Household Leverage and the Housing Wealth Effect Jonathan Chiu, - - PowerPoint PPT Presentation

.

Household Leverage and the Housing Wealth Effect

Jonathan Chiu, Robert Kirkby, Karam Shaar, Fang Yao

Victoria University of Wellington (Chiu, Kirkby, Shaar) Reserve Bank of New Zealand (Yao) December 11, 2017

.

1

Motivation

• Housing debt and housing wealth have become

important components of household balance sheets.

• Great Recession in the US: Falling house prices

led people to cut back on consumption, especially those who were heavily leveraged, leading to recession.

– Dynan (2012), Mian, Rao and Sufj (2013), Kaplan, Mitman and Violante (2016)

• New Zealand: House prices have risen rapidly in

recent years, along with household leverage.

• Rising leverage and house prices may have

implications for monetary and macroprudential policy, as well as the broader economy in New Zealand.

.

2

Household Leverage and the Housing Wealth Effect

Research question

Does household indebtedness (leverage) make consumption expenditure more or less sensitive to changes in housing wealth?

• Empirically, we use microdata from New

Zealand to estimate the consumption elasticity

• ut of exogenous changes in housing wealth,

controlling for the efgect of leverage.

• We build a life-cycle model to interpret the

empirical fjndings.

.

3

Findings

Research question

Does household indebtedness (leverage) make consumption expenditure more or less sensitive to changes in housing wealth? Conventional view:

– high leverage makes consumption more sensitive to housing wealth changes – interpretation: binding borrowing constraints

Our paper:

– high leverage makes consumption less sensitive to housing wealth changes – interpretation: the efgect can be asymmetric; factors other than borrowing constraint matter too, especially in a housing boom.

.

4

Findings

Research question

Does household indebtedness (leverage) make consumption expenditure more or less sensitive to changes in housing wealth?

• Conventional view:

– high leverage makes consumption more sensitive to housing wealth changes – interpretation: binding borrowing constraints

Our paper:

– high leverage makes consumption less sensitive to housing wealth changes – interpretation: the efgect can be asymmetric; factors other than borrowing constraint matter too, especially in a housing boom.

.

5

Findings

Research question

Does household indebtedness (leverage) make consumption expenditure more or less sensitive to changes in housing wealth?

• Conventional view:

– high leverage makes consumption more sensitive to housing wealth changes – interpretation: binding borrowing constraints

• Our paper:

– high leverage makes consumption less sensitive to housing wealth changes – interpretation: the efgect can be asymmetric; factors other than borrowing constraint matter too, especially in a housing boom.

.

6

Outline

• Empirical results
• Intuitions from a two-period model.
• Numerical results from a richer model.

.

7

Empirical Analysis

• Microeconometric study using Household

Economic Survey (HES) data from Stats NZ.

• Repeated cross-sectional data for: 2006-07,

2009-10, 2012-13, 2015-16.

• HES reports consumption spending, income, and

debt for 16,000 households.

• House price data: periodic rateable value at

household level (HES) and house price micro data at local level.

• Also many demographic variables; age, gender,

education, composition of household, etc.

.

8

Regression I

• Baseline regression equation:

log(CEi) = β0 + β1log(Yi) + β2log(HPi,au) + β3Zi + ui

Total exp. Total exp. Non-durable Durable Housing wealth 0.08∗∗∗ (0.03) 0.17∗∗∗ (0.02) 0.15∗∗∗ (0.02) 0.22∗∗∗ (0.05) Income 0.40∗∗∗ (0.02) 0.40∗∗∗ (0.02) 0.36∗∗∗ (0.02) 0.60∗∗∗ (0.08) Regional dummy Yes Yes Yes Yes Time dummy Yes Yes Yes Yes Household dum- mies Yes Yes Yes Yes Observations 5134 4644 4644 3792 Adj-R2 0.56 0.57 0.58 0.14

. . .

9

Leverage Effect: First Impression

Figure : (Clustered) Data Scatter Plot

11 12 13 14 15 expenditure per person 8 9 10 11 12 house value Low Leverage LTV Low High Leverage LTV High

.

10

Regression II

• Regressions with leverage:

log(Ci) = β0 + β1 log(HPau) + β2LTVi + β3LTVi ∗ log(HPau) + β4 log(Yi) + β5Zi + ui

Dependent variable Log non-housing expenditure i ii iii iv Housing 0.14∗∗∗ (0.04) 0.19∗∗∗ (0.05) 0.18∗∗∗ (0.05) 0.14∗∗∗ (0.04) LTV −0.19∗∗∗ (0.05) 2.02 (1.23) 1.62 (1.2)

• LTV*Housing
• −0.17∗

(0.10) −0.16∗ (0.09) −0.04∗∗∗ (0.01) Age*LTV*Housing

• 0.0005∗

(0.00) 0.0006∗∗ (0.00) Controls Yes Yes Yes Yes Obs 1853 1853 1849 1849

• Adj. -R2

0.552 0.554 0.556 0.555

.

11

Summary and Next Step

• We fjnd that consumption of highly leverage

households are less sensitive to house prices.

• Our empirical result is surprising in the light of

the fjndings of Mian, Rao, and Sufj (2013).

• One possible explanation: the interaction

between leverage and housing wealth efgect is asymmetric.

• We build a life-cycle model to understand the

empirical fjndings.

. . .

12

A Two-Period Model

Consider a household who faces the following problem: max

C1,C2,A1 u(C1) + βu(C2),

where: u(C) = C1−ρ 1 − ρ subject to A0 = C1 + A1 A1 ≥ −µVph

1H

ph

2H + A1R

= C2

.

13

Analytical Solution

• Case 1: A1 = −µVph

1H

¯ C1 = A0 + µVph

1H.

• Case 2: A1 > −µVph

1H1

C∗

1 = A0 − ΨA0 − ph 2H

Ψ + R . where Ψ = [βR]

1 ρ .

.

14

Analytical Result

Proposition 1: In this two-period model, the difgerence between MPCs of households with high versus low leverage is given by the following equation: ∆MPC|LTV = ∂¯ C1 ∂ph

1

− ∂C∗

1

∂ph

1

= µVH − dph

2

dph

1

H Ψ + R

• µVH captures the borrowing constraint efgect. A

higher µV makes MPC of constraint households larger.

• dph

2

dph

1

H Ψ+R comes from the optimal

consumption/saving decision of unconstraint households

. . .

15

The Role of House Price Dynamics

∆MPC|LTV = µVH − dph

2

dph

1

H (βR)

1 ρ + R

• When dph

2

dph

1 < 0, ∆MPC|LTV > 0

• When dp

dp

h h 2 1 > 0 and large enough, ∆MPC| LTV < 0

.

16

The Role of Risk Aversion (ρ)

∂∆MPC|LTV ∂ρ = −dph

2

dph

1

ln [βR] HΨ (Ψ + R)2 1 ρ2

• When dph

2/dph 1 > 0 and R > 1 β, then a higher ρ

implies a smaller ∆MPC|LTV.

• When future house price and returns to saving

are both high, an unconstrained agent has an incentive to increase current consumption relative to future consumption. This consumption-smoothing incentive is stronger when agents are more risk averse.

.

17

A Richer Model

• A life-cycle model in which heterogeneous

agents make decision to save/borrow using a combination of housing, debt, and fjnancial assets.

• Housing is desirable as it generates a fmow of

housing-consumption utility and it serves as collateral for borrowing.

• The model is partial equilibrium in terms of

housing and fjnancial markets.

• We solve the model numerically, both for the

stationary equilibrium and transition path given a process of house prices.

.

18

Numerical Results I

• We begin by running the same regression

analysis performed in the empirical analysis on simulated panel data from the model. Dependent variable Consumption i ii iii Housing Wealth 0.07 0.07 0.07 Leverage

• 0.84
• 0.58
• 0.55

Leverage*Housing Wealth

• 0.02
• 0.06

Age*Leverage*Housing Wealth

• 0.00

.

19

Numerical Results II

Figure : Interaction between leverage and housing wealth efgect

A: Efgect of Changing House Prices B: Efgect of Risk Aversion

.

20

Conclusions

• This paper investigates the interaction between

housing leverage and the housing wealth efgect.

• We fjnd a difgerent pattern of the interaction in

the boom phase of the housing cycle.

• We use a theoretical model to interpret the

empirical fjnding.

• We show that interaction between leverage and

the housing wealth efgect can be asymmetric. Besides borrowing constraint, the optimal behavior of unconstrained households also matters.

.

21

Life Cycle of Leverage

Figure : Share of High LTV in Age

.2 .4 .6 share of high LTV households 20 40 60 80 age