Estimating smooth series of housing prices for housings with emerging characteristics . A case study of the opening of the Copenhagen metro

Specialeforsvar ved Maria Andreasen

Titel: Estimating smooth series of housing prices for housings with emerging characteristics
A case study of the opening of the Copenhagen metro

 

Abstract: In 2015 the DREAM group implemented a version of the geographically weighted panel regression which estimates a series of housing prices for each housing in Denmark from 1999 to 2015. Two main problems arise when estimating these series. The first problem is, that the series are discontinuous, which is caused by temporal noise. The noise is removed by smoothing the series, which the DREAM group has done using the LOESS regression. In this thesis, the model is restated as a local level model, making it possible to smooth the series with the Kalman smoother. This smoother is implemented with a variance restriction, restraining the relation between the variance of respectively the structural trend and the cyclical movements of the housing prices. This restriction is tuned across neighbourhood population densities, and an optimal restriction is determined, stating that the variance of the cyclical movements should be twice as large as the variance of the structural trend. The Kalman smoother is compared to the LOESS regression, and the results show that the accuracy of the smoothed series is almost unaffected by the choice of smoothing method. The computation time of the Kalman smoother is almost three times smaller than the time used by the LOESS regression, and hence the Kalman smoother is concluded to be a better choice of smoothing method for smoothing series of housing prices. The second problem with the model implemented by the DREAM group is that characteristics of housings cannot to emerge between 1999 and 2015. This thesis states a model which allows for the opening of the Copenhagen metro to affect the housing prices in Copenhagen. The series of housing prices are smoothed by the Kalman smoother. These smoothed series are tuned to determine the levels of the categorical variable describing the distance to the nearest metro station. The model achieves most accurate results when housings within 200 metres of a metro station are affected by the opening of the metro, while housings more than 200 metres away are not. Due to scarcity of data, this distance is increased to 400 metres for housings close to the stations from Islands Brygge to Vestamager. The tuned model is used to estimate the effect of the opening of the metro. An overall increase in prices is estimated, where housings situated close to stations from Vanløse to Femøren experienced an average increase in prices of 1.37%, while housings close to stations from Islands Brygge to Vestamager experienced an increase of 0.97%. Examining the results near each station, it is shown that the effect caused by the metro differs throughout Copenhagen. The opening of the metro primarily decreased prices of housings in Frederiksberg, and increased prices of housings on Amager. The largest positive effects are found in housings near Amager Strand station, where prices on average increased by 15.57%. The largest negative effect is found in housings near Forum station which on average experienced a decrease in prices of 2.94%.

 

Vejleder:   Mogens Fosgerau, Ø,I,
Censor:     Niels Heldrup, Aarhus Universitet