BANK LENDING, BANK PERFORMANCE AND COMMERCIAL PROPERTY PRICES E Philip Davis NIESR and Brunel...
-
date post
20-Dec-2015 -
Category
Documents
-
view
219 -
download
4
Transcript of BANK LENDING, BANK PERFORMANCE AND COMMERCIAL PROPERTY PRICES E Philip Davis NIESR and Brunel...
BANK LENDING, BANK PERFORMANCE AND
COMMERCIAL PROPERTY PRICES
E Philip Davis
NIESR and Brunel University
West London
www.ephilipdavis.com
groups.yahoo.com/group/financial_stability
Course on Financial Instability at the Estonian Central Bank,9-11 December 2009 – Lecture 9
PAPER 1:BANK LENDING AND
COMMERCIAL PROPERTY PRICES:
some cross-country evidence E Philip Davis and Haibin Zhu
Revise and resubmit in Journal of International Money and Finance
Introduction
• Growing interest in commercial property cycles and link to financial stability
• Likely to be more volatile than residential given no intrinsic reservation value
• Key role of banks in financing commercial property, while CP is also widely used as collateral for non-CP lending
• Little empirical evidence on link from commercial property cycle to credit cycle, notably at international level
Literature review
• Explanations of real estate cycles– Value determined by discounted future rents and
investment by a valuation ratio– Distinctive features of asset market including
heterogeneity, lack of central trading, high transactions costs, supply constraints…
– …and use as collateral for bank loans…– …while external financing needed for construction
and occupancy – generally bank debt
– So optimism raising demand can drive up prices while supply response slow - when supply comes on stream may be excessive relative to demand, driving prices down
– Traditionally such a pattern is seen as requiring not just sticky supplies and rents but also irrationality – basing expected profitability of construction on current prices
– Examples are rules of thumb, myopic expectations, disaster myopia
– Some urge cycles impossible with rational expectations, but following are possible “rational” causes:• No short selling possible to stabilise market• Option value of investment in “anticipated
uncertainty”• Long leases and use of credit• Collateral effects on borrowing capacity,
including the “financial accelerator”• Risk shifting behaviour by banks
– Empirical work in “real estate” literature illustrates interaction of investment, rents and prices, as well as scope for bubbles
• Property prices and bank lending– Background: commercial property price booms
and busts preceding banking crises. Three dimensions of interaction:
(i) Reasons property prices affect credit• Investment channel• Wealth effect on borrowers boosting credit
demand• Banks ownership of property boosting capital
base increases banks’ lending capacity• Financial accelerator effect making lending
procyclical, especially if default risk underestimated in booms
(ii) Reasons lending could affect property prices• Liquidity effect• Credit raising real estate demand; short term
positive effect• Credit raising real estate supply; long term
negative effect• Supply of credit boosted when banks compete,
e.g. after financial liberalisation• Directed to real estate if high quality borrowers
shift to securities market or internal finance• Aggravated by moral hazard
(iii) Common economic factors for lending and real estate prices• Credit affected by shocks to variables such as
GDP and interest rates…• …which also provoke demand and supply
imbalances in real estate(iv) Will changing nature of finance affect the credit-
property price interrelation?• Note in particular that in financially-liberalised
regime, effect of credit on prices is less likely (lending accomodates to demand rather than being rationed, while prices adjust in forward looking manner)
• Extant empirical work – Country-specific studies of interaction with
banking system…– …international studies mainly use residential or
mixed prices, including prediction of financial instability
– But no major academic research project has yet looked at threats to financial stability from the commercial property sector on a systematic, empirical, cross-country basis. This is an important motivation for our own work.
A model of real estate cycles (based on Carey and Wheaton)
Economic environment
– N investors– Heterogeneous valuation of properties, with a distribution of
F(P)– Banks’ lending attitude varies over time wt
– Bank lending function for investors: L(Y, i, P, wt)– Supply K is fixed in short run but adjusts slowly in response
to prices exceeding replacement cost, with separate lending function B(Y,I,P,wt)
– Investment depends on current property prices, for reasons set out above – irrationality, bank capital effects and credit market imperfections
Model
• Market demand function (1), supply adjustment (2), new investment (3) and market clearing (4)
0,0,0,),,,()](1[
PiYt
tttttt LLL
P
wPiYLPFND (1)
11)1( ttt IKK (2)
0,0,0),,,,( 111111 PiYtttttt BBBwPiYBI (3)
tt KD (4)
• Relationship between property prices and bank lending (Lt+Bt)
– Higher current property prices increase bank lending
– Higher Lt (e.g. due to financial liberalisation w) increases current property prices
– Higher Bt reduces future property prices
– Both affected by macroeconomic factors (Y, i)• Simplification – 2 equations, 2 unknowns (K, P)
*
*** ),,,()](1[
P
wPiYLPFNK
(5)
),,,( *** wPiYBK (6)
• Hypothesis I: (collateral/financial accelerator effect) An increase in commercial property prices has a positive impact on bank credit.
• Hypothesis II: (liquidity effect) Bank credit can have offsetting impacts on commercial property prices. New credit to the demand (investor) side may increase property prices in the short run, while new lending to the supply (constructor) side may tend to reduce property prices in the long run.
• Hypothesis III: (macro effect) Commercial property prices adjust to changes in macroeconomic conditions. Their dynamic adjustment depends on the characteristics of the property market in each country. In particular, if the supply is more elastic than the demand, the market reacts to a macro shock in the form of an oscillation around the new steady state; otherwise property prices “overshoot” and then gradually converge to the new steady state.
Empirical analysis
• Data– 17 countries: Australia, Belgium, Canada,
Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, Norway, Spain, Sweden, Switzerland, the UK and the US
– Main focus interrelation of real commercial property prices, GDP, investment, real credit and real short rates
– Most countries’ “true” data is annual – mainly used in our work
– Stationarity as preliminary – all have unit root except real short rate
• Determination of commercial property prices
Error Correction estimation– Panel estimation, GLS, cross section weights, White standard
errors. ECM tends to be highly significant
– For all countries:• Strong short run effect of GDP and credit growth – implies
high cyclical volatility – consistent with model
• Long run positive link to GDP and negative to credit – plausible in terms of model
• Positive real short rate – financial liberalisation?
– Subgroups• G-7, SOEs, bank and market oriented, crisis countries
broadly similar to full panel
• Main contrast is with crisis countries over 1985-95 – long run positive credit and negative investment effect, very high short run elasticities
Results of panel estimation Pooled G-7 Small
open Econ-omies
Bank domin-ated
Market orien-ted
Crisis coun-tries
Crisis coun-tries 1985-1995
All countries 1985-1995
Constant Fixed effect
Fixed effect
Fixed effect
Fixed effect
Fixed effect
Fixed effect
Fixed effect
Fixed effect
DLCREDR 0.75 (6.4)
0.92 (5.5)
0.71 (4.6)
0.84 (5.2)
1.22 (9.9)
0.67 (3.0)
2.4 (5.4)
1.2 (7.7)
DLGDP 1.78 (6.3)
1.13 (2.8)
1.14 (2.3)
1.47 (2.9)
1.8 (4.3)
3.5 (3.8)
2.18 (4.7)
DLI 0.28 (5.0)
0.29 (1.8)
-0.88 (3.7)
LCPPR(-1) -0.09 (5.3)
-0.04 (2.2)
-0.13 (5.0)
-0.16 (6.1)
-0.04 (1.9)
-0.093 (4.0)
-0.18 (4.3)
-0.13 (4.0)
LCREDR(-1)
-0.09 (2.4)
-0.08 (2.2)
-0.073 (2.1)
-0.14 (2.3)
0.4 (2.9)
LGDP(-1) 0.17 (2.3)
0.067 (1.8)
0.21 (2.3)
LI(-1) 0.15 (2.5)
0.096 (1.7)
-0.66 (3.7)
-0.31 (5.6)
RSR RSR(-1) 0.005
(2.4) 0.004
(1.6) 0.005 (1.9)
0.008 (1.8)
R-bar-sq 0.35 0.39 0.32 0.34 0.51 0.38 0.64 0.65 SE 0.11 0.098 0.12 0.11 0.09 0.11 0.11 0.11 DW 1.37 1.23 1.42 1.54 1.00 1.35 1.66 1.52 OBS 439 185 239 285 126 201 88 194
Interaction between bank lending and commercial property prices
• Above evidence gives no view on causality links between credit, commercial property prices and macroeconomic fundamentals
• Granger causality suggests that commercial property prices most commonly precede credit (9 countries) (possibly via effects on collateral and capital), but some reverse causality and interactions (7 countries)
• Granger causality needs supplementing as only bivariate
• Test for dynamic interaction• Method: VECM if there exists cointegration
(Johansen); VAR otherwise (CA, FI, IT, DK, NO, CH)
• Endogeneity issue• Need for choice of recursive ordering in order to
undertake Choleski decomposition• Preferred ordering GDP, commercial property prices,
credit, investment, real short rates• GDP first and interest rate last reflects transmission
mechanism lags• Investment after credit and prices due to supply lags• Prices before credit reflects role of collateral and
price stickiness
• Variance decomposition shows autonomy of commercial property prices (47% in 5 years)
• Link to credit only significant in BE, IT, SE and CH - suggests Granger Causality suffered omitted variables bias
• Wider range of countries show link to GDP – main external influence on commercial property prices
• Credit less autonomous, main influences on variance are GDP (33%) and commercial property prices (20%)
• Overall, confirms influence of external shocks (GDP) on the nexus and of prices on credit
• Variants largely confirm these results
VECM variance decomposition Real commercial property prices Real private sector credit
GDP CPP CRED I RSR GDP CPP CRED I RSR Memo: lags
Australia 40 40 12 1 7 75 9 11 0 5 1 Belgium 41 28 28 1 2 1 2 85 11 1 1 Canada Na Na Na Na Na Na Na Na Na Na Na Denmark 56 34 3 5 1 66 2 20 7 6 1 Finland Na Na Na Na Na Na Na Na Na Na Na France 38 52 3 6 0 55 23 6 13 3 1 Germany 11 83 2 3 1 10 45 11 8 27 1 Ireland 14 44 9 6 26 37 23 3 14 28 1 Italy Na Na Na Na Na Na Na Na Na Na Na Japan 10 76 1 2 11 31 29 4 10 26 1 Netherlands 11 47 13 24 3 14 49 27 1 9 1 Norway 29 66 3 1 2 46 32 21 1 0 1 Spain 9 16 18 53 5 28 3 68 4 0 1 Sweden 32 44 22 0 0 20 19 58 2 1 1 Switzerland 7 40 46 5 2 1 3 94 1 1 1 UK 17 67 1 11 4 31 35 31 4 0 1 US 43 18 1 30 8 42 11 28 13 7 1 Mean level 26 47 12 11 5 33 20 33 6 8 Memo: without RSR
18 58 11 14 34 19 43 4
• Impulse response function– Response of CPP to credit: positive short-term
effect but negative long-term impact in most countries – consistent with theory.
– Response of CPP to GDP: differ by characteristics of national markets. Two types of responses:
– Overshooting in 9 countries (Australia is a typical case)
– Oscillation in 5 countries
Impulse response of prices to creditGERMANY DENMARK
-.05
-.04
-.03
-.02
-.01
.00
.01
1 2 3 4 5 6 7 8 9 10
Response of DELCPPR to CholeskyOne S.D. DELCREDR Innovation
-.07
-.06
-.05
-.04
-.03
-.02
-.01
.00
.01
.02
1 2 3 4 5 6 7 8 9 10
Response of DKLCPPR to CholeskyOne S.D. DKLCREDR Innovation
UNITED KINGDOM UNITED STATES
-.010
-.005
.000
.005
.010
.015
.020
1 2 3 4 5 6 7 8 9 10
Response of UKLCPPR to CholeskyOne S.D. UKLCREDR Innovation
-.015
-.010
-.005
.000
.005
.010
.015
1 2 3 4 5 6 7 8 9 10
Response of USLCPPR to CholeskyOne S.D. USLCREDR Innovation
Impulse response of prices to GDPAUSTRALIA GERMANY
.02
.04
.06
.08
.10
.12
.14
1 2 3 4 5 6 7 8 9 10
Response of AULCPPR to CholeskyOne S.D. AULGDP Innovation
.040
.045
.050
.055
.060
.065
.070
1 2 3 4 5 6 7 8 9 10
Response of DELCPPR to CholeskyOne S.D. DELGDP Innovation
ITALY UNITED KINGDOM
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
1 2 3 4 5 6 7 8 9 10
Response of ITLCPPR to CholeskyOne S.D. ITLGDP Innovation
.02
.04
.06
.08
.10
.12
.14
.16
1 2 3 4 5 6 7 8 9 10
Response of UKLCPPR to CholeskyOne S.D. UKLGDP Innovation
Conclusions• Presented a theoretical model which shows cycles
emerge under plausible assumptions and generating predictions for effects of GDP, interest rates and credit
• Commercial property prices show degree of autonomy, link to GDP but influence on credit
• Predominant direction of causality is from CPP to credit rather than vice versa – collateral/financial accelerator and not liquidity effect; latter effect possibly dampened as financial liberalisation
• Important effect of GDP on both CPP and credit.
• Policy aspects include:– Collateral-based amplification: bank credit policy
• Maximum LTV• Portfolio limits on loan concentration• Valuation method: long run view of valuation vs.
current market value– Financial crises caused by real-estate bubbles– Further research needed
• effects of property prices on bank profitability at micro level – paper 2
• Can commercial property prices predict banking crises – research to be pursued
PAPER 2:COMMERCIAL PROPERTY
PRICES AND BANK PERFORMANCE
E Philip Davis and Haibin Zhu
Published in Quarterly Review of Economics and Finance
Introduction
• Role of asset prices in bank lending and bank performance
• Particular role of commercial property prices, as witness major differences in bank behaviour and performance during the up- and downswings in commercial property prices
• Extensive macro work on commercial property prices and lending (paper 1), but less micro estimation on lending and performance
• Is there a direct impact on the lending decisions, risk and profitability of individual banks?
Table 1
Bank lending and bank performance at different stages of commercial property cycles
(1979-2001)
Growth rate of bank loans (%)
Growth rate of risk-weighted
assets (%)
Return on assets (%)
Provisions on loans as a
percentage of net income (%)
Memo: number of years
Country
Up swing1
Down swing
Up swing
Down swing
Up swing
Down swing
Up swing
Down swing
Up swing
Down swing
Belgium
Canada
Finland
France
Germany
Italy
Japan
Netherlands
Norway
Sweden
Switzerland
UK
US
8.69
6.51
11.02
7.42
7.33
13.02
12.34
13.25
15.00
11.39
8.58
10.48
9.64
4.75
8.16
-1.73
2.67
8.58
7.77
-0.18
10.20
10.03
8.41
4.70
10.45
5.07
7.86
--
--
--
--
9.19
--
13.62
9.59
5.26
3.47
9.74
9.59
3.42
--
--
--
--
3.29
-8.87
5.89
-0.13
8.26
1.17
14.68
3.62
0.38
1.00
0.21
0.44
0.54
1.04
0.48
0.69
0.94
0.73
0.68
1.02
1.39
0.34
1.01
0.32
0.27
0.59
0.70
-0.08
0.58
0.02
0.74
0.57
0.85
1.17
17.13
32.33
37.02
30.63
39.79
25.73
6.98
18.84
23.32
56.10
--
--
22.59
21.36
34.89
27.95
58.25
41.44
37.97
57.02
24.69
145.92
40.87
--
--
39.52
14
9
18
14
14
8
12
15
14
16
11
11
9
9
7
5
9
9
10
11
8
9
7
12
12
14
Average 10.36 6.07 8.54 3.48 0.73 0.55 28.22 48.17
1 “Up (down) swing” refers to the years when real commercial property prices in that country increase (decrease).
Sources: OECD; BIS; authors’ calculations.
• We analyse a sample of 904 banks worldwide over the period 1989-2002.
• Seek to assess the effect of changes in commercial property prices on bank behaviour and performance in a range of industrialised economies, focusing on determination of lending, margins, ROA, bad debts and provisioning
• Consistent with macro-level studies, commercial property prices have a marked impact on the behaviour and performance of individual banks, over and above conventional determinants
• Results have implications for risk managers, regulators and monetary policy makers.
Table 2
Distribution of sample banks
By country Number of banks By specialisation Number of banks
Belgium
Canada
Finland
France
Germany
Hong Kong
Italy
Japan
Netherlands
Norway
Singapore
Sweden
Switzerland
United Kingdom
United States
19
21
4
58
40
13
38
143
8
14
5
5
28
54
454
Bank holding company
Commercial bank
Cooperative bank
Investment bank / securities house
Median and long term credit bank
Non-banking credit institution
Real estate / Mortgage bank
Savings bank
428
269
67
36
12
26
37
29
Total 904
Total 904
• Micro work – empirical analysis– Provisioning (Laeven and Majnoni)– Bank profitability and margins (Demirgüç-Kunt
and Huizinga)– Bad loan ratios (Salas and Saurina) – Lending (Bikker and Hu)– Rare studies looking at CPP and bank
performance• Austria (Arpa et al)• Japan (Gan)• Hong Kong (Gerlach et al)• US (Hancock and Wilcox)
Empirical work• Our advance on earlier literature
– First international study on how commercial property price movements affect individual banks’ lending strategies and performance after we control for the effects of conventional explanatory variables (macro factors, bank-specific variables and country-specific factors)
– Micro-level data allow us to examine whether the determination of bank performance and the role of commercial property prices vary across different groups of banks and across countries.
– Examine whether commercial real estate booms and busts tend to have asymmetric impacts on bank performance.
• Use of panel GLS or GMM (robustness check)
• Control variables– Macro: growth rate of real GDP, inflation and
short-term interest rates – Bank: loan-to-asset ratios, real loan growth
rate, capital strength, net interest margin, bank size dummies
– Country dummies– Growth of real commercial property prices
Issues of endogeneity
• Basic GLS equations ignore dynamic interaction of variables– No lagged dependent variable– Bank specific variables lagged– Nationwide CPP likely to be exogenous to lending
behaviour of individual bank– Previous results showed CPP largely autonomous of credit
even at macro level– Major loss of observations
• Robustness checks– Using lagged CPP– Using difference and levels GMM estimation
Table 3
Summary statistics of regression variables
Variables No. Obs Mean (%) Std. Dev. (%) Min (%) Max (%)
Asset growth rate
5244 8.13 10.90 -49.17 49.72
Loan growth rate
5132 8.54 12.03 -49.98 49.98
Loan to asset ratio
6025 61.07 15.22 11.27 89.86
Net Interest Margin (NIM)
5980 3.39 2.19 -5.88 36.72
Non-Performing Loan ratio
(NPL)
4353 2.44 3.91 0.00 45.79
Return on Assets (ROA)
6056 0.85 0.90 -7.65 8.79
Provisions / Total Assets
5844 0.40 0.65 -2.16 16.36
GDP growth rate
12656 2.44 2.11 -7.85 15.57
Inflation 12656 2.57 1.66 -4.04 10.97
Interest rate 12656 5.22 2.83 0.09 14.76
Growth rate of real commercial property prices
12651 -3.94 10.85 -49.19 35.49
Table 4
Characteristics of banks grouped by sizes1
Large banks Mid-sized banks Small banks Variables
Mean Std dev Mean Std dev Mean Std dev
Loan growth rate 5.91 10.36 5.45 11.90 9.12 12.11
Loan to asset ratio 54.79 14.49 62.33 14.93 61.52 15.19
NIM 1.82 0.86 2.13 1.45 3.67 2.23
NPL 4.58 4.06 4.34 6.23 2.15 3.58
ROA 0.37 0.58 0.44 0.81 0.94 0.91
1 There are 62 large banks, 76 mid-sized banks and 766 small banks.
Pooled regression with random effects
Dependent variables Loan growth rate
NIM NPL ROA Provisions/ Total Assets
Constant 8.8*** (6.1)
1.94*** (7.8)
1.4** (2.4)
0.42*** (3.2)
-0.21** (2.4)
Macro indicators
GDP growth 0.44*** (5.2)
0.05*** (10.6)
-0.046** (2.2)
0.026*** (3.7)
-0.013*** (2.8)
Inflation -0.18 (0.9)
0.007 (0.6)
-0.58*** (10.2)
0.14*** (8.6)
-0.048*** (4.4)
Interest rate 0.42*** (4.0)
0.07*** (11.0)
0.12*** (4.4)
-0.053*** (5.8)
0.007 (1.2)
Bank indicators
Loan/Asset (-1) -0.083*** (5.6)
0.01*** (6.3)
-0.0023 (0.4)
-0.0057*** (4.1)
0.0037*** (4.1)
Loan growth rate (-1) -0.0028*** (3.3)
-0.022*** (6.6)
0.0053*** (4.6)
-0.0043*** (5.6)
NIM (-1) 0.47*** (3.6)
0.14** (2.5)
0.27*** (23.4)
0.007* (8.7)
Capital ratio (-1) 0.084 (1.3)
0.053*** (8.5)
-0.114*** (5.2)
0.052*** (8.9)
0.0066* (1.6)
EBTDA/Total assets (-1) 0.06***
(5.6)
SMALL Insig 0.74*** (3.5)
1.0** (2.5)
-0.25*** (3.2)
-0.11** (2.3)
LARGE Insig Insig Insig Insig Insig
Commercial property sector
D(CPP) 0.16*** (9.4)
-0.0095*** (8.8)
-0.02*** (4.0)
0.0095*** (6.1)
-0.0049*** (4.8)
No. Obs. 5052 4195 3069 4182 4060
Pooled regression with random effects and leveraged size effects
Dependent variables Loan growth rate
NIM NPL ROA Provisions/ Total Assets
SMALL Insig 0.42* (1.9)
1.1** (2.3)
-0.23** (2.0)
-0.29** (2.3)
LARGE Insig Insig Insig Insig Insig
GDP*SMALL Insig Insig -0.24*** (4.1)
Insig 0.022* (1.6)
GDP*LARGE Insig Insig -0.16* (1.9)
Insig Insig
IR*SMALL Insig 0.08** (3.8)
Insig Insig 0.043** (2.3)
IR*LARGE Insig Insig Insig Insig Insig
INF*SMALL -0.94* (1.6)
Insig Insig Insig Insig
INF*LARGE Insig Insig Insig Insig Insig
D(CPP) 0.26*** (5.1)
-0.01*** (3.1)
-0.053*** (3.4)
0.019*** (4.0)
-0.0168*** (5.6)
D(CPP)*SMALL -0.11** (2.2)
Insig 0.04** (2.3)
-0.011** (2.2)
0.014*** (4.4)
D(CPP)*LARGE Insig 0.0082* (1.8)
Insig Insig Insig
No. Obs. 5052 4195 3069 4182 4060
Variants and robustness checks (1)
Dependent variables Loan growth rate
NIM NPL ROA Provisions/ Total Assets
Real residential prices
DRRP 0.22*** (6.1)
-0.0285*** (14.3)
-0.094*** (7.3)
0.019*** (5.6)
-0.014*** (6.5)
Real equity prices
DREP 0.065*** (7.0)
0.00184*** (3.7)
0.01*** (3.8)
0.0028*** (3.5)
-0.0002 (0.3)
Real residential and commercial prices
DRCP 0.149*** (8.0)
-0.004*** (3.8)
-0.002 (0.4)
0.007*** (3.9)
-0.0028** (2.5)
DRRP 0.05 (1.2)
-0.0245*** (11.4)
-0.09*** (6.1)
0.012*** (3.1)
-0.011*** (4.5)
Lagged commercial prices
DRCPP(-1) 0.055*** (5.0)
-0.0012*** (15.0)
-0.032*** (7.7)
0.006*** (4.6)
-0.0047*** (5.5)
Nominal commercial prices
DCPP 0.17*** (10.0)
-0.0092*** (9.6)
-0.02*** (3.9)
0.01*** (6.3)
-0.0055*** (5.3)
Pooled regression with difference specification and lagged dependent variables (GMM-difference estimation)
Dependent variables
Loan growth rate
NIM NPL ROA Provisions/ Total Assets
GMM difference
D.Lagged variable
0.053* (1.8)
0.77*** (6.7)
0.69*** (6.3)
0.35*** (4.5)
-0.014 (0.9)
D.D(CPP) 0.125*** (4.7)
0.0 (0.6)
-0.017*** (2.8)
0.0058*** (3.3)
-0.0037*** (2.7)
Observations 3305 3301 2250 3302 3225
Joint Wald
Sargan
AR(1)
AR(2)
113 [0.0]***
301 [0.47]
-7.7[0.0]***
-0.37 [0.72]
119 [0.0]***
393 [0.41]
-3.7 [0.0]***
0.02 [0.98]
87 [0.0]***
212 [1.0]
-2.3 [0.02]**
0.13 [0.9]
61 [0.0]***
362 [0.22]
-3.3 [0.001]***
-1.8 [0.08]*
54 [0.0]***
351 [0.12]
-1.8 [0.08]*
-1.7 [0.07]*
Pooled regression with lagged dependent variables (2 step GMM-levels estimation)
Dependent variables
Loan growth rate
NIM NPL ROA Provisions/ Total Assets
Lagged variable
0.39*** (8.4)
0.95*** (109.0)
0.813*** (53.4)
0.67*** (14.8)
0.473*** (4.6)
D(CPP) 0.074*** (3.1)
-0.0017* (1.7)
-0.0076* (1.9)
0.0033** (2.0)
-0.0024* (1.6)
No. Obs 4185 4180 2962 4182 4086
Joint Wald
Sargan
AR(1)
AR(2)
330 [0.0]***
427 [0.06]*
-3.07 [0.002]***
1.04 [0.3]
3900 [0.0]***
374 [0.62]
-0.72 [0.4]
-0.84 [0.4]
5577 [0.0]***
333 [1.0]
0.028 [0.98]
-0.83 [0.4]
756 [0.0]***
431 [0.4]
-0.88 [0.4]
-0.37 [0.71]
1323 [0.0]
474 [0.5]
-1.4 [0.15]
0.36 [0.72]
Conclusions
• Results indicate that commercial property prices have a major impact on a wide range of bank performance variables
• Signs found are consistent with a view that commercial property provides important forms of collateral perceived by banks to reduce risk and encourage lending
• Results hold consistently across a number of econometric specifications, as well as for regions.
• Interesting differences in response of small and large banks– Commercial property price movements having a smaller effect
on the loan quality and provisions of small than large banks– Small bank profits less geared to commercial property prices
than are those of large banks. Consistent with large banks being more willing to take risk as a consequence of the safety net and moral hazard.
• Generally, results underline crucial relevance of commercial property prices as macroprudential variable. Need for good data on prices
• Also highlight the need to develop indicators of individual bank exposure to the property market for stress testing (note – wider than CP lending per se given use as collateral)
References
• Davis E P and Zhu H (2004), "Bank lending and commercial property prices, some cross country evidence", BIS Working Paper No 150
• Davis E Philip and Haibin Zhu (2005), "Commercial property prices and bank performance", BIS Working Paper No 175 and Quarterly Review of Economics and Finance, 49, 1341-1359