BANK OWNERSHIP AND COST EFFICIENCY IN RUSSIA, REVISITED MIKHAIL MAMONOV, CENTER FOR MACROECONOMIC ANALYSIS AND SHORT-TERM FORECASTING (CMASF); RUSSIAN

BANK OWNERSHIP AND COST EFFICIENCY IN RUSSIA, REVISITED

MIKHAIL MAMONOV,CENTER FOR MACROECONOMIC ANALYSIS AND SHORT-TERM FORECASTING (CMASF); RUSSIAN ACADEMY OF SCIENCES – INSTITUTE

FOR ECONOMIC FORECASTING; NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS (MOSCOW)

ANDREI VERNIKOVNATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS;

RUSSIAN ACADEMY OF SCIENCES - INSTITUTE OF ECONOMICS (MOSCOW)

World Congress on Comparative Economics (Rome, June 25-27, 2015)

Motivation

Literature on comparative bank efficiency in transition economies and Russia:Caner, Kontorovich, 2004; Bonin, Hasan, Wachtel, 2005; Fries, Taci, 2005; Grigorian, Manole, 2006; Golovan, 2006; Fries, Neven, Seabright, Taci, 2006; Golovan, Karminsky, Peresetsky, 2008; Karas, Schoors, Weill, 2010; Peresetsky, 2010; Fungáčová, Poghosyan, 2011

“Conventional wisdom”: Foreign-owned banks are the best performers (although Mian, 2006 and Lensink et al., 2008

claim otherwise) State-owned banks lag behind in terms of efficiency (is it always true?)

Previous studies on banking in EMs and transition have missed the distorting effect that the revaluations of both securities and the all types of assets and liabilities denominated in foreign currency (Revals) might have on banks’ costs This is especially true for dollarized and commodity countries like Russia Cost efficiency must be re-estimated

The positions held by various bank groups in the efficiency ranking are assumed to remain (i) independent from bank-specific factors and (ii) permanent over the sample period. But this assumption is out of line with empirical evidence. The intra-group causes of ranking change need to be modelled.

2

Revaluations: Is it important to account for them? (1)

3

* The peaks correspond to Sberbank

Frequency distribution of banks according to the materiality of Revals

(a) Distribution by the number of banks (b) Distribution by the share in total assets of the banking system

Revaluations: Is it important to account for them? (2)

The effects of Revals are unevenly distributed among banks at each point of observations, so they do matter for the estimation results.

4

Negative Revals as percentage of total costs

4/1/

2005

10/1

/200

5

4/1/

2006

10/1

/200

6

4/1/

2007

10/1

/200

7

4/1/

2008

10/1

/200

8

4/1/

2009

10/1

/200

9

4/1/

2010

10/1

/201

0

4/1/

2011

10/1

/201

1

4/1/

2012

10/1

/201

2

4/1/

2013

10/1

/201

3

0

10

20

30

40

50

60

70

80

3.90.7

13.1

35.9

20.6

50.3

76.0

68.0

Banking system average 5th percentile 25th percentile

50th percentile 75th percentile 95th percentile

Objective

Perform SFA estimation of Russian banks cost efficiency, comparing state banks to privately owned banks and foreign-controlled banks

Action plan: Show the materiality of Revals for the performance of Russian banks and the

uneven distribution of such revaluation among banks Control for Revals in the stochastic frontier analysis of comparative cost

efficiency of different bank groups in Russia See how bank group ranking changes when Revals are dropped Explain empirically what drives the change in rankings.

5

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: 705-1024;– in adjusted sample: 650-997 after excluding observations below 1st and above 99th percentiles

The number of observations ranges from 20 000 to 29 000 depending on the quarter and the bank’s involvement in different activities (lending, securities, fees & commissions, etc.).

6

Bank grouping

State-1 State-2 Private Foreign

Two sub-groups of state-controlled banks, rather than one group: core state-controlled banks: Sberbank, VTB, Rosselkhozbank (STATE-1) other state-controlled banks: between 37 and 54 banks, depending on the period

(STATE-2)

A refined definition of foreign banks (FOREIGN). Focus on foreign bank subsidiaries, i.e. banks controlled by foreign strategic investors. Filter out:

pseudo-foreign banks (those ultimately controlled by Russian capital); banks owned by portfolio investors and non-banking entities (private individuals,

international institutions, development agencies, car manufacturers, clearing companies, securities firms, etc.).

More detail: Vernikov A. (2015), A guide to Russian bank data: Breaking down the sample of banks // SSRN Working Paper Series No 2600738.

7

Preliminary statistics: Bank groups

8

The breakdown of the sample of banks (the number of banks and the group’s share in total assets of the sample in the respective quarter)

Period

Core state-controlled banks

(State-1)

Other state-controlled banks

(State-2)

Domestic privately-owned banks

(Private)

Foreign subsidiary banks (Foreign)

Total

No. % No. % No. % No. % No. %

4Q2005 3 36.8 28 11.7 745 41.7 27 9.8 803 100.0

4Q2006 3 35.0 30 12.7 865 42.5 27 9.8 925 100.0

4Q2007 3 36.7 33 12.0 891 39.9 34 11.4 961 100.0

4Q2008 3 38.3 45 17.7 871 32.1 37 11.9 956 100.0

4Q2009 3 39.8 46 18.5 920 31.3 46 10.4 1015 100.0

4Q2010 3 39.4 41 17.4 908 33.1 48 10.1 1000 100.0

4Q2011 3 40.8 37 17.5 880 32.0 45 9.7 965 100.0

4Q2012 3 41.5 36 17.0 857 32.4 43 9.1 939 100.0

4Q2013 3 42.6 36 17.7 820 31.7 42 8.0 901 100.0

Empirical strategy (1): two-step vs. one-step

1. Two-step approach (Maudos and Fernández de Guevara, 2007; Solís and Maudos, 2008; Berger et al. 2009; Turk Ariss, 2010): estimate (1) bank-level efficiency scores econometrically and then (2) its determinants

2. Wang and Schmidt (2002): at step (1) efficiency scores are supposed to be one-sided, but at step (2) – two-sided variable, which yields biased estimation results

3. Battesse and Coelli (1995) propose a one-step approach for stochastic frontier analysis (SFA): estimate (1) and (2) simultaneously

4. One-step approach (Bonin et al., 2005; Karas et al., 2010)

5. We choose two-step approach for basic estimations and one-step alternative – for robustness checks

9

Empirical strategy (2): outline

1. Estimate and compare the two alternatives of bank-level cost efficiency via Stochastic Frontier Analysis (SFA) under production approach:

alt=1: with Revals KEPT (as in previous research);alt=2: with Revals DROPPED,using translog cost function with 3 input prices (funds, personnel, physical capital), 3 outputs (commercial loans, deposits, fee & commission), and 1 netput (equity capital).

2. Aggregate bank-level SFA-scores into group-level for the following bank groups: STATE-1, STATE-2, FOREIGN, and PRIVATE (equaled weights)

3. Explain group rankings changes in terms of SFA-scores in a static panel framework:

where X is equity-to-assets ratio (h=1; the risk preferences case) OR loans-to-assets ratio (h=2; the assets composition case), which are, by assumption, responsible for changes in bank or banks’ groups positions in efficiency rankings over time; BSF and MACRO – other bank-specific and macroeconomic control variables; ε – regression error

2-step GMM is employed to address endogeneity and heteroscedasticity concerns in baseline regressions

Determine the distances

10

it

M

mmm

K

kitkkithh

jithjhj

jjhjih

altit MACROBSFXXGROUPGROUPSFA

11,,

3

1,

3

1,

)(

ihhjhjaltijj XSFAGROUPeachfor ,

)(,:

Empirical strategy (3): details

1. Translog cost function: general representation

alt=1: with Revals KEPT (as in previous research);alt=2: with Revals DROPPEDv is two-sided random error; u is one-sided inefficiency term (positive half-N distrib.)

2. Translog cost function: full representation

3. Bank-level cost efficiency score (SFA-score) takes the form:

11

ititititmitalt

it uvEQTrendPYfOC ln;;;lnln ,)(

}ˆexp{ )()( altit

altit uSFA

itqr q

itrrqm

itmmitlk j

itkklj

itjjalt

it PPPYYYOC ,

3

1

3

1,

3

1,,

3

1

3

1,

3

1,0

)( lnln2

1lnlnln

2

1lnln

itm

itpmj

itjjitus u

itssu EQTTTPTYPY lnlnlnlnln 12

21

3

1,

3

1,,

3

1

3

1,

itititm

ititmmj

ititjjit uvEQTEQPEQYEQ

lnlnlnlnlnln3

1,

3

1,

22

Preliminary statistics: translog cost function

12

Descriptive statistics of variables in the cost function (2005 Q1 – 2013 Q4)

Unit Symbol Mean St.Dev Min Max Obs Banks

Dependent variables

Total costs minus interest expenses minus Revals

RUB bn 7.7 69.8 0.0 2904 30784 1196

Total costs minus interest expenses

RUB bn 19.2 207.2 0.0 8886 30753 1196

Explanatory variables

Loans to households and nonfinancial firms

RUB bn 18.2 206.7 0.0 10015 30045 1159

Retail and corporate accounts and deposits

RUB bn 16.6 205.1 0.0 10375 30635 1191

Fee and commission income RUB bn 0.5 5.0 0.0 220.6 30635 1189

Average funding rate % 4.9 2.8 0.0 50.1 29365 1152

Price for personnel expense % 4.1 3.3 0.1 49.5 30784 1196

Price of physical capital % 23.7 22.4 0.2 180.0 30784 1196

Equity capital RUB bn 3.8 40.8 0.0 1954 30745 1196

)1(itOC

)2(itOC

itY ,1

itY ,2

itY ,3

itP ,1

itP ,2

itP ,3

itEQ

Estimation results 1/2: bank-level cost efficiency

13

Aggregate results of cost efficiency estimations: Frequency distribution of banks’ SFA scores as average of 2005Q1-2013Q4 (production approach)

5.60

10.8

3

16.0

6

21.2

9

26.5

2

31.7

6

36.9

9

42.2

2

47.4

5

52.6

8

57.9

1

63.1

4

68.3

7

73.6

0

78.8

3

84.0

6

89.2

9

94.5

2

99.7

5

0

5

10

15

20

25

alt=1: Revals kept alt=2: Revals dropped

Bank-level SFA-scores

% o

f tot

al n

umbe

r of

ban

ks

Estimation results 2/2: group-level cost efficiency in dynamics

alt=1: Revals kept alt=2: Revals dropped

Group-level SFA scores (arithmetic averages within each group)

142

00

5q1

20

05

q42

00

6q3

20

07

q22

00

8q1

20

08

q42

00

9q3

20

10

q22

01

1q1

20

11

q42

01

2q3

20

13

q2102030405060708090

100

Private State-1 State-2 Foreign

Ban

ks' S

FA s

core

s100 - Efficiency frontier

CR

I-S

IS

20

05

q12

00

5q4

20

06

q32

00

7q2

20

08

q12

00

8q4

20

09

q32

01

0q2

20

11

q12

01

1q4

20

12

q32

01

3q2

102030405060708090

100


Ban

ks' S

FA s

core

s

100 - Efficiency frontier

CR

ISIS

GMM estimation results 1/3: what explains the within- and between-group heterogeneity of cost efficiency? Dependent variable: SFA cost efficiency score

15Revals KEPT DROPPED

M1.1 M1.2 M1.3 M2.1 M2.2 M2.3Dummy variables for bank ownership status (Domestic privately-owned banks is a referent group)

State-1 2.780 –1.584 –4.390 2.704* –5.365** 15.123***

State-2 8.559*** 7.387*** 7.055** 1.672*** 1.895*** 5.643***

Foreign –16.222*** –7.939*** –28.756*** –0.021 –4.456*** –20.925***

Bank-specific factors explaining within-group heterogeneity of cost efficiency

Equity-to-assets ratio (ETA) 0.661*** 0.676*** 0.638*** 0.426*** 0.417*** 0.419***

ETA × State-1

0.386

0.569***

ETA × State-2

0.090

–0.021

ETA × Foreign

–0.453***

0.241***

Loans-to-assets ratio (LTA) 0.607*** 0.606*** 0.589*** 0.439*** 0.439*** 0.428***

LTA × State-1

0.107

–0.244***

LTA × State-2

0.023

–0.074***

LTA × Foreign

0.215***

0.371***

No. of obs. (banks) 19546 (967)

19546 (967)

20319 (978)

19573 (967)

19573(967)

20319 (978)

Centered R2 0.337 0.369 0.352 0.557 0.559 0.549No. of endog. vars., excluded instruments 6, 12 9, 15 9, 15 6, 12 9, 15 9, 15P-value for Hansen J-stat 0.558 0.569 0.719 0.143 0.221 0.167Centered R2 0.000 0.000 0.000 0.000 0.000 0.000

***, ** and * – an estimate is significant at the 1%, 5% and 10%, respectively. Standard errors are not reported. Other bank-specific and macro controls are not reported in order to preserve space.

GMM estimation results 2/3: Ranking of group-level cost efficiencies based on heterogeneity factor No.1 (equity-to-assets ratio, ETA)

16

***, ** and * – an estimate is significant at the 1%, 5% and 10%, respectively. Standard errors are not reported

Outcomes:1) If Revals are kept, Foreign banks are always the least cost efficient group. But if Revals are dropped, Foreign with high ETA ratios are more cost efficient as compared to Private and State-2 groups;2) If Revals are kept, State-1 are always less cost efficient than State-2. But if Revals are dropped, State-1 with high ETA ratios outperform all the other groups

Percentile p10 p25 p50 p75 p90

Panel 1: Revals KEPT (Model M1.2)

State-1 1.807 2.331 3.153 4.343* 6.592**

State-27.993*** 8.176*** 8.417*** 9.018*** 10.115***

Foreign–11.670*** –12.967*** –14.981*** –19.120*** –26.742***

Panel 2: Revals DROPPED (Model M2.2)

State-1–0.370 0.403 1.614 3.368** 6.679***

State-21.753*** 1.711*** 1.654*** 1.514*** 1.258***

Foreign–2.471*** –1.781** –0.709 1.493* 5.548***

Panel 3: Percentiles of ETA distributions within particular group of banks

State-18.8 10.1 12.3 15.3 21.2

State-26.7 8.8 11.5 18.1 30.3

Foreign8.2 11.1 15.6 24.7 41.5

Private8.2 11.0 16.5 27.1 44.3

GMM estimation results 3/3: Ranking of group-level cost efficiencies based on heterogeneity factor No.2 (loans-to-assets ratio, LTA)

17


Outcomes: When Revals are dropped, 1) State-1 banks can become the most cost efficient group if they substantially decrease LTA ratio; 2) On the contrary, Foreign banks are the most cost efficient only in case they substantially increase LTA.

Percentile p10 p25 p50 p75 p90Panel 1: Revals KEPT (Model M1.3)

State-1–0.453 0.286 2.141 2.702 3.224

State-27.568*** 7.975*** 8.262*** 8.488*** 8.695***

Foreign–27.380*** –23.573*** –18.699*** –15.599*** –13.536***

Panel 2: RevRevals DROPPED (Model M2.3)

State-16.140*** 4.454** 0.223 –1.058 –2.247

State-24.006*** 2.708*** 1.793*** 1.072*** 0.410

Foreign–18.552*** –11.989*** –3.586*** 1.758** 5.316***

Panel 3: Percentiles of LTA distributions within particular group of banks

State-136.8 43.7 61.1 66.3 71.2

State-222.0 39.4 51.7 61.4 70.3

Foreign6.4 24.1 46.7 61.1 70.7

Private23.3 39.4 54.8 66.7 75.8

Robustness check: Overview

We check the robustness of the findings by re-estimating Translog cost function and respective SFA scores

within the intermediation approach instead of the production approach; by adding more outputs (securities, foreign assets); by estimating total costs rather than operating costs as a dependent variable.

Empirical dependences of group rankings on risk preferences or assets composition within The production approach and Instrumental Variables (IV) Tobit estimation technique

rather than GMM procedure to account for the censored nature of SFA scores (i.e. lower and upper bounds, 0 and 100, respectively);

The intermediation approach and 2-step GMM procedure; The intermediation approach and IV Tobit estimation technique.

Our main outcomes remain qualitatively unchanged.

18

Robustness check: Banks groups’ SFA scores re-estimated 1/3

Adding securities into the translog (operating) cost function:

19

2005

q120

05q3

2006

q120

06q3

2007

q120

07q3

2008

q120

08q3

2009

q120

09q3

2010

q120

10q3

2011

q120

11q3

2012

q120

12q3

2013

q120

13q3

102030405060708090

100



CRI-SIS

Ban

ks' S

FA s

core

s


Adding securities and foreign assets into the translog (operating) cost function:

20

2005

q120

05q3

2006

q120

06q3

2007

q120

07q3

2008

q120

08q3

2009

q120

09q3

2010

q120

10q3

2011

q120

11q3

2012

q120

12q3

2013

q120

13q3

10

20

30

40

50

60

70

80

90

100



CRI-SISB

anks

' SFA

sco

res


Replacing operating costs with total costs in the translog cost function:

21

2005

q120

05q3

2006

q120

06q3

2007

q120

07q3

2008

q120

08q3

2009

q120

09q3

2010

q120

10q3

2011

q120

11q3

2012

q120

12q3

2013

q120

13q3

10

20

30

40

50

60

70

80

90

100



CRI-SIS

Ban

ks' S

FA s

core

s

Robustness check: IV-Tobit estimation results 1/2Ranking of group-level cost efficiencies based on heterogeneity factor No.1 (equity-to-assets ratio, ETA)

22


Outcomes: the same

Percentile p10 p25 p50 p75 p90

Panel 1: Revals KEPT (Model M1.2)

State-11.548 2.028 2.780 3.869 5.924

State-27.921*** 8.128*** 8.400*** 9.078*** 10.314***

Foreign–11.785*** –13.090*** –15.116*** –19.278*** –26.942***


State-1–0.678 0.165 1.488 3.403** 7.018***

State-21.677*** 1.665*** 1.648*** 1.608*** 1.534**

Foreign–2.635*** –1.952** –0.892** 1.286** 5.297***

Panel 3: Percentiles of ETA distributions within particular group of banks

State-18.8 10.1 12.3 15.3 21.2

State-26.7 8.8 11.5 18.1 30.3

Foreign8.2 11.1 15.6 24.7 41.5

Private8.2 11.0 16.5 27.1 44.3

Robustness check: IV-Tobit estimation results 2/2Ranking of group-level cost efficiencies based on heterogeneity factor No.2 (loans-to-assets ratio, LTA)

23


Outcomes: the same

Percentile p10 p25 p50 p75 p90Panel 1: Revals KEPT (Model M1.3)

State-1–0.739 –0.010 1.818 2.371 2.885

State-27.563*** 7.987*** 8.286*** 8.522*** 8.738***

Foreign–27.188*** –23.466*** –18.699*** –15.668*** –13.650***


State-15.517*** 3.951*** 0.023 –1.166 –2.270

State-24.079*** 2.764*** 1.836*** 1.106*** 0.435

Foreign–18.378*** –11.881*** –3.561*** 1.729*** 5.252***

Panel 3: Percentiles of LTA distributions within particular group of banks

State-136.8 43.7 61.1 66.3 71.2

State-222.0 39.4 51.7 61.4 70.3

Foreign6.4 24.1 46.7 61.1 70.7

Private23.3 39.4 54.8 66.7 75.8

Reality check: What’s wrong with foreign banks in Russia?

The share of foreign banks in the Russian banking sector never exceeded 20% (now c. 10%) whereas in the majority of transition countries foreign banks prevail. Small market shares; inability to explore economies of scale

Some of foreign banks entered the Russian banking system just before the 2008-2009 crisis; lost money; became dormant

Risk aversion Institutional misalignment. The greater the difference in institutions between

the home and host countries, the stronger the negative effect that foreign banks can have on cost efficiency of the host banking system (Lensink, Meesters, Naaborg. Bank efficiency and foreign ownership: Do good institutions matter? JBF, 2008)

24

Conclusions 1/2

The effects of Revals are material for, and unevenly distributed among, banks in Russia, so they do affect the efficiency estimation results and upset bank group rankings

Having controlled for Revals in SFA-scores estimations, we find that, on average: efficiency scores become higher and less volatile across the board; the spreads between different types of banks in terms of efficiency shrink; foreign bank subsidiaries appear to be the least efficient market participants; during financial turmoil the efficiency of banks grows as compared to normal

circumstances; the core state banks are more efficient than other state-controlled banks in the post-

crisis period and nearly as efficient as domestic private banks.

Through GMM and Tobit estimations we show that bank-level characteristics such as asset composition and risk preference can help modelling the changes in the bank group rankings

25

Conclusions 2/2

GMM and Tobit estimations within production approach and intermediation approach demonstrate that: the core state banks tend to be the most efficient group in case they hold larger

equity capital and decrease loans-to-assets ratio (pursue non-interest income); conversely, foreign subsidiary banks can outperform other groups in terms of cost

efficiency only when they substantially increase lending These findings only become visible after Revals are dropped.

Some of our results are consistent with previous research in that state-owned banks are not necessarily less efficient as compared to privately-owned banks (Karas, Schoors, Weill, 2010)

Others results challenge the conventional wisdom with regard to the general level of Russian bank efficiency, the performance of foreign-controlled banks (Bonin, Hasan, Wachtel, 2005; Fries, Taci, 2005; Grigorian, Manole, 2006) and bank behavior during crises.

Our approach is applicable to other dollarized emerging markets.

26

THANK YOU FOR YOUR ATTENTION!

MIKHAIL MAMONOVCENTER FOR MACROECONOMIC ANALYSIS AND SHORT-TERM FORECASTING (CMASF); RUSSIAN ACADEMY OF SCIENCES -

INSTITUTE FOR ECONOMIC FORECASTING; NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS (MOSCOW, RUSSIA). EMAIL: [email protected]

ANDREI VERNIKOVNATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS;

RUSSIAN ACADEMY OF SCIENCES - INSTITUTE OF ECONOMICS (MOSCOW, RUSSIA). EMAIL: [email protected]

Documents

BANK OWNERSHIP AND COST EFFICIENCY IN RUSSIA, REVISITED MIKHAIL MAMONOV, CENTER FOR MACROECONOMIC ANALYSIS AND SHORT-TERM FORECASTING (CMASF); RUSSIAN