The Housing Submarkets
6.2 Submarkets of Other Regions
In this section, we present the housing submarkets of other regions of the Czech Republic. We limit the results presented here to only four interesting price determinants for each region and the spatial distribution of the housing submarkets themselves. Notice, that in some regions the differences between submarkets are present, however, their magnitudes are fairly negligible. The entire framework of the housing submarkets we utilized was firstly presented by the study of (Kopczewska & Ćwiakowski 2021) and we expend the framework by providing the continuous distribution of the housing submarkets, which is interpolated via the kriging model.
We would also like to stress out that we are fully aware that the spatial distribu-tion pattern of the housing submarkets is, for certain regions, rather doubtful1. There are, in our opinion, two main sources of these distribution patterns.
Firstly, in the case of some regions, we are working with a (relatively) limited number of observations (see Figure 3.2), which are not covering the entire area of the studied region. Secondly, our framework is based on the linear kriging prediction as opposed to the classification model form and thus the predicted values are rounded to the closes integer.
1For certain reasons, we are not presenting the housing submarkets of the Olomouc Re-gion. We had a problem with the kriging interpolation algorithm convergence because the observations could not be reasonably modeled by any variogram model function. The es-timated variogram models suffered from a rather poor fit. Therefore, we had excluded the Olomouc from the housing submarkets analysis.
79
The Housing Submarkets 6.2 Submarkets of Other Regions
Lastly, as the housing submarkets are not identified beforehand, we may be referring to the housing submarkets identification framework as the unsuper-vised learning task as simply we cannot just simply minimize the grid’s pixels class of the region as simply the "true" class regarding the housing submarket is never known apriori.
Also, in the individual boxplots of the price determinants of the housing sub-markets of each region, we are not explicitly plotting the distributions of price determinants for every single submarket as some submarkets had a relatively small number of observations and/or consist mostly of outlier observations.
These steps are purely for Figure purposes and were conducted with huge at-tention and do not affect any results of our analysis even slightly.
80
Submarket 1 2 3 4 5 6 7 8 9 10
Aussig
Housing Submarkets: Spatial Distribution
Figure 6.6: Housing Submarkets: Aussig
81
The Housing Submarkets 6.2 Submarkets of Other Regions
percentage effect of Room
Submarket
Housing Submarkets: Effect of Room
15.0%
percentage effect of After a Reconstruction
Submarket
Housing Submarkets: Effect of After a Reconstruction
0.80%
percentage effect of Meters
Submarket
Housing Submarkets: Effect of Meters
0.0%
percentage effect of Private ownership
Submarket
Figure 6.7: Housing Submarkets: Aussig
82
Submarket 1 2 3 4 5 6 7 8
Carlsbad
Housing Submarkets: Spatial Distribution
Figure 6.8: Housing Submarkets: Carlsbad
83
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Square Meter
41.75%
Housing Submarkets: Effect of Private Ownership
25.60%
Housing Submarkets: Effect of Very Good Condition
21.700%
Figure 6.9: Housing Submarkets: Carlsbad
84
Submarket 1 2 3 4 5 6 7 8
Pilsner
Housing Submarkets: Spatial Distribution
Figure 6.10: Housing Submarkets: Pilsner
85
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Room
31.50%
Housing Submarkets: Effect of Private OwnershipZlin
18.90%
Housing Submarkets: Effect of Very Good ConditionZlin
31.50%
Housing Submarkets: Private ownershipZlin
Figure 6.11: Housing Submarkets: Pilsner
86
Submarket 1 2 3 4 5 6 7 8
South Bohemian
Housing Submarkets: Spatial Distribution
Figure 6.12: Housing Submarkets: South Bohemian
87
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Square Meter
31.50%
Housing Submarkets: Effect of Private Ownership
18.90%
Housing Submarkets: Effect of Very Good Condition
13.50%
Figure 6.13: Housing Submarkets: South Bohemian
88
Submarket 2 3 4 5 6 7 8 9 10 11 12 13
Capital city Prague
Housing Submarkets: Spatial Distribution
Figure 6.14: Housing Submarkets: Hradec Králové
89
The Housing Submarkets 6.2 Submarkets of Other Regions
percentage effect of Concrete building
Submarket
Housing Submarkets: Effect of Room
23.90%
percentage effect of Garage
Submarket
Housing Submarkets: Effect of After a Reconstruction
0.6200%
percentage effect of Meters
Submarket
Housing Submarkets: Effect of Meters
16.00%
percentage effect of After reconstruction
Submarket
Figure 6.15: Housing Submarkets: Hradec Králové
90
Submarket 0 1 2 3 4 5 6 7
Liberec
Housing Submarkets: Spatial Distribution
Figure 6.16: Housing Submarkets: Liberec
91
The Housing Submarkets 6.2 Submarkets of Other Regions
percentage effect of Concrete building
Submarket
Housing Submarkets: Effect of Room
10.0%
percentage effect of Garage
Submarket
Housing Submarkets: Effect of After a Reconstruction
0.30%
percentage effect of Meters
Submarket
Housing Submarkets: Effect of Meters
0%
10%
20%
1 2 5 6 4 3
Submarket
percentage effect of After reconstruction
Submarket
Figure 6.17: Housing Submarkets: Liberec
92
Submarket 1 2 3 4 5 6 7
Aussig
Housing Submarkets: Spatial Distribution
Figure 6.18: Housing Submarkets: Moravian-Silesian
93
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Room
10%
Housing Submarkets: Effect of Private OwnershipZlin
5.0%
Housing Submarkets: Effect of Very Good ConditionZlin
10%
Housing Submarkets: Private ownershipZlin
Figure 6.19: Housing Submarkets: Moravian-Silesian
94
Submarket 2 3 4 5 6 7
Pardubice
Housing Submarkets: Spatial Distribution
Figure 6.20: Housing Submarkets: Pardubice
95
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Square Meter
27.00%
Housing Submarkets: Effect of Private Ownership
15.80%
Housing Submarkets: Effect of Very Good Condition
6.50%
Figure 6.21: Housing Submarkets: Pardubice
96
Submarket 0 1 2 3 4 5 6 7 8 9
South Moravian
Housing Submarkets: Spatial Distribution
Figure 6.22: Housing Submarkets: South Moravian
97
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Square Meter
7.40%
Housing Submarkets: Effect of Private Ownership
7.00%
Housing Submarkets: Effect of Very Good Condition
9.300%
Figure 6.23: Housing Submarkets: South Moravian
98
Submarket 1 2 3 4 5 6 7 8 9 10 11 12 13
Central Bohemian
Housing Submarkets: Spatial Distribution
Figure 6.24: Housing Submarkets: Central Bohemian
99
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Square Meter
0.0%
Housing Submarkets: Effect of Private Ownership
10.0%
Housing Submarkets: Effect of Very Good Condition
2.5%
Figure 6.25: Housing Submarkets: Central Bohemian
100
Submarket 0 1 2 3 4 5 6 7
Vysocina
Housing Submarkets: Spatial Distribution
Figure 6.26: Housing Submarkets: Vysocina
101
The Housing Submarkets 6.2 Submarkets of Other Regions
Housing Submarkets: Effect of Square Meter
6.50%
Housing Submarkets: Effect of Private Ownership
13.00%
Housing Submarkets: Effect of Very Good Condition
12.50%
Figure 6.27: Housing Submarkets: Vysocina
102
Submarket -1 0 1 2 3 4 5 6 7 8 9 10 11
Aussig
Housing Submarkets: Spatial Distribution
Figure 6.28: Housing Submarkets: Zlin
103
The Housing Submarkets 6.2 Submarkets of Other Regions
percentage effect of Room
Submarket
Housing Submarkets: Effect of RoomZlin
20.50%
20.75%
21.00%
7 1 4 5 6 2 3
Submarket
percentage effect of After a Reconstruction
Submarket
Housing Submarkets: Effect of After a ReconstructionZlin
0.9800%
percentage effect of Meters
Submarket
Housing Submarkets: Effect of MetersZlin
9.20%
percentage effect of Private ownership
Submarket
Housing Submarkets: PrivateZlin
Figure 6.29: Housing Submarkets: Zlin
104
Conclusion
Our empirically oriented study utilized the hedonic theory of the analysis of the estate market. A crucial hedonic component of our study was the factor of location and hence the spatial modeling frameworks were described and applied in our to estimate the hedonic spatial models on the flat estates covering the entire area of the Czech republic. For empirical purposes, we had been extracting the data from the most frequently visited estate server. Assuming a certain representability level of the data of the underlying estates collected.
Before the modeling part, certain filtering and data preprocessing steps were taken to more deepen the representation of the data of the true underlying real estate market trends. For the estimation purposes itself, the multiple model approach was presented and utilized. We firstly estimated the multiple linear regression model (OLS), which provided a certain baseline level for the comparison with the spatially oriented models. The spatial models used in this study are the spatial lag, spatial error, and the geographically weighted regression model (GWR). As we have expected, the empirical results confirm that the spatial models provide considerably better modeling of real estate compared to the baseline OLS models.
The main focus of our study consists of two separate analyses. Firstly, the identification and representation of the ”grandiose-clusters” over the area of the entire Czech Republic for which purposes the spatial prediction kriging technique was utilized and the ’grandiose-clusters” were determined founding interesting spatial patterns within them. Secondly, following the theory of the housing submarkets, we determined and identified housing submarkets for every single region of the Czech Republic individually. Using the GWR models
105
Conclusion
combined with dimensionality reduction techniques and clustering algorithms, the housing submarkets were constructed and the contributions of key price determinants wherein were inspected. Utilizing this framework inspired by the study of (Kopczewska & Ćwiakowski 2021) the hypothesis of the spatially varying price determinants was not only proved but also the spatial patterns of the determinants were inferred.
In the current literature regarding the spatial modeling of the estate market, we frequently see fairly interesting applications of spatial techniques both from the field of statistics as well as machine learning. However, the has not been any consensus regarding the key price determinants and thus most of the studies are utilizing different independent variables. The main downside we frequently see in modern literature is the fact of omitting the time dimension. Which many studies consider as a huge room for improvement. On the other hand, both of our frameworks, which were described in this thesis, allow for the analysis over time. For example, if the housing data are collected at multiple points in time, each housing submarket can be constructed for each time points individually and the effect of time analyzed simultaneously. We believe that housing submarkets are the best approach to investigating the time effects as it is very unlikely to collect exactly the same particular estates at multiple points in time. The housing submarkets, however, represent the aggregated values of each estate wherein and thus are more suitable for time analysis. A similar idea holds for the ”grandiose-clusters” as they also represent the aggregated effect and thus are suitable for time analysis as well.
106
Anselin, L. (2013): Spatial econometrics: methods and models, volume 4.
Springer Science & Business Media.
Anselin, L. & N. Lozano-Gracia (2009): “Spatial hedonic models.” In
“Palgrave handbook of econometrics,” pp. 1213–1250. Springer.
Anselin, L. & S. J. Rey (2012): “Spatial econometrics in an age of cybergi-science.” International Journal of Geographical Information Science 26(12): pp. 2211–2226.
Bhattacharjee, A., E. Castro, T. Maiti, & J. Marques (2016): “En-dogenous spatial regression and delineation of submarkets: A new frame-work with application to housing markets.” Journal of Applied Econometrics 31(1): pp. 32–57.
Bitter, C., G. F. Mulligan, & S. Dallerba (2007): “Incorporating spa-tial variation in housing attribute prices: a comparison of geographically weighted regression and the spatial expansion method.” Journal of Geo-graphical Systems 9(1): pp. 7–27.
Bivand, R. S., E. J. Pebesma, V.Gomez-Rubio, & E. J. Pebesma (2013):
Applied spatial data analysis with R, volume 2. Springer.
Case, B., J. Clapp, R. Dubin, & M. Rodriguez (2004): “Modeling spatial and temporal house price patterns: A comparison of four models.” The Journal of Real Estate Finance and Economics 29(2): pp. 167–191.
Chien, Y.-M. C., S. Carver, & A. Comber (2020): “Using geographically weighted models to explore how crowdsourced landscape perceptions relate to landscape physical characteristics.” Landscape and Urban Planning 203. Chrostek, K., K. Kopczewskaet al. (2013): “Spatial prediction models for
real estate market analysis.” Ekonomia 35. 107
Bibliography BIBLIOGRAPHY Copiello, S. (2020): “Spatial dependence of housing values in northeastern
italy.” Cities 96.
Cupal, M. (2015): “Historical perspective of residential development and its impact on the current market prices of apartments on the czech real estate market.” Procedia Economics and Finance 26: pp. 144–151.
Dubin, R., K. Pace, & T. Thibodeau (1999): “Spatial autoregression tech-niques for real estate data.” Journal of Real Estate Literature 7(1): pp.
79–95.
Dubin, R. A. (1992): “Spatial autocorrelation and neighborhood quality.” Re-gional science and urban economics 22(3): pp. 433–452.
Dubin, R. A. (1998): “Spatial autocorrelation: a primer.” Journal of housing economics 7(4): pp. 304–327.
Elhorst, J. P. et al. (2014): Spatial econometrics: from cross-sectional data to spatial panels, volume 479. Springer.
Formánek, T. (2019): “Spatial econometric analysis with applications to re-gional macroeconomic dynamics.” Habilitation Thesis, Prague University of Economics and Business, Faculty of Informatics and Statistics, Department of Econometrics .
Gelfand, A. E., P. Diggle, P.Guttorp, & M. Fuentes(2010): Handbook of spatial statistics. CRC press.
Gillen, K., T. Thibodeau, & S. Wachter (2001): “Anisotropic autocorre-lation in house prices.” The Journal of Real Estate Finance and Economics 23(1): pp. 5–30.
Goldberger, A. S. (1962): “Best linear unbiased prediction in the generalized linear regression model.” Journal of the American Statistical Association 57(298): pp. 369–375.
Guo, J. & X. Qu (2019): “Spatial interactive effects on housing prices in shanghai and beijing.” Regional Science and Urban Economics 76: pp. 147–
160.
Helbich, M., W.Brunauer, E.Vaz, & P.Nijkamp(2014): “Spatial hetero-geneity in hedonic house price models: The case of austria.” Urban Studies 51(2): pp. 390–411.
108
Hrobař, P. & V.Holý(2020): “Spatial analysis of the flat market in prague.”
In International Conference on Mathematical Methods in Economics 2020 (MME 2020) pp. 193–199.
James, G., D.Witten, T.Hastie, & R.Tibshirani(2013): An introduction to statistical learning, volume 112. Springer.
Jauregui, J. (2012): “Principal component analysis with linear algebra.”
Philadelphia: Penn Arts & Sciences .
Kopczewska, K. & P. Ćwiakowski (2021): “Spatio-temporal stability of housing submarkets. tracking spatial location of clusters of geographically weighted regression estimates of price determinants.” Land Use Policy 103. LeSage, J. P. (2008): “An introduction to spatial econometrics.” Revue
d’économie industrielle (123): pp. 19–44.
Li, X., W. Y.Chen, & F. H. T. Cho (2020): “3-d spatial hedonic modelling:
Environmental impacts of polluted urban river in a high-rise apartment mar-ket.” Landscape and Urban Planning 203: p. 103883.
Lipán, M. (2016): “Spatial approaches to hedonic modelling of housing market:
Prague case.”Bachelor Thesis, Charles University, Faculty of Social Sciences .
Murphy, K. P. (2012): Machine learning: a probabilistic perspective. MIT press.
Nakamura, H. (2020): “Evaluating the value of an entrepreneurial city with a spatial hedonic approach: A case study of london.”Socio-Economic Planning Sciences 71: p. 100820.
Oliver, M. A. & R. Webster (2015): Basic steps in geostatistics: the vari-ogram and kriging. Springer.
Pace, R. K., R. Barry, & C. F. Sirmans(1998): “Spatial statistics and real estate.”The Journal of Real Estate Finance and Economics 17(1): pp. 5–13.
Páez, A., S. Farber, & D. Wheeler (2011): “A simulation-based study of geographically weighted regression as a method for investigating spatially varying relationships.”Environment and Planning A43(12): pp. 2992–3010.
109
Bibliography BIBLIOGRAPHY Pryce, G. (2013): “Housing submarkets and the lattice of substitution.”Urban
Studies 50(13): pp. 2682–2699.
Rosen, S. (1974): “Hedonic prices and implicit markets: product differentia-tion in pure competidifferentia-tion.” Journal of political economy 82(1): pp. 34–55.
Straszheim, M. (1974): “Hedonic estimation of housing market prices: A further comment.” The Review of Economics and Statistics pp. 404–406.
Sun, H., Y.Tu, & S.-M.Yu(2005): “A spatio-temporal autoregressive model for multi-unit residential market analysis.” The Journal of Real Estate Fi-nance and Economics 31(2): pp. 155–187.
Wooldridge, J. M. (2010): Econometric analysis of cross section and panel data. MIT press.
Yoo, E.-H. & P. C. Kyriakidis (2009): “Area-to-point kriging in spatial hedonic pricing models.” Journal of Geographical Systems 11(4): pp. 381–
406.
Zemcık, P. (2011): “Is there a real estate bubble in the czech republic?”
Forthcoming in the Czech Journal of Economics and Finance .
Zhang, L., J.Zhou, & E. C.-m. Hui(2020): “Which types of shopping malls affect housing prices? from the perspective of spatial accessibility.” Habitat International 96.
Zhou, Q., C. Wang, & S. Fang (2019): “Application of geographically weighted regression (gwr) in the analysis of the cause of haze pollution in china.” Atmospheric Pollution Research 10(3): pp. 835–846.
110