RDP 2013-04: Home Prices and Household Spending 1. Introduction

Although home prices and consumption tend to move together, understanding the relationship between the two has proven a vexing task for policymakers and commentators (Figure 1; Dvornak and Kohler 2003; Fisher, Otto and Voss 2010).

Existing estimates for Australia suggest that a 1 dollar change in housing wealth is associated with a 3 cent change in non-housing consumption (see, for example, Dvornak and Kohler (2003)). Likewise, taking the log difference of the series in Figure 1 and regressing non-housing consumption on home prices implies a marginal propensity to consume (MPC) of around 2½ cents per dollar change in home prices.^[1]

Such findings may be useful for policymakers trying to gauge the future pace of consumption growth given a series of housing price shocks. In this setting, the reliability of the MPC is arguably more important than the mechanism driving the result. Nevertheless, aggregate regressions tell us very little about why there is a positive relationship between home prices and consumption. In particular, there are strong arguments against interpreting the aggregate MPC as a ‘traditional wealth effect’, whereby spending rises with unanticipated changes in home prices due to an increase in homeowners' lifetime resources.

Housing assets, like consumer durables, are different from financial assets. A dwelling is both an asset and a consumption item that provides a stream of services over its lifetime. Accordingly, increasing home prices have a distributional impact on wealth, creating winners and losers whose spending responses may differ. For those who own more housing than they foresee needing in the future (for example, an older household looking to trade down), increasing home prices increase lifetime resources; for those who own less housing than they foresee needing in the future (for example, a young family who will need a larger home in the future), increasing home prices decrease expected lifetime resources. In aggregate, therefore, the causal relationship between home prices and spending is ambiguous, and depends on the MPCs of different groups.^[2]

In the context of an overlapping generations model, Buiter (2010) concludes that there is no traditional wealth effect on aggregate consumption. Rather, he argues that changes in home prices affect spending because housing can be used as collateral against which people can borrow to finance consumption. The effect on current spending could be quite large if homeowners were credit constrained before any increase in home prices. In the long run, however, there would be no wealth effect: an increase in home prices would stimulate debt-financed spending in the short run while depressing it in the long run as households repay debt.

Other factors may also affect the relationship between home prices and spending. If older households plan to leave money or even a home to their children or grandchildren, and/or younger households anticipate such bequests, then irrespective of home prices, younger and older households may not perceive any change in their lifetime resources available for spending (see, for instance, Mishkin (2007)). Debelle (2004) suggests the empirical relationship between home prices and consumption could be due to households regarding their homes as saving vehicles, with a rise in home prices increasing perceived savings and thus driving consumption behaviour. Households may also perceive housing wealth as a precautionary saving vehicle against unanticipated future events such as redundancy (Carroll, Dynan and Krane 2003). Finally, the relationship between housing market turnover and consumption – high turnover leading to increased spending on furnishings, audiovisual equipment and the like – could drive co-movements between home prices and spending, since turnover tends to increase when home prices rise.

Against this background, it is perhaps not surprising that there is no clear consensus on the cause of the correlation between home prices and household spending. Broadly speaking there are three hypotheses that have predictions for how households with certain characteristics should respond to changes in home prices. Under hypothesis (1), increases in perceived home prices raise spending via a traditional wealth effect. This channel points to a stronger effect on the spending of older homeowners (who are most likely to own ‘excess’ housing). Under hypothesis (2), increases in home prices loosen credit constraints and therefore raise spending through an increase in the value of collateral, the opportunity for home equity redraws and/or through a reduction in the necessary level of buffer-stock, or precautionary, saving. Younger homeowners are more likely to be credit constrained (Disney, Bridges and Gathergood 2010) as well as buffer-stock savers (Gourinchas and Parker 2001). Accordingly, loosening of credit constraints suggests a stronger link between home prices and spending for younger homeowners. And under hypothesis (3), home prices and spending are influenced by a common third factor such as something that affects expectations regarding future income. A common influence like unexpectedly higher income expectations should have a stronger effect on the spending of younger households, regardless of home tenure status; that is, this hypothesis implies that the spending of young homeowners and young renters should both rise, as both have relatively more years of work ahead of them and so benefit the most from a rise in the wages they may expect to earn in the future.

It is difficult to discriminate between these competing hypotheses based on the aggregate relationship between home prices and non-housing consumption. In light of this, a number of studies have used micro data to understand the co-movement between home prices and consumer spending, but with mixed results. Using a survey of UK households, Attanasio et al (2009) argue that income expectations, as per hypothesis (3), have played an important role, because the association between home prices and spending is stronger for younger households irrespective of home tenure type. Using the same UK survey, Campbell and Cocco (2007) draw the opposite conclusion. They find home-price wealth effects are largest for older homeowners and lowest for renters. They interpret this heterogeneity in home-price wealth effects as being consistent with a traditional wealth effect. Muellbauer (2009) and Duca, Muellbauer and Murphy (2011) disagree, arguing that a housing collateral effect is the key to understanding the role of home prices in explaining consumption fluctuations. While Muellbauer (2009) agrees with the results presented by Attanasio et al (2009), there is disagreement over interpretation. In addition to the common association between home prices, income innovations and spending, Muellbauer finds that credit-constraint effects are significantly positive for young homeowners and negative for the old.

For Australia, Yates and Whelan (2009) examine the variation in spending by home price across households at given points in time after controlling for household demographics and financial conditions. They show that in 2003 the spending of younger households was more sensitive to home prices than that of older households.^[3] They interpret their results as being consistent with the credit constraints hypothesis.

In this paper we use the Household, Income and Labour Dynamics in Australia (HILDA) Survey to examine home-price wealth effects, using household-level data for the eight years to 2010.

Our analysis contributes to the literature in a number of ways. To begin, we fully exploit the panel nature of our dataset that follows individual households through time. That is, we estimate the dynamic response of a household's spending to changes in the perceived price of their home while controlling for unobservable, time-invariant differences between households (such as their level of optimism or thriftiness). To our knowledge this is the first paper to do this using the HILDA dataset.

At the household level, we estimate home-price wealth effects that are larger for younger homeowners, and find that renters exhibit negative home-price wealth effects. We suggest that young homeowners' relatively strong spending response to an increase in home prices supports the hypothesis of credit constraints.

We also examine whether these results can be replicated in a more parsimonious, but less informative, model that relies on a ‘pseudo-panel’ of birth cohorts instead of household-level data. This is done to assess the effect that aggregating may have had on earlier studies using UK data (see Attanasio et al (2009)). For instance, Muellbauer (2007) has argued that some of the studies cited above fail to control for cross-sectional variation across households; our dataset allows us to assess this criticism directly. The results from the cohort pseudo-panel are similar to those obtained from the equivalent actual panel. This suggests that pseudo-panels are a reasonably good substitute for actual panels. However, the necessary use of aggregate home prices rather than self-assessed home prices in pseudo-panels tends to inflate estimated wealth effects.

The remainder of this paper is set out as follows. Section 2 introduces the dataset used in this study and presents some stylised features of the variables of interest. Section 3 presents our methods and Section 4 details results. Section 5 concludes.

Footnotes

Here, the estimated elasticity is converted to a MPC using the sample average ratio of non-housing consumption to home prices of 16 per cent. [1]

A related but distinct argument against traditional wealth effects is that an increase in home prices (not associated with some other factor relevant to future incomes) cannot affect the quantity of goods and services available for consumption, at least in a closed economy. In an open economy, this need not be the case. For example, increases in home prices increase domestic households' collateral values, allowing for increased borrowing from the rest of the world, which would allow for an increase in consumption. [2]

This was not the case in 1998, when the variation in spending to home prices was more sensitive for older households. [3]