RDP 2003-09: Housing Leverage in Australia 5. Conclusions

The results presented here are descriptive, rather than being a fully causal model of household behaviour. Households' leverage at any point in time reflects the cumulation of many past decisions including the decision to purchase, to pay off the original mortgage or to make overpayments when it is feasible to do so. Cross-sectional data can be used to relate current leverage to observable characteristics, but it is not feasible to reconstruct all these past decisions.

The results suggest several descriptive conclusions about the pattern of Housing Leverage in Australia. The graphical results indicate that the households that are most highly leveraged are those most able to bear the debt – mid-life households with high income. Leverage is also higher for households living in areas least vulnerable to reversals in housing prices, the outer suburbs and non-metropolitan regions that have experienced relatively smaller price gains in recent years. Young homeowners are likely to have particularly high leverage, but young households in general are less likely to be homeowners. On the other hand, households that are negatively geared on investment property, and thus declaring a loss on their rental income, are much more likely to have a mortgage, and to have higher leverage when they do have a mortgage. This finding is indicative of a sub-group of the population that is willingly engaging in leveraged asset accumulation, and taking the associated financial risks. The general picture, however, accords with aggregate data in suggesting that leverage on the housing stock remains fairly moderate.

The econometric modelling also shows significant roles for age, life-cycle stage and time at address variables. This points to the importance of the passive paydown of debt as scheduled in the household's mortgage contract, in determining current leverage. Similarly, the pattern of coefficients on the locational variables suggests that increases in housing prices reduce leverage, and are not offset by households increasing their debt in response to the increase in wealth. These determinants of leverage can all be characterised as largely being beyond the control of individual households, suggesting that at least in the short run, households do not necessarily adjust their balance sheet to maintain a desired leverage ratio as predicted by some theoretical models. Against this, however, we find evidence of at least some households making explicit decisions about the end date of their mortgage, which might not be the date specified in the loan contract. As noted earlier, if some households did not make conscious decisions about the desired date on which their loan will be fully paid off, move-in date and payoff date would not both be significant in our estimated leverage equation.

Amongst the households making these explicit decisions about their payoff date, at least, we might expect deliberate portfolio reactions to developments in interest rates and housing prices, rather than simply adhering to a predetermined path for their remaining outstanding debt. Observing such reactions would, however, require tracking households through time. The longitudinal nature of the HILDA Survey will make it uniquely suited, amongst all datasets for Australia, to examination of these household responses. Future waves of the HILDA dataset will therefore be essential for further work on understanding households' decisions about their balance sheets, leverage and debt, and in particular their responses to interest rate changes and housing price movements through these channels.