The wealth effect of housing on aggregate consumption

Abstract

This study measures the effect of changes in net housing and financial wealth on household consumption using Australian data over the period Q2:1988-Q1:2003. It is found a permanent one dollar rise in housing wealth leads to a six cent increase in consumption, three times the effect of financial wealth. The result speaks strongly against the notion of assets fungibility, and suggests that a sharp movement in house prices is potentially more disruptive than a corresponding movement in financial asset prices.

Notes

¹ The International Monetary Fund (IMF, 2002; IMF, 2003), the United States Federal Reserve Bank (Gramlich, 2002) and the Reserve Bank of Australia (RBA, 2003a) are among a number of policy bodies that have expressed concern over the rate of house price appreciation.

² We have also estimated the short run effects. We do not report the results here as they add little to the conclusions. The results are available on request.

³ Employer's contribution to superannuation has been excluded from the measure of income, but enters the empirical estimation via the financial wealth variable. The rationale for the exclusion of superannuation is that labour income measures disposable income rather than total income.

⁴ Income derived from wealth accounts for around 23% of total household income.

⁵ The estimation includes leads and lags of order two. Greater than two lags were found to be statistically insignificant.

⁶ The robustness of this specification is tested by modelling the three alternative long run relationships with a constant term. The value of the constant term is found to be extremely large relative to the exogenous variables’ coefficient values, and consequently resulted in statistical insignificance of both income and net wealth. The results are likely due to the fact that, with aggregate data, all observations are clustered around typical levels of income, housing and financial wealth. In particular, there are no outlying observations at the very high and very low ends, and as such, the estimated constant term takes on the majority of the explanatory power for out of sample ranges.

⁷ The significance of financial wealth variable is particularly sensitive to the data period used. For example, if the first observation is excluded from the estimation, the financial wealth variable becomes significant at the 5% level.