Professional Documents
Culture Documents
https://www.emerald.com/insight/0144-3585.htm
Abstract
Purpose – The Iranian banking industry has been greatly affected by dramatic changes in macroeconomic
conditions over the past several decades owing to volatile oil revenues, changing fiscal and monetary
policies, and the imposition of US sanctions. The main objective of this paper is to estimate potential credit
losses in the Iranian banking sector due to macroeconomic shocks and assess the minimum economic
capital requirements under the baseline and distressed scenarios. The paper also contrasts the applications
of linear and nonlinear models in estimating the impacts of macroeconomic shocks on financial
institutions.
Design/methodology/approach – The paper uses a multistage approach to derive the portfolio loss
distribution for banks. In the first step, the dynamic relationship between the selected macroeconomic variables
are estimated using a VAR model to generate the stress scenarios. In the second step, the default probabilities
are estimated using a quantile regression model and the results are compared with those of the conventional
linear models. Finally, the default probabilities are simulated for a one-year time horizon using Monte-Carlo
method and the portfolio loss distribution is calculated for hypothetical portfolios. The expected loss includes
the loss given default for loans drawn randomly and uniformly distributed and exposed at default values when
loans are assigned a fixed value.
Findings – The results indicate that the loss distributions under all scenarios are skewed to the right, with
the linear model results being very similar to those of quantile at the 50% quantile, but very unlike those at
the 10% and 90% quantiles. Specifically, the quantile model for the 90% (10%) quantile generates estimates
of minimum economic capital requirement that are considerably higher (lower) than those using the
linear model.
Research limitations/implications – The study has focused on credit risk because of lack of data on other
types of risk at individual bank level. The future studies can estimate the aggregate economic capital using a
risk aggregation approach and a panel data (not presently available), which could further improve the accuracy
of the estimates.
Practical implications – The fiscal and monetary authorities in developing countries, specially
oil-exporting countries, can follow the risk assessment approach to assess the health of their banking
system and adapt policies to mitigate the impacts of large macroeconomic shocks on their financial
markets.
Originality/value – This is the first paper estimating the portfolio loss distribution for the Iranian banks
under turbulent macroeconomic conditions using linear and nonlinear models. The case study can be applied to
other developing and emerging countries, particularly those highly dependent on natural resources, prone to
extreme macroeconomic shocks.
Keywords Credit risk, Loss distribution, Capital requirement, Wilson model, Quantile regression, Stress
testing
Paper type Research paper
3. Methodology
Prediction of credit risk is the main interest of financial industry participants, particularly
under a capital adequacy framework (Basel II), which encourages banks to assess their
exposure to credit risk and potential losses. This can be done by estimating the default
probability, which is the primary input in systematic risk assessment and stress testing. In
general, there are two approaches for analyzing credit risk and stress testing: The market-
based approach and macroeconomic- or fundamentals-based approach. In brief, market-
based approach draws on option pricing theory and depends on the prices of traded securities
(Merton, 1974; Lee, 2011). In contrast, macroeconomic approach is useful especially when
firms do not have tradable securities or where prices in secondary markets are not reliable
because of low levels of liquidity. In this study, we adopt the latter to conduct the aggregate
stress test, mainly because of lack of information on financial market data for the Iranian
banks [3].
Our methodology follows Borio et al. (2014), who identify four elements in macro stress
tests: The risk factor, shock scenarios based on a macro model, relation between the macro
model and the risk variable, and prediction of the scenario outcomes. We specify the
probability of default as a proxy for credit risk subjected to stress. Our choice of default
probability for the macro stress test is based on the data availability and the fact that credit
risk remains a significant challenge to the Iranian banking system that has not been
addressed yet. We then set up a macroeconomic model to analyze the impacts of various
scenarios for external shocks on the banking system. In the next step, we feed the output of
the macro model into an auxiliary model that incorporates the credit risk variable. Finally, we
use a credit VaR model to estimate the required capital for banks as outcome to cover credit
losses under distressed scenarios. The details of the process are described below and a
schematic presentation of the process is provided in the appendix Figure A2.
where xt ¼ ðx1;t ; x2;t ; . . . ; xK;t Þ is the set of selected macroeconomic factors specified in
step 1 above, β ¼ ðβ0 ; β1 ; . . . ; βk Þ is the set of regression coefficients which determine the
impact of the explanatory factors on the default probability, and εt is an iid error term with
zero mean and constant variance. The default probability (p) in (2) can be estimated by
evaluating the impacts of shocks to the macroeconomic variables through Equation (3).
In this study, we make two modifications to the basic model above. First, we use a dynamic
system of equations for our macro model, instead of the single equation AR model, to allow for
a richer dynamics and interactions between the macroeconomic variables. Second, as the
stress test scenarios include extreme shocks, a linear approach will be unable to provide an
accurate estimation of the losses caused by extreme events. Therefore, we use a nonlinear
quantile regression model to predict bank losses and compare the results with those of a
linear model as a benchmark.
The quantile regression, as first discussed by Koenker and Bassett (1978), allows to
estimate different quantile functions of the conditional distribution, where each quantile
function is a unique point (the tail or center) of the conditional distribution. This should yield a
more complete distribution of the original distribution. In general, the quantile regression can Macroeconomic
be specified as shocks and
Yi ¼ xi βθ þ εθi ; credit risk
stress
and the conditional quantile θ ∈ ð0; 1Þ of the Y variable, which is given by:
Qθ ðY jxi Þ ¼ xi βθ i ¼ 1; . . . ; n; (4)
where Qθ ðY jxi Þ is the θth conditional quantile of Y given xi, xi ¼ ð1; xi1 ; . . . ; xik Þ represents a
vector of given variables, and βθ ¼ ðβ0 ; β1 ; . . . ; βk Þ is a vector of unknown parameters. With
these two modifications, our base model is thus specified as follows:
1
pt ¼ (5)
1 þ e−yt
Yti ¼ xti βθi þ εθi ; i ¼ 1; . . . ; n; βθ ¼ ðβ0 ; β1 ; . . . ; βk Þ; θ ∈ ð0; 1Þ (6)
X
q
Xt ¼ γ 0 þ γ j Xt−j þ vt ; vt ∼ iid ð0; Σv Þ (7)
i¼1
εt Σε Σε;v
E¼ ∼ N ð0; ΣÞ; Σ ¼ (8)
vt Σv;ε Σv
Equation (6) represents the quantile regression model, Equation (7) the vector autoregressive
(VAR) model, and Σ in (8) the variance-covariance matrix of error terms in (6) and (7).
The VAR model is a rich macroeconomic model that allows for studying the dynamic
relations among macro variables and analyzing the impacts of shocks on the variables
through time. It is also a theoretical model avoiding the endogeneity problem often exists in
macroeconomic models by treating all variables as endogenous. However, the VAR model is
data intensive limiting the number of variables that can be used, especially when the high
frequency data in not available. This leads us to select only a few variables from the list of
variables with potential effects on the probability of default. The list of potential variables
that have also been used in the literature includes GDP, GDP growth, stock price index,
exchange rate, unemployment rate, inflation rate, housing prices, oil revenues and the
liquidity growth rate.
There are different approaches to select the subset of variables from the pool of relevant
variables depending on the objectives of the research. For instance, one might follow a
specific-to-general approach through which the estimation starts with one variable and then
continues with adding other variables stepwise. In contrast, the general-to-specific approach
entails regressions starting with all potential variables and removing variables one at a time.
In either case, different metrics such as the significance level of the estimated coefficients and
measures of prediction errors can be used to find the right model (Jarocinski and Mackowiak,
2011). Since our objective is to assess the impact of macro shocks on credit risk, the significant
relation between the macro variables and the credit risk would be a suitable criterion for the
variable selection. We therefore follow the approach in Boss (2002) to select the variables for
our VAR model. In this method, a series of macroeconomic variables are regressed on the
probability of default in a variety of univariate and multivariate regression models and then
the most effective indicators are selected.
The regression results show that four major macroeconomic variables (GDP growth,
unemployment rate, inflation rate and the exchange rate) are the key factors affecting the
default probability in our data [4]. An increase in economic activities represented by high
economic growth and low unemployment rate enables the borrowers to pay off their debts in
JES due time. Therefore, the relationship between the default probability and the GDP growth is
expected to be negative, whereas with unemployment rate positive. Inflation rate decreases
the real value of the loans making it easier for the borrowers to pay off their debts. However,
higher inflation rates reduce the real income of fixed earners and may reflect uncertainty
about the future economic conditions and policies, leading to higher probability of default.
With fixed interest rate, the higher inflation rate decreases the real interest rates increasing
demand for funds. However, given the supply of loanable funds, the lower real interest rate
paves the road to rent seeking behavior and discretionary allocation of loans, with often high
probability of default.
The depreciation of real exchange rate may have a negative impact on the default
probability, because of improving the net exports (Kalirai and Scheicher, 2002), but it may
also generate opposite effect if domestic production heavily relies on imports of intermediate
goods or if the loss of domestic currency value is perceived as weak economic conditions and
prospects. The changes in the real exchange rates also reflect the changes in oil prices and
revenues as one of the key factors in the Iranian economic performance, a well-known
phenomenon as Dutch disease. Higher (lower) oil prices are often associated with the
appreciation (depreciation) of the real exchange rates and a boom in the non-traded sectors,
such as construction and real estate activities. Therefore, the probability of default effect of
the changes in real exchange rate will depend on the relative strengths of these opposing
factors, which will be determined by the data.
EAD and LGD are often assumed to be constant and, therefore, PD is the key input into the
portfolio models that should be estimated by the banks. In this paper, the loss distribution
merely considers the losses arising from default state and not the losses arising from any
unexpected changes in credit quality.
Credit portfolio risk is usually determined by two important parameters, namely, expected
loss (EL) and unexpected loss (UL), as specified by the probability distribution of losses.
Expected loss is the average of the loss distribution, and usually covered by financial
institutions with provisions and write-offs. Unexpected losses relate to potentially large
losses that occur at different frequencies and are therefore covered by bank capital, where UL
is the standard deviation from the mean at a certain confidence level. It is also referred to as
credit Value at Risk (VaR). For example, the UL of a portfolio at a 99% confidence level will be
expressed as follows:
ULð99%Þ ¼ VaRð99%Þ EL
Credit capital is the amount of economic capital (EC) that a bank must maintain to cover Macroeconomic
unexpected losses caused by default. Therefore, at a given confidence level, the EC required shocks and
for banks is equal to the unexpected loss.
To derive the loss distribution, the probability of default estimated in the base model is
credit risk
simulated one-step ahead. That is, substituting the current values for time t by the respective stress
forecasts for time tþ1 after a time step and repeating the simulation steps, a forecast of the
default probability for time tþ2 is obtained. This procedure can be repeated until the desired
time horizon H is reached. This method thus yields a path of future default probability
ðptþ1 ; ptþ2 ; . . . ptþH Þ. Monte Caro simulation steps iterated several times and the new default
probability obtained at each time. Finally, the entire loss distribution can be obtained over all
simulated paths of the default probabilities. The determination of the loss distribution
involves the following steps (Ruben and Manuel, 2014):
(1) Construction of a hypothetical portfolio consisting of 7,000 loans.
(2) The exposure at default (EAD) for each loan is determined randomly by a uniform
distribution. Random exposures and the number of loans are drawn until the
aggregated amount of exposure of the hypothetical portfolio equals the aggregate
nonperforming loans held by the banks (about USD 31 billion in 2016).
(3) Focusing on the IRB approach, the loss given default (LGD) is set to 0.45.
(4) The simulated default probabilities are used in the arbitrary portfolio and the loss
distribution is estimated over a one-year time horizon.
(5) The expected and unexpected losses and VaR are calculated from the loss
distribution.
A Monte Carlo simulation with 10,000 simulation steps is performed for the baseline
scenario (with no macroeconomic shocks) and the distressed scenarios to determine the
loss distribution and to estimate the required capital as outcome. The stress test is
performed by comparing the credit loss distributions of the distressed scenarios to those
of the baseline scenario. In this paper, we create the macro scenarios for the second
quarter of 2016 and calculate their results for the second quarter of 2017 (a one-year time
horizon).
4. Results
4.1 Data structure
Our quarterly date represents the Iranian banking industry and macroeconomic conditions
from 2004: Q1 to 2016: Q2. The purpose of the aggregate credit risk stress test is to measure
structural vulnerabilities and the risk situation in the entire financial system by identifying
losses resulting from borrower default. Since the aggregate nonperforming loans (NPL) ratio
is the only data made available by the Central Bank of Iran, we use it at the end of each quarter
as a proxy of the default rates. The data on unemployment rate, GDP growth, inflation rates
and the exchange rates are collected from the Statistical Center of Iran and the Central Bank
of Iran.
Table 1 presents a summary of the data used in the model. The mean default probability is
15% during the period 2004–2016 with nearly a symmetric distribution. As Figure 1 shows,
among the macroeconomic variables, the unemployment rate is most stable, while the
inflation rate, the growth of real exchange rate and GDP growth exhibit sharp fluctuations,
particularly in recent years. In evidence, the inflation rate, the growth of GDP and real
exchange rates during the sample period have varied from 7% to 43%, 6.6% to 12%, and
13.9% to 32.88%, respectively. Changes in oil prices along with the US and the UN
JES sanctions on the oil and banking sectors in Iran were the main causes of the recent
fluctuations in the economy. We use a dummy variable to control for the sharp changes in the
real exchange rates in 2012.
GDP growth
Inflation rate
Unemployment rate
Growth of real exchange rate
40
30
20
10
0
Figure 1.
-10
Trends of
macroeconomic 2005 2007 2009 2011 2013 2015
variables (%), 2004:Q1- Year
2016:Q2
Source(s): Central Bank of Iran and authors’ calculation
roots are inside the unit circle and, therefore, the estimated model is stable. An F-statistic Macroeconomic
confirms the significance of all four regressions. shocks and
To link the macro model to the credit risk factor, we estimate the quantile regression as
specified in (4). We also run a linear model used in the literature to compare the results with
credit risk
the quantile model. Table 3 provides the results of the linear model and quantile regressions stress
for the 10%, 50 and 90% quantiles. The coefficients have the expected signs for all variables
in both models but their levels of significance differ. The Jarque–Bera test results also
indicate that the hypothesis of normal distributions cannot be rejected at 5 percent
significance level and the Durbin–Watson test result leads us to reject the presence of
autocorrelation in residuals.
The unemployment rate is positively related to the probability of default as it provides a
measure of the income state of households. Higher unemployment implies less disposable
income which in turn indicates borrowers are more likely to default on their debts. GDP
growth indicating the state of economy is inversely related to the default probability. Higher
GDP growth suggests that the economic outlook is favorable leading to greater confidence in
the economy and increasing economic activities. Therefore, when borrowers are in better
financial positions, loan default decreases, but when GDP growth declines, borrowers are
more likely to default due to adverse economic conditions.
The inflation rate is positively related to loan default, although the effect is not significant
statistically. Generally, the relationship between the inflation rate and the credit risk is
Variables UR GDPR ER IR
Scenario UR GDPR ER IR
0.95
0.95
wilson wilson
QR(0.1) QR(0.1)
QR(0.5) QR(0.5)
F(PD)
F(PD)
0.85
0.85
QR(0.9) QR(0.9)
0.75
0.75
0.10 0.15 0.20 0.25 0.10 0.15 0.20 0.25
PD PD
0.95
0.95
wilson wilson
QR(0.1) QR(0.1)
QR(0.5) QR(0.5)
F(PD)
F(PD)
QR(0.9)
0.85
0.85
QR(0.9)
0.75
0.75
0.05 0.10 0.15 0.20 0.10 0.15 0.20 0.25
PD PD
0.95
wilson
QR(0.1)
QR(0.5)
F(PD)
QR(0.9)
0.85
0.75
0.10 0.15 0.20
PD
Note(s): All scenarios comprise one - std. dev. shock for each variable using Linear model
and QR (10%), QR (50%) and QR (90%)
credit risk
shocks and
Macroeconomic
Figure 2.
estimates
PD cumulative
stress
JES 4.3.2 Expected and unexpected credit losses. In the next step, using the estimated probability
of default above, we calculate the expected and unexpected losses resulting from credit risk under
the different scenarios. To estimate the economic capital (EC), the capital required to cover
unexpected losses, we take the difference between the value at risk at a given confidence level and
the expected losses. Given that the probability of default is simulated for a one-year time horizon,
the loss distribution also corresponds to the same time period.
The results in Table 5 show that the loss distributions for all scenarios are skewed to the
right. The amount of loss at the 50% quantile is close to the linear model, but unlike those at
the 10 and 90% quantiles. In fact, the QR (90%) model generates a requirement for economic
capital that is considerably higher than those generated by the linear model, and the QR
(10%) model estimates are less than those generated by the linear model. For example, under
the GDP growth shock, the predicted EC by the linear model is about US$5,050 million, which
is greater than US$4,296 million implied by QR (10%) and less than US$5,448 million
estimated by QR (90%).
All distressed scenarios, excluding the growth of real exchange rate scenario, involve
greater loss than the baseline scenarios. The reason is the negative effect of the exchange rate
on the probability of default. Comparing the unexpected losses in the distressed scenarios and
the baseline scenario using the linear model, a GDP growth shock is the most influential factor
on credit risk followed by the exchange rate shock. QR (90%) and QR (50%) also largely
confirm these findings. It is also worth noting that the unexpected losses are considerably
greater than the expected losses. Although unexpected losses are less likely, their large
magnitudes suggest that banks should maintain higher level of capital to cover any
unexpected losses.
As noted earlier, one-standard-deviation shock scenarios do not exemplify the worst
possible economic situation in Iran. In fact, in recent years, the Iranian economy has
experienced three-standard deviation shocks to the unemployment rate, GDP growth and
exchange rate and one-standard-deviation shocks to the inflation rate. Here, we reflect these
empirical realities and perform stress tests based on the following historical scenarios.
5. Conclusion
In this paper, we analyze the credit risk stress test and estimate credit loss distributions to
calculate the minimum capital requirement that banks should hold to absorb macroeconomic
shocks in Iran. In the first step, we estimate the relationship between the selected
macroeconomic variables using a VAR model. The output of this stage is the macro scenarios,
which are then fed into an equation linking the probability of default with macroeconomic
variables. We simulate the path of default probability using the Monte-Carlo method under
Notes
1. Iran’s current macroeconomic problems, including high inflation rates, have been exacerbated
following the resumption of sanctions by the US and its unilateral exit from the G5þ1 nuclear
agreement with Iran in mid-2018. Our study, however, does not include this most recent
development.
2. Economic Indicators of Central Bank of Iran (www.cbi.ir).
3. Only a few banks have recently joined the stock market and the balance sheet and the financial data
on many banks are still not publicly available.
4. To save space, the regression results used to select the variables are not reported here but they are
available upon request.
5. We also run the model with one lag, but the main results remained unchanged.
References
Allen, D.E., Kramadibrata, A.R., Powell, R. and Singh, A.K. (2010), “Using quantile regression to
estimate capital buffer requirements for Japanese banks”, Proceedings of the Globalization,
Monetary Integration and Exchange Rate Regimes in East Asia Conference, Perth.
Basel Committee on Banking Supervision (2005), An Explanatory Note on the Basel 2 IRB Risk Weight
Functions.
Borio, C., Drehmann, M. and Tsatsaronis, K. (2014), “Stress-testing macro stress testing: does it live up
to expectations?”, Journal of Financial Stability, Vol. 12, pp. 3-15.
Boss, M. (2002), A Macroeconomic Credit Risk Model for Stress Testing the Austrian Credit Portfolio,
Financial Stability Report 4, Oesterreichische National Bank, Vienna.
Breuer, T., Jandacka, M., Rheinberger, K. and Summer, M. (2009), “How to find plausible, severe and
useful stress scenarios”, International Journal of Central Banking, September, pp. 205-224.
Deming, W., Fang, M. and Wang, Q. (2018), “An empirical study of bank stress testing for auto loans”,
Journal of Financial Stability, Vol. 39, pp. 79-89.
Drehmann, M., Patton, A., J. and Sorensen, S. (2007), “Non-linearities and stress testing”, Proceedings
of the Fourth Joint Central Bank Research Conference on Risk Measurement and Systemic Risk,
pp. 213-301, European Central Bank, Frankfurt, Germany.
Financial Sector Assessment Program (FSAP) (1999), “International Monetary Fund (IMF)”,
Washington, available at: https://www.imf.org/external/np/fsap/fssa.aspx.
Financial, Tribune (2016), “$6.3 b reduction in Iranian banks NPLs”, 19 November, available at https://
financialtribune.com.
Francisco, V., Benjamin, M., Tabak and Souto, M. (2012), “A macro stress test model of credit risk for
the Brazilian banking sector”, 2012, Journal of Financial Stability, Vol. 8 No. 2, pp. 69-83.
Francisco, C., Ben, R. and Egon, Z. (2014), “Stress-testing US bank holding companies: a dynamic panel
quantile regression approach”, International Journal of Forecasting, Vol. 30 No. 3, pp. 691-713.
Gambetta, N., Garcia-benau, M.A. and Zorio-Grima, A. (2019), “Stress test impact and bank risk
profile: evidence from macro stress testing in europe”, International Review of Economics and
Finance, Vol. 61, pp. 347-354.
JES Heidari, H., Zavarian, Z. and Noorbakhsh, E. (2010), “The effect of macroeconomic indicators on
bank’s Non-performing loans”, The Journal of Money and Economics No. 4.
Hemati, A. and Mohebi Nezhad, S. (2009), “Investigating the effects of macro-economic variables on
credit risk of banks”, Economic Research Letter, Winter, pp. 33-59.
Jarocinski, M. and Mackowiak, B. (2011), Choice of Variables in Vector Autoregressions, European
Central Bank and CEPR, Frankfurt, September.
Kabir Hassan, M., Unsal, O. and Emre Tamer, H. (2016), “Risk management and capital adequacy in
Turkish participation and conventional banks: a comparative stress testing analysis”, Bursa
Instanbul Review, Vol. 16 No. 2, pp. 72-81.
Kalirai, H. and Scheicher, M. (2002), Macroeconomic Stress Testing: Preliminary Evidence for Austria
in: OeNB, Vol. 3, Financial Stability Report, Vienna, pp. 58-74.
Koenker, R. and Bassett, G. (1978), “Regression quantiles”, Econometrica, Vol. 46 No. 1, pp. 33-50.
Kok, ch., Mirza, H. and Pancaro, C. (2019), “Macro stress testing euro area bank’s fees and
commissions”, Journal of International Financial Markets, Institutions and Money, Vol. 61,
pp. 97-119.
Lee, W.-C. (2011), “Redefinition of the KMV model’s optimal default point based on genetic algorithms:
evidence from taiwan”, Expert Systems with Applications, Vol. 38, pp. 10107-10113.
Merton, R., C. (1974), “On the pricing of corporate debt: the risk structure of interest rates”, The
Journal of Finance, Vol. 29, pp. 449-70.
Noroozi, P. (2015), “The impact of macro variables on bank’s credit risk in Iran”, Journal of Monetary
and Banking Research, Vol. 20, pp. 237-257.
Paivar, A., (2015), “Iran banks unprepared for post-sanctions world”, 20 November, available at: http://
www.bbc.com/news/business-34882309.
Ricardo, S. and Wanger, P. (2011), “Macro stress testing of credit risk focused on the tails”, Banko
Central do Brasil, Working Paper Series.
Ruben, G.C. and Manuel, M. (2014), “Estimating the distribution of total default losses on the spanish
financial system”, Journal of Banking and Finance, Vol. 49 No. C, pp. 242-261.
Simons, D. and Rolwes, F. (2009), “Macroeconomics default modeling and stress testing”, International
Journal of Central Banking, September 2009.
Strohecker, K. and Saul, J. (2015), “Iran to overhaul banking sector”, looks to Eurobond issue, 9 March,
available at: http://uk.reuters.com/.
Virolainen, K. (2004), “Macro stress testing with a macroeconomic credit risk model for Finland”,
Vol. 11, Bank of Finland Discussion Papers.
Wilson, T.C. (1997a), “Portfolio credit risk (I)”, Risk, September.
Wilson, T.C. (1997b), “Portfolio credit risk (II)”,, Risk, October.
Further reading
Basel Committee on Banking Supervision (2006), International Convergence of Capital Measurement
and Capital Standards: A Revised Framework Comprehensive Version.
Bharath, S. and Shumway, T. (2008), Forecasting Default with the KMV-Merton Model, University of
Michigan.
CGFS (2005), Stress Testing at Major Financial Institutions: Survey Result and Practice, Bank for
International Settlements.
Chan-Lau, J.A. (2006), “Fundamentals-based estimation of default probabilities: a survey”, IMF
Working Paper, No. 06/149, Washington.
Chan-Lau, J.A. (2013), Market-Based Structural Top-Down Stress Tests of the Banking System,
International Monetary Fund, Washington, WP/13/88.
Crouhy, M., Galai, D. and Mark, R. (2000), Risk Management, Mc-Graw Hill, Washington. Macroeconomic
Drehmann, M. (2007), “Macroeconomic stress-testing banks: a survey of methodologies in stress- shocks and
testing the banking system”, in Mario Quagliariello (Ed.), Cambridge University Press, London,
pp. 39-29.
credit risk
Foglia, A. (2009), “Stress testing credit risk: a survey of authorities approaches”, International Journal
stress
of Central Banking, Vol. 5 No. 3.
Hemmati, A. and Mohebinejad, Sh (2011), “Assessing the impact of macroeconomic variables on
bank’s credit risk”, Journal of Economics, Vol. 6.
Koenker, R. and Xiao, Z. (2002), “Inference on the quantile regression process”, Econometrica, Vol. 70
No. 4, pp. 1583-1612.
Quagliariello, M. (2009), Stress-testing the Banking System: Methodologies and Applications, Cambridge
University Press.
Wei, L. and Yang, Z. (2012), “Stress testing of commercial banks’ exposure to credit risk: a study
based on write-off nonperforming loans”, Asian Social Science, Vol. 8 No. 10.
Wong, J., Choi, K.-F. and Fong, T. (2008), “A framework for stress testing bank’s credit risk”, The
Journal of Risk Model Validation, Vol. 2 No. 1, pp. 3-23.
Appendix
Levels Growth
ADF PP ADF PP
PD GDPR IR ER UR
Optimal lag 1 2 3 4 5
VAR model
Figure A1.
stability test
OLS-CUSUM of equation gdpr OLS-CUSUM of equation ER1
-1.0
-1.0
-1.0
OLS-CUSUM of equation IR
Model
Model linking macro and banking
variables
(Linear and Quantile Regression)
Stressed PDs
Hypotheses
on LGD and
VaR and Stressed EL EAD
Outcome
using simulation
Figure A2.
The structure of macro
stress-testing
Impact on credit risk
capital
Corresponding author
Saeed Moshiri can be contacted at: smoshiri@stmcollege.ca
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com