1

HONG KONG MONETARY AUTHORITY

Bank Lending and Property Prices in Hong Kong

Motivation

• Hong Kong has experienced a number of property price “cycles” since 1980.– Severe and frequent!– Have been associated with movements in bank

lending.

• International experience:– Credit expansion and property price booms. – Banking sector fragility.

2

0

20

40

60

80

100

120

0

20

40

60

80

100

120

82 84 86 88 90 92 94 96 98 00

Property Prices Loans

Residential property prices and loans(1997:3=100)

3

-.6

-.4

-.2

.0

.2

.4

-.6

-.4

-.2

.0

.2

.4

82 84 86 88 90 92 94 96 98 00

Property prices Loans

Growth rate of residential property prices and loans in real terms(Over 4 quarters)

4

• Several issues arise:

– What was the role of bank lending in residential property price cycles?

• Did bank lending “trigger” the cycles?• Did banks merely expand lending in response to a

growing demand for loans?

– Understanding the correlation may be important for policy.

5

– Given the currency board, monetary policy can not be used to guard against asset price cycles.

• Little evidence that interest rates drive property prices.

– Focus shifts to regulatory policy:• “Loan-to-value” ratio of ≤ 70% in 1991.• January 1997 ≤ 60% for luxury properties (withdrawn).• Limit on share of property lending to 40% 1994 (withdrawn).• The 70% ratio can be exceeded if the excess is covered by

mortgage insurance (and, recently, to facilitate refinancing of loans with negative equity).

– How have these policies impacted on bank lending?6

2

Empirical Work

• Focus on three variables:

– Bank lending.• Total domestic loans.• Alternative measure: mortgage loans.

– Property prices.• Residential.• Alternative measure: commercial.

– Real GDP.

7

-.1

.0

.1

.2

.3

-.1

.0

.1

.2

.3

82 84 86 88 90 92 94 96 98 00

Domestic loansMortgage loans

Growth rate of domestic and mortage loans in real terms(Over 4 quarters)

8

9

cial property prices in real terms(Over 4 quarters)

-.8

-.6

-.4

-.2

.0

.2

.4

.6

.8

-.8

-.6

-.4

-.2

.0

.2

.4

.6

.8

82 84 86 88 90 92 94 96 98 00

ResidentialCommercial

Growth rate of res idential and commer

• Two step analysis:

– Cointegration (common trend) analysis.• Unit root tests.

– Dynamic (short-run) analysis.

• Cointegration analysis:

– Real bank lending.

– Real property prices.

– Real GDP.10

Appendix — Table 2

Trace Tests for Cointegration

( 1982:1 - 1998:4 )

Null hypothesis ofr = 0 r = 1 r = 2

Trace test statistics 29.80 12.90 4.10p-value 0.05 0.12 0.04

11

Appendix — Table 3

Cointegration Tests: β and α Vectors

( 1982:1 - 1998:4 )

β α

Real bank lending 1.00 -0.09(0.035)

Real GDP -0.90 -0.02(0.034)

Real property price -0.43 0.14(0.093)

Note: Numbers in parentheses are standard errors for α.12

3

Table 1. Long-run Relationship

( 1982:1 - 2001:4 )

CI vector Loading coefficientβ α

Real bank lending 1.00 -0.13(0.03)

Real GDP -1.00 0.00Real property price -0.36 0.00

Note: Number in parentheses is the standard error for α .

13

Cointegration

• Cointegration results:

– One CI vector:• Lending - RGDP - 0.36*Property Price.

– Weak exogeneity:• Loans adjust.

14

• Dynamic (short-run) analysis:

– Models for quarterly changes in:• Real bank lending.• Real property prices.

– Strongly contemporaneously correlated.• ρ = 0.43 in the data.• ρ = 0.41 in the VAR system.

– What explains the correlation?• “Lending driving property prices”.• “Property prices driving lending”.• Simultaneity! 15

• Strategy:

– Use general-to-specific modelling to obtain models for:• ∆l = g(∆p, CI, … ).• ∆p = f(∆l, CI, … ).• Expect CI to be insignificant in the second relationship.

– Obvious simultaneity bias:• Use Hausman tests to see whether one or both regressions are

subject to simultaneity.

16

• Results as expected:

– ∆p and CI are both significant.

∆l = + 0.244*∆l_2 - 0.313 + 0.239*∆y + 0.176*∆p - 0.078*CI_1 + 0.357*(∆r_1 - ∆r_2)

(SE) (0.083) (0.107) (0.101) (0.034) (0.026) (0.142)

R2 = 0.57; Sample period: 1984:1 - 2001:4

17

• Results as expected:

– ∆l significant.

– CI insignificant.

∆p = + 0.283*∆p_1 - 0.008 + 0.900*∆l - 0.034*∆umr - 0.041*∆unit_2

(SE) (0.097) (0.007) (0.248) (0.013) (0.024)

R2 = 0.49; Sample period: 1984:1 - 2001:4

18

4

• Simultaneity bias:

– Use predetermined variables as instruments.

– ∆l-equation: • p-value = 0.55.• Property prices enter equation as structural variable.

– ∆p-equation:• p-value = 0.01.• Lending growth subject to simultaneity.

– “reverse causation”.

– Estimate with IV:• Lending growth not structural determinant of property prices.

19

• Property prices appear to be driven by:

– State of the economy.

– Supply of new housing.

∆p = + 0.397*∆p_1 + 0.006 - 0.045*∆umr - 0.047*∆unit_2

(SE) (0.100) (0.006) (0.014) (0.026)

R2 = 0.39; Sample period: 1984:1 2001:4

20

• Further analysis of lending equation:

– Restriction on loan-valuation ratio introduced in 1991.

– Did this constrain the growth of lending?• Did the parameter on ∆p decline around 1991?

– Note, Hansen test does not point to instability.• Assumes unknown breakpoint (lacks power).

– Recursive estimates.

211990 1995 2000

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7dp × +/-2SE

22

• Suggests break.

• Test:

– Assumes knowledge of break date.

– Add dummy*∆p.

– Dummy = 0 before 1991:2.

– β = -.27 (se = 0.09).

– β declined from 0.40 to 0.13.

23

1990 1995 2000

0.2

0.4

0.6dp × +/-2SE

1990 1995 2000

-0.75

-0.50

-0.25

0.00

dumdp2 × +/-2SE

2424

5

• The restriction on banks’ ability to lend is likely to have given them some price making power.

• Some evidence that spread between BLR and interbank rates rose around 1990.– Much less volatile.

• Spread significant and negative if included.– Signs of instability.

25

1980 1985 1990 1995 2000

-0.005

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0.035

0.040 spread

26

∆l = + 0.211*∆l_2 - 0.269 + 0.214*∆y + 0.196*∆p + 0.349*(∆r_1 - ∆r_2)

(SE) (0.081) (0.105) (0.099) (0.034) (0.137)

- 0.070*CI_1 - 0.419*spread

(0.026) (0.178)

R2 = 0.61; Sample period: 1984:1 - 2001:4

27

Conclusion

• Strong co-movements between lending and property prices.

• Appear to reflect reactions of bank lending to demand for credit rather than impact of lending on property market.

• Evidence that the introduction of loan-to-valuation ratio in 1991 reduced the impact of property prices increases on bank lending.

28