Modelling Spatial Scale and Heterogeneity in the Rotterdam Housing Market using Multiscale Geographically Weighted Regression
This study investigates the spatial heterogeneity of the Rotterdam housing market, as well as the spatial scale at which the processes that explain housing prices operate. I use data on 3632 residential properties that were sold in Rotterdam in 2018 provided by the NVM. Using a recently (2017) developed technique called Multiscale Geographically Weighted Regression (MGWR), I show that certain factors explaining residential housing prices operate on diﬀerent spatial scales, indicating the existence of relatively local and more global eﬀects. Lot size and the state of maintenance of the house exterior seem to operate on a local scale, whereas the degree of isolation and number of rooms appear to be global. Comparison of the MGWR model with a regular GWR with ﬁxed bandwidths and a global hedonic model with location ﬁxed eﬀects, show that the MGWR best ﬁts the data, i.e. a model that allows the bandwidths to vary for each parameter outperforms the models that do not allow this.
|Thesis Advisor||N Cortinovis|
NA Feddes. (2020, July 28). Modelling Spatial Scale and Heterogeneity in the Rotterdam Housing Market using Multiscale Geographically Weighted Regression. Economics. Retrieved from http://hdl.handle.net/2105/52241