Transcript The pessimistic scenario

The Academy of Economic Studies
The Faculty of Finance, Insurance, Banking and Stock Exchange
Doctoral School of Finance and Banking
The impact of macroeconomic factors on the
quality of household loans: an empirical
approach for some CEE countries
MSc student: Tatarici Roxana Luminiţa
Supervisor: PhD Professor Moisă Altăr
July 2010, Bucharest
1
Contents:

Motivation and objectives

Literature Review

Data

Results

Conclusions

Further research

References
2
Motivations and objectives





“Macroeconomic based models are motivated by the increase of default rates during recessions”
(Chan-Lau, IMF (2006)).
This has been proven in the recent period when the ability of debtors to cover their debt has been
shaken leading to an increase of non performing loans.
Increase of losses for banks and potential risks for financial stability
High importance of credit risk for the banking sector.
The necessity of financial authorities to identify and monitor the risks and imbalances for the entire
banking system (macroprudentiality).
Objectives:
-
-
-
To identify the macroeconomic variables that influence debtors capacity of repayment in order to
efficiently asses the strengths and vulnerabilities of household loan portfolios;
To identify the period of transmission from the explanatory variables into the quality of household
loans;
To define a multivariate framework which allows the development of top down stress testing
exercises.
3
Brief literature review

Rinaldi and Arellano (2006) using a panel cointegration find an important impact of the
debt burden on the borrowers capacity of repayment. For the short run relationship
relevant factors for explaining NPL ratio are the unemployment rate, nominal interest rate,
household wealth indices, house price index and the ratio of owner occuppied dwellings to
stock of dwellings (to account for collateralized loans).

Pesola (2007) introduces non-linearities by means of a multiplying impact of financial
fragility (loans to private sector/GDP). He proposes as explanatory factors a income
surprise variable and interest rate surprise and tests for the joint impact of financial fragility
and the mentioned factors.

Kalirai(2002), Boss(2002) and Boss et al. (2009) propose a classification of variables into
six categories, which are further on tested in an univariate and multivariate framework.

Jakubik and Schmieder (2008) find that different factors affect the household portfolio in
Czech Republic and respectively Germany. While for the first unemployment rate and real
interest rate are the most important macroeconomic drivers, for Germany credit to GDP
ratio and household income showed the strongest impact.

Marcucci and Quagliariello (2008) analyze the asymmetries in the relationship between
credit risk and business cycles and conclude that: (i) riskier banks’ portfolios are more
cyclical (more sensitive to business cycles) than less risky ones and (ii) cyclicality is more
pronounced in bad/severe economic conditions.
4
Data: Quantifying the quality of portfolio by using NPL
ratio
Non performing loan ratio is one of the core financial soundness indicators proposed by the IMF
and considered as an appropriate measure for the quality of loan portfolios.
Consensus on the construction of the dependant variable (quarterly data 2005Q1: 2009Q4)
Romania: the ratio of loans with more than 90 days past due /outstanding amount of household
loans;
Hungary: the ratio of loans with more than 90 days past due /outstanding amount of household
loans;
Poland: (substandard + doubtfull+loss)/ outstanding amount of household loans;
Czech Republic: (substandard + doubtfull+loss)/ outstanding amount of household loans;
Estonia : the ratio of loans with more than 60 days past due/ outstanding amount of household
loans;
Slovakia: bad loans/ outstanding amount of household loans.
Advantages of using this measure for credit risk:
- Facilitates the monitoring of credit risk at aggregate or individual levels;
- Allows the development of macro-based models;
- Starting to gain comparability: used as a macroprudential indicator by national regulators (Financial
Soundness Indicators, IMF).
Shortages in using this type of indicator:
Few data available and lack of information on types of loans;
Legislative changes may influence the volumes of non-performing loans within countries;
Backward indicator of credit risk;
5
Data: Explanatory variables proposed
Category
Motivation and expected impact
Cyclical indicators
Variables that describe the « state of the
economy». The assumption for their
use: credit risk is pro-cyclical, it
accumulates in periods of economic
growth and quantifies or amplifies in
“distress cycles”, Pesola (2005).
Banking system
indicators
Created in order to account for the specificity
of aggregate portfolios. A higher
indebtedness ratio is expected to
determine an increase in NPL ratio.
Household indicators
External indicators
Cost factors
Variable and transformation applied
Source
Real GDP, s.adj, speed of changes of annual
growth rate
Eurostat
The index of industrial production, quarterly
changes
Eurostat
The ratio of household loans to GDP, proxy
for households` financial fragility, level
ECB,Eurostat
The structure of household portfolios,
quarterly changes
ECB
Unemployment deteriorates debtors capacity
of repayment.
Harmonized unemployment rate, first
difference
Eurostat
Leading indicator for the financial situation
of households
Households final real consumption, speed of
changes in annual growth
Eurostat
Behavioral indicator: considered as an index
of households’ risk aversion.
Consumer confidence indicator, first
difference
GfK, European
Comission
An increase of debt burden in case of
depreciation.
Exchange rate , quarterly changes
Eurostat
Indirect cost factors: an increase of interbank
rate will be transferred in the costs
supported by the debtors.
3 Months money market interest rates, first
difference
Eurostat
Increase in inflation decreases the disposable
income and the real cost of credit.
Annual inflation rate, level
Eurostat
6
Data: Similarities and differences among countries
-
20
10
8
6
0
20
-20
10
-40
4
60
2
50
-60
0
40
0
30
-80
20
-10
10
-20
0
RO
HU
CZ
PL
NPL ratio (right scale)
Indebtedness (HH loans/GDP)
EE
SK
RO
HU
CZ
PL
EE
SK
Consumer confidence index (right scale)
Annual GDP growth
 Household indebtedness has grown considerably for all countries;
 Similar path of NPL ratio (except for Czech Republic and Poland);
 Severe contraction in economic growth, except for Poland and deterioration of consumer confidence;
 Heterogeneity of portfolio structure: high proportion of consumer loans for Romania, while for Estonia the
majority is represented by mortgage loans.
7
Results: Exploring for the individual impact of
explanatory variables
Fixed effects panel model accounts for country specific factors: financial education of
debtors, socio-economic factors, legislation etc;
∆NPLt = f(∆xit), where xit:
Variable
Expected
sign
Real GDP
-
Coefficient
Std.Error
Represent
ative
lag
Rsquare
d
Adjusted
Rsqua
red
-0.089058***
0.02193
1
0.32
0.29
Industrial production
-
-0.052629***
0.009932
2
0.38
0.35
Unemployment rate
+
0.230557***
0.047633
1
0.34
0.30
Households' final consumption
-
-0.073693***
0.017982
1
0.30
0.26
Consumer confidence
-
-0.015339**
0.006599
3
0.25
0.21
3M nominal money market rate
+
0.115864***
0.043285
3
0.25
0.21
3M real money market rate
+
0.054056*
0.029481
3
0.21
0.17
Inflation rate
+/-
0.056093***
0.016488
3
0.25
0.21
Exchange rate
+
0.037599***
0.012537
2
0.29
0.26
Percentage of consumer loans
+
2.756848
4
0.20
0.14
Household loans/GDP
+
1.091221
3
0.22
0.179
-0.081
-2.054339*
Significance levels: * -10%, **-5%, ***-1%.
8
Multi factor credit risk model (1)
NPL ratio = f( GDP, unemployment rate, interest rate, exchange rate)
Accounting for legislative changes: dumsk_06 (Slovakia 2006).
Similar changes in Poland 2005 but uncertainty surrounding the
robustness of the control variable.
9
Multi factor credit risk model (2)
Coefficient
CROSSI
D
Exchange rate
Fixed Effects
RO
0.195981
HU
0.108029
CZ
-0.172963
PL
-0.219503
EE
0.026598
SK
0.061858
Unemployment
rate
Final consumption
Real GDP
Consumer
confidence
R-squared
Adj.R-squared
Consumer loans
5.314443***
0.24
0.2
Mortgage loans
1.974629**
0.16
0.12
Consumer loans
0.345991***
0.27
0.23
Mortgage loans
0.156463***
0.29
0.25
Consumer loans
-0.067833**
0.21
0.17
Mortgage loans
-0.060835***
0.31
0.28
Consumer loans
-0.032866***
0.25
0.21
Mortgage loans
-0.021274***
0.26
0.22
Consumer loans
-0.030989**
0.26
0.22
Mortgage loans
-0.021347***
0.27
0.24
Lower probability of growing NPL ratio in Czech Republic and Poland (other factors affecting NPL
ratio: restructuring process, well developed non performing market).
Increased probability of growing non performing loans for Romania (large proportion of consumer
loans) and Hungary (increasing proportion of consumer loans).
10
Multi-factor credit risk model (3)
NPL ratio is considered as the PD for the household
sector.
Hypothesis: In a better state of the economy the rate
of default is lower (Credit Portfolio View)
Based on the specification of Virolainen (2004):
modeling default rate for country i by a logistic
functional form:
1
p i ,t 
1  e yit
y i ,t
1  p i ,t
 ln(
)
p i ,t
Where: country specific macroeconomic index:
y,it  I ,0 i,1 x1,t   i,2 x2,t  ......  i,n xi,n  i,t
11
Scenario Analysis (1)

-
-


-
The baseline scenario relies on official projections for changes in real GDP and for the
unemployment rate at the end of 2010.
-Around 1% average annual growth for Romania, Hungary, Czech Republic and Estonia,
while for Poland and Slovakia the perspectives for economic growth are more optimistic.
- Small increase of unemployment rate. The difficulties could be however amplified by the
risks associated by the problems that may appear on the reintegration in the labor market.
Average values from Bloomberg survey were considered for the third and last quarter of
2010 and for the second quarter the average daily values.
Because the transmission lag for money market rate was identified as three quarters we
haven’t considered any change in interest rates, although in the case of an economic
recovery the banks’ risk aversion might decrease and the deterioration of the quality of the
portfolio might imply a reconsideration of the contractual agreements with the existent
clients with problems to repayment (through restructuring their loans);
The pessimistic scenario is mainly based on a shock from the labor market:
An increase of the unemployment rate by the same amount as in 2009;
The worst case results of Bloomberg survey regarding the evolution of exchange rate;
Lower economic growth;
The optimistic scenario
A better evolution for the economic growth in 2010 (a symmetric increase of growth rates
with 1 pp).
The changes in unemployment rate were considered at half the movements in moderate
scenario.
The best case from Bloomberg Survey.
12
Scenario Analysis (2)
8.5
% Baseline Optimistic
Scenario
% Baseline Pessimistic
Scenario
% Pessimistic Optimistic
Scenario
2010 Q4
2010 Q4
2010 Q4
NPL - Baseline Scenario
NPLPessimistic Scenario
8.0
7.5
7.0
Romania
1.96
-3.72
5.89
Hungary
2.26
-4.27
6.84
Czech Republic
2.81
-5.44
8.74
6.0
Poland
3.89
-5.24
9.63
5.5
Estonia
1.38
-12.62
16.01
5.0
Slovakia
1.08
-7.99
9.86
( NPLstressed  Nˆ PL)
%deviations
Nˆ PL
6.5
4.5
RO
HU
CZ
PL
EE
SK
-Highest deterioration for the pessimistic scenario for Estonia;
-Slovakia would suffer also from an increase of approximate 8% at the end of the year in face of a slower
economic growth (around 2,3% on an annual basis) and an unemployment rate of almost 18% at 2010Q4.
- Similar increases for Czech Republic and Poland.
- Increase of NPL ratio for Romania, Hungary and Slovakia, but with slower dynamics;
13
Conclusions

Shocks are rather rapidly transmitted in case of cyclical indicators (one or two lags),
but it takes longer for the cost factors to affect the quality of loans (three quarters).
For the linear specification:
- the largest impact from the unemployment rate.
- the aggregate indebtedness ratio is not significant and does not exhibit the
appropriate sign;
- the structure of the portfolio does not explain the behavior of NPL ratio;


Consumer and mortgage loans are sensitive to the same factors, but the impact of
macroeconomic shocks is different; They may also account for different explanatory
variables such as real estate prices in case of mortgage loans.

Estonia and Slovakia are the most affected in case of the pessimistic scenario;

No single best fit model is desirable. Consensus among different specifications and
if possible inclusion of individual specific factors.
14
Further research

Introduction of non-linearities conditional on debt burden, Pesola (2007);

As it has been shown the impact of some of the proposed variables is different for
each type of loan, but the transmission period remains similar. Further improvement
of the framework for sub-categories which can account for factor specific: LTV, real
estate prices etc;

Incorporation of feedback effects from the financial sector into the real economy,
Foglia (2008);


Indentification of asymmetries in debtors’ behavior is of great interest for both
financial authorities, but also for credit institutions to ensure the efficiency of credit
risk models. A reasonable length of time series is required;
Analysis at bank level data to account for differences in lending policies strategies
(dynamics of credit growth), specificity of banks’ portfolios etc .
15
References (1)











Baboucek, I. and M. Jancar (2005) , “A VAR Analysis of the Effects of Macroeconomic Shocks to the
Quality of the Aggregate Loan Portfolio of the Czech Banking Sector” , Czech National Bank, Working
Paper Series 1/2005.
Boss.M (2002), “A macroeconomic credit risk model for stress testing the Austrian credit portfolio”,
Financial Stability Report No.4.
Boss, M., G. Fenz, J.Pann, C. Puhr, M.Schneider and E.Ubl (2009), “Modeling Crdit risk through the
Austrian business cycle: An update of the OeNB model”, Article published in the Financial Stability Report
No.17
Chan-Lau, J.(2006), “Fundamentals–Based estimation of default probabilities: A survey”, IMF Working
Paper, WP/06/149.
Foglia,A (2008).”Stress testing credit risk: a survey of authorities’ approaches”, Bank of Italy Occasional
Papers, No.37.
Gasha, J.G and R. A. Morales (2004), “Identifying threshold effects in credit risk stress testing”, IMF
Working Paper WP/04/150.
Girault,G. and M.Alfredo (2008), “Modeling extreme but plausible losses for credit risk:a stress testing
framework for the Argentine Financial System” MPRA paper No.16378.
Glogowski,A.(2008), “Macroeconomic determinants of Polish banks’ loan losses –results of a panle data
study”, National Bank of Poland Working Paper, No. 53.
Jakubik, P. (2007), “Macroeconomic Environment and Credit Risk” ,Czech National Bank and the Institute of
Economic Studies of Charles University.
Jakubik,P. and C.Schmieder, (2008),”Stress Testing Credit Risk:Comparison of the Czech Republic and
Germany”, Financial Stability Institute.
Jimenez,G. and J.Saurina ,(2005), “Credit Cycles, Credit risk and prudential regulation”, Bank of Spain,
Working Papers, No. 0531.
16
References (2)












Katai,R. (2010), “Credit risk model for the Estonian Banking Sector”, Bank of Estonia Working Papers,
1/2010.
Kalirai, H and M. Scheicher (2002), “Macroeconomic stress testing: Preliminary evidence for Austria”,
Financial Stability Report No.3.
Kern, M. and B.Rudolph (2001), ‘Comparative Analysis of Alternative Credit Risk Models –an Application on
German Middle Market Loan Portfolios-“, Center for Financial Studies WP No.2001/03.
Marcucci, J. and M. Quagliariello (2008), “Credit risk and business cycle over different regimes”, Bank of
Italy Working Papers, No. 670.
Pesola, J. (2005a), “Banking fragility and distress: An econometric study of macroeconomic determinants”,
Bank of Finland Research, Discussion Papers 13.2005.
(2007b) “Financial fragility, macroeconomic shocks and banks’ loan losses: evidence from Europe” , Bank
of Finland Research, Discussion Papers 15.2007.
Rinaldi,L. and A.Sanchis-Arellano (2006), “Household debt sustainability. What explains household nonperforming loans? An empirical analysis”, ECB Working Paper Series, No. 570.
Van den End,J.W, M.Hoeberichts and M.Tabbae (2006), “Modelling Scenario Analysis and Macro StressTesting”, DNB Working Paper, No.119.
Virolainen, K (2004), “Macro stress testing with a Macroeconomic Credit Risk Model for Finland”, Bank of
Finland, Discussion Papers, No.18.
(2008)” Macro-model-based stress testing of Basel II capital requirements”, Bank of Finland, Discussion
Papers, No.17.2008.
Zeman,J. and P.Jurca, (2008), “Macro stress testing of the Slovak banking sector”,Bank of Slovenia
Working Paper 1/2008.
Wooldridge, J. (2001), “Econometric Analysis of cross section and panel data”.
17