On the Spillover of Exchange-Rate Risk into Default Risk! Miloš Božović! Branko Urošević! Boško Živković!
2 Motivation Globalization and inflow of foreign capital Dollarization in emerging economies o On average, 40 to 45 percent of bank deposits in emerging markets are denominated in or indexed to a foreign currency (IMF, 2004). o Foreign-currency liabilities typically offset by domestic foreigncurrency loans. Positive aspects of dollarization: o Exchange rate risk fully hedged. o Banking system acts as an important source of foreign-currency liquidity. Negative aspect of dollarization: o Credit risk of foreign-currency loans increases in case of depreciation of the local currency.
3 This paper Studies the interaction between currency and credit risk in foreign currency loans. Analyzes the likelihood of default of a foreign currencyindexed/denominated loan when domestic currency depreciates. Using a very simple model explains the following negative feedback mechanism: domestic currency depreciates default risk increases supply of credit decreases GDP decreases
4 Related literature Duffie & Singleton (2003): o A thorough review of credit risk models. Jarrow & Turnbull (2000): o One of the first models of interaction between credit and market risk. Alessandri & Drehmann (2007): o Study issues related to separate estimation of market and credit risk. Breuer et al. (2008): o Compare regulatory capital based on integrated vs. separated approach to estimation of currency and credit risk. o Key result: regulatory capital for foreign-currency loans may be significantly underestimated if interactions between risks are ignored.
The model 5 Partial equilibrium setup with rational expectations. The economy lives for one period (dates t = 0 and 1). There is a single foreign currency with exogenous exchange rate X t. A single commercial bank: o At t = 0 borrows an amount B in foreign currency at the exogenous foreign interest rate r. o Immediately places B locally at higher interest rates. o At t = 1, the borrowers repay their loans, in part or in full. o No central bank; domestic currency in fixed supply; no inflation. There are i = 1, 2,..., n rational borrowers, each with its own compensated demand for money. At time 0, the bank lends a foreign currency amount b i to agent i, at the interest rate r + s i.
6 The model Payment ability of borrower i in local currency: A t,i Agent i pays the loan back in full only if Payment ability follows (cf. Breuer et al., 2008): where Y t is GDP at time t and are i.i.d. normal random variables. Total foreign currency amount repaid at time 1 by agent i:
7 The model Demand for loans m i has standard Hicksian elements: o Responds positively to expected foreign currency income at t = 1. o Responds negatively to interest rates. (cf. Goodhart et al., 2005). GDP endogenized by assuming its positive dependence on aggregate credit supply: where is exogenous normal shock to log GDP.
8 Bank's profit maximization The bank solves: where is the payment ability of borrower i in foreign currency at t = 0, 1, while is the foreign currency amount due at time 1.
9 The equilibrium The bank maximizes the expected future profit by chosing the optimal interest rates (i.e. spreads). The borrowers have compensated credit demand levels that give the optimal expected utility. In equilibrium, the aggregate demand for foreign currency loans has to be equal to the supply at any moment: where
10 Probability of default Borrower i defaults at time 1 with probability where
11 Proposition 1 Depreciation of the value of the local currency, ceteris paribus, leads to an increase in borrowers probabilities of default This increase is non-linear
Probability of default 12
13 Proposition 2 A decrease in the expected growth rate of GDP, ceteris paribus, leads to an increase in borrowers probabilities of default. I.e. if borrowers expected economic decline they are more likely to default on their loans
14 The expected profit The expected bank profit can be expressed as follows n S 0,i e E ln ( Y 1 /Y 0 ) ln ( X1 / X0 )+σ 2 Y +σ 2 A / 2 ( 1 N(d 1,i )) + K i N(d 2,i ) B(1+ r) i=1 where ( ) + E ln Y 1 / Y 0 d 1,i = ln S 0,i / K i ( ) ( ) + E ln Y 1 / Y 0 d 2,i = ln S 0,i / K i ( ) ( ) + σ Y 2 + σ A 2 / 2 ln X / X 1 0 σ A 2 + σ Y 2 σ A 2 + σ Y 2 ( ) σ A 2 / 2 ln X / X 1 0
15 The interpretation The expected revenue is the sum of the expected payment abilities of agents when they default at time 1 S 0,i e E ln ( Y 1 /Y 0 ) ln ( X 1 / X 0 )+σ 2 Y +σ 2 A / 2 ( 1 N(d 1,i )) and the face values of their loans multiplied by the probability of survival, when they do not default ( ) K i N(d 2,i ) = K i 1 N d 2,i = K i ( 1 PD i ) On the other hand, the expected cost is, simply B(1+ r)
16 First order conditions With respect to s i (cannot be solved in closed form) n j=1 S 0, j e g ln ( X 1 / X 0 )+σ 2 Y +σ 2 A / 2 1 N(d 1, j ) ( ) n S 0, j e g ln ( X 1 / X 0 )+σ 2 Y +σ 2 A n(d / 2 1, j ) σ 2 2 j=1 A + σ Y 1 φ β 1 e γ s n i γ bn(d ) s i 2,i e γ j j=1 K j n(d 2, j ) σ A 2 + σ Y 2 1 φ 1 B(1+ r) = 0,
17 Proposition 3 A weakening of the exchange rate leads to an increase in equilibrium spreads charged to a borrower, which, in turn, leads to a higher default probability
18 A numerical example Annualized interbank rate r = 0.05 Volatility for payment ability σ A = 0.2 Volatility of GDP growth rate σ Y = 0.04. The money demand function α = 0.1, β = 0.5, γ = 0.8 Endogenous growth parameters φ 0 = 0, and φ 1 = 0.1. Individual borrower s initial payment ability 10 percent above the borrowed amount, i.e. S 0,i = 1.1b i
19 Equilibrium total supply of credit as a function of exchange rate
20 The intuition Relative to the case when there is no change in the exchange rate, X 1 = X 0, the total supply of credit decreases when local currency depreciates, and viceversa. A higher exchange rate implies higher optimal spread, and hence lower equilibrium demand for credit. Foreign currency loans are more difficult to afford. Anticipating that, banks restrict the supply of credit
A decrease in credit supply feeds backs negatively onto the GDP 21
22 Spillover effect Local currency depreciation increases the default risk of a foreign-currency borrower: o This is a first-order effect that would exist even if GDP is assumed to be exogenous (Breuer et al., 2008). Higher default risk implies higher interest rates, combined with a lower aggregate supply of credit. Lower aggregate supply of credit has a negative effect on future GDP. Borrowers reduce their credit demand in a rational expectation of a decrease in GDP. The economy ends up in an equilibrium with reduced supply of credit, higher interest rates, higher default rates and reduced growth.
23 Spillover effect an illustration Serbian currency depreciated from 77.09 RSD/EUR on September 1, 2008 to 95.46 RSD/EUR on March 31, 2009. o X 1 / X 0 = 1.24 Baseline case no interaction of currency and credit risk: o Fixed supply of credit. o Exogenous GDP growth rate set to 0. Model equilibrium interaction included: o Parameter values set to r = 0.05, σ A = 0.2, σ Y = 0.04, α = 0, β = 0.68, γ = 0.8, φ 0 = 0, φ 1 = 0.1, and S 0,i = 1.1b i, o PD increases approximately by around 50% from its baseline value. o Equilibrium supply of credit decreases by 18%. o Annualized GDP growth rate drops by 4%.
24 Possibility of regional spillovers EU banks in the past 20 years rapidly expanded to Central, Eastern and South Eastern Europe (CESE) Clusters of countries dominanted by banks from relatively few countries Baltics dominated by Sweden (in Estonia 98% of assets in the hands of foreign banks) South East Europe dominated by Austrian and Italian banks (in Serbia also Greek, Slovenian, Hungarian, and a Russian bank) Even if fundamentals of a country is sound, a currency crisis in one country can cause the common lender to reduce loans to healthy countries, spreading the crisis (IMF 2009).
Alternatives to foreign-currency indexing 25 Denominate loans in domestic currency, but tie the interest payments to the consumer price index (CPI). Rather successfully done in Poland Inflation is typically positively correlated with the exchange rate. Why not used more often: Firstly, exchange rates are established on a daily basis and always involve some market mechanisms. The official CPI data are published by government-sponsored statistical offices once every month, and usually with a lag. Secondly, CPI is often calculated using a methodology not best representing the prevailing payment abilities of the borrowers.
Alternatives to foreign-currency indexing 26 Offer loans in domestic currency Use currency derivatives (swaps, for example) to hedge the exchange rate risk This would reduce the credit risk (this risk is harder to hedge or to measure) Burnside et al (2001) argue that banks, facing the significant likelihood of a bailout in case of serious difficulty may not hedge all of their exchange rate risk away (since that would be costly). This is particularly true for operating in countries with an explicit deposit insurance schemes.
Manipulating obligatory reserves 27 Regulators can manipulate the obligatory reserve requirements. Provide incentive to banks to either increase the supply of foreign currency-denominated credit or decrease the interest rates. In a floating or managed floating regime, one way is to reduce the required reserves for the foreign currency placements whenever domestic currency weakens.
Manipulating obligatory reserves 28 This decreases the regulatory costs for banks, which leaves them the opportunity to reduce the spreads in order to be more competitive while remaining profitable. Lower interest rates on foreign-currency loans should then lead to a higher demand for credit. The reduction of reserve requirements would, therefore, result in equilibrium where the credit supply and GDP growth rate are higher, probabilities of default are lower, while the bank remains at optimum
More accurate determination of risk-based capital 29 Traditionally, market and credit risk are treated separately (Pillar 1 of the Basel II). The overall regulatory capital is simply the sum of regulatory capitals for credit, market and operational risks. Some forms of market risk in the banking book, such as exchange rate and interest rate risk, do carry additional capital charges under Pillar 2 of Basel II. Erroneous belief that this is a conservative approach, driven by prevalence of VaR methodology in Basel II In spite of being a de-facto industry and regulatory standard, VaR is not a coherent risk measure. There might be situations in which VaR(X + Y) > VaR(X) + VaR(Y).
More accurate determination of risk-based capital 30 In practice, banks assets are simultaneously sensitive to both market and credit risk factors. Hence, any change in value of the total portfolio that is separated into a pure-market and a pure-credit risk change may lead to a serious underestimation of the total risk. This is the case whenever market and credit risk factors exhibit a positive non-linear co-dependence or whenever they are driven by other, common, risk factors in a non-linear fashion. The negative feedback described in this paper clearly reinforces the necessity for a careful treatment of market risk in the banking book.
Preventing regional contagion by a common lender 31 Countries that have a common significant lender that is heavily exposed to a region are facing a possibility of a regional contagion. Regulators of potentially affected host and home countries should conduct joint stress tests and exchange data Improvement of RM practices in potentially affected countries, not just home country needed Currently, at least in the South East European region, banks typically do not transfer much risk management skills to their subsidiaries. While this initially saves money, in the long run it creates potential trouble for the home country as well
32 Ongoing research Introduce the central bank into the model Test the model empirically (preliminary results on Russian data encouraging) Model (simulate) multi-country contagion in Eastern Europe Carefully address capital charge for loans with foreign currency clauses Make distinction between bank investing in productive (export-generating) vs consumption (importgenerating) projects and endogenize the exchange rate