RDP 2014-09: Predicting Dwelling Prices with Consideration of the Sales Mechanism 6. Conclusion

This paper analyses whether the mechanism of sale is useful for forecasting average dwelling prices. Using hedonic price indices for Sydney and Melbourne, we show that auction prices and private-treaty prices have different statistical properties, including significant differences in their momentum, ability to forecast each other, and ability to forecast average growth of prices overall.

Our results suggest that growth in auction prices is much less autocorrelated than growth in private-treaty prices – indeed, we could not reject the null of a random walk with drift in Sydney auction prices. This surprisingly strong result, which holds even though the two measures share the same common price trend, suggests that auction prices incorporate new information more quickly than private-treaty prices. Consistent with this, we find that including lagged information on auction prices improves forecasts of average dwelling prices growth overall.

In addition, we find that auction prices Granger cause private-treaty prices, but that the reverse is not true. When combined with the assumption that these two price series are cointegrated, this result is useful for separating prices into their transitory and permanent components and for forecasting the evolution of prices at short-term forecasting horizons.

Empirically, we find that auction prices are driven by permanent shocks, whereas private-treaty prices are affected by both permanent and transitory shocks. If permanent shocks are an accurate measure of changes in the common trend in all prices, as should be the case when there is cointegration, our results suggest that auction prices are likely to be a better reflection of this common trend. Furthermore, the presence of transitory shocks in private-treaty prices implies that they take longer to incorporate changes in the common trend. This also helps to explain why private-treaty prices are less useful when forecasting.

We interpret our empirical findings using two models of price formation – an English auction with linearly affilated values and a bilateral Nash bargain. We show that these two models imply a VECM approximation that is consistent with our data. These two models of price formation, when combined with our empirical findings, suggest that the key issues at hand are: the extent to which the average dispersion of valuations and the average bargaining strength of buyers and sellers affect prices; whether the relative importance of buyers' and sellers' valuations differs according to the mechanism of sale; and whether buyers and sellers behave differently in response to shocks.

The question we are interested in is whether distinguishing prices by the mechanism of trade can assist in the forecasting of dwelling prices and in understanding the dwelling price cycle. On both empirical and theoretical grounds, we argue that the answer is yes. Separating prices by the type of sale, and more specifically focusing on auction prices, can improve forecasting and in identifying the persistence of shocks. We believe that these results are interesting both narrowly, for those concerned with forecasting prices, and more broadly for understanding how price formation can provide alternative insights into dwelling price cycles.