Currency Induced Credit Risk in a Dollarized Economy1

Abstract

This study analyzes the effects of exchange rate fluctuations on corporate credit default in a dollarized
economy. The application of a new regulation on currency induced credit risk (CICR) in the Peruvian
banking system, and a posterior episode of exogenous exchange rate shock, created the condition of a
natural experiment to compare exposed and not-exposed corporate borrowers. This study uses two
definitions of currency exposition: debt dollarization and a self-reported indicator of CICR produced by
private banks as established by the new regulation. Contrary to literature using aggregated or bank-level
data of debt dollarization as a proxy for CICR, this study uses firm-level data to find that the effects of
CICR and debt-dollarization on credit risk are opposed. While CICR increases default, debt dollarization
reduces it. Our results suggest that banks transfer exchange risk as a hedging mechanism by lending in
dollars only to high quality borrowers. Consequently, debt-dollarization and CICR become disconnected.

JEL : E4, E5, C5 and C8.

Keywords : Duration Models, Credit Risk, Exchange Rate Fluctuations.

1
The authors thank Enrique Mendoza, Christopher Cummings, Stephen Kay, Tapen Sinha, Myriam Quispe-Agnoli
and the people from the Federal Reserve Bank of Atlanta for their comments during the visit of one of the authors
to their Research Department. We are also grateful to Javier Poggi and Michel Canta from the Superintendencia
de Banca y Seguros del Peru for letting us use their database and for their support to this research.
1. Introduction

One major issue discussed in the literature on dollarization is the exposure of firms to exchange

rate volatilities due to currency mismatches (see for example, Ennis, 2000 and Broda & Levy-Yeyati,

2000). In a dollarized economy, banks face two options: to keep the currency risk and lend in local

currency, or to transfer the currency risk to firms by lending in dollars. In the former option, the bank is

transforming currency risk into credit risk. This risk transferred to firms is the Currency Induced Credit Risk

(CICR) which captures the impact of exchange rate fluctuations on firms’ credit default.

In dollarized economies2, transferring currency risk to firms will be an optimum hedge mechanism

only if the CICR is smaller than the currency risk held by the bank. That would be possible only if firms can

fully or partially hedge CICR, which means they have mechanisms to offset currency fluctuations. In

emerging economies, market mechanisms for hedging like futures, derivatives or swaps are very limited

and only available for big corporations, resulting in the use of self-hedging mechanisms to cover risks.

One of these mechanisms is currency matching that requires firms to generate income in dollars and

match their dollar liabilities to assets. Another mechanism is to keep enough reserves to cover for

unexpected income fluctuations, including currency volatility. This mechanism does not require currency

matching or dollar income generation for the firm, but only explicit or implicit reserves to hedge against

currency fluctuations.

Some attempts to measure the effect of currency fluctuations on CICR have considered

aggregate country data (Cayazzo, et al., 2006; Reinhart and Kaminsky, 1999; Gulde et al., 2004; De

Nicoló, et al., 2003; Magud, 2009) as well as firm-level data (Bleakley and Cowan, 2006 and Carranza,

Cayo and Galsón-Sanchez, 2003, Cowan et al., 2005, Galindo, et al., 2003). However, in all cases debt-

dollarization has been considered as proxy for CICR. The rationality for this approximation rests on the

assumption that CICR can only be hedged by currency mismatching when market mechanisms are

absent. Consequently, highly debt-dollarized firms are more likely to have currency mismatches.

Using aggregated country data, Cayazzo et al. (2006) studied dollarized vulnerabilities in small

economies using growth of nonperforming loan ratios and estimates of annual loan provisions. The study

2 Machicado (2008) shows how liquidity uncertainty may increase the degree of dollarization in the banking system.
determined that the impact of credit risk devaluation is higher in countries with higher dollarization.

According to the findings of Cayazzo, et al., the quantitative effect of a shock on exchange rate credit risk

varies substantially across countries with a threshold effect that depends on the degree of dollarization. In

particular, Cayazzo et al. articulated a supervisory framework that addressed partially dollarized banking

systems’ vulnerabilities, which loosely follows the Basel Committee for Banking Supervision (BCBS).3

Additionally, the effect of exchange rate fluctuations has also been discussed by Ize and Powell (2004)

who identified how partial dollarization increases the vulnerability of financial systems’ liquidity risks. Few

studies have used firm-level data to analyze the impact of currency fluctuations. Bleakley and Cowan

(2006) used micro data of firms for Argentina, Brazil, Colombia, Chile, México, Peru, and Venezuela from

1990 to 1999. They studied the firm’s mismatches between short and long term debt during financial crisis

episodes. Bleakley and Cowan found that capital flight that emerging countries bear during crisis did not

affect the structure of firm’s liabilities. This result contrasts a previous study by Carranza, Cayo and

Galsón-Sanchez (2003).4 Also, Serieux (2009) find in a Paraguayan sample of firms that partial

dollarization led to negative balance sheet effects, in the form of reduced access to investment credit due

to depreciation-induced reduction in firms’ net worth as a result of currency mismatches on their balance

sheets5.

Using firm’s debt-dollarization as a proxy for CICR imposes an important limitation in applied

studies. First, it omits other hedging mechanisms that firms may use to reduce CICR, mainly additional

reserves. Second, and most importantly, it could understate the real effect of currency fluctuations on

credit default by assuming that banks are naïve and passive. Banks can reduce overall risk by transfering

currency risk to the most solvent firms--those who have more reserves to hedge CICR. In that way, banks

will replace a high currency risk with a smaller CICR. This risk management strategy would have two

empirical consequences. First, the most solvent firms--which are not necessarily the currency matched

firms--will also be the most debt-dollarized. Second, currency fluctuations would do less harm to the most

3 The Basel Committee for Banking Supervision (BCBS) established a comprehensive framework for the oversight of
banking activities. This framework was revised in the context of Basel II and it specifies several recommendations to
align the capital measurement with sound and contemporary practices in banking and promotes further
improvements in risk management. The specific documents on the management and supervision of the main banking
risks, including credit market and liquidity risks are, in principle, applicable to all banking systems.
4
These authors concentrated their analysis on the Peruvian banking system.
5
Same result is obtained by Fuentes (2009) for a Chilean sample data of firms.
 debt-dollarized firms. Consequently, debt-dollarization will not be a good proxy for CICR, and its use

could bias the real impact of CICR.

The first contribution of our study is to overcome the limitations of using debt-dollarization as a

proxy for CICR. We take advantage of unique firm-level data from a dollarized emerging country that

includes separate estimates for debt-dollarization and CICR. CICR was internally estimated by banks

following a new regulation that was implemented in Peru in 2005.

The second contribution of this study is that it estimates the impact of debt-dollarization and CICR

on credit risk separately in a period of an exogenous currency shock. The implementation of the CICR

regulation in a period of highly stabilized currency fluctuation, and a posterior temporal currency shock

originated by a political event, created the condition of a natural experiment to evaluate the impact of

exposed and not-exposed corporate borrowers on credit default, using debt-dollarization and self-

reported CICR as measures of CICR exposition.

Our study shines more light on the discussion of financial dollarization and balance sheet effects

in dollarized economies. It also provides more information on how banks transfer and hedge currency risks

by lending in dollars to the most solvent firms. Debt-dollarization could be the result of an efficient risk

management strategy when markets for financial hedging products are underdeveloped. Consequently,

policies that put limits to debt-dollarization may produce more harm than benefits.

The paper is organized as follows: Section 2 describe the CICR regulation in Peru. Section 3

present the data and methodology performed. Section 4 shows the results and the last section concludes.

2. The Supervision of Currency Induced Credit Risk

During the 1980´s, the Peruvian economy experienced hyperinflation. Large government

spending driven by populism and ambitious investment programs produced fiscal deficits and

unsustainable foreign debt. The government started to print money to accomplish their obligations with the

international financial markets, thus creating high rates of inflation. This induced a recomposition of

currency portfolios toward dollars (IMF, 2001). After Peru’s stabilization, the preference for foreign

currency is still high. The share of dollar debt from 2005 to 2006 was two thirds of total loans.
 In 2005, Peru’s banking system regulator, the Superintendencia de Banca y Seguros (SBS),

issued the regulation SBS N°041-2005, which authorized the adoption of a surveillance monitoring

system to identify, measure, control and report accurately loan level exposure. The reports were required

to include evaluation of control mechanisms as well as the corrective actions or mandatory improvements.

At the same time, the legislation compelled the Director to implement policies and procedures to identify

and manage CICR.

The regulation establishes three different forms of CICR’s regulation: limit of exposure,

provisions, and capital requirement. If a bank does not identify its exposure then it is penalized with

additional provision. Exposed debtors can be identified by an assessment of their capacity to pay versus

flow of payment. Banks are allowed to use their own internal methodologies to identify debtors with CICR,

and their methodologies are subsequently reviewed by the SBS.

Regarding a debt’s qualification, banks should take corrective actions in case the CICR

affects the payment capacity of a firm. The regulator requires doing an annual stress test of the bank´s

liabilities.

3. Data and Methodology

3.1 Data

After a long period of low volatility in the exchange rate (see Figure 1), the nuevo sol plummeted

from 3.25 to 3.45 per dollar between August 2005 and mid January 2006, which represented a devaluation

of 6%. The electoral environment in Peru triggered the devaluation of the local currency when national

surveys reported that the leftist candidate, Ollanta Humala, placed second in the polls for the presidential

race. There was a financial turmoil in the Peruvian financial markets because of the possibility of having a

leftist party win the presidency.

In spite of the significant devaluation of the Peruvian nuevo sol, it did not produce a currency

collapse in the financial system, and it is clear that this devaluation was caused by political expectations

rather than current macroeconomic conditions, implying that this was a truly exogenous shock

independent of credit conditions. Coincidentally, the CICR regulation passed in January 2005, eight

months before the dollar volatility. The law provided that banks should start reporting CICR indices in June
2005, 3 months before the devaluation, suggesting that the first reports were exogenous to the exchange

rate shock. This temporal exogenous shock in the exchange rate created the condition of a natural

experiment to test the role of CICR and debt-dollarization in the impact of exchange rate shocks on

corporate credit default.

[Insert Figure 1]

The data in the analysis include detailed debt information of 42,059 small and medium firms

during the period June 2005 – July 2006 (see Figure 1). The data was reported by the four biggest

Peruvian banks, which concentrate more than eighty percent of loans in the system and have the most

reliable internal methodologies for CICR. The information was obtained from the Credit Bureau

administered by the SBS.

Our analysis search to examine the empirical relationship between the hazard rate and the factor

that induces the firms to default. Default was defined as a binary variable that equals 1 when the loan is

past due for more than 90 days. In order to investigate these factors, we applied a parametric duration

model where default is the event that defines the hazard. Similar strategies to estimate credit default have

been considered in the literature6. A parametric model was considered over a semi-parametric

specification after rejecting the hypothesis of proportionality of hazard function. Using a score test based

on scaled Schoenfeld residuals, we rejected the null hypothesis of proportionality at 1% of significance

level.

3.2 Methodology

Let T denote a continuous non-negative random variable representing survival time, with

probability density function (pdf) f(t) and cumulative distribution function (cdf) F(t) = Pr{T ≤ t}. We focus on

the survival function S(t) = Pr{T > t}, the probability of being alive at t, and the hazard function (t) =

f(t)/S(t).

6
See for example Belloti and Crook (2007) who analyzed the macroeconomic determinants of credit card defaults
using a logistic duration model.
 ∫ t0 λ (u )du
Let (t) = denoted the integrated hazard and recall S(t) = exp{- (t)} (see Wooldridge,

2002). We assume “t” distributed as exponential. In an exponential distribution, we could let the

parameter λ depend on a vector of covariates x, in the following way:

log λ = x' β ……. (1)

Where X is a vector of variables that may determine the variable λ . In our case λ is the time span until a

firm has a debt past due 90 days (default) and the X consider the variables presented in table 1.

[Insert Table 1 about here]

For the estimation we considered the following variables listed in Table 1. Basic descriptive

statistics are included.7

Spells in our duration model were right censored. We selected firms that were not in default at

baseline, and we followed them for 14 months until July 2006, when 13.9% of firms defaulted in or before

July 2006 (completed spells), and 86.1% remained solvent (right censored spells).

[Insert Table 1]

We expected the variable CICR increased the probability of credit default during this period of

temporal shock in the exchange rate. If a firm has CICR then its probability of default should be higher

because it is more exposed to unexpected devaluation. For the firm’s debt dollarization, however, we

expect a negative relationship with default if our hypothesis that banks lend in dollars to the most solvent

firms is valid. A positive relationship would indicate that banks do not reduce total risk when they lend in

dollars, and therefore debt dollarization conveys additional CICR.

We control for observable firm’s characteristic. Because default is defined using the number of

past due days, we control by the delayed days in payment (DAYS), which is, by construction, less than 90

7
DEFAULT variable is the Time Span until a firm declares in credit default for firm I which is under assessment.
days. Firms with a higher delay in days should have a higher probability of default. We also include a

dummy variable for size of the firm (SIZE_SMALL), the number of banks where the firm holds debt

(BANKS), the ratio of collaterals to debt (COVERAGE), the ratio of non-performing loans to total debt

(NPL), and the total debt in the banking system (DEBT).

4. Results

We have considered four different estimations. Each specification is presented in Table 2. All

estimations reported robust standard errors, and a Variance Inflation Factor (VIF) test eliminated the

possibility of multicollinearity problems. As a robustness check, different parametric specifications were

tried, without major impacts on our results. Although the semi-parametric Cox proportional hazard model

was ruled out after testing for proportionality, the results were also similar to the parametric counterpart.

[Insert Table 2]

The first model replicates the traditional approach of using dollarization as a proxy for CICR

without controlling for observable factors. The hazard ratio (HR=0.682) describes a negative and

statistically significant impact on probability of default. When CICR is included (model II), the hazard ratio

for dollarization remains significant at 1% but becomes even smaller (HR=0.669). On the other hand, the

hazard ratio for the CICR indicator is statistically significant, and it suggests a positive relationship

between CICR exposure and probability of default (HR=1.18).

Adding controls to the regression (Models III and IV) reduces the impact of dollarization and

CICR, but both variables remain statistically significant. In the model with only dollarization (Model III), the

hazard ratio suggests a lower but still negative impact on default (HR=0.825). As before, adding the CICR

amplifies the impact of dollarization on probability of default (HR=0.808). In this model, CICR is still

significant, and, although its impact default is still positive, its magnitude is smaller than in the model

without controls (HR=1.067).

The impact of control variables reported in Models III and IV were as expected, with the

exceptions of firm’s size and coverage, which were not statistically significant. The number of banks where
the firm has loans was also statistically significant. The results show that the larger the number of banks,

the higher the probability of default, a relationship that deserves further exploration, but that is out of the

scope of this paper.

5. Discussion and Implications

We estimate the impact of debt dollarization on credit default in a period of a temporal exogenous

shock in the exchange rate. Our results show that firms with a higher debt dollarization are 0.825 times

less likely to default than firms with lower debt dollarization. This result confirms our hypothesis that banks

lend in dollars to the most solvent firms as a mechanism to hedge currency risk. By selectively lending in

dollars, banks are efficiently transforming a certain level of currency risk into a lower level of CICR. This

result is similar to what Cowan et al. (2005) found for Chilean non-financial firms. The authors used a firm

level study to explore determinants and consequences of currency mismatches in Chile, finding that in

periods following a depreciation firms with higher dollar debt did not underperform their lower dollar debt

counterparts.

This result suggests that dollarization is not a good proxy for CICR, and that empirical papers that

rest on this assumption are over-estimating the real impact of CICR on creditworthiness. We can also see

this when we include the measure of CICR estimated internally by banks. When CICR is included in the

regressions, the negative impact of dollarization on default becomes even higher. Our paper benefits from

this unique measure of CICR, which proved to be consistent with the data. As expected, firms that are

exposed to CICR as reported by this measure are 1.067 more likely to default than non-exposed firms

during a devaluation period. Although the effect is small in magnitude after controlling for a firm’s

observable characteristics, this impact is statistically significant. Most importantly, the sample design used

in this study guarantees that the CICR indicator was exogenous to the exchange rate shock experienced

during the period of analysis.

Our study sheds light that CICR and DOL have different effects on creditworthiness in periods of

exchange rate turbulence in a dollarized economy. Our results contrast with the standard claim in the

literature that dollar denominated debt poses excessive risk to financial institutions, and that firms with
high levels of debt dollarization cannot reduce currency risk, imposing an equal level of credit risk to

banks. Indeed, our results suggest that solvent firms have a cushion to assimilate part of the currency risk,

lowering the level of credit risk to banks. In this way, banks can partially hedge their own currency risk by

lending in dollars to the most solvent firms. This process creates a gap between debt dollarization and

CICR, making these two variables disconnected. Although our study supports this hypothesis, it is not

conclusive. Better firm level data is required to explore and understand this process.

This study is relevant to policymakers in dollarized economies. It presents a successful regulation to

control CICR in a highly dollarized developing country. Banks’ internal methodologies to identify CICR

were consistent in predicting a positive impact on default during a shock to the exchange rate. To our

knowledge, there is no other country that has implemented a similar mechanism to control the effect of

exchange rate volatilities on credit default. Most countries (Argentina and Chile, for example) use limits to

loans in foreign currency (US dollars), so rather than focusing on CICR, these regulations focus on debt-

dollarization as a proxy for CICR. As is shown in this study, these two terms are disconnected. Most

importantly, restricting the ability to lend in dollars would impose an important restriction to banks that use

dollar lending as a hedging mechanism to reduce currency risk. A regulation that focuses on debt

dollarization rather than on a measure of CICR would produce more harm than benefits, because it would

limit the banks’ ability to reduce risk by transforming currency risk into lower levels of CICR. In that regard,

the Peruvian experience, which focused on CICR, deserves more attention from other bank regulators.

6. References

Belloti, T., and Crook J. 2007 “Credit Scoring with macroeconomic variables using survival Analysis”

Working Paper 07-15. Wharton School of Business. Financial Institutions Center.

Bleakley, H., and Cowan, K. 2006. Maturity Mismatch and Financial Crises: Evidence from Emerging

Market Corporations. Federal Reserve Bank of San Francisco, WP.

Cowan K., Hansen E., Herrera L., 2005. “Currency Mismatches, Balance Sheet Effects And Hedging In

Chilean Non-Financial Corporations” WP 346. Banco Central de Chile.

Broda, C., and Levy-Yeyati, E. 2003. “Endogenous deposit dollarization.” Number 160. Federal Reserve

Bank of New York. February.

Carranza, L., Cayo J. and Galsón-Sanchez J. 2003. “Exchange rate volatility and economic performance

in Peru: a firm level analysis.” Emerging Markets Review, pages 472-496.

Cayazzo, J., Garcia Pascual, A., Gutierrez, E., and Heysen, S. 2006. “Toward an effective supervision of

partially dollarized banking systems. Working Paper 0632. Interamerican Development Bank

(IADB). January.

Ennis, H. 2000. “Banking and the political support for dollarization.” WP 00-12. Federal Reserve Bank of

Richmond.

Fuentes, Miguel. 2009. “Dollarization of debt contracts: Evidence from Chilean Firms” Developing

Economies Journal. Vol 47. Issue 4. Pages 458-487.

Galindo, A., Paniza, U., and Schiantarelli, F. 2003. “Debt composition and balance sheet effects of

currency depreciation: a summary of the micro evidence”. Emerging Markets Review, pages 330-

339.

Gulde, A., et al. 2004. “Financial stability in dollarized economies” Ocasional Paper 20. International

Monetary Fund (IMF).

International Monetary Fund (IMF, 2001) “The Decline Of Inflation In Emerging Markets: Can It Be

Maintained?” World Economic Outlook. Chapter 4.

Ize, A. and Powell, A. 2004. “Prudential Responses to De facto Dollarization” WP 04/66. Torcuato Di Tella.

Centro de Investigación en Finanzas (CIF).

Machicado C. 2008. “Liquidity shocks and the dollarization of a banking system”. Journal of

Macroeconomics. Vol. 30. pp 369-81

Magud N. 2009. “Currency mismatch, openness and exchange rate regime choice” Journal of

Macroeconomics. Forthcoming. Pages 1-22 (In press)

Reinhart, C. and Kaminsky, G. 1999. “The twin crisis: The causes of banking and balance of payment

problems.” American Economic Review. No 89, pages 473-500.

Serieux, John. 2009. “Partial Dollarization, Exchange Rates, and Firm Investment In Paraguay”

Developing Economies Journal. Vol 47. Issue 1, pages 53-80.

Wooldridge, J., 2002. “Econometric Analysis of Cross Section and Panel Data”, Cambridge (MA) and

London (UK): The MIT Press.

 Appendix
Figure 1: Evolution of the PEN
(Soles for a Dollar)
3.50
3.45 Banks start
to implement
3.40 the CICR

3.35
3.30
3.25
3.20
3.15
jul-05
jun-05
ene-05
feb-05
mar-05
abr-05

ago-05
sep-05

dic-05
ene-06
feb-06
mar-06
abr-06

jun-06
ago-06
may-05

nov-05
oct-05

may-06
Note: PEN is the Peruvian Nuevo Sol denomination in the stock market. The highlighted area indicates the
period of high volatility due to electoral turbulence.

Table 1: Description of variables

Variable Description Standard

Mean
deviation
CICR Dummy that takes the value of 1 (one) if it has Currency Induced Credit Risk. 0.158 0.365

DOL Ratio of dollarization for the firm 0.238 0.378

DAYS Delayed days in payment for firm i 2.967 35.547

SIZE_SMALL Dummy that takes the value of 1 (one) if it is a small firm 0.974 0.160

BANKS Number of banks where the firm has a debt 2.391 1.381

COVERAGE Ratio of coverage (guarantee/debt within the base month) for the firm 50.284 8166.611

NPL Ratio of nonperforming loans in the financial system during the period 0.0191 0.116

DEBT Total debt in the financial system 986170.4 1.14e+07

Table 2: Hazard Rate Regressions Results (Dependent Variable: Time Span until a firm declares in credit default)

Explanatory variables Model I Model II Model III Model IV

CICR 1.18 *** 1.067 *

(0.040) (0.036)
DOL 0.682 *** 0.669 *** 0.825 *** 0.808 ***
(0.026) (0.025) (0.033) (0.032)
DAYS 1.000 *** 1.000 ***
(0.000) (0.000)
SIZE_SMALL 0.931 0.904
(0.173) (0.161)
BANKS 1.425 *** 1.422 ***
(0.013) (0.013)
COVERAGE 0.999 0.999
(0.000) (0.000)
NPL 65.260 *** 61.972***
(1.981) (1.847)
DEBT 0.999 *** 0.999 ***
(0.000) (0.000)
Number of subjects 42,059 42,059 42,059 42,059
Number of failures 5,846 5,846 5,846 5,846
Log pseudo-likelihood -24288.604 -24277.289 -16157.644 -17131.749
Robust standard errors in parenthesis.
*** Significant at 1%, ** Significant at 5%, * Significant at 10%