National Instituteof Economic and Social Research

International developments in housing markets – lessons for Sweden

E Philip Davis

Introduction

• In this presentation we seek to give an overview of recent developments in housing markets for 12 OECD countries, via data and relevant research

• We begin by noting key structural differences, before looking at developments in the crisis

• We proceed to put these in a longer term context• And finally look at key implications of house prices

for investment, consumption, public finance and financial stability/financial regulation

Background – structural features

• Housing markets cannot be treated as homogeneous

• Population density is correlated with dwelling size and availability of land, although the latter is also affected by planning restrictions

• Dwelling size and inhabitants per dwelling is indicators of living standards in terms of housing

• Interest rate risk may affect both supply and demand for housing, and demand may also be affected by the prevalence of fixed rate loans

Structure of housing markets

Population density (2005)

Housing density (2001)

Dwelling size (2001)

(persons per sq km)

(per 1000 inhabitants)

average m2 per capita 1980s 1990s 2000-2006

Australia 2.6 405 81.0 4.33 0.88 0.33Canada 3.2 403 69.7 0.64 0.60 0.07France 108.4 490 43.9 0.63 0.12 0.02Germany 231.0 469 42.1 0.88 0.30 0.07Italy 194.5 368(b) 35.0(b) 1.99 1.09 0.02Netherlands 391.5 417 41.2 1.35 0.16 0.02Spain 85.8 510 47.6 1.71 0.68 0.08United Kingdom 246.3 431 36.4 1.08 0.55 0.09United States 30.8 428 70.8 0.91 1.12 0.11

long real interest rate volatility(a)

Personal sector borrowing cost vulnerability

Net

hs

Ger

man

y

Den

mar

k

Fra

nce U

S Sw

eden UK

Ital

y

Irel

and

Fin

land

Spa

in

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

1

Flo

atin

g ra

te d

ebts

as

a pr

opor

tion

of d

ispo

sabl

e in

com

es

Recent house price developments

• Boom-bust cycle in housing in a number of countries…

• US housing credit linked to global crisis directly via falling CDO prices (note principal agent problem in US mortgage securitisation)

• Since crisis, house price falls less marked than widely expected, recovery in some countries

• Evidence of further financial distress with high level of arrears and repossessions in countries such as the US

• Less so in those such as the UK as interest rates low and house price falls modest – and loans recourse based

Recent house price developmentsHouse prices since 2001

-20

-15

-10

-5

0

5

10

15

20

25

Quarters

Ann

ual p

erce

nt c

hang

e AUphi

CNphi

FRphi

GEphi

IRphi

ITphi

Recent house price developmentsHouse prices since 2001

-15

-10

-5

0

5

10

15

20

25

30

Quarters

An

nu

al p

erc

en

t ch

an

ge

JPphi

NLphi

SDphi

SPphi

UKphi

USphi

Latest “Economist” Data

Country Year on year change

Country Year on year change

Australia 18.4 Japan -3.4 Canada 4.5 Netherlands 4.2 France 6.0 Sweden 8.9

Germany 4.8 Spain -4.0 Ireland -17.0 UK 3.0

Italy -2.8 US -4.9 (FHFA) 3.6 (CSNI)

A longer term perspective• Boom in house prices following liberalisation in

the 1980s, often leading to banking crises…• Long term rise in real house prices (higher

income, shortage of land) – implicit intergenerational transfers (Weale 2007)

• Rise in debt-income ratios to households correlated with rise in house prices (house purchase but also equity extraction)

• But credit should not drive house prices in liberalised financial system (conventional determinants are income, interest rates, supply conditions, demographics, see e.g. Muellbauer and Murphy 1997))

House price inflation since 1980House price inflation

-20

-10

0

10

20

30

40

50

Years

An

nu

al p

erc

en

tag

e c

han

ge

AUhpinfl

CNhpinfl

FRhpinfl

GEhpinfl

IRhpinfl

IThpinfl

House price inflation since 1980House price inflation

-20

-10

0

10

20

30

40

50

Years

An

nu

al p

erc

en

t ch

an

ge

JPhpinfl

NLhpinfl

SDhpinfl

SPhpinfl

UKhpinfl

UShpinfl

Real house pricesReal house prices (1980=100)

0

50

100

150

200

250

300

350

Years

Ind

ex

AUrph

CNrph

FRrph

GErph

IRrph

ITrph

Real house pricesReal house prices (1980=100)

0

50

100

150

200

250

300

350

400

Years

Ind

ex

JPrph

NLrph

SDrph

SPrph

UKrph

USrph

Debt-income ratiosHousehold debt/income ratios

0

50

100

150

200

250

Years

Perc

en

t

AUdyr

CNdyr

FRdyr

GEdyr

IRdyr

ITdyr

Debt-income ratiosHousehold debt-income ratios

0

50

100

150

200

250

300

Years

Perc

en

t

JPdyr

NLdyr

SDdyr

SPdyr

UKdyr

USdyr

Housing and investment

• Housing investment typically a small proportion of the stock, given houses are long-lived assets

• Overall housing investment has tended to decline as a proportion of GDP in a number of countries, even before 2008/9 when sharp falls especially Spain and Ireland

• Key determinant Q ratio (house prices/housing investment deflator) (Jud and Winkler (2003), Berg and Berger (2005))

• Correlation of house price change to investment/GDP change in 2008/9 is 0.73

Housing investment/GDPHousing investment/GDP ratios

0

2

4

6

8

10

12

14

16

Years

Perc

en

t

AUiy

CNiy

FRiy

GEiy

IRiy

ITiy

Housing investment/GDPHousing investment/GDP ratios

0

2

4

6

8

10

12

Years

Perc

en

t

JPiy

NLiy

SDiy

SPiy

UKiy

USiy

Q ratios for housing in upturn

-6

-4

-2

0

2

4

6

8

10

Aus

tral

ia

Can

ada

Fra

nce

Ger

man

y

Net

herl

ands

Spa

in

UK

US

ave.

an

nu

al p

erce

nta

ge

chan

ge

1991-1995 1996-2000 2001-2006

House prices and consumption• Research on wealth effect shows strong link

from house prices/housing wealth to consumption (e.g. Barrell and Davis (2007), Case et al (2005)), although effect on non-homeowners should partly offset

• Simple cross section regression shows house prices discriminated the falls in consumption between 2008/3 and 2009/4 better than equity prices, although RPDI and lagged debt/income also relevant

• Collateral effect likely intensified by credit rationing, but banking crisis dummy not significant

Change in consumption 2009/4 over 2008/3

• Dependent Variable: DC• Method: Least Squares• Date: 10/12/10 Time: 13:37• Sample: 1 12• Included observations: 11••• Variable Coefficient Std. Error t-Statistic Prob.  ••• DPH 0.258924 0.092604 2.796035 0.0267• DRPDI 0.490681 0.224692 2.183794 0.0653• LDY -0.011408 0.005645 -2.020971 0.0830• DEQP -0.026005 0.058024 -0.448169 0.6676••• R-squared 0.752215     Mean dependent var -0.942513• Adjusted R-squared 0.646022     S.D. dependent var 2.813637• S.E. of regression 1.674003     Akaike info criterion 4.143600• Sum squared resid 19.61600     Schwarz criterion 4.288289• Log likelihood -18.78980     Durbin-Watson stat 2.391862•••

•

Housing and fiscal position

• NIER (2010) decomposed deterioration in fiscal position into cycle, policy changes, bank support and residual

• Residual linked partly to revenue from financial sector but also housing market

• Falling house prices affect tax revenues directly (via stamp duties and profits of construction/realtor sector) and indirectly (via consumption taxes – to the extent consumption fell more than in a normal cycle)

House prices and deficits

Ireland

Italy

Canada

Spain

United Kingdom

Greece

Denmark

Finland Portugal

Belgium

Netherlands

Japan

France

Sweden Germany

US

-14

-12

-10

-8

-6

-4

-2

0

2

4

6

-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1

Residual category of the deficit (% of GDP)

Ho

use

pri

ce g

row

th 2

009

House prices and banking crises

• Barrell, Davis, Karim and Liadze (2010) in JBF and subsequent work were first to find role for bank capital and liquidity in OECD crisis models

• House price bubbles matter• Sustained deficits matter

• Using logit model together with a banking sector sub-model of NiGEM global macro model enabled assessment of overall costs and benefits of regulation in the UK – optimal level of tightening (Barrell et al (2009) FSA OP)

• Recent work looks at the split between on balance sheet and other revenues (OBS)

• Level of OBS does not matter as it varies across countries a lot• Faster growth of OBS activity boosts crisis probabilities

Calibrating macroprudential surveillance

• In “Calibrating macroprudential surveillance” we put in all ‘normal’ variables including lagged house price rises and test down with 14 OECD countries, 12 crisis and data for 1980 to 1997 (vastly shorter sample than earlier work)

• As in earlier work, found that “traditional” variables such as credit growth, output growth and M2/reserves less relevant to OECD – artefact of dominance of global samples by emerging markets

• A researcher undertaking this work in the late 1990s could have picked the same equation

Explaining OECD Financial Crises

• We explain crisis probabilities (logit) in OECD 1980-1997

Box 1: List of Variables (with variable key)

1. Real GDP Growth (%) (YG) 2. Real Interest Rate (%) (RIR) 3. Inflation (%) (INFL) 4. Fiscal Surplus/ GDP (%) (BB) 5. M2/ Foreign Exchange Reserves (%) (M2RES)

Variables used in previous studies:

Demirguc-Kunt and Detragiache (2005);

Davis and Karim (2008). 6. Real Domestic Credit Growth (%) (DCG) 7. Liquidity (%) (LIQ) 8. Leverage (%) (LEV)

Variables introduced in JoBF.

9. Real Property Price Growth (%) (RHPG) This paper 10 Current Balance as % GDP (CBR)

it

it

X'

X'

itite1

eXF1YobPr

Nested testing of the crisis model, 1980-1997

Step (1) (2) (3) (4) (5) (6) (7)

LEV(-1) -0.339 (1.7)

-0.339 (1.8)

-0.348 (1.9)

-0.347 (1.9)

-0.417 (2.9)

-0.345 (2.7)

-0.384 (3.2)

NLIQ(-1) -0.106 (1.8)

-0.106 (1.9)

-0.108 (2.0)

-0.113 (2.2)

-0.126 (2.7)

-0.104 (2.5)

-0.105 (2.6)

RHPG(-3) 0.091 (1.9)

0.091 (1.9)

0.089 (1.9)

0.095 (2.4)

0.09 (2.4)

0.086 (2.3)

0.081 (2.1)

CBR(-2) -0.434 (2.3)

-0.434 (2.3)

-0.441 (2.4)

-0.438 (2.4)

-0.418 (2.3)

-0.3 (1.9)

-0.333 (2.2)

DCG(-1) -0.101 (1.5)

-0.101 (1.6)

-0.1 (1.6)

-0.1 (1.5)

-0.108 (1.7)

-0.053 (1.0)

YG(-1)) 0.277 (1.5)

0.277 (1.5)

0.274 (1.4)

0.279 (1.5)

0.29 (1.5)

RIR(-1) -0.054 (0.3)

-0.055 (0.6)

-0.055 (0.6)

-0.06 (0.7)

BB(-1) 0.022 (0.2)

0.02 (0.2)

0.023 (0.2)

M2RES(-1) -1.51E-05

(0.2) -1.52E-05

(0.2)

INFL(-1) -0.0012

(0.1)

Model character• Up to four lags tried in house prices, credit

growth, current account and GDP growth– Cyclical variables drop out– Lending growth drops out

• Lending quality matters with house price growth and current balances as indicators

9 out of 12 crises calledAlmost half of false calls precede crises

Estimated Equation

Dep=0 Dep=1 Total P(Dep=1)<

0.057 143 3 146 P(Dep=1)>

0.057 55 9 64

Total 198 12 210

Correct 143 9 152

% Correct 72 75 72

% Incorrect 28 25 28

Using the model in macroprudential surveillance setting• Forecasts over 1998-2008, using actual

for RHS (bold exceeds sample mean)

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008BG 0.005 0.004 0.003 0.004 0.009 0.005 0.007 0.014 0.025 0.048 0.070CN 0.032 0.054 0.056 0.033 0.018 0.022 0.026 0.037 0.030 0.036 0.042DK 0.015 0.041 0.060 0.046 0.048 0.029 0.043 0.030 0.042 0.030 0.113FN 0.004 0.006 0.011 0.007 0.000 0.000 0.000 0.004 0.002 0.007 0.008FR 0.025 0.018 0.012 0.014 0.040 0.028 0.032 0.053 0.100 0.193 0.218GE 0.026 0.027 0.029 0.045 0.058 0.031 0.016 0.020 0.007 0.007 0.007IT 0.001 0.002 0.002 0.009 0.017 0.020 0.026 0.039 0.034 0.054 0.019JP 0.071 0.025 0.009 0.010 0.007 0.007 0.003 0.002 0.001 0.001 0.002NL 0.020 0.018 0.050 0.049 0.157 0.141 0.079 0.028 0.017 0.019 0.007NW 0.011 0.006 0.039 0.016 0.001 0.001 0.006 0.003 0.002 0.001 0.001SD 0.019 0.016 0.034 0.048 0.039 0.058 0.017 0.006 0.009 0.011 0.008SP 0.005 0.006 0.010 0.028 0.043 0.044 0.047 0.096 0.266 0.516 0.580UK 0.049 0.060 0.088 0.173 0.203 0.201 0.115 0.207 0.282 0.277 0.254US 0.025 0.032 0.044 0.074 0.081 0.067 0.103 0.064 0.075 0.097 0.125

Using the model in macroprudential policy setting • We can invert the probability model to calculate

the additional levels of liquidity and leverage required for the probability of a crisis to be 0.01 in each country and year– Re-estimate each year from 1997, predict one year– Raise capital and liquidity to get probability 0.01

• Capital and liquidity form the defences, while house prices and current balances are the problems we need to provision against, not cycles or credit.

• Separate result shows credit does not Granger cause OECD house prices either (except Belgium, Canada and Finland)

Country and aggregate targets• Country max

reduces probability to 0.01 in worst year

• The average of these could be used as a criterion

• Major cross country differences in warranted tightening

Column 1 2 3 4

Under or overshoot

Additions to country specific levels of

liquidity and leverage to reduce all prob. to 0.01 or

below*

(column 1 - 3.7)

(column 2 -4.59)

Top Panel lev+nliq lev alone lev and nliq lev

Belgium 2.11 2.56 -1.59 -2.03 Canada 3.31 4.15 -0.39 -0.44 Denmark 3.35 4.15 -0.35 -0.44 Finland 0.00 0.00 -3.70 -4.59 France 5.08 6.25 1.38 1.66 Germany 3.12 3.79 -0.58 -0.80 Italy 1.74 2.14 -1.96 -2.45 Japan 3.96 5.19 0.26 0.60 Neths 4.72 5.80 1.02 1.21 Norway 2.34 2.87 -1.36 -1.72 Sweden 2.38 2.90 -1.32 -1.69 Spain 9.32 11.48 5.62 6.89 UK 6.08 7.63 2.38 3.04 US 4.35 5.34 0.65 0.75

Mean (International Benchmark) 3.70 4.59

SD 2.24 2.77

Countercyclical provisioning

• Has to be calibrated on house prices and current account and not credit or output gap – example of 5% higher house price growth

1998 0.0 -3.2 0.0 -0.1 0.6 -2.5 0.0 0.8

1999 0.0 1.3 0.0 -1.7 1.0 0.2 0.0 1.52000 0.0 0.8 0.0 0.0 1.9 6.2 0.4 1.82001 0.0 1.7 0.0 3.8 3.5 8.9 1.5 4.22002 0.1 7.5 0.3 5.3 3.8 9.5 1.7 3.12003 0.0 6.2 0.4 4.7 3.8 13.7 1.2 4.0

2004 0.0 6.1 0.5 6.1 2.5 6.1 2.2 5.42005 0.7 7.2 2.1 12.5 3.9 14.3 1.1 4.92006 2.2 10.0 4.6 14.1 4.8 13.9 1.5 4.32007 3.7 13.4 6.8 13.4 4.7 10.1 2.1 6.72008 4.0 13.6 7.3 10.2 4.4 3.0 2.7 8.4

Regulatory adjus tment

Actual RHPG (-3)

Regulatory adjustment

France

Actual RHPG (-3)

Ac tual RHPG (-3)

Regulatory adjustment

Actual RHPG (-3)

Regulatory adjustment

USUKSpain

Decomposing changes in crisis probabilities

France

2004 0.006 na na na na na na na2005 0.046 0.007 0.005 0.004 0.006 0.036 0.017 0.0402006 0.131 0.011 0.030 0.023 0.006 0.044 0.030 0.0852007 0.106 0.016 -0.008 0.024 0.027 -0.126 -0.043 -0.025 2008 0.895 -0.000 0.013 0.001 0.001 0.775 -0.000 0.789

Sum of changes na 0.034 0.040 0.051 0.039 0.729 0.004 0.889

Probability NLIQ CBR DOFFTOONChange in probability

Contribution to change in probabilityAdj for

InteractionLEV RHPG

Spain

2004 0.035 na na na na na na na2005 0.067 0.022 -0.012 0.026 0.004 -0.009 -0.001 0.0322006 0.173 0.031 0.040 0.017 0.055 -0.007 0.030 0.1062007 0.398 0.065 0.044 -0.013 0.120 0.045 0.037 0.2252008 0.479 -0.021 0.016 -0.063 0.098 0.050 -0.001 0.081

Sum of changes na 0.097 0.089 -0.033 0.277 0.079 0.064 0.444

Change in probability

Contribution to change in probability

Probability NLIQ LEV RHPG CBR DOFFTOONAdj for

Interaction

Decomposing changes in crisis probabilities

UK

2004 0.116 na na na na na na na2005 0.241 0.002 0.010 0.099 -0.007 0.035 0.015 0.1252006 0.442 0.009 0.057 -0.008 0.031 0.127 0.016 0.2012007 0.292 -0.002 0.024 -0.066 0.026 -0.135 -0.003 -0.151 2008 0.253 0.010 0.026 -0.119 0.036 -0.007 -0.015 -0.038

Sum of changes na 0.020 0.118 -0.094 0.087 0.020 0.013 0.137

RHPG CBR

Contribution to change in probability

Probability NLIQ LEVAdj for

InteractionChange in probability

DOFFTOON

US

2004 0.074 na na na na na na na2005 0.045 -0.017 -0.015 -0.002 0.004 0.003 0.001 -0.029 2006 0.043 0.002 -0.000 -0.002 0.006 -0.009 -0.001 -0.002 2007 0.064 0.002 -0.008 0.011 0.008 0.010 0.002 0.0212008 0.087 0.004 0.004 0.010 0.002 0.004 0.002 0.023

Sum of changes na -0.008 -0.019 0.017 0.019 0.008 0.004 0.013

Probability NLIQ LEV RHPG CBRAdj for

InteractionChange in probability

Contribution to change in probability

DOFFTOON

Conclusion

• Caution needed in directly comparing housing markets due to structural differences

• Housing finance clearly at core of recent financial crisis – US housing loans packaged into CDOs and recent defaults following price falls

• Falls in house prices can be related inter alia to changes in consumption, housing investment and fiscal deficits since the crisis

• And clear relation of lagged house price increases to OECD banking crises – relevant to ongoing bank regulation reform also

• Key lesson for Sweden is to avoid boom-bust cycle in housing, given macroeconomic volatility and systemic financial risk it generates – and long term intergenerational implications

• Control could be via appropriate monetary and macroprudential policies (including control of LTVs) – possibly also planning regulations

• If using securitisation ensure system is transparent and incentive compatible

• Ensure banks have sufficient capital as well as countercyclical reserves based on trends in house prices (not credit per se)