Comparative Analysis and Stress-Testing
Corporate and Retail Segments of
Russian Credit Market
Oleg Kozlov
National Research University – Higher School of Economics,
London School of Economics and Political Science
HSE, Moscow, 2013
Motivation and objective
photo
photo
photo
2
0
5
10
15
20
25
20
04
-08
20
05
-05
20
06
-02
20
06
-11
20
07
-08
20
08
-05
20
09
-02
20
09
-11
20
10
-08
20
11
-05
20
12
-02
Size of segments, trln Rub
-20
0
20
40
60
80
100
An
nu
al lo
an g
row
th, %
Annual growth rates in retail and corporate loans
Retail loan growth Corporate loan growth
5
10
15
20
25
20
04
-08
20
05
-02
20
05
-08
20
06
-02
20
06
-08
20
07
-02
20
07
-08
20
08
-02
20
08
-08
20
09
-02
20
09
-08
20
10
-02
20
10
-08
20
11
-02
20
11
-08
20
12
-02
20
12
-08
Ave
rage
CA
R, %
Average capital adequacy ratio across banks by segment
Many experts were concerned that since 2011 Russian credit market has grown promptly
while average ratio of non-performing loans (NPL) remained rather high and capital adequacy
of major banks decreased
The objective of the paper is to investigate and compare risk patterns in retail
and corporate segments and assess the potential impact of macroeconomic
shocks on loan quality
photo
photo
photo
Literature overview
Recent IMF recommendation papers (IMF, 2012) call for comprehensive framework, paying
special attention to systemically important institutions and treating feedback effects
Stress-testing using VAR approach allows to analyse interconnection between
financial and real sector variables, assess effects of shocks Examples: UK (Hoggarth et al., 2005), Italy (Marcucci, Quagliariello, 2005), Czech Republic
(Simeckova, 2011)
Stress-testing Russian banking system:
3
CBR Foreign studies CMASF
CBR (2011, 2012) Financial
stability Reports
Fungсovа Z. and Jakubík P.
(2012)
Pestova (2010), Solntsev,
Pestova, Mamonov (2010, 2012)
Sample ~1000 banks –
reduced efficiency
Out-of-date methods (system
OLS) subject to technical
issues e.g. omitted variable
bias and auto-correlated errors
Too strong assumptions to
apply Basel formulas and
average estimators for Russia
Use cross-country data, do not
distinguish between different
sectors of credit market
Data and sampling
photo
photo
photo
Data:
Micro data on banks: Mobile’s “Banks and Finance” (2004-2012, monthly)
Macro data: CBR, Rosstat, Joint Economic and Social Data Archive (HSE)
Sampling:
1) For retail and corporate sector list all credit organizations as of 01.09.2004 by the amount of loans
they provided in descending order
2) First N banks which cumulatively account for 85% of the market are included in the samples (allow
each bank to enter both segments samples) get 97 banks in retail, 104 banks in corporate sector
4) Distinguish between State banks (Vernikov, 2011), foreign (>50% capital) and private domestic
(rest)
4
78
80
82
84
86
88
90
92
20
04
-08
20
05
-01
20
05
-06
20
05
-11
20
06
-04
20
06
-09
20
07
-02
20
07
-07
20
07
-12
20
08
-05
20
08
-10
20
09
-03
20
09
-08
20
10
-01
20
10
-06
20
10
-11
20
11
-04
20
11
-09
20
12
-02
20
12
-07
sam
ple
co
vera
ge, %
Sample coverage by segments
Retail loan segment Corporate loan segment
0%
20%
40%
60%
80%
100%
200
4-0
8
200
5-0
1
200
5-0
6
200
5-1
1
200
6-0
4
200
6-0
9
200
7-0
2
200
7-0
7
200
7-1
2
200
8-0
5
200
8-1
0
200
9-0
3
200
9-0
8
201
0-0
1
201
0-0
6
201
0-1
1
201
1-0
4
201
1-0
9
201
2-0
2
201
2-0
7
Market share of sampled banks by ownership
State Foreign Prvate Domestic
NPL trends in Russian banking
photo
photo
photo
5
0
2
4
6
8
10
12
14
16
20
04
-08
20
04
-11
20
05
-02
20
05
-05
20
05
-08
20
05
-11
20
06
-02
20
06
-05
20
06
-08
20
06
-11
20
07
-02
20
07
-05
20
07
-08
20
07
-11
20
08
-02
20
08
-05
20
08
-08
20
08
-11
20
09
-02
20
09
-05
20
09
-08
20
09
-11
20
10
-02
20
10
-05
20
10
-08
20
10
-11
20
11
-02
20
11
-05
20
11
-08
20
11
-11
20
12
-02
20
12
-05
20
12
-08
Ave
rage
NP
L, %
Non-perfoming loans (NPL)
Retail loan segment Corporate loan segment
0
2
4
6
82
00
4-0
8
20
05
-01
20
05
-06
20
05
-11
20
06
-04
20
06
-09
20
07
-02
20
07
-07
20
07
-12
20
08
-05
20
08
-10
20
09
-03
20
09
-08
20
10
-01
20
10
-06
20
10
-11
20
11
-04
20
11
-09
20
12
-02
20
12
-07
Ave
rage
NP
L, %
NPL in corporate segment by ownership
State Foreign Prvate Domestic
0
5
10
15
20
20
04
-08
20
05
-01
20
05
-06
20
05
-11
20
06
-04
20
06
-09
20
07
-02
20
07
-07
20
07
-12
20
08
-05
20
08
-10
20
09
-03
20
09
-08
20
10
-01
20
10
-06
20
10
-11
20
11
-04
20
11
-09
20
12
-02
20
12
-07
Ave
rage
NP
L, %
NPL in retail segment by ownership
State Foreign Prvate Domestic
Credit Risk – Return Stability: derivation
photo
photo
photo
Chebyshev theorem (inequality):
Z(NPL)-score: new measure of credit risk
6
Cihak, (2007, p.26): back-testing
(EWS): take NPL and CAR of a
sample of banks and plot the on
the plane
The framework is modified:
instead of CAR take Roy’s (1952)
Z-score and instead of NPL an
indicator to measure institution’s
credit risk is developed
Advantage: captures NPL
dynamics and reserve buffer
Enables to track bank’s position
over time in credit risk – return
stability trade-off
Credit Risk – Return Stability: results
photo
photo
Median Z(NPL) is two times lower in retail segment, meaning banks are exposed
to higher credit risk there
Same banks tend to exhibit similar but not exactly the same patterns in retail and
corporate sectors
VTB group banks have high credit risk exposure
7
Sberbank
VTB 24
HKF Bank
Unicredit
SKB-Bank
MDM
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
0 1 2 3 4 5
Z(N
PL
)
Z(ROA)
Credit Risk –Return Stability diagram for
Retail segment
Unicredit
MDM
Rossiya
VTB
Sberbank
-7
-6
-5
-4
-3
-2
-1
0
1
2
0 1 2 3 4 5
Z(N
PL
)
Z(ROA)
Credit Risk – Return Stability for Corporate
segment
VAR analysis: model specification
photo
photo
8
Goodness of Fit
Equation Y M0 INV Loans NPL
R2, corporate
model
0.95 0.42 0.54 0.65 0.31
R2, retail model 0.95 0.43 0.56 0.63 0.40
Granger causality tests results:
Corporate Segment
Y growth NPLc growth
NPLc growth CL growth
Retail Segment
Y, M0, RL growth NPLR growth
Y, M0 growth RL growth
photo
photo
photo
VAR analysis: FEVD analysis
9
Forecast Error Variance Decomposition
(FEVD) indicates how much variation in
one variable can be attributable to
changes in other variables
Variation attributable to macro factors in
Corporate segment
Retail segment
Loan growth 17% 43%
NPL growth 23% 40%
Retail segment is more sensitive to
macroeconomic conditions
M0 3%
INV 10%
NPL_C 7%
Y 4%
CL 76%
FEVD of CL
M0 9%
INV 3%
NPL_C 64%
Y 10%
CL 14%
FEVD of NPL_C
M0 5%
INV 5% NPL_R
2%
Y 33%
RL 55%
FEVD of RL
M0 17% INV
2%
NPL_R 52%
Y 21%
RL 8%
FEVD of NPL_R
photo
photo
photo
VAR analysis: output growth shock
Output growth shock (1 quarter SD) results in prolonged negative
response of NPL in both segments (pro-cyclicality of credit risks)
For similar studies of the UK, Italy, Czech banking systems the effect fades
within 3-4 quarters, in Russia – 2 years
10
-.4
-.3
-.2
-.1
.0
.1
.2
2 4 6 8 10 12 14 16 18 20 22 24 26
Response of NPL_R to Y shock
months since shock
-.5
-.4
-.3
-.2
-.1
.0
.1
2 4 6 8 10 12 14 16 18 20 22 24 26
Response of NPL_C to Y shock
months since shock
photo
photo
photo
VAR analysis: stress-testing
Generally, banking system is prone to shocks (recovers within a year)
Magnitude of shock to NPL_R is less but adaptation period is longer and it
is associated with higher uncertainty
11
Variable M0 INV Y CL RL NPL_C NPL_R
August 2008 shock, p.p. -0,103 14,199 -0,554 1,658 50,457
September 2008 shock, p.p. -1,67 2,38 -1,220 -2,102 31,401
-.3
-.2
-.1
.0
.1
.2
.3
.4
.5
.6
1 2 3 4 5 6 7 8 9 10 11 12
Response of NPL_C to shock
months since shock
-.1
.0
.1
.2
.3
.4
1 2 3 4 5 6 7 8 9 10 11 12
Response of NPL_R to shock
months since shock
photo
photo
photo
Concluding remarks
12
Average level and sensitivity of credit risk to
macroeconomic shocks is higher in retail segment
Average NPL in retail segment has been twice as high as in corporate
Variation in NPL explained by macroeconomic shock: 23% in corporate,
40% in retail
On bank-specific level VTB group banks were identified as ‘risky’
The effect of repetition of 2008 scenario on NPL fades within a year,
hence the system is generally prone to shocks
Policy implications: focus supervisory control on retail segment, expand
and implement Credit Risk – Return Stability framework for systematic
distant risk-based monitoring