Forecasting house prices: the role of market interconnectedness

While the existing research uncovers interconnections between various housing markets, it largely ignores the question of whether such linkages can improve house price predictions. To address this issue, we proceed in two steps. First, we forecast disaggregated house price growth rates from Australia and China to determine whether incorporating price links can improve out-of-sample predictions. We find that accounting for within-city house price interconnectivity in Sydney and Melbourne can indeed improve house price predictions. However, when forecasting city-level prices from China, univariate models produce superior predictions. Second, in order to shed light on our empirical findings, we conduct simulation experiments calibrated to reflect the connections estimated from the data. The predictive ability of house price connectivity hinges on the sparsity and strength of the connections between interconnected markets. In the presence of stronger and denser connections, connectivity information is crucial for improving short-term forecasts. On the other hand, when the connections are sparse and weak (as in the Chinese housing data), the univariate models outperform. Our study shows that finding evidence of significant price interconnections does not always lead to forecasting gains.