MICROECONOMIC MODIFICATION OF AN INDUSTRY-WIDE BOONE INDICATOR: NEW ESTIMATES OF THE MARKET POWER OF RUSSIAN BANKS April 8, 2015; Moscow, Russia XVI April International Academic Conference on Economic and Social Development Mikhail Mamonov, Center for Macroeconomic Analysis and Short-term Forecasting (CMASF) at the Russian Academy of Sciences – Institute for Economic Forecasting; National Research University “Higher School of Economics”
16
Embed
MICROECONOMIC MODIFICATION OF AN INDUSTRY …€¦ · microeconomic modification of an industry-wide boone indicator: new estimates of the market power of russian banks april 8, 2015;
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
MICROECONOMIC MODIFICATION OF AN INDUSTRY-WIDE
BOONE INDICATOR: NEW ESTIMATES OF THE MARKET
POWER OF RUSSIAN BANKS
April 8, 2015; Moscow, Russia
XVI April International Academic Conference on Economic and Social Development
Mikhail Mamonov,
Center for Macroeconomic Analysis and Short-term Forecasting (CMASF) at the Russian Academy of Sciences – Institute for Economic Forecasting;
National Research University “Higher School of Economics”
Motivation
The recent multi-country research on non-linear relationship between the bank market power and risk exposure provides conflicting results:
(Berger et al., 2009; Beck et al., 2013): the inverse U-shaped form if Z-score=f(Lerner index; Lerner index2) whereas
(Tabak et al., 2012): the U-shaped form if Z-score=f(Boone; Boone2);
This could be a problem of measuring market power on the bank level rather than a problem of samples’ differences
The literature suggests only the Lerner index for the bank-level estimation of banks’ market power so far
Problems with the Lerner index (price mark-up over marginal costs): (1) based on averaged price rather than on actual price; (2) catches price competition, ignores quality competition; hence, it cannot describe competition completely
If so, at least one more competition indicator on the bank level needed
Introducing a new indicator: creating smth. new or modifying smth. old?
2
Literature review: just modification or smth. “new”?
Boone (2008), EJ: the “efficiency-structure” hypothesis of Demsetz (1973) Market share = f(Marginal costs…) a new competition proxy for industry
Recent contributed studies on the evaluation of bank-level market power:
1. Carbo, Humphrey , Maudos, Molyneux (2009), JIMF: translog revenue function with the interactions of input prices H-statistic
Shortcoming: other BSF (risk preferences, assets/liab. composition, outputs) are assumed to have no effect on the heterogeneity of H-statistic
2. Bolt, Humphrey (2010), JBF: estimation of competition frontier via SFA
3. Delis, Tsionas (2009), JBF: local optimization of the conjectural parameter (CP) from the Bresnahan model the first bank-level modification of CP
4. Brissmis, Delis (2011), JOR : local optimization a more broad alternative for H-statistic as compared to the CHMM(2009)
5. Delis (2012), JDE: local optimization of the Boone equation the first bank-level modification of the Boone indicator
Shortcomings of 3-5: bank-level heterogeneities in each case «fall from sky»
Beck, De Jonghe, Schepens (2013), JFI: Z-score=f(Lerner, Lerner*Institutions…)
3
Contribution
1. Propose a new version of bank-level Boone indicator (BI)
The external variables approach in a Beck et al. (2013)’s manner: BI is a linear function of bank-specific factors (business models – retail vs. corp. lending, loans vs. non-lending activities, risk profiles; and more);
Define and estimate this function within static panel framework using 2-step GMM;
It is an alternative to the Delis (2012)’s version of bank-level BI obtained through local optimization technique;
Use quarterly data bank-level for Russia (single country study) whereas Delis (2012) exploits Bankskope yearly data on 84 banking systems worldwide;
Contrasts to the bank-level H-statistics obtained through the internal variables approach by Carbo et al. (2009);
2. Divide the credit market into 4 niches by intersecting loan quality indicator with the share of retail loans in total loans. Analyze the BI values of:
largest banks both within and across these niches;
the other banks only within these niches, not across.
3. Compare the BI values with the Lerner indices estimated for Russian banks and find no conflict between them as opposed to the recent research
4
Methodology 1/3
Specify the following translog operating cost function (similar to Turk Ariss, 2010; Fiordelisi et al., 2011):
where for bank i at quarter t OC – operating costs; Yh – h-th output (j=1…3 – commercial loans; deposits; fee and commission); Pm – m-th factor input price (m=1…3 – funds; personnel; physical capital); EQ – equity capital as a netput to control for risk preferences
Estimate it using ML (within SFA), OLS and GMM with currency and securities revaluations being dropped from or kept in OC – obtain 6 alternatives for marginal costs (j=1…6):
5
itq
r q
itrrq
m
itmmitl
k j
itkkl
h
ithhit PPPYYYOC ,
3
1
3
1
,
3
1
,,
3
1
3
1
,
3
1
,0 lnln2
1lnlnln
2
1lnln
2
21
2
21
3
1
,
3
1
,,
3
1
3
1
, lnlnlnlnln itit
m
itpm
h
ithhitu
s u
itssu EQEQTTTPTYPY
itit
m
ititmm
h
itithh EQTEQPEQY
lnlnlnlnln3
1
,
3
1
,
LNS
it
it
j
LNS
it
it
j
LNS
it
itLNS
itjY
OC
Y
OC
Y
OCMC
ln
ln,
Methodology 2/3
A set of multivariate modifications of the market shares equation:
where for bank i at quarter t MS – market share; MCj – j-th proxy for marginal cost (j=1…6); Xk – k-th bank-specific factor; DUMq – a dummy for bank group (q=1…2); Cycle – macro control: a deviation of loans-to-GDP ratio from its HP-filtered values
2-step GMM estimator is employed
The underlying bank-level Boone indicator is:
The more negative values – the stronger a competition on quantity
The more positive values – the stronger a competition on quality
6
K
k
itkkj
K
k
itk
LNS
itjkj
LNS
itjjtjijit XXMCMCMS1
,,
1
,,,,,, lnlnlnlnln
itjtjt
LNS
itjj
q
qqj
q
q
LNS
itjqj CycleCycleMCDUMDUMMC ,,
2
1
,
2
1
,, lnln
tj
q
qqj
K
k
itkkjjLNS
itj
ititj CycleDUMX
MC
MSBOONE
2
1
,
1
,,
,
, lnln
ln
Methodology 3/3
Compare the Boone indicator (competition on quantity or in quality) with the Lerner index (in Koetter et al. (2012) modification):
vs.
where for bank i at quarter t r – Interest income from Loans / Total loans; AFR – Interest expenses / Total Funds; MCj – j-th proxy for marginal cost (j=1…6)
7
LNS
it
LNS
itjit
LNS
it
itjr
MCAFRrLerner
,
,
tj
q
qqj
K
k
itkkjjLNS
itj
ititj CycleDUMX
MC
RBOONE
2
1
,
1
,,
,
, lnln
ln
Data
1. Bank-specific factors (BSF): The Bank of Russia web-site (www.cbr.ru) – monthly balance sheets of banks (Form 101); – quarterly profit and loss accounts (Form 102).
2. Macroeconomic controls (MACRO): The Federal State Statistics Service web-site (www.gks.ru)
3. Time period: Q1 2005 – Q4 2013 (40 quarters)
4. Number of banks (depending on the quarter): – in original sample: up to 1254;
– in adjusted sample: up to 971 after excluding the observations below 1st and above 99th percentiles on relative indicators. It provides up to 21446 observations
8
Preliminary estimation results: univariate Boone (Marginal costs obtained from SFA; revaluations were dropped from operating costs)
9
Outcome: omitted variables do matter ⇒ multivariate Boone indicator needed
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
-0.2
7
-0.2
5
-0.2
2
-0.1
9
-0.1
7
-0.1
4
-0.1
2
-0.0
9
-0.0
6
-0.0
4
-0.0
1
0.0
2
0.0
4
0.0
7
0.0
9
0.1
2
0.1
5
0.1
7
0.2
0
1. The ratio of retail loans in total loans, %
2. Loans-to-assets ratio, %
3. The share of overdue loans in total loans, %
4. Loans-to-deposits ratio (LTD), %
5. The share of reserve assets in total assets, %
6. Loan loss reserves to interest income ratio, %
7. Net deposits to total assets ratio, %
Bank-level Boone indicator (market for loans)
% o
f to
tal nu
mb
er
of
ban
ks
1.
2.
3.
4.
5.
6.
7.
Market shares equations with one of the following bank-specific factor (X):
Final estimation results: Market shares equations (GMM) Explanatory variables (in logs)
Dependent variable – market share (in logs) M1 Model M2 Model M3 Model
MC1 (via SFA; Revals are dropped from OC) 0.010 (0.013) 0.035*** (0.014) 0.135*** (0.013)
MC1 × The share of retail loans in total loans 0.105*** (0.011) 0.103*** (0.011) 0.101*** (0.010)
The share of retail loans in total loans –0.112*** (0.010) –0.108*** (0.010) –0.112*** (0.009)
MC1 × Loans-to-assets ratio 0.001 (0.028)
Loans-to-assets ratio 1.026*** (0.023)
MC1 × Equity-to-assets ratio –0.007 (0.021) 0.002 (0.021) 0.021 (0.019)
Equity-to-assets ratio –0.869*** (0.017) –0.842*** (0.017) –0.662*** (0.015)
MC1 × The share of overdue loans in total loans 0.028*** (0.006) 0.029*** (0.006) 0.036*** (0.006)
The share of overdue loans in total loans –0.060*** (0.004) –0.059*** (0.004) –0.044*** (0.004)
MC1 × The share of non-interest income in total income –0.001 (0.014) 0.009 (0.014) –0.004 (0.014)
The share of non-interest income in total income –0.024** (0.011) –0.020* (0.011) 0.007 (0.010)
MC1 × Loans-to-deposits ratio 0.117*** (0.031)
Loans-to-deposits ratio 0.685*** (0.024)
MC1× The share of reserve assets in total assets 0.079*** (0.018) 0.076*** (0.017) 0.019 (0.015)
The share of reserve assets in total assets –0.100*** (0.015) –0.112*** (0.015) –0.159*** (0.013)
MC1 × Loan loss reserves to interest income ratio –0.0043*** (0.0006)
Loan loss reserves to interest income ratio –0.014*** (0.000)
MC × Federal state banks 0.281*** (0.108) 0.246*** (0.109) 0.294*** (0.103)
MC × Foreign-owned banks –0.107 (0.080) –0.116 (0.079) –0.088 (0.072)
Final estimation results: the distributions of alternative versions of Boone indicator
11
The final versions of bank-level Boone indicator: 1. MC is estimated within SFA; Revaluations of currency and securities were dropped from costs 2. Russian banks are mostly competing on quality, not quantity 3. Sberbank dominated in terms of bank-level Boone in all regressions 4. The M1 Model is preferred (provides the largest – 30% – part of banks competing on quantity)
0
5
10
15
20
25
30
35
-0.3
0-0
.26
-0.2
1-0
.16
-0.1
1-0
.07
-0.0
20
.03
0.0
80
.12
0.1
70
.22
0.2
70
.31
0.3
60
.41
0.4
60
.50
0.5
5
1. The M1 Market share Model
2. The M2 Market share Model
3. The M3 Market share Model
Bank-level Boone indicator (market for loans)
% o
f to
tal sa
mple
ass
ets
Competition
on quantity
(M1) 30% of
total sample
assets
Competition
on quality
(M1) 70% of
total sample
assets
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18-0
.31
-0.2
2
-0.1
4
-0.0
6
0.0
3
0.1
1
0.2
0
0.2
8
0.3
7
0.4
5
1. The M1 Market share Model
2. The M2 Market share Model
3. The M3 Market share Model
Bank-level Boone indicator (market for loans)
% o
f to
tal num
ber
of
banks
1.
2.
3.
Boone indicator in the 4 niches of the market for loans
12
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
The share of overdue loans in total loans, %
Th
e r
ati
o o
f re
tail
lo
an
s
in t
ota
l lo
an
s, %
SBER-
BANK
VTB24 RUSSIAN STANDARD HOME
CREDIT OTP
ROSBANK
URALSIB
MDM
TRUST
MTS-BANK
GAZPROMBANK
VTB
ROSSELKHOZBANK
BANK OF MOSCOW
20.0
CITY
ROSGOSSTRAKH
UNIASTRUM
PETROCOMMERZ
SOYZ
ALFA-BANK
VOZROZDENIE
Notes: The competition map as of the end of 2013 Circle area corresponds to the value of Boone indicator interacted with respective bank size (in terms of assets)
Boone vs Lerner: no conflict revealed 1/2
13
-0.02
-0.03
0.01 0.01
0.01
0.04
0.05
0.06
0.14 0.13
0
20
40
60
80
100
120
140
160
180
200
-0.04
-0.02
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
-53 -37 -20 -4 13 29 46 62 78 95
Boone indicator: averaged value in respective Lerner decile
Share of total sample assets (secondary axxis), %
Number of banks (secondary axxis)
Lerner index (market for loans), %
Ban
k-le
vel B
oo
ne
ind
icat
or
(mar
ket
for
loan
s)
Nu
mb
er
of
ba
nk
s a
nd
th
eir
siz
e
Pairwise correlation of Boone and Lerner is 93% at the decile level
Boone vs Lerner: no conflict revealed 2/2
14 All banks in the sample
Boone
Lerner, lag = 0 q 0.222
Lerner, lag = 1 q 0.223
Lerner, lag = 2 q 0.221
Lerner, lag = 3 q 0.211 Lerner, lag = 4 q 0.205
The sample’s median
Boone
Lerner, lag = 0 q -0.03
Lerner, lag = 1 q 0.12 Lerner, lag = 2 q 0.28
Lerner, lag = 3 q 0.44 Lerner, lag = 4 q 0.56
0.00
0.06
-0.02
0.03
30.6
11.3
19.2
0
5
10
15
20
25
30
35
-0.03
-0.02
-0.01
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
20
05
q1
20
05
q3
20
06
q1
20
06
q3
20
07
q1
20
07
q3
20
08
q1
20
08
q3
20
09
q1
20
09
q3
20
10
q1
20
10
q3
20
11
q1
20
11
q3
20
12
q1
20
12
q3
20
13
q1
20
13
q3
Bank-level Boone indicator (market for loans), sample's median
Lerner index (market for loans), sample's median; secondary axxis
Conclusion
Within the panel framework, propose to use the external variables approach in order to disaggregate an industry-wide Boone indicator of banking competition to the bank level
For the sample of Russian banks, show that the key role in such bank-level modification belongs to (i) The share of retail loans in total loans, (ii) The share of overdue loans in total loans, (iii) The share of reserve assets in total assets, (iv) Loan loss reserves to interest income ratio
Having estimated the bank-level Boone indicator (BI) for Russian banks, show that
(1) Russian credit market is a monopoly with Sberbank being the dominant player and possessing highest value of BI;
(2) Russian banks mainly compete on quality rather than quantity in credit market. This effect becomes larger during the upward phase of credit cycle or when loan quality deteriorates;
Having divided Russian credit market into 4 niches using risk and loans compositions criteria and having analyzed the BI of banks in that niches, find that
(1) Competition is strongest within the “low risk – corporate lending” niche and
(2) Competition is weak within the “high risk – retail lending” niche
For Russian banks, bank-level Boone indicator and Lerner index have no conflict