. 59 (March), pp. 32347.

    Mishkin, F. and K.Schmidt-Hebbel (2005) ‘Does Inflation Targeting Make a Difference?’, paper prepared for the Monetary Policy Conduct under Inflation Targeting Conference, 20-1October, Santiago, Central Bank of Chile.

    Other Resources Citing This Publication