RDP 2019-01: A Model of the Australian Housing Market 6. Conclusion

Our main results, which we do not repeat, were stated in the introduction. Several limitations of our model and directions for further research are worth noting.

Some of our equations are better than others. The stability and precision of our estimates give us confidence in the responsiveness of construction to interest rates and prices and in the response of rents to vacancies. We are much less certain about the response of vacancies to completions and household formation. More broadly, we know more about the response of quantities to prices than we do about the response of prices to quantities.

To keep our analysis relatively transparent and manageable, we have ignored some minor variables (like taxes, lending restrictions, the approvals pipeline and first home owner grants). And we have assumed that many relevant variables are exogenous. A more complicated analysis would relax those assumptions. A priority is to allow household size to depend on income and rents. This would then affect household formation and housing quality. It is not clear that adding taxes to the user cost would substantially change the behaviour of the model, but it would be useful for policy discussions.

We have also assumed that most behaviour is backward looking (though house prices depend on expectations of interest rates and capital gains). That assumption seems consistent with the stability of our equations. The conditions in which it would hold in the future would be interesting to explore.

Although our equations represent a long search over alternative specifications, this was far from exhaustive. Further examination of aggregate time series would doubtless improve on our model. However, we suspect that a more fruitful line of enquiry might involve examining alternative approaches and datasets. Cross-section, panel and event studies provide information that should corroborate or modify our estimates, particularly of long-run parameters.