HOUSE PRICES AND CONSUMPTION INEQUALITY

• DOI: http://doi.org/10.1111/iere.12404
• Date: 01 November 2019
• Published date: 01 November 2019
• Author: Ben Etheridge

INTERNATIONAL ECONOMIC REVIEW

Vol. 60, No. 4, November 2019 DOI: 10.1111/iere.12404

HOUSE PRICES AND CONSUMPTION INEQUALITY∗

BYBEN ETHERIDGE1

University of Essex, U.K.

I characterize how house price shocks affect consumption inequality using a life cycle model of housing and

nonhousingconsumption with incomplete markets. I derive analytical expressions for the dynamics of inequalities

and use these to analyze large house prices swings seen in the United Kingdom. I show that movements in

consumption inequality were large, that they correspond with the theoretical predictions qualitatively, and that

the model explains a large fraction of the movements quantitatively. I demonstrate the accuracy of this analysis

using an extended model’s full nonlinear solution. Finally, accounting for house price shocks alters estimates of

labor–income risks using cross-sectional data.

1. INTRODUCTION

Across much of the world, house prices have fluctuated substantially in the past 30 years,

seemingly in connection with the wider macroeconomy. Accordingly, the effect of house price

shocks on aggregate consumption has been extensively researched.2Yet, given that housing

wealth and home ownership are unevenly distributed, it seems important to ask not only how

house price shocks affect mean consumption, but also how they affect consumption inequality.

I contribute to the literature on house prices and consumption by addressing this question.

Of course, an extensive literature has examined how house prices affect consumption across

different groups. These groups include age and housing tenure (for example, Li and Yao, 2007)

and, perhaps most saliently, region (for example, Mian et al., 2013). Nevertheless, in this article,

I find that the effect of house prices shocks on inequality is both starker and richer. I show that

even a common house price shock affects inequality within groups defined by age, education,

and region. This within-group measure of inequality is important because it explains the bulk of

total inequality.3Specifically, I find that a positive common house price shock generally causes

such consumption inequality to increase.

This article therefore builds on two literatures: first, on that assessing the effect of house

prices on consumption and, second, on that explaining the evolution of consumption inequality.

This second literature has typically focused on the relationship between consumption inequality

∗Manuscript received February 2016; revised December 2018.

1I thank Richard Blundell, Tom Crossley, and Hamish Low for invaluable advice and encouragement. Sincere thanks

also to the editor, Dirk Krueger, and to three anonymous referees, whose comments helped substantially improve the

article. Further thanks to Michael Graber and to Cormac O’Dea for comments. I also thank numerous individuals

and seminar audiences for further comments. Financial assistance was provided by the ESRC through the Institute for

Fiscal Studies. All errors are my own. Please address correspondence to: Ben Etheridge, Department of Economics,

University of Essex, Wivenhoe Park, Colchester, UK, CO4 3SQ. Phone: 01206 872889. E-mail: bsethe@essex.ac.uk.

2For an early literature using data similar to that used here, see Campbell and Cocco (2007) and Attanasio et al.

(2009) for the United Kingdom and Carroll et al. (2011) for the United States. See also the literature discussed below

in the dedicated literature review.

3On within-group inequality in labor earnings, the literature is vast. It has focussed in particular on the inequality

“boom” in the 1980s. Perhaps most relevant in the present context is Moffitt and Gottschalk (2002). For consumption

inequality, see, for example, Cutler and Katz (1992), Krueger and Perri (2006), Blundell et al. (2008), and Aguiar

and Bils (2015). Meyer and Sullivan (2017) provide a wide-ranging overview of inequality trends in both income and

consumption in the United States since the 1960s. They informally discuss the role of asset prices in the joint evolution

of these inequality measures.

1781

C

(2019) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

1782 ETHERIDGE

and income inequality, and particularly the role of shocks from the labor market.4I contribute

to this second literature by exploring the role of shocks to asset prices: in this case house prices.

I organize the analysis around a relatively standard partial equilibrium model of household

choices over (nonhousing) consumption, housing tenure, and housing size in an incomplete-

market setting. This benchmark model includes exogenous aggregate house price shocks, id-

iosyncratic income shocks, transactions costs, and borrowing restrictions determined by collat-

eral constraints.

I use this framework to make three specific contributions. First, I provide an analytic charac-

terization of how house price shocks affect consumption inequality. To do this, I simplify the

benchmark model by abstracting from transactions costs and collateral constraints. I also use

inequality measures that decompose appropriately: These are the variance-of-the-logarithm

measure, and associated covariances, as used since Deaton and Paxson’s (1994) paper. By sim-

plifying the benchmark model, and then by approximating the shock to consumption in terms

of the house-price and labor-market shocks, I track how house price shocks, in particular,

drive the inequality measures. More precisely, I decompose the consumption response to a

house price shock into three factors: an income effect, a substitution effect, and an endow-

ment/wealth effect.5As such, my decomposition relates to that in Berger et al. (2018), who

derive an interpretable sufficient statistics formula for the approximate effect of house price

shocks on consumption. My formula is similar to theirs but is more easily used to study hetero-

geneity and inequality movements because each component of my formula is observable at the

household level.

In terms of the three factors, I argue that consumption inequality is driven mainly by het-

erogeneity in the endowment effect. This heterogeneity arises not only because of the division

between owners and renters, but also, importantly, because of dispersion in leverage. Ulti-

mately, I express the effect of house price shocks on inequality in terms of key statistics, which

can be computed from cross-sectional microdata. At the level of the household, the key statis-

tic is an endowment, or housing wealth share: the proportion of gross housing wealth in total

lifetime wealth. In terms of inequality movements, the key statistic is how this housing wealth

share covaries with the consumption and income distributions. It is important to remember

that, although inequality is driven by the house price shocks, it is also, of course, driven by

idiosyncratic shocks from the labor market. This is the first article therefore to characterize how

these shocks affect consumption inequality in combination.

By simplifying the model, I am able to derive this analytic characterization. However, this

characterization comes at the cost of abstracting from important features of the housing and

mortgage markets, such as collateral constraints. I therefore test the accuracy of my approxi-

mations by solving the full benchmark model and disciplining it using U.K. data. I show that

the approximations driven by a housing wealth effect are close to the true house price elas-

ticity, on average, and that the approximations pick up variation in the true elasticity well in

several dimensions.

My second contribution is to apply the framework quantitatively to observed movements in

inequality. I examine the United Kingdom from the late 1980s until the financial crisis around

2008. This is a suitable period for the application. House prices displayed large swings: a large

decline in the early 1990s followed by a boom of unprecedented length. Over this period, the

labor market was also comparatively stable.6More practically, during this period, detailed data

are available on household wealth in 1995, 2000, and 2005 from the British Household Panel

4In addition to those papers focusing more explicitly on the inequality boom in the 1980s, see also Storesletten et al.

(2004), Krueger and Perri (2005), Blundell et al. (2013), Heathcote et al. (2014), and Blundell et al. (2016).

5Following a house price increase, the income effect captures the result of the reduction in real lifetime resources

owing to the increase in the cost of housing services; the substitution effect captures the shift away from housing

services holding fixed lifetime utility; the endowment effect captures the pure wealth effect from the increase in value

of housing assets.

6The United Kingdom experienced a mild recession in the early 1990s during which the unemployment rate rose by 3

percentage points to a little over 10%. Thereafter, unemployment declined and stayed below 6% from 1999 onward. On

HOUSE PRICES AND INEQUALITY 1783

Survey (BHPS). This data set also contains data on incomes and on food consumption. My

headline inequality measures come from the Family Expenditure Survey (FES), which has

detailed cross-sectional data on incomes and on broad expenditures.7

I first show that the approximated model is highly consistent with broad movements in the

data. Specifically, I analyze fixed groups of households given by decade-of-birth cohorts. As

discussed, the inequality movements depend most importantly on how the housing wealth

share is distributed. The data show that, for each group, this share is positively correlated with

consumption, and increasingly so over time. Because of this positive correlation, the model

implies that house price increases should cause the gap between consumption inequality and

other measures to grow. Similarly, house price declines should cause the gap to fall. For both

the groups I study, this gap fell in the 1990s when house prices declined and grew strongly

thereafter consistently with the model.

I then quantify these movements, first using the wealth share approximation. Focusing on the

peak years of the house price boom over 1997–2004, the gap between consumption inequality

and the other measures grew by 4.1 log points for the 1950s cohort and by 5.8 log points for

the 1960s cohort.8These are large changes compared with previous inequality movements. By

computing the statistics described above, I show that the simplified model explains around 50%

of the observed gap for the 1950s cohort, although for the 1960s cohort, the model explains less.

A key strength of my approach, here, is that I can impose the cross-sectional distributions of

the relevant variables into the calculations directly.

I then extend this analysis by examining the role of several additional features, most impor-

tantly collateral effects.9To this end, I use the full benchmark model’s nonlinear solution and

measure the strength of collateral constraints at the household level using housing leverage.

I use the model to compute a simulation-based consumption elasticity that I can then take to

the data. Qualitatively and empirically, I find that leverage covaries positively with the income

and consumption distributions. Therefore, the housing wealth and collateral channels work in

the same direction: Through both channels, a positive house price shock causes consumption

inequality to increase. Quantitatively, collateral effects are less important in driving inequality

because households with high leverage do not tend to lie at the extremes of the consumption dis-

tribution.

My third and final contribution is to assess how accounting for house price shocks affects

estimates of income risks. As discussed above, researchers have used the cross-sectional distri-

butions of income and consumption to quantify risks and risk sharing at least since papers by

Deaton and Paxson (1994) and Blundell and Preston (1998). In my model, ignoring house price

increases causes estimates of the variance of permanent income shocks to be biased upward.

This is because the researcher misattributes growth in consumption inequality to income shocks

instead of to shocks from asset prices. Likewise, I find that accounting for the house price

declines in the first part of the sample period increases the estimated variance of permanent

shocks during the recession in the early 1990s.

This article proceeds as follows: I first provide a dedicated discussion of the literature on

housing, which has recently developed rapidly. Section 2 presents the benchmark model. I

discuss the simplified model’s approximated solution and show the effect of house price shocks

on inequality measures in Section 3. I discuss the various sources of data in Section 4, where I

also discuss the method used to estimate income risks. Section 5 presents the primary results

based on housing wealth shares. I examine the accuracy of the wealth-share approximations

in Section 6, where I also extend the quantitative analysis to include collateral effects and two

additional features: a time-varying rent–price ratio and an elasticity of substitution between

the other hand, inequality was also influenced by redistributive reforms beginning in 1997. These reforms do not affect

the present analysis. See Etheridge (2013), an earlier working paper version of the current study, for a fuller discussion.

7As is relatively standard in the literature, I drop the distinction between consumption and expenditure in this article.

8In other words, the variance of log consumption grew by 0.058 more than the variance of log income for the 1960s

cohort. I use the term “log points” similarly throughout the article.

9The collateral effect captures the result of relaxing or tightening borrowing constraints, holding house prices fixed.