Measuring bank competition in China: A comparison of new versus - - PowerPoint PPT Presentation

Measuring bank competition in China: A comparison of new versus conventional approaches applied to loan markets.

Bing Xu (Universidad Carlos III de Madrid) Adrian van Rixtel (Bank for International Settlements) Michiel van Leuvensteijn (APG) Michiel van Leuvensteijn (APG)

VIII Annual Seminar on Risk, Financial Stability and Banking Banco Central do Brasil/Journal of Banking and Finance São Paulo, 8-9 August 2013.

Restricted

Summary: What do we analyze?

 How competitive are loan markets in China according to

conventional and new measures?

 How did competition in these markets develop over time?  What explains the differences between the results of the

i ( i i ll d th ti ll )? various measures (empirically and theoretically)?

 How to interpret the results over time (financial reform)?  How to interpret the results over time (financial reform)?

Restricted

2

Summary: How do we analyze it?

 First paper that combines conventional and new measures.  Conventional

approaches to measure bank competition (Panzar-Rosse H statistic and (elasticity-adjusted) Lerner index).

 Relatively new approach to measure bank competition (Profit

El ti it (PE) i di t ) Elasticity (PE) indicator).

 PE indicator: competition rewards efficiency and punishes  PE indicator: competition rewards efficiency and punishes

• inefficiency. If markets are more competitive, difference in

profitability between efficient and less efficient banks is larger.

Restricted

3

Summary: What are our results?

R l f i l i di d li i i i

 Results of conventional measures indicate declining competition

• ver time in Chinese loan markets.

 Relatively new PE indicator shows improving competition.  Results PE indicator show that financial reform (based on three

different indicators) is associated with improving competition (reverse finding for conventional measures) => not discussed.

 Results for both conventional and new measures robust for

alternative specifications.

Restricted

4

Summary: What explains the different results?

 Conventional measures are biased due to regulation of interest

rates:

• Panzar-Rosse H-statistic will overestimate competition when

due to interest rate regulation lending and deposit rates move in step move in step.

• We demonstrate theoretically that the Lerner index is biased

when interest regulation is binding. g g

• The

elasticity-adjusted Lerner index will

• verestimate

competition when regulation of lending rate is binding.

 We prove theoretically that the model underlying the PE

indicator is robust under interest rate regulation.

Restricted

5

Structure presentation

I.

Institutional framework: Chinese banking.

II.

Methodological framework PE indicator.

III.

Results conventional vs. new measures.

IV.

Conclusions.

Restricted

6

• I. Chinese banking: financial reform.

Fi i l f (1978 1993 1994 2002 2003 )

 Financial reform (1978-1993, 1994-2002, 2003-present). I.

November 2001: membership WTO approved.

II

Interest rate deregulation:

II.

Interest rate deregulation:

• October 2004: elimination deposit rate floor and lending rate

ceiling. g

• Remaining deposit rate ceiling (binding) and lending rate

floor (non-binding) (Feyzioğlu et al., 2009; He and Wang, l 2011; Ma et al., 2011).

• Deposit rates capped at reference rates (until June 2012: 1.1x

RR) RR)

III.

Reform of credit control system (binding credit plan system abolished in 1998) and other reforms.

Restricted

7

• I. Empirical studies on Chinese banking

 Most on efficiency (Berger et al., JBF 2009; Fu and Heffernan, JBF

2009 and AFE 2010; Matthews et al., HKMA 2009). i i l f l d i

 Financial reform (Berger et al., JBF 2010; Hasan and Xie, BOFIT 2012).  Few on competition, both with H-stat (Yuan, JAE 2006; Bikker et al.,

DNB 2007; Fu BFR 2009) and Lerner index (Fungáčová et al BOFIT DNB 2007; Fu, BFR 2009) and Lerner index (Fungáčová et al., BOFIT 2012; Soedarmono et al., JIFM 2013): declining competition.

 Fungáčová et al. (2012): “… It is somewhat remarkable that China’s

accession to WTO has not led to greater competition in the banking industry.”

 P

l d it f fi i l f ti l

 Puzzle: despite process of financial reform, conventional measures

show bank competition in China declining over time.

 Empirical evidence EMEs: financial reform improved competition.

Restricted

8

 Empirical evidence EMEs: financial reform improved competition.

• I. Our sample of banks in China

 State Owned Commercial Banks (SOCBs): Bank of China,

Agricultural Bank of China, China Construction Bank and Industrial and Commercial Bank of China (“Big Four”) Provide Industrial and Commercial Bank of China ( Big Four ). Provide nationwide wholesale and retail banking services.

 Joint Stock Commercial Banks (JSCBs) (13): partially owned by

Jo t Stoc Co e c a a s (JSC s) ( 3): pa t a y o ed by local government, state owned enterprises and private sector. Operate at national and regional level.

 City Commercial Banks (CCBs): operate at city level. In recent

years though, this geographic restriction has been loosened.

 Foreign banks (FBs): foreign exchange services nationwide  Foreign banks (FBs): foreign exchange services nationwide,

renminbi services at city level.

 In total 127 different banks, 714 observations.

Restricted

9

• II. New measure: Profit Elasticity (PE) indicator.

 Boone or Profit Elasticity (PE) indicator (Jan Boone, Tilburg

University, CEPR); underlying theoretical model of Relative University, CEPR); underlying theoretical model of Relative Profit Differences (RPD).

 Boone et al. (2007), Boone (EJ 2008): RPD/PE indicator are more

robust both from a theoretical and an empirical point of view than PCM-based indicators (eg. Lerner index).

 Our earlier work on competition in various countries and  Our earlier work on competition in various countries and

impact competition

• n

pass-through interest rates (Van Leuvensteijn et al., AE 2011, AE 2013).

 Widely accepted: ECB, IMF, other central banks; academic

publications (Tabak et al., JBF 2012).

Restricted

10

• II. PE indicator.

 Competition rewards efficiency and punishes inefficiency: linkage

between efficiency (relative marginal costs) and performance (i.e. relative profits or market share) relative profits or market share).

 Efficient banks are in general more profitable than inefficient

• banks. The more competitive the market, the larger the difference.

 To obtain the PE indicator, estimate the relationship between

profits and marginal costs (MC) for each bank i: ln (profitsit) = α + βt ln (MCit)+ t dt+ uilt (1)

 βt is the PE indicator, dt is a time dummy.  β

h ld b ti if k t titi fit i

 βt should be negative: if markets are competitive, profits increase

for banks having lower marginal costs relative to their

• competitors. The smaller βt, the stronger is competition.

Restricted

11

p βt g p

• III. Results: conventional measures, Panzar-Rosse H-

statistic and (elasticity-adjusted) Lerner index.

 Panzar Rosse

H statistic: Results show broadly declining

 Panzar-Rosse

H-statistic: Results show broadly declining competition over time.

 Lerner index and elasticity-adjusted Lerner index: Results show

y j that competition declined after 2002.

Restricted

12

• III. Results: new measure, PE indicator.

Step 1: Estimation Translog Cost Function (TCF). Step 1: Estimation Translog Cost Function (TCF). Step 2: Calculate marginal costs (MCs): MCs of loans are the first derivative of the TCF for loans. Step 3: Estimate PE indicator:

 Estimate relationship between profits and MCs:

ln πilt = α +t=1,..,(T-1)tdt + t=1,..,T βt dt ln mcilt + uilt where πilt stands for profits, dt is a time dummy, mcilt for marginal costs i refers to bank i l to output type “loans” marginal costs, i refers to bank i, l to output type loans .

 βt = PE indicator for year t.  βt is negative in theory: profits increase for banks having lower  βt is negative in theory: profits increase for banks having lower

marginal costs relative to their competitiors.

 The more competitive the market, the lower the value of βt.

Restricted

13

• III. Results: new measure, PE indicator.

 Competition improved after WTO accession in 2001.

5 7 1 1 3 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 PE indicator

• 5
• 3
• 1

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

 Sub-sample estimates: competition in post-WTO sign. higher.

• 7

1996-2008 1996-2001 2002-2008

 Results robust for alternative specifications.

PE Indicator

• 2.388***
• 1.514
• 3.570***

(-5.78) (-1.43) (-7.74)

Restricted

14

• IV. Conclusions: H-stat less appropriate for China.

 Interest rate regulation in China.  H-stat: if input and output prices move in step, higher value of

H-stat and interpretation of higher competition.

 H-stat picks up the co-movement of regulated deposit and

lending rates in China (Bikker et al DNB 2007; Feyzioğlu et al lending rates in China (Bikker et al., DNB 2007; Feyzioğlu et al., IMF 2009) =>it measures the degree in which the regulator sets deposit and lending rates jointly.

 Lending rate ceiling abolished in 2004: may have reduced this

bias somewhat => O fi di f h l i i f i i h

 Our findings for the elasticity of interest revenue with respect to

the funding rate are much higher in pre-WTO period than in post-WTO period.

Restricted

15

p p

• IV. Conclusions: Lerner-index less appropriate for China.

W h i ll h f bi di d i d l di

 We prove theoretically that for binding deposit and lending rate

regulation the Lerner index produces biased results.

 If binding deposit rate regulation (binding ceiling) reallocates  If binding deposit rate regulation (binding ceiling) reallocates

sufficient output (loans) from efficient to less efficient banks => Lerner index biased towards increasing competition. This is due f f ff f to the fact that the PCM of less efficient firms is generally smaller than that of more efficient firms.

 Reallocation effect stronger in concentrated markets (such as  Reallocation effect stronger in concentrated markets (such as

Chinese loan markets).

 If binding lending rate regulation => elasticity-adjusted Lerner

index will overestimate competition (Salvo, RAND JE 2010).

 Pre-WTO period in China: lending rate ceiling was binding =>

L i d h hi h l l f titi i Chi

Restricted

16

Lerner index shows high level of competition in China.

• IV. Conclusions: PE indicator better suited for China.

 We prove theoretically that the theoretical model of the PE

i di t (RPD) i i t t f titi i f indicator (RPD) is a consistent measure of competition in case of binding interest rate regulation.

 Intuition:

RPD measures the link between efficiency and

 Intuition:

RPD measures the link between efficiency and performance in relative terms (relative MCs and relative profits).

 If interest rate regulation affects the levels of efficiency and

performance of all banks in the same way, it does not affect them in relative terms.

Restricted

17

Annex (not to be presented, for reference only)

Restricted

18

Interest rate deregulation in China

 Lending rate ceiling: ranged from 1.1 to 1.7 times lending

benchmark rate. Abolished October 2004. di fl d f i l di

 Lending rate floor: ranged from 0.9 to 1 times lending

benchmark rate (mostly 0.9 times).

 Deposit rate ceiling: capped at deposit reference rate until June  Deposit rate ceiling: capped at deposit reference rate until June

2012 when banks were allowed to set one-year deposit rate 10% above the benchmark rate.

 Deposit rate floor: set at deposit reference rate. Abolished

October 2004.

Restricted

19

Model PE indicator: Relative Profit Differences (RPD) Model PE indicator: Relative Profit Differences (RPD).

 RPD is defined as the ratio of the profit differences between  RPD is defined as the ratio of the profit differences between

any three banks in the market (Boone, EJ 2008).

 Banks can be ordered by efficiency (marginal costs or MCs).  Suppose we take three banks

A, B and C, with MC(A) < MC(B)< MC(C), then RPD is defined as: RPD (P fi P fi ) / (P fi P fi ) RPD = (ProfitA – ProfitC) / (ProfitB – ProfitC)

 Competition increases due to increased interaction between

banks and due to lower entry costs banks and due to lower entry costs.

 RPD is an increasing function of interaction and a decreasing

function of entry costs.

 RPD

increases when competition intensifies, ie. fiercer competition decreases (increases) profits of more efficient firms by smaller (larger) amounts than for less efficient firms

Restricted

20

firms by smaller (larger) amounts than for less efficient firms.

PE indicator is estimated in relative terms PE indicator is estimated in relative terms

 Suppose the benchmark bank has profits derived from loans

pp p profitlt and MCs of loans mclt. Regress relative profits on relative MCs to obtain the PE indicator, resulting in: l ( fit / fit ) β l ( / ) ln(profitilt / profitlt)= α +βt ln(mcilt / mclt )+ uilt

 This equation can be transformed into:

ln(profit )= α +β ln(mc )+ ln(profit ) β ln(mc )+u ln(profitilt)= α +βt ln(mcilt)+ ln(profitlt) - βt ln(mclt)+uilt

 Replacing ln(profitlt) - βt ln(mclt) with tdt, we have equation (1) on

previous slide: p ln(profitilt)= α +βt ln(mcilt)+ tdt +uilt

 The relative concept of the PE indicator (and RPD) is captured by

time dummy dt (Boone et al., 2007).

 The exact identity of the benchmark bank does not matter as it

will end up in the time dummy effect

Restricted

21

will end up in the time dummy effect.

Estimation Translog Cost Function (TCF)

S 1 i l l i PE i di E i i f T l C Step 1 in calculation PE indicator: Estimation of Translog Cost Function (TCF):

 Standard approach (Jorgenson 1986; Bikker 2004)  Standard approach (Jorgenson, 1986; Bikker, 2004)  Costs explained by:

• Bank input prices (wage rates, deposit rates (price of

p p ( g , p (p funding) and price of other expenses) in logs.

• Bank
• utput

components (loans, securities and

• ther

i i l services) in logs.

• Control variable: equity ratio which controls for differences

in loan portfolio risk across banks (Berger and Mester 1997) in loan portfolio risk across banks (Berger and Mester, 1997).

 Estimate different TCF for each bank group (SOCBs, JSCBs, CCBs

and FBs) to improve accuracy of calculation marginal costs.

Restricted

22

Robustness tests

P R H i i

 Panzar-Rosse H-statistic:

• Unscaled H-statistic
• Parametric approach
• Parametric approach
• 3-year rolling-window regressions

 Lerner index:  Lerner index:

• Conjectural variation
• Alternative marginal costs

g

 PE indicator:

• Alternative marginal costs
• Alternative definition PE indicator (variable profits)
• Alternative estimation techniques (2SLS, LIML)

Restricted

23

PE indicator Japan PE indicator Japan

12 14 16 6 8 10 12 dicator 2 4 6 Boone-ind

Japan

• 6
• 4
• 2

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 B time

Restricted

24