FINANCIAL DOLLARIZATION AND DEDOLLARIZATION

Eduardo Fernández Arias

Inter-American Development Bank

First Draft: March 2005

This Draft: April 2005

Abstract: Financial dollarization is a major problem in many Latin American countries

concerning systemic financial fragility; blunt dedollarization measures simply repressing
dollarization may easily fail to solve fragility, at best fostering risky short-term debt and
at worst provoking massive financial disintermediation and crisis. This paper analyzes
the sources of liability dollarization and identifies the failures leading to excess that call
for policy intervention in a portfolio framework. It then derives an analytically sound
multipronged domestic dedollarization program that takes into account the risks of
misdiagnosis and the experience, both successful and failed. This program centers around
the development of good local currency substitutes for dollar debt, such as CPI-indexed
debt.

JEL codes: F31, O16, G11, G28

Preliminary draft
Please do not quote
 FINANCIAL DOLLARIZATION AND DEDOLLARIZATION

There are many reasons why financial dollarization (FD), understood as the holding by
residents of assets and liabilities denominated in foreign currency, has recently been
placed at the forefront of the policy debate. They spring from the fact that substantial FD
inevitably leads to currency mismatches throughout the economy (i.e., residents holding
peso assets and dollar liabilities).

The role of dollarized foreign savings leading to aggregate dollarized liabilities for
residents of borrowing countries and inducing currency mismatches domestically is by
now well understood (“original sin”). However, a more important contributor to FD in
the countries most affected by it, and arguably a problem much easier to fix, which has
not received due attention so far is the dollarization of domestic savings. In fact, the
domestic intermediation of dollarized savings by residents amounts to dollarized
liabilities of domestic agents and currency mismatches for non-tradable firms, the public
sector and families. Alternatively, if dollarized domestic savings are invested abroad
(“capital flight”) then the domestic financing demand they fail to satisfy is met with
additional foreign savings, which being dollarized (“original sin”) reproduce similar
currency mismatches among domestic agents. This paper will pay due attention to the
domestic sources of FD and rebalance the policy battery to address FD where the
problem is more severe.

Why is FD a critical issue? First, the dollarization of liabilities of producers of non-

tradables that accompany any significant FD is a source of concern that has been
highlighted in most of the recent financial crises. Under these conditions, real
depreciation may financially choke firms en masse leading to the demise of viable firms
and deep systemic crisis in the real sector due to extreme instability. Second, to the extent
that public debt is dollarized, the public sector would suffer a similar shock. If debt
sustainability is compromised, fiscal and monetary policy would need to turn procyclical
and its capacity of lender of last resort would be weakened. Third, FD puts the banking
sector at risk of systemic crises (as well as other components of the financial system).
This risk results either from currency risk in the banks’ balance sheets through local
currency on-lending of dollar liabilities (typically deposits) or, otherwise, from increased
credit risk through foreign currency on-lending to producers of non-tradables noted
above.

The above factors are so powerful that end up dominating policy concerns and inducing
perverse reactions that only fuel risks. While the massive currency imbalances throughout
the economy determine the financial system’s exposure to large real devaluations as well
as the willingness of the monetary authorities to use the exchange rate as a shock
absorber, the authorities’ reluctance to let the real exchange rate fluctuate may in turn
foster FD, as dollar debtors anticipate either a stable real exchange rate or, if this strategy
becomes unsustainable, a government bail out.

Due to the above considerations of deleterious macroeconomic effects of FD and that a

2
financially dollarized economy suffers from important external vulnerabilities, there is a
growing concern that unregulated financial markets tend to produce excessive FD. The
center of the FD debate appears to be shifting from a generally passive stance (a
“learning-to-live-with-it” type of approach, focused on the overall strengthening of
prudential norms or the conduct of monetary policy in both currencies) to a more
proactive agenda, oriented to control the incentives that favor dollarization and to foster
the development of alternative local currency intermediation. The debate is now centered
not on whether FD is a problem that would merit strong policy intervention but on policy
feasibility: whether there are effective policy treatments with acceptable side effects.

By and large, those pessimistic on the feasibility of dedollarization policy base their view
on country experiences, rather than theoretical elaboration. The experience is in fact less
than auspicious. Proposals to entice foreign investors with peso claims to address the
original sin problematic have been met with indifference in the market and little traction
in international financial institutions (IFIs): the problem of original sin appears
insurmountable.1 With the exception of Chile and Israel, previous attempts to dedollarize
by introducing local-currency indexed deposits have failed. On the other hand, countries
that avoided dollarization by discriminating against, or directly banning, on-shore dollar
deposits (e.g. Brazil) have seen average duration of domestic financial claims shortening
and off-shore (dollar) intermediation growing. While many Latin American countries
with high FD would be delighted to have Brazil’s milder problems instead, it remains
true that even these seemingly successful experiences exhibit troublesome symptoms of
repressed FD.

Those who believe that dedollarization is in most cases a feasible and worthwhile
economic policy proposition, or at least a policy agenda worth a serious try, contend that
failed dedollarization experiments are no proof of impossibility and should not block the
development of a dedollarization policy agenda because they have been typically carried
out under unfavorable economic conditions by means of forced conversions. Neither
should the paucity of carefully crafted dedollarization experiments be an indication that
dedollarizing is a sterile idea, because success in this endeavor requires a long-term
investment (in costs of setting up a new market and, like with all insurance policies, in
up-front costs for future risk reduction) that governments may be unwilling to make on
their own under favorable conditions, when the status quo is acceptable and politically
safe. The same inertia bias holds true for IFIs, which have not made enough efforts to
promote and share the burden with potentially interested governments.

Analytical and policy research on whether dedollarization should be pursued and, if so,
how, has gained momentum only recently.2 This paper takes stock of the state of the
debate on FD reviewing the theoretical arguments and empirical evidence pertaining to
the causes and consequences of FD, and explores the soundness and possibilities of a
dedollarization agenda, still in its infancy. Concerning dedollarization, it analyzes the

1
See Eichengreen and Hausmann (2003) for such proposal.
2
See, e.g., the papers presented in the IDB Conference on Financial Dedollarization: Policy Options
(downloadable from http://www.iadb.org/regions/re1/events/index1), or, for a more skeptic view, Reinhart
et al. (2003).

3
soundness of dedollarization as a policy objective in light of the underlying drivers of
FD, extracts the lessons from success and failure in country experiences, and derives a
consistent policy program centered on fostering domestic savings in local currency (as
opposed to redressing original sin).

The paper is organized in five sections. The first two sections are an introduction to the
question of dedollarization policy this paper addresses. I first elaborate the concepts of
FD used in this paper and their analytical relevance for diagnosis and policy action
(Section I) and then take stock of what we know about the basic factors driving FD in the
developing world and about the consequences (both positive and negative) of FD on
economic performance (Section II). The theme covered in this second section is by far
the one that has received the most attention in the literature. The remainder of the paper
analyzes FD as a policy concern. Section III makes the case of dedollarization as a
policy objective to reduce excessive FD. Section IV discusses the experience of
dedollarization policies (or the prevention of FD) in developing countries. Finally,
Section V proposes a theoretically grounded financial dedollarization agenda that takes
into account the experience.

I. WHAT FINANCIAL DOLLARIZATION?

The central concern of financial dollarization has to do with dollar assets and liabilities in
the residents’ balance sheets. In contrast with currency substitution, its concern is not
inflation but in solvency, not flows but stocks. Let us start by distinguishing among
alternative concepts of financial dollarization that will be utilized throughout the paper.

Liability dollarization encompasses liabilities of residents denominated in foreign

currency (L*). Liability dollarization is the most interesting concept to analyze in relation
to the consequences of FD to the extent that it is on the head of firms in the non-tradable
sector and on public debt, because in these cases liability dollarization entails a currency
mismatch between assets and liabilities that would give rise to a negative balance sheet
effect in the case of real depreciation. The degree of liability dollarization (l) in this
concept is measured as a share of total debt liabilities (L), be it in the aggregate (more on
this later) or for specific economic sectors (e.g. non-tradable private firms’ debt, public
debt).

l = L*/ L

It is clear that the existence of domestic currency derivatives markets to help allocate
opposing currency hedging needs or capacities would reduce the problem of currency
mismatch for any given overall dollarization l (by separating credit from currency risk it
would reduce liability dollarization in non-tradable sectors and increase it in tradable
sectors). Ideally, liability dollarization would be measured net of currency hedges and
only when it entails a currency mismatch; this crude liability dollarization index l is best
interpreted as a proxy of such ideal measure.

4
Liability dollarization may originate in dollar claims held by residents or by foreigners.
As we will see, the analysis of this lending side is key to understand the drivers of FD
and design dedollarization policies. Domestic financial dollarization encompasses dollar
claims held by residents (D*). The degree of domestic financial dollarization (d) will be
measured as a share of total debt assets D held by residents (the remainder of which is of
course debt assets in local currency). In turn, debt assets D can be claims against other
residents or held abroad (flight capital, amounting to a fraction f of total debt assets D).

d = D*/D
fD = country’s debt assets abroad;

Alternatively, external financial dollarization refers to dollar claims (against residents)

held by foreigners (E*). In line with the “original sin” hypothesis (Eichengreen and
Hausmann, 1999), developing countries do not have access to external financing in pesos
and therefore E* also amounts to total external debt. (Correspondingly, we also assume
that flight capital fD is also held in dollars).

E* = country’s debt exposure;

It is useful at this point to frame original sin as a particular concept of liability

dollarization associated with a particular form of aggregation. In fact, if the whole
country is aggregated for the purpose of computing liability dollarization, thus netting out
dollar assets and liabilities of residents, then liability dollarization would amount to E*-
fD, the net debt position of the country. This is not quite the concept of original sin,
which refers to the debt liability position E*. To obtain that concept the assumption is
that domestic dollar claims against residents can be netted out from their dollar liabilities
but not assets abroad. This exception is consistent with a literal interpretation of capital
“flight”: capital flew and the government cannot lay its hands on it. But it is not clear
why it could or it would have an incentive to offset liabilities with domestic dollar
claims.3 Theory and experience suggest that any individual or sectoral currency
mismatch is a potential concern and that therefore netting out is not a good
methodological approach to the policy issues of liability dollarization in countries where
domestic financial dollarization is significant. Our definition of liability dollarization l
aggregates debtors without any netting out.

FD is widespread in developing countries no matter how it is measured, especially in

Latin America. The typical country in Latin America (simple average across countries)
has a bank FD of around 40% (either in deposits or loans) and a dollar public debt share
of 75%, ratios almost twice as high as in the rest of emerging countries; after the 1998
Russian crisis aftermath bank FD has been increasing while the degree of dollarization in
public debt has been slowly receding (Galindo and Leiderman, 2004). Reinhart, Rogoff
and Savastano (2003) construct FD indexes on the basis of the share of dollar bank
deposits in broad money, the share of dollar public debt in domestic public debt and

3
Witness the recent Argentinean crisis to see how difficult it is to pesify dollar debts at the expense of
domestic dollar lenders (bank depositors).

5
external debt as a proportion of national income. FD in Latin American countries appears
extremely high on every count, especially concerning domestic FD (see table 1).

Equating borrowing and lending in the country, total and in each currency, we can derive
identity relationships linking the various concepts described above:

(1a) L = (1 − f ) D + E *
(Total borrowing = Lending from residents + Lending from foreigners)

(1.b) L* = ( D * − fD ) + E * = D * +( E * − fD )
(Dollar borrowing = Domestic dollar lending + External lending =
= Domestic dollar savings + Net foreign savings)

(1.c) L − L* = D − D *
(Peso borrowing = Domestic peso lending)

(2.a) (1 − l ) = (1 − d )( D / L)

At this point it is worth noticing that the above asset/liability balances apply to any on-
shore financial intermediary which does not take a dollar position. This is typically the
case of the banking sector, which would therefore match peso and dollar deposits to peso
and dollar loans. A popular measure of domestic financial dollarization and of liability
dollarization is precisely obtained from d and l applied to bank deposits and loans,
respectively. As long as D/L remains constant, that is the macroeconomic relationship
between (accumulated debt-intermediated) domestic savings and investment is stable,
liability dollarization l moves along with domestic financial dollarization d (given
original sin as assumed, of course). (For any given degree of domestic financial
dollarization, liability dollarization decreases with domestic savings and increases with
domestic investment.)

It is clear that another source of change in liability dollarization would be the reduction of
the dollarization rate of foreing lending from 100% (making a dent to original sin). It is
useful to consider foreign financial dollarization net of foreign currency hedging offered
(foreigners willing to receive pesos in exchange for dollars would alleviate original sin).
Let e be the dollarization rate of foreign lending (net of swaps), assumed 1 above, and E
be total foreign debt ( E * = eE ) . Then the liability dollarization expression can be written
as:

(2.b) (1 − l ) = (1 − d )( D / L ) + (1 − e)( E / L ) = (1 − d )( D / L ) + (1 − e)(1 − (1 − f )( D / L))

Assuming again that the aggregate stock ratio D/L remains constant, now changes in
liability dollarization l come from changes in domestic financial dollarization d or in
external financial dollarization e. This paper mainly focuses on policies concerning d,
while Eichengreen, Hausmann and Panizza (2002) propose policies to redress original sin
and focus on e. It is clear then that both initiatives are complementary to reduce liability
dollarization.

6
Liability dollarization as a share of GDP (l’=L*/GDP), which is arguably a better
measure of systemic currency mismatch because it incorporates the degree of debt
leverage of debtors, can also be expressed in terms of the same parameters. In fact:

(3) l ' = dD '+ e( L '− D ' ) − (1 − e) fD ' = dD '+ ( L '− D ' ) when e = 1 (original sin)
where D'= D / GDP and L'= L / GDP

In this formulation, liability dollarization l’ also moves with domestic financial

dollarization d (and external financial dollarization e). As long as the (accumulated)
domestic savings and investment ratios D’ and L’ remain constant, it still holds that
liability dollarization depends linearly on domestic financial dollarization d (under
original sin e=1).4 For this reason, in this paper I examine aspects of the empirical
evidence on either one almost interchangeably. Note that constant domestic savings and
investment D’ and L’ imply a constant excess absorption L’-D’, and therefore a constant
net debt asset position of the country ( fD − E *) / GDP , or put differently, that debt
capital outflows fD and inflows E* move in tandem to bring about a given net debt
capital account (as a share of GDP). I will come back to this point.

Both domestic and foreign sources of liability dollarization are important in most
countries, although their relative importance varies. Table 2 shows countries ordered by
their degree of overall FD from high to low (measured as liability dollarization as share
of GDP, l’) and the contribution of domestic and external savings to it based on (1.b). It
considers two bounds for the contribution of domestic dollar savings: a lower bound in
which the foreign source is proxied by gross external savings, thus ignoring the indirect
contribution of domestic savings abroad, and an upper bound in which foreign savings is
proxied by net external savings, which assumes that domestic savings abroad induce
borrowing from abroad for the same amount (consistent with a constant net debt asset
position discussed above). This upper bound is actually underestimated because the
estimate of this net position does not include errors and omissions and other debt assets
abroad not included in reported off-shore deposits.

The impact of alleviating domestic financial dollarization d on liability dollarization l

appears very substantial. A simple regression of l on d across countries (shown in table 3
ordered by overall FD from high to low) reveals that as much as 85% of the cross-country
variation of liability dollarization l can be accounted for the variation in domestic
financial dollarization d, which suggests that the indirect effects of domestic FD through
induced higher exposure to external dollarization (via domestic savings abroad and lower
overall domestic savings) are important. Table 3 shows, for each country, the marginal
elasticity of liability dollarization l to changes in domestic dollarization d based on that
regression. It also shows estimates of the same marginal elasticity to changes in domestic
dollarization d and original sin e based on (2.b).

4
Also as before, for a given degree of domestic financial dollarization, liability dollarization decreases with
domestic savings and increases with domestic investment.

7
It is clear that the domestic contribution to FD is as important as the external
contribution, and actually more important in a number of high FD countries, especially
those with large unreported capital outflows not reflected in table 2. Furthermore, the
high correlation between liability dollarization l and the degree of domestic financial
dollarization d suggests that its reduction can be highly effective in alleviating liability
dollarization. While in many countries the effectiveness of changes in the degree of
original sin e would predominate over changes in the degree of domestic dollarization d
as measured by the elasticities in table 3, policy effectiveness needs to factor in the
feasibility of these objectives. There is consensus in that the degree of domestic FD
appears significantly more amenable to policy treatment and easier to crack than that of
external FD, and arguably a precondition in terms of sequencing for reasons explained
later. Therefore, this paper, centered on domestic financial dollarization, addresses a key
contributing factor to liability dollarization in high FD countries, which, I believe, is the
first node in the critical path for a comprehensive solution.

II. WHAT DO WE KNOW ABOUT FINANCIAL DOLLARIZATION?

This section gives a brief and selective account of what we know about the drivers and
the consequences of FD in order to support the policy analysis in the remainder of the
paper. The reader interested in more details is encouraged to read the excellent review of
Levy-Yeyati (2004) and references contained therein.

What drives domestic financial dollarization? The basic portfolio model

There is wide consensus in that domestic financial dollarization is a coping strategy on

the part of agents to obtain insurance against surprise changes to peso prices, which make
real returns of peso nominal debt very risky. To the extent that the main fear is a surprise
inflationary surge of the kind many developing countries have experienced in the past,
one can interpret this fear as domestic lenders eager to defend the real value of their
savings from dilution by inflation.5 In summary, domestic financial dollarization is a
market adaptation to cope with low quality currencies.

The high and volatile inflation that has been observed in many developing countries over
the years, including episodes of hyperinflation, jibe well with the widespread
development of financial dollarization as financial globalization established itself with a
strong footing and increasingly facilitated it by offering opportunities to save abroad.

However it has been observed that over the past decade there has been a remarkable
reduction in inflation in Latin America while at the same time financial dollarization has
tended, if anything, to increase. Of course, what matters is the expectation about future
inflation (strictly speaking, surprise changes in inflation), rather than current or past
inflation. One way to reconcile this apparent anomaly is to consider that despite the
reduction in observed inflation rates, currencies and monetary policies lack credibility:
bad memories are not easily forgotten and it will take a long time for credibility building

5
This approach calls for a policy analysis of other forms of defense, of which financial dollarization may
be just one, to which we will return in the last section.

8
policies and institutions to be effective (see Savastano, 1996). Another way to explain
the evidence is to consider costs of reswitching to local currency instruments due to
network externalities, although this argument appears more relevant for currency
substitution (Guidotti and Rodriguez, 1992) than for asset substitution. These hysterisis
views of the persistence of FD do not bode well for the prospects of dedollarization based
on monetary discipline (see for example Reinhart, Rogoff and Savastano, 2003).

This paper favors an alternative, more optimistic, view that can be derived from the
portfolio approach introduced in the foundational paper by Ize and Levy-Yeyati (2003).
This portfolio optimization approach allows the translation of the identities in the
previous section to behavioral equations, from which implications can be examined.
While I will return in the next section to the drivers of FD and go behind some of the
proximate causes, it is useful at this point to cover the basics that this basic portfolio
model lays out.

Let us start with the case of domestic lenders (say depositors) and borrowers (say firms)
choosing between contracting in pesos or dollars for the repayment of a given loan.
These choices entail different real return profiles: a peso loan delivers a real return rp and
a dollar loan a real return rd. Let P be the debt repayment for one unit loan from borrower
to lender, S the borrower’s revenue, and π = S − P the borrower’s profit, everything in
real terms. They choose the fraction d of dollars (and 1 − d of pesos) of their debt
financing portfolio, so that

P = (1 − d ) r p + drd .

Being domestic agents, they discount nominal payoffs with the domestic price index, so
that returns are subject to price risks: peso loans are subject to currency risk (surprise
inflation levels) and dollar loans are subject to exchange rate risk (surprise changes to the
real exchange rate). Besides price risks, all loans are subject to default risk, that is the risk
of the borrower failing to pay the contractual amount to the lender (this failure amounts
to a transfer from lender to borrower, which for the moment is assumed costless to the
contracting parties). These underlying risks to real returns correspond to the following
variances: Vcc (currency or inflation risk), Vxx (exchange rate risk), Vdd (default risk).
Currency and exchange rate risks may be correlated (covariance Vcx) but for simplicity it
is assumed that both price risks are uncorrelated with default risk (Vcd=Vxd=0).

Lenders (depositors) are risk averse and are concerned with the return volatility of their
asset portfolio. For simplicity I assume that borrowers (firms) are risk neutral.6 In this
basic model the currency composition of loan repayment, d , is determined by:

(4.a) Max E ( P ) − aV ( P ) (Lenders)

d

(4.b) Max E ( S ) − E ( P ) (Borrowers)

d

6
In this simple model this simplification is not needed for the result; borrowers can be also risk averse.

9
Under the assumption that real revenue S is exogenous (in the next section I will relax
this assumption and consider the case in which revenue depends on FD), joint
maximization efficiency of objective functions (4.a) and (4.b) determines dollarization d
such that the variance of the portfolio is minimized:

(5) Min V (P )
d

Private efficiency leads to choosing the dollar fraction d* ratio that minimizes the
variance of payment (see Appendix):

(6) d * = (Vcc + Vcx ) /(Vcc + V xx + 2Vcx ) = Vcc /(Vcc + V xx ) if price risks are uncorrelated

The main intuition of the equilibrium result is that what matters for FD is not just
inflation risk but the relative risk between inflation and real exchange rates: FD may
coexist with low currency risk if real exchange rate risk is also low.

Could this portfolio analysis be applied to foreign lenders to explain original sin (i.e. a
corner solution d * = 1 )? While some see original sin as a reflection of the same
phenomenon of weak (or outright abusive) monetary policy, Eichengreen and Hausmann
(1999) forcefully contend that other factors such as country size (the currency’s value for
international diversification purposes) are more important and explain why countries with
strong monetary and fiscal policy like Chile also suffer from original sin. However, the
portfolio approach offers an additional explanation: because foreign savers would
discount with a different factor (their own foreign consumption basket), they are not as
aligned as domestic savers are concerning the borrowers’ appreciation of the relative
volatility of real returns of lending instruments. This misalignment with foreigners
induces home bias, as noted by Thomas (1985). From the point of view of foreigners’
real returns, dollar lending would not contain price risk (V’xx=0, abstracting from surprise
international dollar inflation) and peso lending would contain not only currency risk but
also exchange rate risk involved in the conversion of the peso deflator into dollar
deflator. In the context of this basic portfolio model, optimal foreign FD is in fact 100%
and original sin is the natural result of this home bias.

Credit risk imposes costs on borrowing firms which, once factored in, lead to a special
preference for lower risk debt instruments (i.e. those with low probability of putting the
firm’s balance sheet in the danger zone). Short of modeling actual liquidation costs or
near bankruptcy costs to reflect the inefficiency of the financial crunch caused by risky
debt, I assume that firms are risk averse as a reduced-form preference for safer financing.
Assuming for simplicity that they are as risk averse as lenders, joint contracting
efficiency in the basic portfolio model of the previous section would be:

(xx) Max E ( S ) − a(V ( S ) + 2V ( P) − 2Cov( S , P)) , which being S exogenous boils

d
down to:
Min V ( P ) − Cov ( S , P ) instead of Min V (P ) as in (5).
d

10
Whether this additional factor contributes to FD depends on how Cov ( S , P ) behaves as
debt becomes more dollarized; if this covariance is not a function of currency choice as
implicitly assumed above then benchmark dollarization d* would still hold. It is possible
to construct models in which peso debt is riskier than dollar debt from the point of view
of firm’s solvency7. However, the literature on the balance sheet effect of dollarized debt
in non-tradable firms and recent crises in financially dollarized countries strongly suggest
that the relevant assumption to make is Cov d ( S , P) < 0 , consistent with the strong
counterciclicality of the dollar ( Cov (S , P ) becomes more negative as debt is more
~
dollarized). This favored assumption leads to some moderation of dollarization, d < d*,
so that costly credit risk would not be a factor explaining high FD. Each individual firm
anticipates lower revenues in the case of real depreciation and in so doing moderates its
dollar demand in order to hedge that scenario. 8 Ize and Parrado (2002) and Rajan
(2004) assumes in contrast that lenders’ income, which would be positively associated
with revenues S, is hedged by dollar claims (and debtors do not hedge), which would lead
to the opposite result, that is higher dollarization. Whether or not this additional
covariance factor of a consumption-based portfolio model would dominate the previous
demand effect and partly explain high FD is an empirical matter. De Nicolo et al. 2003
finding that GDP hedging makes a negative contribution to dollarization appears to
confirm the moderating hedging effect.

A simple model to analyze this question of dollarization and costly credit risk, to which I
will come back, is the following (solved in the Appendix). There is high inflation with
probability (1 − p ) and high real depreciation with probability e so that returns are as
follows: Pesos, with real return rp (r with probability p and rp l , p l < 1, with probability
(1 − p ) ) and Dollars with real return rd ( r * e h , e h > 1, with probability e and r* with
probability (1 − e)) . For simplicity I assume that returns on peso and dollar claims are
independent (which implies that shared default risk is left aside). Importantly, real
depreciation is positively correlated with recession; for simplicity they occur jointly (with
probability e). Let revenue S be R with probability (1 − e) and R − φ with probability e,
φ ≥ 0 . Then Cov (S,P)=- φedr * (1 − e)( e h − 1) and in fact market dollarization is
~
moderated due to hedging ( d < d* ).

Let us now consider a financially open economy (but subject to original sin). Domestic
lenders have a choice of off-shore investment (in dollars) not subject to country risk. Its
return, rf, is therefore only subject to currency risk (with variance V XX ). Then the
portfolio problem entails three choices: domestic peso, domestic dollar, and, now, off-
shore dollar. Let d denote the fraction of dollar claims in the portfolio as before (now the
7
Jeanne (2002) is able to construct a special model in which this is so because of a large peso premium.
Ize and Powell (2004) posit the assumption that real project returns for the representative firm increase with
real exchange rate depreciation, an assumption which appears optimistic, if not down right counterfactual.
Empirical work is needed to settle the relevance of these assumptions to explain high FD.
8
The microeconomic evidence in Galindo, Panizza and Schiantarelli (2003) confirms the existence of
(partial) hedging via the currency composition of the stock of debt.

11
sum of domestic and off-shore dollar) and f the fraction of off-shore claims in the
portfolio:

P = (1 − d ) r p + ( d − f ) rd + fr f (with d ≥ f , so that no short positions exist).

On the other hand, borrowers have a choice to borrow from foreign lenders (also in
dollars). Foreign lenders are not explicitly modeled, but they are assumed to add a risk
premium k on top of their alternative return at home on account of country default risk.
Let r denote the (exogenous) expected real cost of foreign financing for a domestic
borrower. Then non-arbitrage implies that the expected return of off-shore investment
(for a domestic lender) is r − k . Furthermore, the non-arbitrage conditions across
financing options for borrowers imply that E ( r p ) = E ( rd ) = r . In these models borrowers
always pay r in expected value to all lenders, domestic and foreigners. Therefore,
presumably, overall domestic savings increase with r and overall domestic investment
decrease with r.

Then the lenders’ portfolio problem to determine dollarization d* and offshorization f*

becomes:

(7) Max E ( P ) − aV ( P ) = r − kf − aV ((1 − d ) rp + ( d − f ) rd + fr f )

d, f

It can be checked (see Appendix) that the optimal dollarization d* to minimize the
variance of payments is the same as before, the key result in Ize and Levy-Yeyati (2003),
that is to say, the previous result continues to hold for the degree of dollarization d* of
total deposits, but part of them are made off-shore.

This portfolio model has been empirically tested by the authors to satisfaction to explain
domestic financial dollarization using historical variances and covariances to proxy
expected volatilities for the computation of d*. (Empirically, they further note that in the
countries in their sample in which indexed peso instruments provided a good alternative
to nominal peso instruments (namely Chile and Israel), actual financial dollarization was
far below d*. I will return to this prediction of an extended portfolio model encompassing
indexed debt instruments in the last section).

The key insight in this basic portfolio model is that the decline in inflation volatility need
not lead to a decline in FD if there is a concomitant decline in real exchange rate
volatility. Exchange-rate based stabilization in Latin America in the 1990s illustrates
such a case. More generally, fear of floating associated with high FD appears to be
behind real exchange volatility declining even more than inflation volatility, which led to
an increase in latent dollarization d* partly explaining the persistence of observed FD.
This more elaborate explanation of the observed trends assigns better prospects to a
dedollarization strategy based on monetary discipline (say inflation targeting) combined
with flexible exchange rates.

12
As to offshorization, the optimal fraction of total deposits made off shore is f*, which
depends on country default risk (see Appendix):

(8) f * = 1 − k /(2aVdd )

(The above results for d* and f* hold for an interior solution in which there is domestic
dollar lending, that is d*>f*. Given the high demand for domestic dollar savings in
countries with high FD, this appears to be the most interesting case to focus on.
However, I will come back to this to review the implications of the alternative case of a
corner solution.)

Impediments to Financial Dollarization: Restricted portfolio models

What would it happen with portfolio allocations, that is dollarization and osffshorization,
if some of the choices in the model above are impeded or restricted, exogenously or by
policy? I now show how the above model would be altered.

An increase in the real cost of borrowing r, be it because dollar real international interest
rates rise or because there is an expectation of real exchange depreciation, would have no
portfolio effect on either dollarization d* or offshorization f*. Restrictions to foreign
savings (a tax on capital inflows or, in the extreme, a sudden stop) increase on-shore
returns of domestic savings (it is equivalent to a subsidy to both peso and on-shore dollar
savings in the model below) and therefore amount to widening of the return gap between
on-shore and off-shore dollar deposits. In the context of this portfolio model, from an
incentive point of view a tax on foreign savings amounts to a tax on off-shore domestic
savings. In terms of the model above, it amount to an implicit increase of the risk
premium k, which would reduce offshorization f* (assuming that underlying default risk
Vdd remains constant), but would not alter dollarization d*. Similarly, direct restrictions to
domestic savings abroad would not have any effect on dollarization d*, which holds for
any level of offshorization f. In this model, (effective) restrictions to capital flight would,
as in the previous case, only shift dollar savings from abroad to domestic lending without
producing additional peso savings.

Liability dollarization would be altered only to the extent that the level of domestic
savings and investment changes, which remains outside this portfolio model (presumably,
higher cost of borrowing and restrictions to capital inflows would reduce FD through less
investment and more domestic savings, and restrictions to capital outflows would
increase FD through less savings). In this model, therefore, changes to the real cost of
foreign or dollar borrowing and restrictions to international financial flows would not
have an impact on FD on portfolio grounds, beyond their macro effects on the overall net
financial position of the country. (In all the cases reviewed in the last paragraph there
would be less reliance on foreign savings and less foreign debt, but that is not particularly
relevant or advantageous in this simple model.)

By contrast, discouraging domestic dollar claims (say taxes on on-shore dollar deposits)
and encouraging domestic peso claims (say subsidies on peso deposits) would have a

13
beneficial portfolio effect on FD. A tax t on domestic dollar deposits imposes a wedge in
the expected return between lender and borrower: the equilibrium borrowing cost r
translates into an expected real return on domestic dollar claims of r-t. A subsidy s on
peso deposits also imposes wedge: the expected real return on peso claims is r+s. The
portfolio problems then become:

(9.a) Max E ( P ) − aV ( P ) = r − td − ( k − t ) f − aV ( P ) with tax t on-shore dollar deposits

d, f

(9.b) Max E ( P ) − aV ( P ) = r + s − sd − kf − aV ( P ) with subsidy s on peso deposits

d, f

It is easy to check (Appendix) that a tax on domestic dollar claims and a subsidy on peso
claims of the same rate (t=s) have the same effect in reducing FD:

(10a) d’(t)= xxx

(10b) d’’(s)=xxx

Taxes and subsidies to domestic claims are not equivalent in this portfolio model,
however. A subsidy on peso claims does not alter the degree of offshorization f*, which
depends on the relative return between on-shore and off-shore dollar claims (assuming a
subsidy small enough so that an interior solution still obtains, that is d’’(s)>f*). A tax on
on-shore dollar claims, however, does alter this relative return and therefore increases
offshorization (again, assuming a tax small enough for an interior solution, that is
d’(t)>f’(t)):

(11.a) f ' (t ) = 1 − (k − t ) /( 2aVdd ) > f *

(11.b) f ' ' ( s ) = f *

The conclusion is that from a portfolio incentive viewpoint both a subsidy on peso
savings and a tax on on-shore dollar savings are equivalent concerning FD if they are not
too large. On-shore taxes on dollar claims, however, encourage off-shore dollar savings
and, therefore, higher external debt. Note that off-shore dollar savings f sets a floor to
dollar savings d (all capital flight is in dollars), and therefore a ceiling to the degree of
domestic financial dedollarization. This difference between the two policy instruments in
relation to incentives to capital flight implies that while a subsidy on peso savings may
minimize dollarization to d’’(s)=f* with a sufficiently large subsidy, a tax on on-shore
dollar savings is more limited as an instrument to achieve financial dedollarization.9 This
dedollarization limit when d’(t)=f’(t) can be easily found by considering the case of a ban
on domestic dollar savings, obtained by imposing the condition d=f (offshorization is the
only way to save in dollars), solved in the Appendix.

9
Nevertheless, a tax on foreign savings amounts to a tax on offshorization, and therefore could be used as a
policy instrument to enlarge the scope of dedollarization up to the theoretical d=0. Of course if effective
impediments or direct restrictions to capital flight were imposed at the same time, then in this model
dedollarization could also be pushed to the limit. But in the context of this portfolio model, policies on
international capital flows play at most a supporting role in dedollarization policy.

14
A ban or prohibitive tax on on-shore dollar savings yields the following minimum
dollarization level (equal to the new offshorization level):

(12) d * * = f * * = (Vcc + Vcx + f * Vdd ) /(Vcc + V xx + 2Vcx + Vdd )

It can be checked that in fact f * < f * * = d * * < d * (under the maintained assumption
that the basic portfolio model entails positive on-shore dollar savings, d * > f *). In other
words, a ban on onshore dollarization leads to some dedollarization, but part of the
onshore dollars find their way into increased offshorization. This theoretical result
appears to be confirmed in IMF (2005), where it is shown that in countries with banned
onshore dollarization there are sizable offshore deposits but, overall, bank dollarization is
lower. This result is also consistent with Cowan, Kamil and Izquierdo (2004), who
found that the variance of the currency mix (and a fortiori any restriction to minimum
variance mix d*) negatively affects financial development (EXPAND to show that a ban
has more side effects the more valuable dollarization is in terms of variance
minimization).

Some of the qualitative implications emerging from these augmented models have been
further confirmed econometrically by De Nicolo et al. (2003), where they control for
administrative restrictions to domestic dollar deposits and test their effects. One
important finding is that credibility, proxied by institutional and regulatory variables as
well as the macroeconomic environment, is relevant and in fact explains half of the cross-
country variation. In my view this additional driver should not detract from the relevance
of the portfolio model, but rather should be interpreted as a correction term to adjust the
measured historical volatilities to arrive at expected volatilities.

Finally, it is important to highlight that this portfolio approach also allows us to think
about other factors connected to FD in an integrated fashion. For example, other portfolio
choices such as the choice of maturity can be studied alongside. Being a short maturity
peso loan an alternative way to obtain protection against surprise inflation, it is a natural
substitute to a dollar loan. The evidence in IDB (2004) showing that dollarization is
prevalent in long maturity bank loans and deposits clearly confirms this conjecture. A
fortiori, short maturity financing, with all its attendant price and rollover risks, can be
expected to expand if dollar deposits were banned. This is a concern rightly emphasized
by De la Torrre and Schmukler (2003) as a potential pitfall of a narrow dedollarization
strategy. At the same time, equity financing is another hedge against surprise inflation
and therefore, presumably, also a subsitute of dollar financing without the obvious
downside of short-term debt. Empirical work on this link between dollar debt and equity
financing is sorely missed.

What are the aggregate effects of financial dollarization?

There is the widespread concern that financial dollarization is socially costly. There has
been a traditional concern with the effectiveness of monetary policy in the context of high
FD. Recent crises, however, have pushed the risk of crisis and prudential concerns to the
forefront. The empirical evidence strongly supports the concerns, especially that found

15
in De Nicolo, Honohan and Ize, NHI, (2003), Reinhart, Rogoff and Savastano, RRS,
(2003); and Levy Yeyati, LY, (2003). A summary of the main empirical findings
follows:

FD complicates countercyclical policies. There is a well established link between fear of

floating and liability dollarization, that is fear of a real exchange depreciation due to
currency mismatches that may bring insolvency and financial crisis (see Calvo and
Reinhart, 2002; Levy Yeyati, Sturzenegger and Reggio, 2002). RRS empirically confirm
this link. This fear limits the scope of countercyclical monetary and fiscal policy.
Furthermore, the monetary transmission based on peso rates is weakened by the
prevalence of dollar rates (Balino et al. 1999).

FD substantially contributes to financial fragility and the risk of systemic crisis.

Widespread currency mismatches, largely unmitigated by hedging derivatives whose
markets are fairly underdeveloped for most local currencies, increase solvency risk of
debtors (including the public sector) and, consequently, of the banking system even if
currency matched. Liquidity risk in the banking sector is further increased by the lower
ability of the Central Bank to perform as a lender of last resort in foreign currency.
Systemic crisis may also result from fiscal risk due to dollar public debts, which are quite
high in a good number of countries. The reader is referred to NHI for strong evidence
concerning solvency and liquidity effects on banks, as well as Gulde et al. (2004),
Goldstein and Turner (2002), and IMF (2005). Calvo, Izquierdo and Mejia (2003) find
that liability dollarization is a predictor of sudden stops in capital inflows, which is an
indirect indication of the systemic fragility it induces.

FD has an overall negative effect on output volatility. To the extent that financial
dollarization complicates countercyclical policies and contributes to crises, a negative
effect on output volatility is to be expected. This prediction is confirmed by the
econometric analysis in LY and agrees with the evidence shown by RRS concerning the
association between their overall dollarization index and output volality. This effect is
also confirmed by microeconomic evidence in a number of Latin American countries
which shows the destabilizing effect of liability dollarization in a good fraction of firms
in terms of profits and investment, possibly leading to contractionary devaluations (see
Galindo, Panizza and Schiantarelli, 2003 and references contained therein).10

FD appears to have some negative effect on the inflation level and volatility. By and
large, cross-country evidence shows that the higher the degree of dollarization, the higher
and more variable the inflation rate (notwithstanding the widespread improvements in
inflation performance over the past decade as financial dollarization increased). However
these effects do not appear to be strong. Furthermore, FD does not appear to have
complicated disinflation policies. The reader is referred to RRS for information and a
complete non-econometric analysis of FD and monetary policy and to LY for some
econometric evidence pointing in this direction.

10
Depreciations are expansionary with low levels of dollarization and contractionary with high levels of
dollarization.

16
At the same time there is a widespread premise that financial dollarization contributes to
deeper financial systems (provides a credit instrument of good quality) and, hence, the
growth level. Surprisingly, recent econometric evidence is mixed and suggests that
impediments to financial dollarization need not have adverse effects if carried out with
adequate policies under the right circumstances. The following two empirical findings
stand out:

FD contributes to financial depth only under high inflation. Being financial dollarization
a coping strategy for market agents to be willing to agree on credit contracts, in the
absence of such instrument many of these contracts may not be carried out. A reasonable
presumption is that part of these savings would be frustrated or would be invested abroad,
legally or illegally, and therefore there would be financial disintermediation and a smaller
domestic financial system. However, NHI established the important econometric finding
that FD does not appear to contribute to financial depth unless under circumstances of
high inflation, which is confirmed with a variety of econometric methods to control for
endogeneity bias that would result from adverse unknown conditions causing both higher
FD and less intermediation. LY also obtains this key finding with his own FD measures
and specification using a dollarization restrictions index as instrument, providing
confirming evidence that the negative effects of impeding dollarization with restrictions
depend on inflation circumstances.

FD does not appear to contribute to faster average growth. This finding results from LY
econometric analysis of the issue on the basis of a Barro-type regression enlarged with
financial dollarization and is also consistent with the simple association analysis in RRS.
This result can be explained in part because as shown FD does not necessarily improve
financial depth and in part because the increased output volatility associated with FD may
in itself depress average growth (Ramey and Ramey, 1995). Nevertheless, the fact that
LY controls for investment opens the possibility that FD contributes to faster growth
through larger investment on the basis of better access to credit. It appears advisable to be
cautious at this point with this preliminary finding and conclude that there is no evidence
that FD contributes to faster long-run growth.

III. DEDOLLARIZATION AS A POLICY OBJECTIVE: IS FINANCIAL

DOLLARIZATION EXCESSIVE?

The empirical analysis of the consequences of FD strongly suggests that some degree of
dedollarization is a valid policy objective, since the evidence clearly points to important
costs which do not appear to be offset by any clear advantage. In fact, such consensus is
visibly emerging. Still, we should be cautious with this conclusion because most of the
costs are ex-post (say losses after a crisis) and therefore quite noticeable, while potential
advantages such as financial depth are ex-ante, less visible. Furthermore, if the only
basis for analysis is a set of empirical regularities, it is easy to end up attacking symptoms
or to miss the incidental costs associated with the benefits of dedollarizing. For all the
aggregate costs FD may entail, it may still be the best deal feasible.

17
If we accept the basic portfolio model as the key driver of FD, being a market outcome
between lenders and borrowers it does represent the best deal in individual contracts
given the circumstances. Conditional on the volatilities the agents perceive, they are in
fact arriving at the best portfolio (privately and socially). The empirical evidence
concerning the relationship between estimations of benchmark dollarization d* and
observed dollarization across countries leaves ample room for externalities and suggests
that they may be very substantial (Ize and Levy-Yeyati 2005), but there is little
empirical work identifying the factors behind these deviations. For dedollarization to be a
valid policy objective requires an explanation of the failures that distort the relevant
relative volatilities, either in agents’ perceptions (externalities) or in actuality (unduly
weak policy-sensitive fundamentals), and make private contractual efficiency to deviate
from the social optimum. Furthermore, an analytical approach is needed for identifying
the type of policies which appear to be most suitable in each case, that is to say, most
effective to eliminate socially excessive FD at minimum cost. This section makes the
case for these failures and explores the nature of the corresponding policy instruments.

Failures come in two forms: by “error” and by “omission”. Failures by “error” consist in
the presence of interferences that distort the signals perceived by agents. Adequate
remedies in this case consist in the removal or the compensation of the distortion. In
what follows I distinguish between market externalities and what I will call policy
externalities. Failures by “omission” consist in the absence of a sufficiently fertile
contractual environment for agents to arrive at better (private and social) outcomes.
Adequate remedies in this case are the provision of the missing market infrastructure to
make the environment functional, such as the provision of missing markets and, more
generally, the reform of policies and institutions underlying financial markets.

Market Externalities

High FD leads to systemic financial fragility because it is both powerful (real exchange
rate depreciations can be substantial, especially in this era of international financial
turmoil and sudden stops) and widespread across agents, which are then hit at once.
Recent crises have clearly illustrated the potential for aggregate economic collapse when
financial and real networks among firms and banks are disrupted systemically (Calvo and
Fernandez-Arias, 1998). In fact, it is this coordination mechanism what has substantially
increased the risk of liquidity crisis in highly dollarized economies, as the ripple effect of
a solvency shock on the balance sheets quickly becomes sufficient precondition for a
self-validating liquidity run. The failure of contracting private agents to internalize their
contribution to aggregate FD and the consequent systemic inefficiency costs is arguably
the most important reason for excessive FD in the market and yet it has received very
little attention in the dedollarization debate (most of the enquiry has focused on policy
failures that induce excessive FD).

The risk of systemic crisis produced by aggregate FD as a vehicle for a systemic shock
and a coordinating device is not internalized by private agents in our simple portfolio
model. One way to see the excess dollarization that results from this externality is to posit
firms whose revenues decrease when the other firms suffer financially, so that, ceteris

18
paribus, in the case of real depreciation the revenue of each firm declines more the higher
aggregate FD. This assumption would also correspond to the empirical regularity that in
developing countries the real value of the dollar tends to be countercyclical (increase
during crises).

Let us come back to the model in the previous section in which the firm has a preference
~
for safer financing and Cov (S,P)=- φedr * (1 − e)( e h − 1) , which led to dollarization d . I
now assume further that because individual revenues are negatively affected by the
financial fragility of other borrowers, including contagion induced by systemic banking
crisis, the depth of the recession depends on overall FD: φ = φ (d ), φ ' > 0 , where d is
aggregate financial dollarization. For example as in Aghion, Bacchetta and Banerjee
(2001), where a self-fullfilling currency and banking crisis obtains through the dynamics
of incomplete pass-through once FD reaches a critical level. The question is whether
~
observed dollarization d is excessive.11 The central planner realizes that revenue S is
endogenous, internalizes aggregate FD d and considers:

(14) Min (V ( P ) − Cov ( S , P )) − (1 / 2)( E ( S ) − aV ( S ))

d

The marginal condition to determine dollarization in (14) contains new terms further
moderating dollarization in order to protect the economy from recession, diminish the
variance of output, and improve cyclical hedging (see Appendix). Optimal dollarization
~ ~
is dˆ < d (market dollarization d exceeds optimal dollarization d̂ ). Individual
moderation is not enough because it does not internalize its contribution to aggregate
FD.12 In this case the best suited policy instrument would be a Pigouvian tax wedge in
favor of peso financing and against dollar financing to offset the marginal effect of the
externality and bring the market to pick lower socially optimal dollarization d̂ ). The rate
of this tax, however, ought to depend on the firm’s individual default risk margin across
currencies, which is relevant for the size of its potential contagion externality.

Another aspect of market externalities associated with FD to which little attention has
been paid relates to the risk of turbulence and disruption in international financial
markets, or risk of sudden stop for short. This concern is at the center of the debate of
recent crises. Sudden stop or external credit rationing would be realized at the time of
rolling over debts, so it is a risk absorbed by borrowers that cannot be explicitly modeled
in this simple static framework. We can implicitly model this link by positing that φ
above is a function of aggregate foreign debt stock E*: the higher this stock the more
likely it is that there would be a financial disruption leading to output losses throughout
11
In the previous section I obtained that observed dollarization would be lower than the benchmark d*. If
lenders’ hedging dominate, as in Rajan (2004), the opposite would be true. This empirical question is not
important for the issue at hand; what matters for policy is not how dollarization compares with the basic
benchmark but how it deviates from the social optimum on account of the externality analized in this
section.
12
Notice that with a crisp diagnosis like this of where is the first best we would not need to evaluate “side
effects” of dedollarization such as disintermediation. Intermediation is excessive.

19
the economy.13 This loss risk associated with foreign debt results from the increasing
probability of a sudden stop as country exposure mounts and from the increased damage
in the event of a sudden stop from a higher exposure to debt not being rolled over

To highlight the specific aggregate and foreign characteristics of this risk (see Tirole,
2002 for an elaboration in terms of dual agency theory), let’s neglect any link with
individual loans (in reality, borrowers contracting with foreign lenders may be more
affected) and with domestic lenders (in reality, the liquidity of capital flight abroad may
cushion the blow of a sudden stop). We are then left with the pure externality of
individual borrowers not internalizing their contribution to the aggregate risk of sudden
stop when contracting foreign debt. This problem has been studied by Fernandez-Arias
and Lombardo (2002) and more recently by Wright (2004) in dynamic models with
identical agents where the externality induces intertemporal competition for borrowing,
leading to excessive borrowing due to a commons problem. Caballero and
Krishnamurthy (2003) have shown that in the presence of heterogeneous agents,
imperfections in the domestic financial system not allowing the allocation of the
aggregate risk to those more able to bear it also leads to excessive foreign borrowing.

This simple static model would call for a tax on foreign borrowing to internalize the
aggregate risk. These tax wedges would increase the home bias premium for domestic
savers who would invest less abroad (less capital flight) and absorb more country risk.
However, while the degree of offshorization f* of domestic savings would diminish (due
to the increase in domestic returns favored by the tax on foreign savings), in principle
domestic FD would not change. If domestic savings and investment remain constant, then
smaller capital outflows and inflows would go hand in hand, that is there would be no
change in the capital account of the balance of payments, and both liability dollarization
l* and domestic financial dollarization d* would remain. Of course, the country would
have a smaller foreign debt and, to the extent that there is a specific risk associated with
aggregate foreign debt, then the structure of resulting FD would be not as harmful.

Finally, there is the intriguing possibility raised in Chamon (2001) and Broda and Levy-
Yeyati (2003) that dollar financing displaces peso financing because of a coordination
failure among creditors to share bankruptcy partial payments. In fact, limited liability
implies that in a state of default partial payments are shared according to a rationing rule.
According to bankruptcy law, within each creditor class receiving partial payment the pie
is shared according to total due to each creditor. Since dollar debts tend to increase in
peso terms due to (nominal) surprise depreciation associated with default, then dollar
financing would have an advantage over peso financing in default states. This asymmetry
would not lead to any distortion if the financial terms of lending take full account of the
borrower’s balance sheet, in particular of the currency mix of its liabilities, in what would
amount to contingent contracts. In fact, the debtor has an interest in enabling this efficient
solution by coordinating creditors through commiting to a given borrowing currency mix.

13
Even the mere misallocation of capacity to pay foreign debt within the domestic economy due to
inefficiencies in the domestic financial system (e.g. lack of derivatives market to allow exporting firms to
“lend” their internationally valuable collateral to non-tradable firms taking on dollar debts) leads to
excessive aggregate foreign debt (Caballero and Krishnamurthy, 2003).

20
However if this coordination is not feasible, as it appears realistic to assume, then it is
easy to show that in equilibrium there is full dollarization, because otherwise a switch
from peso to dollar lending at going (uncontingent) rates leaves a rent to dollar lenders
from the sunk peso lending (in the same and more junior classes) carrying less default
payments after the switch.

The key question is how relevant is this distortion, that is to what extent the pricing does
not internalize ex-post disparities and restore ex-ante parity across currencies. Specific
studies ought to look at the influence of default risk on FD, controlling for reverse
causation. In their absence, the available evidence suggests that this effect may be
substantial. Microeconomic data suggest that more leveraged firms (a proxy for riskier
firms) are also more liability dollarized (see for example Galiani, Levy-Yeyati and
Schargrodsky, 2003). Macroeconomic data suggest the same thing: across countries in
table 3, the degree of liability dollarization significantly increases with total liabilities (as
a share of GDP) according to an estimating equation l=0.50+0.27L’. In theory, a tax
wedge in favor of peso financing would erode the monopoly of dollar financing once it
gets to a critical level, and if appropriately calibrated as a function of default risk could
redress this externality. However, it appears much simpler to change laws and regulations
pertaining to bankruptcy eliminating the opportunistic advantage of dollar financing in
default states by some appropriate pre-bankrupcty rate of conversion to pesos for the
purpose of bankruptcy. In this sense, this market coordination failure can also be
interpreted as a policy failure concerning limited financial liability.

Policy Externalities and Moral Hazard

The literature has mainly focused on policy failures that induce excessive FD, including
biased regulation, which implies that government is the problem. An excellent analysis
of most of the cases reviewed in this subsection (except liquidity issues) in the context of
an integrated formal banking model can be found in Ize and Powell (2003).

A first case is that of arbitrage across currencies in financial intermediation costs unduly
favoring dollar intermediation. For example, costs in intermediating local currency may
be higher due to deficiencies in the payment system or monetary management that may
lead to higher or more expensive liquid reserves in local currency. Unremunerated (or
not adequately remunerated) reserves is also a regulatory disadvantage for peso
intermediation in the context of inflation (Catão and Terrones, 2000). Unequal
competition with less regulated off-shore banks, which operate in dollars, is yet another
reason for currency regulatory arbitrage favoring FD. The solution in these cases is to get
the systems and regulations right.

The argument has also been made that currency-blind safety nets such as deposit
insurance or policies of lending of last resort result in an undue advantage to dollar
instruments, for the same reasons that this was so under partial bankruptcy recovery
(Broda and Levy Yeyati, 2003). This is because payment events are associated with real
depreciation (again the countercyclicality of dollar value), in which case the face value of
dollar debts capture a larger share of the insurance (or more of the insurance under full

21
insurance). In this case the solution would be in the hands of policymakers, which ought
to design a currency-sensitive insurance to undo the distortion (or a currency-sensitive
premium, which being an ex-ante discrimination may be easier to implement).14

Finally, there is a large literature on how free public insurance, explicit or implicit, causes
moral hazard. In general, such hazard favors excessive risk taking on the part of the
parties that stand to benefit from the insurance as a way to capture the expected wealth
transfer. While banks are generally currency matched, they are still exposed to exchange
rate risk through credit risk (of mismatched borrowers). In relation to FD, given that the
value of dollar claims is countercyclical in developing countries, free insurance is better
captured by contracting in dollars (McKinnon and Pill, 1997; Dooley, 2000; Schneider
and Tornell, 2000). Furthermore, once FD is beyond some threshold, dollar borrowing
becomes privately less risky because in the case of devaluation some form of financial
rescue is to be expected (Burnside et. al., 2001).

In terms of the simple portfolio model, free insurance opens a wedge between the
effective borrowing and lending rate offered by banks. If such wedge is uniform, the
same across currencies, then it would not have an effect on portfolio choices concerning
dollarization (it would only have macroeconomic effects expanding debt, which may
cause excessive debt but not excessive dollarization). However, the dollar wedge is
larger and the equilibrium portfolio would end up being excessively dollarized.
Currency-specific capital requirements, in the spirit of a value-at-risk approach, suggests
itself to eliminate the FD bias induced by moral hazard. Similarly, currency-specific
liquidity requirements may also be needed to further level the playing field taking into
account the differential risks of liquidity crises and the fact that it is more costly for the
Central Bank to hold adequate reserves to cover dollar deposits; otherwise FD is
excessive (Ize, Kiguel and Levy-Yeyati, 2005). Since dollar liquidity is costly, the
Central Bank ability to provide liquidity support is a source of moral hazard (Dooley
2000).

On top of this generic banking moral hazard argument, it has been suggested that dollar
pegs represent an implicit guarantee to borrowers in case of breaking the peg (De la Torre
et al., 2003). Fixed exchange rate regimes encourage excessive FD by its insistence on
currency-blind regulation as a way of deriving credibility for the sustainability of the peg.
Argentina 2001 is a good example of ex-post rescues of dollar debtors and, probably, of
ex-ante pro-FD distortions (on the latter, see microevidence in Galiani et al., 2003).
Moral hazard driven FD calls for safeguards that limit the expectation of implicit
guarantees and bailouts, in particular concerning dollar borrowing. More generally,
Caballero and Krishnamurthy (2004) show how indexing inflation targeting and foreign
exchange interventions to negative external shocks (committing to a countercyclical
monetary policy) may also diminish incentives for FD.

Moral hazard also plays in reverse, from FD to public incentives to pursue certain
policies that fit a highly dollarized economy but may be suboptimal overall. The
detrimental effect of aggregate FD on public policies amounts to an externality, similar to

22
the market externality analyzed at the beginning of this section. Anticipating the
constraints that FD may impose on policies ex-post it could be justified to discourage FD
in the first place. While the traditional time inconsistency leading to incentives to dilute
nominal peso debt points in the opposite direction, dual-agency distortions by which
governments may want to help debtors with dollar liabilities through confiscatory
measures in the case of real depreciation make the case for less FD (see De la Torre and
Schmukler, 2003, for an interesting analysis of systemic risks in emerging markets debt
contracting).

This case is also very relevant in the case of exchange rate policy. High FD induces fear
of floating which in turn reinforces the incentives to FD for portfolio reasons (low
exchange rate risk). On the contrary, a floating exchange rate regime would encourage
less FD, which in turn would favor freer floating.15 These mutually reinforcing
influences may result in two equilibria; to the extent that the low FD is superior there is a
case for policy intervention to dedollarize (Chamon and Hausmann 2003; Ize and
Powell 2004). Chang and Velasco (2003) discuss a general equilibrium portfolio model
with these characteristics. This two-directional causality, and in the extreme multiple
equilibrium, makes clear that dedollarization policies crucially depend on maintaining
floating exchange rate policies, which in turn may be risky if there is not rapid success in
bringing down FD.

Missing Markets and Policy Reform

Markets other than dollar markets enabling protection against surprise peso inflation, the
crux of the matter in high FD countries, are sorely missed. There are many missing
markets whose existence would lead to better contracting with less FD, but the key one is
a market for inflation-indexed peso instruments, which is well established only in a few
countries such as Chile and Israel. In the context of the basic portfolio model the
possibility of contracting in inflation-indexed terms would eliminate all the volatility in
real terms concerning price risks, and would therefore dominate all nominal instruments,
both peso and dollar, for domestic lending (it would still be subject to country risk). In a
more sophisticated model it would also achieve the desirable objective of delinking real
returns from changes in circumstances or perceptions that would affect the expected
return profiles of other instruments concerning the evolution of their prices.

Even in this idealized context, this new market may not eliminate FD altogether because
of original sin (from the perspective of foreigners, inflation-indexed peso terms are better
than nominal peso terms but still carry currency risk, and therefore do not dominate
nominal dollar terms). Furthermore, it would also be subject to some disadvantage vis a
vis the dollar with respect to recovery values in case of credit and bank default mentioned
above, since the dollar would retain a competitive edge in private contracting for hedging
default with real (instead of nominal) appreciation in value. But it would clearly go a
very long way.

15
Following the same logic, flexible exchange rate regimes would induce an inflationary bias as a way to
dilute debts of debtors in difficulties. This case is the private sector generalization of the inflationary bias
of public debt in Guidotti and Rodriguez, 1992 in the context of dual agency theory.

23
Other missing markets could help alleviating the adverse consequences of FD, or the fear
of sudden stops (Caballero, Cowan and Kearns, 2004). Domestic currency derivatives
markets would allow a better allocation of currency risk among residents (and would also
reduce the ex-ante incentives for excessive FD in Caballero and Krishnamurthy, 2003)
Contracts contingent on negative external shocks, correlated with real depreciation and
sudden stops, can also serve the purpose of hedging against liability dollarization
(without the burden of mistrust or moral hazard to the extent that the contingency is
exogenous).

More generally, policy and institutional reform related to the local currency may reduce
actual risks associated with lending and borrowing in pesos, inflation indexed or not, and
in so doing make more attractive this safer form of liability. This appears to always be a
good policy initiative, so, if feasible, presumably there is some cost (legitimate or not)
that impedes it, so it would be naïve to discuss this kind of welfare-improving
dedollarization policies in a vacuum. Current circumstances, however, appear to be
functional. The realization of the large costs associated with FD that the experience is
showing and the shift to more flexible exchange rate regimes in many countries call for a
reassessment of the merits to invest in policies and institutions that strengthen the local
currency. For example, inflation targeting, the independence of the Central Bank, the
credibility of the inflation index, the strength of the fiscal institutions to back all of the
above, are all policy concerns that belong to a policy objective of dedollarization and
may pass the political cost/benefit test once the gains from such objective are realized.
On the other hand, it is important to keep in mind that policies and institutions that
diminish country risk (e.g. the respect for contracts and, if not possible, the maximum
recovery value) are a good idea in its own right but it is not clear whether they help to
reduce FD. In fact, in the basic portfolio model it would reduce offshorization but not FD
(returned capital flight would be saved domestically but still in dollars).16 I will return to
this issue in the last section with a new wrinkle.

IV. LESSONS FROM DEDOLLARIZATION EXPERIENCES

In all likelihood, policies aimed at fulfilling the objective of some dedollarization will
entail costs in terms of implementation and of undesired side effects (e.g. undue financial
disintermediation), so it is important to face this policy question with a second-best mind.
Notwithstanding the importance of the previous analytical approach to identify failures
and the nature of appropriate remedies, attempts at dedollarizing (or at preventing FD)
contain key information as to these costs. In a second-best framework the careful
evaluation of costs and benefits is inescapable, so it is crucial that these experiences
inform policy recommendations. The two main reviews of the experience are Reinhart,
Rogoff and Savastano, RRS, (2003), and Galindo and Leiderman, GL, (2004). In fact,
16
Furthermore, the case can be made that more efficient bankruptcy proceedings enlarge the distorted
competitive advantage of dollar claims vis a vis peso claims mentioned above, because deadweight
bankruptcy costs affect both claims uniformly (in the extreme, a gunboat bankruptcy technology would
eliminate default payments and with them the advantage of dollar financing). De la Torre and Schmukler
(2003) make the point that, ironically, better institutions allowing for larger recovery values would then
induce even more FD.

24
RRS think that recent proposals towards dedollarization ignore history and should not be
taken seriously for that reason.

Dollarization is pervasive and persistent. As of 2001, there were about 35 developing

countries in which bank dollar deposits exceed 20% of broad money, including 9 in Latin
America.17 A few countries saw substantial reductions in FD in recent years after
recovering from high inflation, such as Bosnia, Egypt, Slovenia and Poland, but even in
these cases dollar deposits remained substantial. A number of Transition economies
(Albania, Armenia, Czech Republic, Estonia, Georgia, Lithuania, Mongolia,
Mozambique, Yemen) saw falling FD after stabilization but then increasing again. RRS
report that 20 developing countries saw the bank dollar deposit ratio decline substantially
(at least 20 percentage points to a level below 20%) in 1980-2001, but that this
dollarization ratio rebounded in the vast majority of cases (16 countries of the 20).

Dedollarization is difficult. Not that the failure rate is 80%, because the 20 cases of
significant reductions in FD were not necessarily the result of policy attempts to achieve
an objective of dedollarization. But the fact is that only in the other 4 cases in RRS
reporting the gain was sustained and can be considered dedollarization experiences with
this definition (Israel, Mexico, Pakistan and Poland). Alternatively, GL define
dedollarization as a situation in which dollar bank deposits or loans exceeding 40% are
reduced to less than 20% for a period of at least 5 years. With this definition, only 3
countries dedollarized: Chile, Israel and Poland. The differences between the two sets are
easily explained: Chile’s high dollarization was not in bank deposits but loans (not
considered by RRS), Mexico’s initial level of dollarization did not reach 40%, and
Pakistan’s dedollarization too place less than 5 years ago. More recently, Argentina has
emerged from its crisis with a dedollarized banking system. In what follows we consider
as dedollarization experiences the union of the various sets of countries: Argentina,
Israel, Mexico, Pakistan, Poland and Chile.18

These six dedollarization experiences have some commonalities. First, there was some
shock treatment to bank dollar deposits and loans. In Chile bank dollar loans, mostly
financed by capital inflows, were converted to inflation-indexed loans in a market-
friendly way using the sweetener of a generous fiscal package designed to solve the 1982
banking crisis, which was precisely produced by liability dollarization.19 Others
delivered shocks not with carrots but with sticks. Israel imposed a mandatory holding
period for dollar deposits valued at administrated rates (an implicit tax). In Argentina,
Mexico and Pakistan dollar deposits were forcibly pesified inflicting capital losses in the
conversion.

Second, monetary policy induced a favorable evolution of relative real returns at the
beginning. Israel and Poland dedollarization success started with a disinflation program
17
Argentina, Bolivia, Paraguay, Peru, Costa Rica, Ecuador, Jamaica, Nicaragua and Uruguay. Since then,
Argentina has largely dedollarized.
18
Interestingly, Chile dedollarized twice, the first time (not considered here) in the 1960s after debtors
realized the risks of dollar debts and financial regulations were introduced.
19
This is reminiscent to the recent pesification of dollar debts in Argentina but with the twist of attaching
inflation indexation, a proposal that was actually considered at the time for all dollar contracts.

25
based on the exchange rate, which being successful tilted ex-post real returns against
dollar deposits. Mexico dedollarization took hold after 1988 as the process of real
exchange rate appreciation made peso instruments more attractive (in the years after the
forced conversion in 1982 it suffered severe credit contraction and capital flight that
authorities weathered).

And last, but certainly not least, a suitable substitute to dollar instruments was made
available. Chile had indexed peso instruments from before and Israel offered indexed
peso assets from the start of dedollarization. Mexico created an inflation indexed unit of
account in the 1990s which is offered by banks. (Poland’s perspective with euro?)
Argentina has also created an inflation indexed unit which has been used in private and
public debt restructurings in lieu of dollar debts. (On this account, as I discuss below, the
availability of indexed instruments has also allowed Brazil and Colombia to control FD
satisfactorily.)

What do unsuccessful dedollarization attempts have in common? Attempts so

unsuccessful that led to policy reversals because of very undesirable consequences are
characterized by blunt prohibitions of domestic deposits in foreign currency, followed by
massive financial desintermediaton, capital flight, and the undoing of the attempt. In
Latin America, Bolivia and Peru are examples of countries that prohibited bank dollar
deposits in the early 1980s only to see extreme macroeconomic instability that led them
to allow again dollar deposits (remaining very highly dollarized to this day). What
distinguishes the “successful” cases of Mexico and Pakistan which were equally blunt?
Not much. While Mexico finally succeeded by these standards, it suffered dearly and
pressures for dollarization subsided only in the 1990s. As to Pakistan, it may be too early
to tell. Be it the bad signal or its sheer size, a cold-turkey ban on dollar deposits appears
extremely risky, especially with perception of high country risk and in the absence of
easy access to foreign financing that may offset a surge in capital outflows.

The evidence on avoidance of dollarization of bank deposits on the basis of explicit

policies offers interesting clues too. In Israel banks are required to actively hedge
currency risks or impose higher collateral in the case of dollar lending in the non-tradable
sector. Success in Latin America has been accompanied by strong financial prudential
policies favoring local currency lending such as legal restrictions to dollar bank deposits,
as opposed to just indirect policies aimed at improving economic fundamentals. The main
cases of avoidance of dollarization of bank deposits in Latin America are Chile
(prohibition to lend to borrowers in the non-tradable sector), Mexico (quantitative
limitations on lending and prohibition for households to hold dollar deposits) and Brazil
(prohibition of dollar lending except on-lending of foreign credit and prohibition of dollar
deposits), Colombia and Venezuela (ban or strong legal restrictions on dollar deposits).
However, even in countries where banning onshore dollar bank can be considered an
overall success, like Brazil, Colombia and Mexico, there is a substantial degree of
offshore dollar bank deposits (see GL). IMF (2005) provides details on the strict
regulations on dollar financial transactions in these countries and the extent to which
repressed dollarization onshore is partly reflected in dollarization offshore.

26
However, policy sticks penalizing FD may have achieved low bank dollarization but have
not been sufficient for overall success concerning FD. Success in avoiding or undoing FD
in bank deposits is only half the victory against FD; for example, Venezuela does not
suffer from FD in banking due to regulatory prohibitions but cannot be considered a
success overall. The policy problem is that the stronger the disincentives to FD in the
banking system, the stronger the incentives for existing substitutes, most of them
harmful. Within the banking system, local currency deposits and loans tend to become of
shorter durations (nominal rate instruments at short maturities or floating rate instruments
that allow frequent repricing) as a way of recuperating some of the protection against
surprise inflation that dollar instruments offered (Brazil is a clear example). This form of
“repressed dollarization” avoids exchange rate risk at the cost of interest rate risk and
roll-over risk, which share some of the harmful characteristics of exchange rate risk.
Outside the banking system there is an overall tendency for FD to substitute for the
repressed dollar deposits, either through dollarization of domestic public debt (Brazil),
through increased borrowing abroad (Colombia private borrowing is a clear example), or
massive capital flight and instability (the case of Venezuela).

Successful countries in Latin America concerning FD had not only the stick of bank
regulation (or controls on capital outflows to repress dollar savings) but also a reasonable
substitute for domestic dollar savings that could be used as a carrot to sway depositors to
local currency savings. These are countries that in the past had severe macroeconomic
instability and resorted to the promotion of indexation to protect savings from
inflationary erosion, which have survived and allowed them to avoid dollarization
permanently at a reasonable cost. Chile innovated decades ago with an indexed unit of
account, the Unidad de Fomento (UF), which proved a very effective substitute of dollar
units as a financial asset and is clearly the most successful experience. Brazil and
Colombia are moderately success stories. Brazil also has a long history of indexation;
currently, the use of the SELIC index of overnight interest rates in local currency is
widespread. In fact, Brazil’s local currency debt indexed to the overnight SELIC rate
plays now a key role in its public debt dedollarizing strategy to match its higher exchange
rate flexibility.20 Colombia has also experimented with various types of local currency
indexes and has now adopted a UF-type index used in the mortgage market. By contrast,
Venezuela resorted to financial repression and ineffective capital controls that have
caused recurrent capital flight and disintermediation. Argentina has recently created an
inflation index used in debt stock restructurings but has not yet been made available for
bank deposits and other debt financing flows; its future use remains uncertain.

The particularly successful case of Chile is revealing as to the combination of policy

sticks and carrots (and favorable preconditions) that allowed it to dedollarize and then
build a solid financial system around indexation for 20 years. Herrera and Valdes (2003)
highlight that favorable preconditions were important, especially the existence of a
consolidated inflation index supported by favorable conditions for market development.
Even under these favorable conditions for avoiding dollarization, pegging of the
exchange rate in the 1990s was an implicit incentive for dollarization; only after flotation
the forward foreign exchange market and the local bond markets have started to develop.
20
Nevertheless, this interest-rate indexation has its own fiscal risks due to large interest rate risk.

27
(Israel, the other clearly successful case, has also actively pursued the development of
financial derivatives markets and made efforts to deepen local currency bond markets.)

As to policy actions, they highlight two. First, the embracing of indexation in policy
making, not only encouraging it in financial markets but also engineering the banking
crisis resolution with the objective of switching to these instruments and designing the
entire macroeconomic framework around it (e.g. exchange rate and monetary policy in
real terms). And second, and seldom mentioned, Chile’s long experience and legal
tradition with capital controls which helped the avoidance of on-shore dollarization and,
especially, limited off-shore dollarization, effectively repressing demand for dollar
savings. Capital controls took the form not of prohibitions but of an allowed foreign
exchange black market that segregated favored transactions (with access to the official
(yet realistic) exchange rate) from disfavored ones, and were largely dismantled only in
the 1990s once access to capital inflows was ample.

Among Latin American countries with high FD, there is a sharp divide between countries
banning dollar deposits and countries having weak or non-existent bank prudential
regulation dealing with FD, which appears hard to rationalize (see GL). On the one hand,
substantial levels of off-shore dollar deposits in countries banning (in-shore) dollar
deposits, including reasonably successful cases like Brazil, Colombia and Mexico,
suggests that there is scope for gains in financial depth at little or no cost of FD by
softening this extreme bank prudential policy. On the other hand, the widespread
passivity of bank regulation in most of the other countries concerning credit risk
associated with borrowers’ currency mismatch suggests that there is ample scope for
improvements in controlling the risks of FD through the introduction of tighter prudential
requirements on foreign currency loans at little or no cost in terms of financial depth.

Successful experiences with policy attempts to dedollarize public debt also point in the
direction of purposeful policies and the development of new non-bank local currency
markets, especially indexed.21 As a debt management policy, purposeful in this context
means to internalize the indirect costs of FD in public debt in terms of fiscal and country
risk and assign a shadow cost to dollar debt in the financial comparison of sources of
funding, in the spirit of redressing market externalities discussed in the previous section
(see Hausmann, 2003 for an analysis). For example, Mexico has been paying down
foreign (dollar) public debt and increasingly relying on peso domestic public debt since
1995 after the Tequila crisis as it reduced its overall indebtedness ratios. Brazil has also
followed a purposeful debt management policy of reducing dollar and dollar-indexed
liabilities (debt and guarantees) for local currency debt at a substantial cost.22 It is not
clear, however, how fiscal risk is balanced with risks induced to the private sector, an
issue I will discuss in the next section. Finally, Argentina has also drastically reduced its
share of dollar public debt in the context of its recent debt restructuring by offering a
valuable (that is, expensive) conversion option for inflation-indexed bonds.

21
Dollar public sector debt in Latin America is high, exceeding. 75%, but shows some tendency of
containment and reduction over the past decade (Calvo, Izquierdo and Mejia, 2003).
22
Unfortunately this has resulted in a significant increase in the share of SELIC-indexed debt, a poor
substitute of dollar-indexed debt.

28
The second and fundamental characteristic of successful public debt dedollarization,
apart from diminishing fiscal risk, is the introduction and promotion of non-bank local
currency debt instruments that offers some of the portfolio advantages of bank dollar
deposits. Again, the Mexican example is very illustrative. The strategy included gradually
issuing fixed-rate debt at various maturities in order to create a yield curve that can help
the development of long-term private debt markets and liquid derivative markets. Israel
followed a similar strategy to develop a yield curve in nominal local currency and to do
so willingly paid a financial cost over inflation-indexed debt (real rates on nominal debt
were high due to disinflation during the period). In this context, the promotion of markets
for local currency indexed instruments appears extremely useful, especially when based
on indexes that allow long durations, such as inflation (of consumption baskets or
particular products) or income. Mexico has also developed inflation-indexed securities
(mainly demanded by domestic institutional investors) and Uruguay has an active
program of inflation-indexed public debt issues both domestic and abroad which has met
success despite its recent public debt restructuring. In fact, several other Latin American
countries issue inflation-indexed public debt, such as Argentina, Brazil, Bolivia, Chile,
Colombia and Costa Rica.

Before excessively raising our hopes with indexation, however, it is important to recall
that within Latin America inflation-indexed bank deposits are only substantial in Chile.23
While it is true that indexation has been key to control FD in countries where the
objective was successfully achieved, it is also true that it has failed to take hold in some
of the countries which tried to incorporate this innovation despite the presence of
inflationary concerns (see Shiller, 1998 for an analysis). Credibility of the index appears
to be an important factor. Uruguay’s wage-indexed unit has found important regulatory
(pensions) and contractual (rents) uses for decades but was not favored over dollars for
savings purposes; in Argentina experiences with indexation were also failed, including
after the recent crisis in the pesification of dollar mortgages after the collapse of the
currency board. In both cases it appears that the obstacle was lack of confidence that the
indexation would survive high inflation rather than technical reasons of market size, an
alternative reason sometimes argued.24 In fact, the necessary credibility for indexation to
work was hard to secure even in countries where it worked (Brazil privatized the agency
computing the inflation index out of concerns of manipulation; in Chile altering the index
was seriously considered in the 1970s, and fortunately for Chile dismissed). The usual
difficulties in setting up new financial markets, such as imbalances between demand for
assets and liabilities and lack of development of secondary markets are also referred in
this literature.

What did it take for developing strong inflation indexed markets where they were
successfully established? The successful experiences of Chile and Israel in developing
inflation-indexed instruments throughout the economy were based on the credibility of

23
Inflation-indexed deposits are marginally significant in Argentina after pesification. Inflation indexed
loans are also significant in Mexico and Colombia.
24
In fact, in Argentina there was arbitrary de-indexation in the 1970s and change of index base in the
aftermath of the recent crisis, although it is fair to say that pesified dollar-indexed claims fared even worse.

29
monetary and fiscal policies, which lends credibility to the index itself.25 The early
importance of long-term institutional investors with a natural demand for inflation
indexed assets, such as pension funds, was also very functional to the development of this
market in Chile (in the case of life insurance companies demand is not only natural but
mandated by regulation). Interestingly, after having reached low levels of inflation for a
long time both countries are making some efforts to substitute indexed instruments for
nominal instruments, and are finding that the indexed instruments have strong demand,
especially at longer maturities. This reveals in practice the theoretical value of inflation-
indexed markets alongside fixed nominal markets once markets are set up and fully
developed, and suggest that indexation remains fundamental as a tool in a dedollarization
strategy even in the case of solid advances in inflation control.

Finally, the review of the experience would not be complete without an explanation of
why haven’t most countries afflicted by high FD attempted to fight against it, or at least
not in a resolute way along the lines reviewed in this section? One key explanation is that
it is only after recent experiences that the staggering costs of high FD have been revealed,
so we can expect more policy action in the future.26 Furthermore, only now with more
flexible exchange rates and low inflation levels are conditions set for a substitution of
peso for dollar instruments within the reach of policy. Yet, this is only part of the
explanation. In fact, GL report that their survey reveals that many Latin American
countries with high FD do not have too many specific policy initiatives aimed at reducing
FD other than developing capital markets in domestic currency or in indexed units. The
rest of the explanation is likely to be that dedollarization policy, like all investments,
impose up-front costs but delivers benefits only in the future; it is possibly too costly or
risky to launch in bad economic situations and too unappealing to short-sighted
politicians in good economic situations. The implication is that it will be important to
think in domestic institutions or outside influence, for example conditionality by
multilateral organizations, that can help compensate this status quo bias.

V. FINANCIAL DEDOLLARIZATION STRATEGY

The analysis of failures leading to excessive FD in Section III is a good starting point for
the discussion of dedollarization policy. That analysis suggests a two-pronged approach
aimed at addressing the failures by “error”, which result in the private mispricing of
dollar contracting relative to social pricing, and the failures by “omission”, that is the
absence of markets and institutions that would enable market participants to arrive at
more (private and social) satisfactory outcomes by choosing to replace dollar financing.
The former calls for policies to constrain and alter private incentives to discourage the
choice of FD under the circumstances and the latter for policies to change the
circumstances by promoting attractive substitutes of FD.
25
There is the related concern that due to inevitable lags hyperinflation would still dilute inflation
indexation, thus weakening its attactiveness. This concern appears theoretical, however, since in such a
extreme situation any form of indexation or contracting would also be at risk, including dollar indexation
(witness Argentina’s pesification of dollar contracts).
26
In fact, Herrera and Valdes (2003) conclude that learning from the dollarization-driven crises of the early
60s and 80s helped policymakers and market agents to converge in the healthy financial system developed
afterwards.

30
Pitfalls of a Dedollarization Focus

There are two pitfalls of a dedollarization focus that makes it extremely risky: the risk of
going overboard, beyond the elimination of excessive FD, and the risk of fighting
symptoms, favoring other harmful forms of debt financing springing from the same
disease as an unintended consequence of narrowly fighting FD. Consequently, I suggest
that dedollarization policy would benefit from focusing not on bringing down the dollar
but on competing with the dollar with a better peso substitute.

The experience is sobering in relation to the risk of going overboard and provoking
massive capital flight, instability and financial disintermediation. It is easy to see how
policies intended to alter incentives or constrain financial choices to correct mispricing
due to externalities may go awry. Even in the (first) best of circumstances, there is the
concern that the outcome of such policies may not bring a substantial improvement to
welfare. In fact, it is not clear that externalities are responsible for high FD; in other
words, it is not clear that after excessive FD is eliminated, (socially) optimal FD will not
be high and will cease to be the menace it is today.27 Not that externalities are not
important, but the legitimate reasons for avoiding peso savings in the basic portfolio
model may be even more important. In my view, it may well be the case that high
domestic FD results mainly from poor fundamentals (in particular, unreliable pesos)
rather than externalities generating socially excessive domestic FD, in which case even
successful elimination of excessive FD would leave high FD. If so, then a dedollarization
policy attempt which would not stop until succeeding in substantially bringing down FD
would go too far, be counterproductive, and lead to capital flight and financial
disintermediation. This theoretical risk clearly resembles many of the failed
dedollarization experiences.

We know little about the size of the interventions required to deal with the most
important externalities, so the risk of keep increasing the dose of the treatment until the
desired results in terms of a substantial reduction in FD show is very real. In fact, the
case has been made that quantities rather than prices may be the most suitable policy
instrument given the nature of the problem (Levy Yeyati, 2003), so policy may take the
form of substantial (and arbitrary) quantitative ceilings of FD from the start. There is a
serious risk of ending up with policies that relate more to bringing down high FD (with
the understandable intention of avoiding its well-established undesirable consequences)
than to offsetting externalities to avoid excessive FD. There is the risk that incentive
policies be derived more from the empirical regularities on the consequences of high FD
discussed in Section II than from the causes of excessive domestic FD discussed in
Section III. Sound analysis may end up producing a knee-jerk reaction policy
prescription.

As to the risk of fighting symptoms, there is the risk that incentive and regulation policy
to discourage FD may cause dollar savings to shift to other forms of savings with
problems of their own. Domestic FD results from the need of residents to cope with a
27
To my knowledge, there has been no attempt to quantify these factors.

31
weak currency unreliable for saving purposes. The basic portfolio model brings this
clearly to light but has the unfortunate implication that impediments to domestic dollar
savings would make them shift to peso savings, the only alternative, all too easily. For
example, a ban on bank dollar deposits would substantially increase bank peso savings
(eq. xx), and if coupled with impediments to capital flight would be completely effective
in corralling domestic savers and bring them back to the no-alternative-other-than-peso
portfolio model. The consideration of induced off-shore savings introduces an important
realistic alternative in the basic portfolio model but its impact on economic activity tends
to be muted because capital flight is easily offset by external debt. In actuality foreign
financing may be unavailable, especially after a surge in capital flight, and even if
available, mounting foreign debt may increase risks not accounted for.

In actuality there are other portfolio alternatives to be considered. The bigger picture is
that domestic FD is only one manifestation of this financial adaptation to a weak peso;
short duration of peso contracts (either short maturity or floating rates) is another
protective device against surprise inflation and it is likely to be seen as a preferred
alternative to dollar savings (De la Torre and Schmukler, 2003). Policies specifically
addressing excessive FD will likely produce shrinking financial duration and heighten the
attendant financial fragility risks. The positive correlation between dollarization and
maturity confirms this prediction. Repressed FD in Brazil also attest to this prediction.
The problems associated with short duration debt, including externalities of the kind
discussed in this paper, are cousins of the problems associated with high FD. It would be
irresponsible to go too far in altering incentives, or imposing constraints, specific to
dollar financing without taking into consideration this second-best complications and
balancing the costs and benefits of each. The literature is lacking on this.

The development of substitutes in a competing with dollarization frame of mind can

overcome this fighting the symptoms risk. Better fundamentals to back local currency
financing and the development of healthy peso instruments, for example inflation-
indexed, compete favorable not only with dollar financing but also with other risky
adaptations such as short duration financing. In fact, the development of substitutes such
as inflation-indexed peso instruments should be successful not only to drastically reduce
FD but also to lengthen maturities because it is immune to both exchange rate risk and to
interest rate risk. Public debt management can also be cast in a more functional role in the
overall policy agenda, and instead of focusing exclusively on how to reduce fiscal risk it
can be refocused as an important tool to help the development of missing markets
through appropriate issuing and pricing decisions. The very prudential policy aimed at
discouraging FD may very well be more effective preferentially supporting good peso
substitutes (e.g. focusing on subsidies to desirable peso instruments instead of taxes on
regulated dollar instruments).

Financial Dedollarization Strategy: A Substitution Approach

For these reasons I would like to suggest a dedollarization strategy that replaces the focus
on bringing down FD for one of displacing FD with better substitutes. By focusing on
favoring healthy forms of peso financing rather than narrowly discouraging FD, this

32
approach gives primacy to the development of good substitutes for dollar financing, the
second component mentioned above. On the basis of the well-established deleterious
consequences of high FD there is a clear case for investing resources to develop an
environment in which low FD emerges as a market outcome. This policy agenda has no
risk attached (other than losing the investment, which can be well quantified).
Furthermore, the overall success from setting the right incentives as a solution to high FD
will in large part depend on the success of this second component to facilitate the
transition. Good substitutes would give traction to a change in incentives by facilitating
substantial dedollarization without financial disintermediation, removing the temptation
to overdo incentives and regulation in order to see results. In particular, it may be needed
to overcome fear of floating and make flexible exchange rates a credible policy.

The development of good substitutes appears to be an eminently good idea, be it a better

environment for domestic savings and existing peso markets or the development of new
markets with risk profiles that are competitive with dollar markets but do not share its
most harmful characteristics, so why not do it alone without taking the risks of tinkering
with incentives or adopting intrusive regulation to try to address externalities causing
excessive FD? The answer is that these market and institutional development initiatives
have large set-up costs and any help to push them past the first hump is important.

In fact, the case can be made for going somewhat beyond the appropriate incentives to
eliminate excessive FD under current markets conditions in order to impel the switch to
dollar substitutes as a way of diffusing set-up costs, although such move has to be
carefully weighed against the risk of instability. If it were not for this risk, shock
treatment of dollarization to dislodge the support of the high FD equilibrium, including
forced conversions, would be recommendable as a way of saving transition costs and
ensuring obtaining needed critical mass in cases of multiple equilibria and corner
solutions. As we saw, the evidence on shock treatment is mixed: it was an element in
some successful experiences but it was also an element in some of disasters. While it
remains a judgment call, it appears that shock treatment may be reasonably used as a
device to speed up a transition that looks certain but may be too risky otherwise.28 In any
event, once the whole program has critical mass and credibility some of the ancillary
incentives could be discontinued if they appear doubtful.

The proposal is then to pursue addressing externalities by setting the right incentives and
addressing lack of an enabling contracting environment by developing good dollar
substitutes at the same time, in a coordinated fashion. A full court pressing effort to jump
start the process. Successful dedollarization experiences combine elements of both tracks.

However the development of substitutes is the main character and get the spot light: The
development of good substitutes addresses the basic driver of domestic FD, appears to be
a necessary condition for overall success, may go a long way alone, and poses no risk.
The development of good substitutes for dollar financing are a constant in successful

28
Argentina exemplifies an interesting case in which shock treatment was the unintended result of crisis
resolution. If a coherent policy is designed around this new initial condition, this structural change may be a
silver lining of crisis.

33
dedollarization experiences, especially inflation-indexed; unsuccessful experiences are
characterized by financial dedollarization achieved with blunt prohibitions and controls
disregarding the improvement of markets and institutions.

Altering incentives through prudential regulation is assigned a supporting role because it

will likely not succeed in making a substantial dent in FD alone and, unless regulation
defects can be eliminated directly rather than offset through compensatory incentives,
poses great risks if the part is overplayed. According to experience, developing
substitutes such as inflation-indexed markets will likely entail set up costs to jump start
and time to consolidate. In this context, the announcement of a package of clear signals
with regulations to address externalities may be an important boost to creating
momentum for dedollarization and reducing set up costs of substitutes.

This coordinated policy effort should be launched when circumstances are most
conducive to success, such as an expected relative low real return of domestic dollar
savings (e.g. expected real appreciation) to sweeten the exit from dollar assets and
macroeconomic stability to lend credibility to the peso substitutes. Current circumstances
appear right for this. The experience shows that macroeconomic circumstances and the
short and medium-term evolution of real returns played an important role for the ultimate
success of the effort. Experience also shows that there may be a risk of dramatic
disintermediation, especially if policymakers decide to run the risk of shock treatment of
existing domestic dollar claims, so it is useful to have impediments to capital flight in
place to dampen temporary instability in the transition and to have good access to foreign
financing to guarantee financial intermediation if the attempt to reduce domestic financial
dollarization is not sufficiently successful and capital flies. Many of the costly attempts
to dedollarize in the 1980s did not count with the backing of access to foreign financing
and that may have sealed their destiny.

Development of Dollar Substitutes: Inflation-indexed instruments

Domestic FD emerged as an adaptation to an environment of weak local currencies.

Dollar terms amount to a form of indexation that protects residents from peso inflation at
the cost of exposure to real appreciation; it trades inflation risk for currency risk. The
most direct way to address the demand for inflation protection behind the choice of
domestic FD is through inflation-indexed instruments. In theory it amounts to the
creation of a strong local currency for the purpose of storing value, the most direct
solution to the problem. This form of market completion is important because it
facilitates the substitution away from FD in a comprehensive fashion, without the pitfalls
of a piecemeal approach that tends to shift and create new risks.

The portfolio model enables us to speculate how dollarization could change with the
introduction of inflation-indexed domestic peso loans alongside all the other alternatives.
These loans would contain no currency risk and would therefore dominate nominal peso
loans. Not being subject to price risk, they would be also superior to domestic dollar
loans. The only risk they would contain is risk of default (country risk), and therefore it
would only face competition from off-shore dollar deposits, which are free from country

34
risk. Therefore all dollarization would be offshore (d’) like in the case of a ban on
domestic dollar deposits (d**), in turn smaller than the unconstrained level of
dollarization d* in the basic model:

d ' = f * Vdd /(V xx + Vdd ) < d * * = (Vcc + Vcx + f * Vdd ) /(Vcc + V xx + 2Vcx + Vdd ) < d *

Two points are worth making. First, dollarization (and offshorization) is smaller than
under a total ban on domestic dollar debt ( d ' = f ' < d * * = f * *) . Let’s remember
however that a reduction in FD is not an objective in itself; in fact the level of
dollarization d** obtained with a total ban is too low unless all domestic dollar lending is
due to an externality! The second and fundamental point to make about inflation-indexed
loans is that the even lower dollarization level d’ is the unconstrained market outcome,
and therefore this enormous reduction in FD does not carry any hidden cost burden.

Being a pro-peso policy rather than anti-dollar, its success would provide a bona fide
substitute that would also compete with harmful dollar alternatives such as short-term
peso debt. In fact, being real interest rates constant over time, it is to be expected that
inflation-indexed instruments would partly substitute short-term peso instruments and,
overall, lengthen average maturity of savings and lending.

Too good to be true? The above model abstracts from a number of factors which are
behind the difficulties in introducing this kind of indexed instrument (see Shiller xxx).
Complexity is one. But the successful examples show that this is a fixed switching cost.
Lack of trust in the computation of the index (moral hazard), maybe bred by complexity,
is another important factor. Another factor is externalities: if the incentive framework
favors risk taking in the form of dollar debts, then the dollar would continue to have the
upper hand and agents would not be willing to buy the implicit insurance of inflation-
indexed instruments at fair rates. Most of the externalities we analyzed in Section III
would also apply in this new environment to some lesser extent. Nevertheless, it stands
to reason that if the gain in efficiency is so enormous, this instrument should have very
wide acceptance after it reaches critical mass despite all these imperfections.

For this instrument to dominate the scene it should be prominent both in banking and in
capital markets. How to jump-start this missing market? Besides its acceptance in
banking and the role regulation may play in this to favor its development while ensuring
the balance between bank assets and liabilities, two players stand out. One is institutional
investors with a natural demand for these instruments, such as pension funds and life
insurance companies whose liabilities are indexed similarly. These players would
demand long-term inflation-indexed claims. This demand can be strengthened by
regulation, and in fact that has been the case in Chile for example. The other is the
government in the context of active public debt management policy. I will return to this
issue. However it is jump-started, low liquidity will probably be the weak point of this
market, especially at the beginning. This and the existence of switching costs imply that
public policy may need to subsidize the initial development phase directly or by
regulation.

35
I note in passing that although inflation-indexed loans would eliminate currency risk for
foreigners too it would not be enough in this model to alter original sin. In fact, because
foreigners discount with their own consumption basket deflator, dollar lending does not
contain any price risk (abstracting from surprise dollar inflation) while peso lending does
contain price risk. Inflation-indexed lending removes currency risk from peso lending but
still contains exchange rate risk due to the difference in price deflators, so the peso-dollar
gap closes somewhat but persists, and so would original sin ( d ' = 1 for foreigners). The
stubbornness of original sin in this model even if currency risk is eliminated altogether
suggests that even in the context of a portfolio approach fixing original sin may be an
uphill battle.

Nevertheless, this analysis also shows that this inflation-indexed peso market would be a
step in the right direction to make local currency attractive to foreigners, and with other
factors outside this model it may be part of the solution. Therefore the development of
inflation-indexed instruments is not only key for addressing domestic financial
dollarization (the proposal in this paper) but also a first step to address original sin (the
proposal in Eichengreen, Hausmann and Panizza 2002).

Development of Dollar Substitutes: Supportive Markets, Policies and Institutions

One important foundation of inflation-indexed instruments is the reliable calculation of

the index, i.e. that is not manipulated opportunistically. This moral hazard problem would
arise directly in the case of indexed public debt but can easily arise indirectly in private
debt if the government has an incentive to rescue debtors in certain circumstances. This
moral hazard problem ought to be much less severe than with dollar debts precisely
because it contains no price risk causing the acute need for rescue. At the same time,
however, the dilution of these debts is much easier for a government in control of the
index than it would be, for example, to pessify dollar debts (an open breach of contract).
Therefore, it is important to back the credibility of the index with a stable monetary and
fiscal position and it is fundamental the creation of strong supporting institutions that
help gather credibility in areas such as independence in the production of the index and
outside oversight. In this regard, indexation to exogenous variables such as commodity
prices would be much easier to implement, but at the same time less useful as dollar
substitute29.

Flexible exchange rates coupled with an inflation anchor, as in inflation targeting

regimes, are the ideal setting for supporting peso instruments as substitutes of dollar
instruments. The basic portfolio model and the empirical evidence both suggest that this
environment, increasingly prevalent in Latin America, is much more conducive to
dedollarization, an attribute to be factored in the pros/cons analysis of monetary and
exchange rate policy. Controlled inflation favors nominal peso debt but should be
irrelevant for inflation-indexed peso debt; in practice, however, it is likely to also benefit
the credibility of this new instrument. Flexible exchange rates in this context implies real

29
On the other hand, multiple inflation indexes to cater to agents’ preferences and the hedging properties of
GDP-indexation would be particularly useful but have practical drawbacks (Shiller 1998, Borenzstein and
Mauro, 2002). A full discussion on optimal indexation exceeds the scope of this paper.

36
volatility, which discourages dollar financing. There is the question of whether high FD
countries can sustain such regime until substantial dedollarization takes place; the current
tendency for real exchange rate appreciation following dramatic peg adjustments in some
of the countries and the tendency for real depreciation of the US dollar, the currency of
denomination of most foreign currency debt, is facilitating this transition because
inoculate dollar debts.

The evidence on the importance of institutional variables in explaining cross-country

dollarization (De Nicoló et al., 2003) suggests that policy and institutional reform is
fundamental for dedollarization and should be factored in policy evaluation. One
interpretation of this result is that these variables reflect the degree of confidence in the
future of the local currency, beyond observed inflation in the recent past. Under this
interpretation, policy and institutional reform would be an important signal to jump start
the process of substituting domestic FD. I subscribe to this interpretation.

Another interpretation of the same evidence is that country risk, which is largely a
function of these variables, is in itself a determinant of domestic FD. This interpretation
is consistent with the evidence but not with the portfolio model as it stands, in which
country risk, or default risk, is relevant for the degree of off-shorization (f*) but not
dollarization (d*). This results holds under the maintained assumption that f*<d*, that is
to say, the case in which in equilibrium there are domestic dollar loans d*-f*. This
assumption is realistic on average, certainly in high domestic FD countries. However, it
could be the case that particularly risk-averse asset holders choose to have a high
proportion of their savings offshore to avoid country risk; they would like to have some
of it in pesos but because that possibility is not available, they are forced to hold all their
offshore holdings in dollars and consequently minimize their dollar claims onshore to
zero. For these savers, dollarization (and offshorization) are determined by the
expressions d**=f** found before:

d * * = f * * = (Vcc + Vcx + f * Vdd ) /(Vcc + V xx + 2Vcx + Vdd )

Under the new assumption that f * > d *, it can be checked that then
d * < d * * = f * * < f * . It is clear that less country risk is welfare improving. It can be
checked in the Appendix that the above expression falls as default risk Vdd decreases; in
other words, less country risk would lead in this case to less domestic FD d** (and less
offshorization and foreign debt).30

In the context of developing new markets, Levy-Yeyati (2004) asks the question of what
would it happen under this corner assumption if an off-shore peso market is created (for
example, a foreign institution like a multilateral development bank (MDB) issues peso
bonds bought by residents). This instrument would allow residents to avoid default risk
without being forced to hold dollar claims. Levy-Yeyati shows that with this new market
portfolio allocations become again unconstrained and domestic FD would fall, again to
the level d*. The same kind of result would hold for inflation-indexed peso instruments

30
This assumes that the country risk premium by foreigners k does not fall too fast.

37
issued off-shore. Not surprisingly effects on domestic FD are larger than with nominal
peso debt, but still there is a lower bound to domestic FD determined by country risk.

A more ambitious and exciting question is what would it happen if this experiment is
performed once inflation-indexed peso instruments are already established onshore, as in
the previous subsection. In other words, what would it happen if inflation-indexed
instruments are available both onshore and offshore. First notice that in this experiment
the doubtful corner assumption f * > d * necessary to obtain the previous result actually
holds, since all dollar claims would be due to country risk (in fact, with onshore indexed
instruments, d*=0<f*). It is easy to check that with the availability of inflation-indexed
instruments both onshore and offshore, domestic FD would be eliminated: dollar holdings
would be dominated by inflation-indexed peso holdings both onshore and offshore! And
original sin would be essentially solved: offshore local currency intermediation would
apply the entire stock of domestic savings D to domestic financing in local currency and
only net liabilities L-D would be financed in dollars. This feat would take foreigners to be
willing not to hold peso claims but just to intermediate peso claims. It would not take
Belgian dentists as in Eichengreen, Hausmann and Panizza (2002), but rather resident
dentists saving in Belgium.

It is clear that MDBs have an important role to play in supporting this financial
dedollarization strategy based on the development of dollar substitutes. Apart from the
involvement in operations supporting policy actions (including a role for conditionality)
on the part of multilateral organizations, MDBs can play an important role in their
capacity of lenders without incurring currency mismatches themselves. Their ability to
issue investment grade local currency instruments would complete markets and may be
highly effective in creating a local currency claim that can substitute for dollar claims. In
turn they would on-lend in local currency to their traditional official clients (or through
their private sector windows), thus reducing the overall FD in the economy.31 It follows
that the best candidates are countries with high FD, high offshorization and high country
risk. A good example is provided by Levy Yeyati: pension funds, whose liabilities are
matched to the peso, save abroad not to avoid currency risk but to avoid default risk.
MDB intermediation in this case is also useful to allow pension funds to indirectly
finance the government without fiscal risk and moral hazard.

Last but not least, another suitable substitute of dollar debt financing frequently forgotten
is equity financing. So far domestic savings channeled through debt claims have been
taken as given in the portfolio models, but it is clear that leverage is a variable subject to
policy influence. In fact, equity financing is another natural protection against inflation
risk. A full court pressing dedollarization strategy ought to include incentives on the
margin for the promotion of vehicles of equity financing. The briefness of this reference
reflects the scant analytical and best practice research on the subject, not the importance
of this policy area. The tradeoffs merit further study, but this is a very relevant factor to
take into account in policy evaluation.

31
Notice however that this intermediation uses up lending capacity to the country; in other words, foreign
funds for the same amount are not intermediated by MDBs.

38
Development of Dollar Substitutes: Public Debt Management

Public debt dedollarization is usually portrayed as an indicator of progress concerning FD

and governments pursuing an active debt management policy of dedollarization are set as
examples. Governments are advised to switch dollar debt for long-term (nominal) peso
debt. Hausmann (2004) proposes the use of risk weights for the assessment of financing
choices in public debt management to estimate the true cost of dollar debt on account of
lower debt carrying capacity (due to higher volatility). The justification of dedollarizing
in this fashion is amply justified as a fiscal concern because from a financial viewpoint
the public sector is non-tradable (revenues are a fraction of GDP). However, this may be
another case in which the focus of bringing down FD leads to narrow, piece-meal policies
of doubtful value overall.

The beneficial impact of dedollarizing public debt management policies is diluted by risk
shifting: the public sector switching of dollar to peso financing may substantially crowd
peso financing of the private sector. In this case, overall FD would remain high in the
“national balance sheet”. Worse yet, the public sector may be giving up foreign financing
that the private sector may be unable to access, thus leading to a financial crunch. In
general, the fiscal risk of FD would be partly shifted to the private sector. This shift may
or may not be welfare improving. The point is that, here too, the focus on bringing down
FD is a risky guide.

Public debt management, however, can play a fundamental role in the overall
dedollarization strategy supporting market development. It is one large player which may
issue inflation-indexed debt and make it appetizing for domestic institutions and the
public at large.32 It can join efforts with MDBs to develop new peso markets by
borrowing from these institutions in these new markets as explained above, and perhaps
further onlending them to support financial intermediation in these markets domestically.
It may also take care of the technical aspects of these markets, such as building a yield
curve. In the financial dedollarization strategy proposed, these contributions are the
important ones at the end of the day, more than dedollarizing as a fiscal concern.

Excessive FD: Aligning Incentives and Prudential Regulation

An important role of prudential regulation is to support the development of dollar

substitutes, in harmony with the rest of the effort. Dollar substitutes require appropriate
prudential regulation that help them develop. There is therefore need for thinking how to
incorporate dollar substitutes, such as inflation-indexed instruments and equity-like
financing (or leverage), in a coherent prudential regulation.

Concerning the more traditional elements of prudential regulation, the key and clear
message is that it ought to be currency-sensitive, disfavoring dollar intermediation, to
compensate market externalities and policy externalities. Most suggestions refer to the
banking system, which is at the core of FD and of the policy externalities. I will provide

32
At the same time, by acquiring a direct interest in the index it detracts from its credibility, which will
need to be strengthened by other means.

39
a brief summary of the kind of prudential policy that the literature has identified as most
suitable for the purpose. Nevertheless, the market externalities I emphasized in Section
III relate to dollar borrowing by non-tradable firms, not necessarily bank deposits or bank
lending; the implications for prudential regulation ought to extend to the internalization
of the social cost of liability dollarization of non-tradable firms beyond the implications
for credit risk of the banking system. Furthermore, as noted, in many instances optimal
policy requires taking into account the leverage and financial risks of firms, not only the
nature of their revenue. If some of the externalities identified are substantial, this
approach calls for a rethinking of the limits of prudential regulation well beyond banking
system risk.

Before reviewing prudential regulation, let’s remind ourselves of the two risks that a
dedollarizing prudential policy package may run into. First is the risk of going overboard.
There is very little guidance as to how to quantify currency discrimination, be it prices or
quantities, to match the underlying externality to eliminate excessive FD. If
antidollarization policy is set such that it has a significant effect on aggregate FD, for
example if regulation is set on the basis of aggregate quantitative ceilings, then there is
substantial risk of going overboard and trigger instability and excessive
disintermediation. The second risk is the risk of fighting the symptoms. Currency-
sensitive regulation may lead to shrinking duration of peso deposits and loans, with the
attendant risk consequences. Similarly, bank regulation may push financing to
unregulated institutions. There is a hole in the literature concerning how to best address
these issues of risk shifting which needs urgent filling; it is simply not possible to arrive
at sound currency-sensitive bank regulation in practice without taking into account these
substitution to harmful dollar alternatives.

It is tempting to refer to successful experiences of prudential regulation without realizing

that good regulation crucially depends on the availability of suitable substitutes for dollar
financing. Chile’s prohibition of bank dollar lending to non-tradable firms may be
appropriate for Chile given the availability of long-term UF-financing (in this case the
ban would remove a small level of latent FD that would arguably be excessive) but would
be inappropriate for a banking system without suitable peso alternatives, where a ban
would remove far more than excessive FD and would mean a collapse of financing,
especially long-term, to the non-tradable sector.

With these fundamental problems in mind, there appears to be enough analysis and
evidence to venture some key directions. If moral hazard is an important distortion for
some of the reasons identified, then dollar lending to non-tradable firms ought to be
subject to higher requirements, such as higher capital and liquidity requirements, or
higher deposit insurance and liquidity provision premia. Alternatively, requirement could
take the form of quantity ceilings on dollar lending. Even if the underlying problem is
not moral hazard, market externalities leading to excessive dollar financing would merit
similar prudential responses, in that case as a tool of macrofinancial policy rather than a
disciplining device for banks and other financial intermediaries. However, to repeat, the
actual implementation of currency-sensitive regulation will require taking into account

40
the alternatives to dollar financing, both harmful and functional; these models are not
available at this time.

Finally, an intriguing and promising way to introduce currency sensitivity in prudential

regulation to avoid excessive FD is to fine tune the regulatory treatment of dollar claims
in default states. If as discussed in section III, FD is excessive because bankruptcy law
gives an opportunistic advantage to dollar financing due to its likely increase in value
under default leading to the reduction of the share assigned to peso financing, then the
policy proposal is to modify bankruptcy laws and regulations to convert dollar claims to
pesos for the purpose of bankruptcy according to a formula that eliminates the ex-ante
advantage and restores currency neutrality (and efficiency) to the financing currency
choice on this account. The similarity with Argentina’s surprise pessification of dollar
claims is superficial because this proposal is intended to be known ex-ante, at contracting
time.

In the same vein, another way to deal with moral hazard currency disparities in banking
would be to specify a penalized conversion formula for dollar claims for the purpose of
granting official help. This discriminatory approach at crisis time could be also used for
both currency of denomination and maturity in order to protect the core of the banking
system when official resources are not enough in systemic crises, as it was done in the
recent Uruguayan banking crisis in which sight deposits were fully insured and long-term
dollar deposits were reprogrammed for banks that needed help. Again, this proposed
approach to curtail excessive FD, and similar to the circuit breaker proposal in Ize,
Kiguel and Levy-Yeyati (2005) for managing liquidity risks under FD, is intended to be
formally stipulated ex-ante, not as a surprise ex-post resolution mechanism. In this sense,
the Chilean approach to banking crisis resolution is an example of pre-programming
selective interventions in a transparent way. More generally, the existence of a special
crisis regime in which contracts (in our case dollar contracts) are altered in pre-specified
ways bears a clear parallel with the automatic debt roll over triggers (what was the name
of Buiter’s proposal?) and collective action clauses being proposed for international
bonds, which would enable or facilitate contingent debt restructuring.

41
 REFERENCES

Baliño, Tomás, Adam Bennet and Eduardo Borensztein (1999). “Monetary Policy in
Dollarized Economies,” IMF Occasional Paper 171.

Bordo, Michael, Christopher Meissner and Angela Redish (2002). How “Original Sin”
was Overcome: The Evolution of External Debt Denominated in Domestic Currencies in
the United States and the British Dominions 1800-2000. Mimeo.

Borensztein, Eduardo and Paolo Mauro (2002). “Reviving the Case for GDP-Indexed
Bonds”. IMF Policy Discussion Paper 02/10.

Broda, Christian and Eduardo Levy-Yeyati (2003). “Endogenous Deposit Dollarization”.

Federal Reserve Bank of New York Staff Papers, No. 160.

Burnside, Craig, Martin Eichenbaum and Sergio Rebelo, (2001). “Hedging and Financial
Fragility in Fixed Exchange Rate Regimes”. European Economic Review 45 .

Caballero, Ricardo, Kevin Cowan and Jonathan Kearns (2003). “Dollar-risk, banks, and
fear-of-sudden stop: Lessons from Australia and Chile”. IDB Working Paper, RE1-04-
014.

Caballero, Ricardo and Arvind Krishnamurty (2003). “Excessive Dollar Debt: Financial
Development and Underinsurance”. Journal of Finance Vol. 58, No 2 , pp. 867-893.

Caballero, Ricardo and Arvind Krishnamurty (2004). “Smoothing Sudden Stops. Journal
of Economic Theory” Journal of Economic Theory Vol. 119, pp. 104-127.

Calvo Guillermo and Eduardo Fernández-Arias (1998). “The New Features of Financial
Crisis in Emerging Markets” in E. Fernández-Arias and R. Hausmann, eds. Wanted:
World Financial Stability, Inter-American Development Bank.

Calvo, Guillermo and Pablo Guidotti (1989). “Credibility and nominal debt: exploring
the
role of maturity in managing inflation”. IMF Working Paper 89/73.

Calvo, Guillermo, Alejandro Izquierdo and Luis-Fernando Mejía (2004). “On the
empirics of sudden stops: the relevance of balance-sheet effects”. IDB Working Paper
#509.

Calvo, Guillermo and Carmen Reinhart (2002). “Fear of Floating”. Quarterly Journal of
Economics 117.

Chang, Roberto and Andrés Velasco (2003). Endogenous Dollarization and Monetary
Policy. Mimeo, IDB.

42
Chang, Roberto and Andrés Velasco (2000). “Exchange Rate Policy for Developing
Countries”. American Economic Review 90.

Catão, Luis and Marco Terrones (2000). “Determinants of Dollarization – The Banking
Side”. IMF Working Paper 00/146.

Claessens, Stijn, Daniela Klingebiel and Sergio Schmukler (2003). “Government Bonds
in Domestic and Foreign Currency: The Role of Macroeconomics and Institutional
Factors”. Mimeo, The World Bank.

De La Torre, Augusto and Sergio Schmukler (2003). “Coping with risk through
mismatches: domestic and international financial contracts for emerging economies”.
Mimeo, IDB.

De Nicoló, Gianni, Patrick Honohan, and Alain Ize (2003). “Dollarization of the Banking
System: Good or Bad?” Mimeo, IMF.

Dooley, Michael (2000). “A Model of Crises in Emerging Markets”. The Economic

Journal, Vol. 110, no. 460, pp. 256-272.

Eichengreen, Barry, Ricardo Hausmann, and Ugo Panizza (2003). “Currency

mismatches, debt intolerance and Original Sin: why they are not the same and why it
matters”. NBER Working Paper 10036.

Eichengreen, Barry, Ricardo Hausmann, and Ugo Panizza (2002). “Original Sin: The
Pain, the Mystery and the Road to Redemption”. Mimeo, IDB.

Eichengreen, Barry and Ricardo Hausmann (1999). “Exchange Rates and Financial
Fragility”. NBER Working Papers 7418.

Fernández-Arias, Eduardo and Davide Lombardo (2002). “Market Discipline and

Exuberant Foreign Borrowing” in L. Hernández and K. Schmidt-Hebbel, eds. Banking,
Financial Integration, and International Crisis, Santiago, Chile. Central Bank of Chile.

Galiani, Sebastián, Eduardo Levy-Yeyati and Ernesto Schargrodsky (2003). “Financial

dollarization and debt deflation under a currency board”. Emerging Markets Review, Vol
4, No.4.

Galindo, Arturo and Leonardo Leiderman (2005). “Living with dollarization and the
route to Dedollarization”. Mimeo, IDB.

Galindo, Arturo, Ugo Panizza and Fabio Schiantarelli (2003). “Debt composition and
balance sheet effects of currency depreciation: a summary of the micro evidence”.
Emerging Markets Review, Vol 4, No.4.

43
Garcia, Gillian (1999). “Deposit Insurance - A Survey of Actual and Best Practices”. IMF
WP/99/54.

Goldstein, Morris and Philip Turner (2002). “Currency mismatching in Emerging

Economies”. Bank for International Settlements, Basel

Guidotti, Pablo and Carlos Rodriguez (1992). “Dollarization in Latin America:

Gresham’s Law in Reverse?” IMF Staff Papers, Vol. 39.

Gulde, Anne-Marie, David Hoelscher, Alain Ize, David Marston, and Gianni De Nicoló
(2004). “Financial Stability in Dollarized Economies”. IMF Occasional Paper 230.

Hausmann, Ricardo (2003). “Good credit ratios, bad credit ratings: the role of debt
structure”. Presented at Conference on “Rules-Based Fiscal Policy in Emerging Market
economies”, February 14 to 16, 2002, Oaxaca, Mexico.

Herrera, Luis Oscar and Rodrigo V. Valdés (2003). “Dedollarization, Indexation and
Nominalization: The Chilean Experience”. IDB Working Paper, RE1-04-015.

Ize, Alain and Andrew Powell (2003). “Prudential Responses to De Facto Dollarization”.
IDB Working Paper, RE1-04-013.

Ize, Alain and Eduardo Levy-Yeyati (2003). “Financial Dollarization”. Journal of

International Economics, 59.

Kaminsky, Graciela and Carmen Reinhart (1999).”The Twin Crises: The Causes of
Banking and Balance of Payments Problems”. American Economic Review.

La Porta, Rafael, Florencio López-de-Silanes, Andrei Shleifer, and Robert W. Vishny,

(1997). “Legal Determinants of External Finance”. The Journal of Finance, LII(3):1131-
50.

Levy-Yeyati, Eduardo (2004). “Dollars, Debt and the IFIs: Dedollarization Multilateral
Lending”. Mimeo.

Levy-Yeyati, Eduardo (2003). “Financial Dollarization: where do we stand?”. Mimeo,

IDB.

Levy-Yeyati, Eduardo and Federico Sturzenegger (forthcoming). “Classifying Exchange

Rate Regimes: Deeds vs. Words”. European Economic Review.

Licandro, Gerardo and José Antonio Licandro (2002). “Anatomía y Patología de la

Dolarización”. Revista de Economía del Banco Central del Uruguay, Vol. 9 (2).

Licandro, Gerardo and José Antonio Licandro (2003). “Building the Dedollarization
Agenda: Lessons from the Uruguayan Case”. Mimeo, Central Bank of Uruguay.

44
McKinnon, Ronald and Huw Pill (1997). “Credible Economic Liberalizations and
Overborrowing”. American Economic Review, Vol. 87:2, pp. 189-193.

Rajan, R. (2004). “How Useful Are Clever Solutions?”. Finance & Developments.

Reinhart, Carmen, Kenneth Rogoff and Miguel Savastano (2003). “Addicted to Dollars”.
NBER Working Paper No.

Ramey, G. and V. A. Ramey, (1995), "Cross-country evidence on the link between

volatility and growth," American Economic Review.

Savastano, Miguel (1996). “Dollarization in Latin America: Recent Evidence and Some
Policy Issues”. IMF Working Paper 96/4.

Shiller, Robert (1998). “Indexed Units of Account: Theory and Assessment of Historical
Experience” in F. Lefort and K. Schmidt-Hebbel, eds. Indexation, Inflation and
Monetary Policy, Santiago, Chile. Central Bank of Chile.

Schneider, Martin and Aaron Tornell (2000). “Balance Sheet Effects, Bail-out
Guarantees, and Financial Crisis”. Mimeo, UCLA

Thomas, L.R. (1985). “Portfolio Theory and Currency Substitution”. Journal of Money,
Credit, and Banking, Vol. 17.

Tirole, Jean (2002). “Financial crises liquidity and the international monetary system”.
Princeton, NJ: Princeton University Press

Wright, Mark (2004). “Private Capital Flows, Capital Controls, and Default Risk”.
Mimeo.

45
 Table 1 - Financial Dollarization in Developing Countries (1996-2001 average)

Index
Foreign currency Foreign currency
bank deposits domestic public External Debt
Country Composite
(% of broad debt (% domestic (% of GNI)
money) public debt)
(1) (2) (3) ( 4 )=( 1 )+( 2 )+( 3 )
Ecuador 70 90 90 250
Bolivia 80 70 70 220
Uruguay 90 90 40 220
Argentina 60 90 50 200
Nicaragua 70 0 100 170
Peru 60 40 60 160
Paraguay 50 60 40 150
Honduras 30 0 100 130
Jamaica 40 0 70 110
Guyana 0 0 100 100
Costa Rica 40 20 30 90
El Salvador 20 30 30 80
St. Kitts and Nevis 30 0 50 80
Brazil 0 30 40 70
Guatemala 0 40 30 70
Chile 10 10 50 70
Belize 0 0 70 70
Haiti 30 0 30 60
Trinidad and Tobago 20 0 40 60
Venezuela 10 0 50 60
Dominica 0 0 60 60
Grenada 0 0 60 60
St. Vincent and the Grenadines 0 0 60 60
Colombia 0 10 40 50
Mexico 10 0 40 50
St. Lucia 0 0 40 40
Dominican Republic 0 0 30 30
Mean of LAC 26.7 21.5 54.4 102.6
Mean of non LAC 17.4 1.9 63.1 82.4
75-percentile LAC 45 35 65 140
75-percentile non LAC 30 0 90 100
Source: Reinhart, Rogoff and Savastano (2003) and author's calculations
Note: Percents rounded to the closest ten (and truncated at 100%)

46
 Table 2 - Asset sources of liabilities dollarization (end 2001 as a % of GDP)
(non-industrial economies excluding offshore centers)

Domestic
Domestic External Net foreign Liability Domestic liability
dollar
Country dollar lending lending savings dollatization dollarization (%)
saving
( D*-fD ) ( E* ) (E - fD ) L*
( D* ) Min Max
(1) (2) (3) (4) (5) = (1)+(2) = (3)+(4) ( 6a ) = (1)/(5) ( 6b ) = (3)/(5)

Nicaragua 51 228 84 195 279 0.18 0.30

Uruguay 46 47 91 2 93 0.49 0.98
Croatia 42 45 74 13 88 0.49 0.85
Philippines 16 68 31 53 84 0.19 0.37
Indonesia 11 67 20 57 77 0.14 0.26
Argentina 17 58 26 50 76 0.23 0.34
Jamaica 9 59 36 32 68 0.13 0.53
Turkey 14 53 25 41 67 0.21 0.38
Moldova 8 56 20 43 64 0.12 0.32
Peru 17 38 31 24 55 0.31 0.57
Egypt 17 33 38 13 51 0.34 0.75
Bulgaria 16 31 35 12 47 0.33 0.75
Malaysia 4 43 18 28 46 0.08 0.39
Chile 5 41 26 20 46 0.10 0.56
Hungary 8 36 16 28 44 0.18 0.36
Thailand 1 43 15 30 44 0.03 0.33
Slovak Republic 11 27 20 17 37 0.28 0.53
Lithuania 9 23 16 16 32 0.29 0.51
Dominican Republic 7 25 18 14 32 0.21 0.55
Guatemala 0 31 15 16 32 0.01 0.48
Latvia 10 20 21 9 30 0.33 0.69
Czech Republic 8 20 32 -4 28 0.29 1.13
Estonia 7 19 20 6 26 0.28 0.77
Romania 5 21 11 15 26 0.18 0.42
Kazakhstan 7 18 12 14 26 0.29 0.47
Poland 7 18 18 7 25 0.30 0.71
South Africa 4 18 18 4 22 0.16 0.81
Mexico 2 19 12 9 21 0.09 0.57
Venezuela 0 18 24 -5 18 0.00 1.30
Source: Eduardo Levy-Yeyati (2004) and author's calculation

47
 Table 3 - Elasticity of liability dollarization ( l ) to domestic FD ( d )
and external FD ( e ) ( end 2001 )
(non-industrial economies excluding offshore centers)

Liability Domestic Domestic Domestic External

Country dollarization FD elasticity elasticity elasticity
(l) (d)
(εˆ dl ) (ε dl ) (ε el )
(1) (2) (3) (4) (5)
Nicaragua 0.93 0.80 0.75 0.30 0.82
Argentina 0.92 0.81 0.76 0.34 0.77
Uruguay 0.92 0.92 0.87 0.98 0.51
Moldova 0.89 0.71 0.70 0.32 0.88
Turkey 0.87 0.72 0.72 0.38 0.79
Peru 0.86 0.78 0.79 0.57 0.69
Romania 0.84 0.69 0.71 0.42 0.82
Croatia 0.84 0.81 0.85 0.85 0.51
Kazakhstan 0.82 0.68 0.72 0.47 0.71
Bulgaria 0.80 0.75 0.82 0.75 0.67
Lithuania 0.75 0.60 0.70 0.51 0.71
Latvia 0.70 0.62 0.77 0.69 0.67
Jamaica 0.70 0.55 0.69 0.53 0.87
Philippines 0.70 0.46 0.58 0.37 0.81
Indonesia 0.65 0.32 0.43 0.26 0.86
Guatemala 0.62 0.43 0.61 0.48 0.99
Hungary 0.58 0.34 0.50 0.36 0.82
Mexico 0.56 0.42 0.65 0.57 0.91
Dominican Republic 0.55 0.40 0.64 0.55 0.79
Chile 0.54 0.40 0.64 0.56 0.90
Venezuela 0.54 0.60 0.97 1.30 1.00
Egypt 0.49 0.42 0.74 0.75 0.66
Estonia 0.49 0.42 0.75 0.77 0.72
Poland 0.44 0.36 0.71 0.71 0.70
Slovak Republic 0.43 0.29 0.58 0.53 0.72
Malaysia 0.33 0.16 0.43 0.39 0.92
Czech Republic 0.33 0.36 0.94 1.13 0.71
Thailand 0.32 0.14 0.37 0.33 0.97
South Africa 0.29 0.25 0.74 0.81 0.84
Source: Eduardo Levy-Yeyati (2004) and author's calculation

Note: d
εˆdl = (0.89) from regression-based estimating equation lˆ = 019 + 0.87d
l

d D 1E
ε dl = and ε el =
l L lL

48