Urban modelling——the price of housing areas

1.Overview

1.1.Research object

We choose the urban area of Jiaxing as the research area and focus on the housing price, aiming to explain why the price varies around the city and how this phenomenon forms.

1.2.Data source

Necessary data is collected to identify problems and verify the feasibility of the model. It includes road network and POI. All the data is GIS-based, and most of the procedures should be suitable for mainstream GIS software.

2.Introduction

2.1.The key issue: counterintuitive housing prices

The urban spatial structure of Jiaxing is monocentric, which is similar to Beijing. But the assumption that the housing price of Jiaxing urban area should be the same has been proved to be wrong. We selected POI of housing areas and used IDW to visualize the level of housing price around the urban area (Figure 1). It’s obvious that Jiaxing’s high-priced housing area is away from the urban center and the center’s housing price is relatively low.

As Households’ utilities increasingly influenced by the absolute location of the house (Sheppard, 1997; Malpezzi, 2003). There exist three possible reasons for this phenomenon: 1. The real urban center is offset from the geographical center, which means the high-priced housing area is where the urban center locates. 2. The distribution of the important public facilities plays a decisive role, like school and hospital. 3. Urban expansion encourages the functional differentiation, which provides a better living environment outside the urban center. The three reasons form the basis of our hypothesis, and specific models are applied to verify them and uncover the phenomenon.

2.2.Methodology

Based on the hypothesis above, three specific models are incorporated to solve the issue. The first one is Central Place Model, which is used to identify hierarchical system of the urban area. The second one is Potential Model, which is used to evaluate every housing area on the basis of the transportation connection. The third one is Cellular Automata Model, which shows and simulates the evolution of the urban area to reveal the future development trend. We try to use ArcGIS to operate these models and verify their feasibility. Several potential errors are also expected to be detected in the process. 

3.Key steps

3.1.Hierarchical system identification——Central Place Model

Gravity model is a kind of model which embodies the interaction strength of multiple space objects. In order to build Central Place Model, we introduce Gravity model to help us identify and analyze the spatial form of urban polycentric structures.

Based on Huff’s model, three factors are chosen to be measured: population & economy, social activity, spatial accessibility. AHP is adopted to calculate their weights.

1. population & economy: Data comes from the statistical yearbook. To avoid the effect of administrative boundary, we extract the geometric center of the region and use IDW to get the distribution of population and GDP over the urban area.
2. social activity: Data comes from POI (provided by Amap.com). Give different weights to the selected aspects of POI and use kernel density estimation to process data.
3. spatial accessibility. Construct topological structure of the road network and use Spatial Syntax by means of DepthmapX software to analyze its integration degree. Then calculate the spatial accessibility (SSA) of every grid unit.

Combine the three factor evaluations with the field investigation and perform the spatial weighted superposition. The comprehensive result will highlight the urban centers. Compare the urban polycentric structures with the housing price distribution and judge whether those high-priced housing areas locate at the urban centers.

3.2.Value evaluation——Potential Model

There are several kinds of public facilities like hospital and school which can significantly lead to a rise in housing prices. As for the housing areas are relatively fixed under the influence of the urban planning, we eliminate the interference of other factors and assume that the housing prices largely depend on the influence of the essential public facilities.Therefore, we adapt Potential Model to describe the potential value of housing areas

Considering the characteristic of the research area, four factors including school, hospital, commercial zone and attraction are selected. Data comes from POI (provided by Amap.com) and AHP is adopted to calculate the weights.

The Potential model should follow the formula:

According to the formula, the distance between facilities and housing areas is also required. We use the length along the road network instead of euclidean distance to ensure the accuracy of data. Compare the final result with the housing price distribution and explore whether there exists relevance between the housing prices and the value.

3.3.Urban growth——Cellular Automata Model

Cellular automata (CA) are suitable for modelling dynamic spatial phenomena. We try to use CA to simulate the urban growth of Jiaxing and reveal the future development trend.

However, most of the CA models are raster-based, including the most typical one——SLEUTH model. Vector-based CA models are relatively more suitable for the city simulation. In 2016, STEVEN proposed a CA model based on cadastral blocks and developed a GIS-based software named iCity. In this research, we try to use this GIS-based CA model to simulate the process of the urban growth.

CA model can help us forecast the future development trend of the city, which tends to meet people’s positive expectation on the prospect of the housing area, leading to a rise in housing price. The hypothesis above will be proved to be true if those high-priced housing areas overlap the fast-growing areas.

4.Potential errors

4.1.Model distortion

In the process of identifying hierarchical system, we use Gravity Model to help to build Central Place Model. The population and GDP generated by IDW is also one of the essential factors. But the result can’t represent the actual distribution, especially in a relatively small scale. Considering the live-work commuting behavior, there will be a large distortion in the evaluation. And the distortion could reduce the credibility of the Central Place Model we built.

4.2.Homogenization

In the process of evaluating the value of housing areas, the influence of important public facilities is the decisive factor. However, all facilities in the same category are viewed as the same. Actually, facilities of different characteristic like scale and quality will show different level of influence degree and range. The homogenization in the evaluation could mislead the final result and reduce the accuracy.

References

[1]Huff D L, Haggerty J W. Determination of intra-urban retail trade areas. Revista Salud Uninorte, 1962, 28(3):411-418.

[2]李欣.基于引力势力范围的郑州城市多中心空间形态识别[J].地理学报,2022,77(06):1475-1489.

[3]Daniel, Suzana. A GIS-based irregular cellular automata model of land-use change. Environment and Planning B: Planning and Design 2007, volume 34, pages 708-724

[4]Melissa Boyle & Katherine Kiel (2001) A Survey of House Price Hedonic. Studies of the Impact of Environmental Externalities, Journal of Real Estate Literature, 9:2, 117-144, DOI: 10.1080/10835547.2001.12090098

[5]Marco Helbich, Wolfgang Brunauer. Spatial Heterogeneity in Hedonic House Price Models: The Case of Austria. Tinbergen Institute Discussion Paper. TI 2013-171/VIII

[6]John Geanakoplos, Robert Axtell. Getting at Systemic Risk via an Agent-Based Model of the Housing Market. American Economic Review: Papers & Proceedings 2012, 102(3): 53–58