RDP 2003-07: Housing Wealth, Stock Market Wealth and Consumption: A Panel Analysis for Australia 5. Robustness of the Results Through Time

Previous studies of consumption functions for Australia have found considerable parameter instability through time (e.g. Debelle and Preston (1995) and Tan and Voss (2003)). One possible reason for this is the process of economic reform in Australia during the last two decades. Financial liberalisation and deregulation can create structural breaks in the underlying relationships that we are estimating and these are not necessarily ‘one-off’ breaks but could occur slowly over time. Naturally, this raises the question of how robust our results are to changes in the time period chosen.

We have therefore re-estimated the model over different time periods. Similar to other studies, we find that the income effect rises considerably over the 1990s, irrespective of the estimation method chosen. In comparison, the wealth coefficients, which are our parameters of interest, appear to be more stable through time. Figure 1 presents the results for the wealth MPCs using a rolling window of 10 years and the fixed-effects IV model.

Figure 1: Rolling Fixed-effects Estimates
Figure 1: Rolling Fixed-effects Estimates

While there appears to be some variation in the estimated coefficients over time, it is hardly excessive considering that a 10-year window is a rather short sample for this type of regression. For the entire time period, the stock market effect is greater than the housing wealth effect. The confidence intervals indicate that for most of the period this difference is statistically significant. When we use the measure with direct equity holdings only (Figure 1 shows the stock market wealth effect for all equity holdings), the two MPCs are not significantly different from each other over most of the time period. Both wealth effects are significantly different from zero at the 10 per cent level of significance for every window. Thus, whatever structural changes have occurred over the estimation period, they apparently do not alter the finding that both stock market wealth and housing wealth affect consumption.

It should also be noted that the coefficient on stock market wealth has changed somewhat over time, perhaps reflecting – at least in part – ongoing changes in the structure of the financial markets. Similarly, the increase in the housing wealth coefficient towards the end of the sample period might be consistent with the observed increase in housing equity withdrawal over the recent years. These changes, although they appear to be occurring gradually, caution also that the past may not always be an indication of the future, especially during periods of rapid financial innovation.