Disentangling Housing Density Effects in a Compact Urban Form : A Novel Spatial Clustering Approach

Achieving higher urban density has become a core principle of the contemporary planning profession as a part of sustainable city development. Both positive and negative externalities of housing density are well-recognised in urban studies. However, very little is known about the net housing density effects. In planning research, density effects are usually studied qualitatively through surveys or interviews. However, the results are hard to make cross-city comparisons. The few quantitative studies considered housing densities of administrative area units, such as census traits or statistical area units, prone to boundary bias. This paper seeks to empirically identify the overall housing density and specific housing density effects (public housing and Airbnb) on house prices using a revealed preference approach. We develop a novel spatial clustering method for the number of houses at either 0.5 km, 1.0 km or 1.5 km radial circle of each home. Based on 18,000+ housing transactions in 2020 in Auckland, New Zealand, the results confirm the adverse density effects of both overall housing and public housing in a neighbourhood, ceteris paribus. In contrast, the density effects estimated using the statistical area units are found to be positive and insignificant, which shows the problem of using administrative area units in estimating neighbourhood effects. The results will have far-reaching implications for urban planners and government officials in formulating development intensity controls