This work looks at the clustering of neighbourhood groups into classified zones based on a comprehensive spatio-temporal longitudinal database capturing high-resolution LSOA monthly price dynamics along with a series of generated spatial and temporal lags to characterise the persistence of all neighbourhood prices.
The preferred specification is a 7 K-means clustering algorithm using a parsimonious set of inputs to characterise the neighbourhoods. The clustering identifies three large collections of neighbourhoods with positive price growths with auxiliary clusters capturing declining neighbourhoods or those which are spatially isolated. The three primary clusters each represent a different level of prices and growth ranging from ultra- high priced, moderately-high priced and the ambient price level. For each of these broad groupings iterative sub-clustering breaks the cluster into two and is able to identify distinctive within cluster sub-groups with marginally different characteristics.
The clustering output is able to identify those neighbourhoods in the urban area which deviate from the ambient price levels over time. As ambient neighbourhoods transition into moderately or ultra-high priced zones, we are able to study the evolution of changing boundary extents. This is further useful for highlighting housing pressures outside of the city centre as more peripheral areas move into structurally higher growth categories, potentially isolating other areas or leading to segregation or displacement.