Emerging Polycentricity in Beijing: Evidence from Housing Price Variations,2001

Abstract

This paper examines the emerging multiple centre urban spatial structure in Beijing using housing price variation as an indicator. A random sample of 3783 apartment units was used. These apartments were recent sales in 2001, 2003 and 2005. The dataset included transaction prices and main housing attributes gathered from the Beijing Construction Committee. A hedonic price model was calibrated to investigate the importance of the different urban centres to housing price variations. The results show that multiple urban centres (such as Tian’anmen, CBD, Zhongguancun and the Olympic Centre) explain more of the variations in housing price differences in the metropolitan space than any centre does alone. The findings also reveal changes in impacts from the individual centres in the study period. These outcomes confirm that Beijing is moving towards a polycentric urban form. The emerging multiple urban centres are key factors in understanding the spatial restructuring of Beijing, especially in modelling its emerging housing market.

1. Introduction

A remarkable characteristic in China’s metropolitan spatial restructuring is the co-existence of forces that encourage the development of a central business district and that push towards a spatial equalisation of new developments. Rapid economic development and restructuring in the 1990s led to the recognition that Chinese cities need CBDs in order to accommodate the emerging business activities and to organise urban space. Many cities spelt out clearly in their development plans a CBD for immediate construction. In fact, the turn of the century witnessed a rush towards establishing a CBD in many cities, commonly known as the ‘CBD fever’ (21st-century Economic Report, 2 April 2003). At the same time, local governments in metropolitan regions compete with each other for development opportunities, leading to projects that could be located in the emerging CBDs but appear in many non-CBD districts. These contradictory forces towards concentration and dispersion of businesses in metropolitan space are manifested in the city plans and raise empirical questions as to whether a monocentric or polycentric spatial pattern has emerged in large cities in transitional China where market forces have been released gradually.

The purpose of this paper is to address this empirical question using housing price variations and a hedonic model that counts in the influence of the various urban centres defined in Beijing’s Master Plan (2005). This methodology is novel as it takes advantage of the newly established housing market and the established theory in the study of housing price models to examine the impact of urban centres directly, rather than indirectly through population densities. Past studies using similar datasets reveal great variations of housing price distribution in Beijing (Han, 2004) and the negative relationship between distance to city centre and housing prices (Yang, 2001; Yu et al., 2008). However, no attempt has been made to evaluate the significance of the emerging urban centres in determining housing prices over time using a consistent model. In theory and empirical research, such a model is available in the literature, stemming from the bid–rent concept (Alonso, 1964) and the empirical findings that single centre influence on housing price modelling declined in the multiple centre urban environment of market economy countries (Yeates and Garner, 1971; Heikkila et al., 1989). Many researchers questioned the traditional CBD-oriented urban model and tested the polycentric urban model, which was found to be more representative of modern urban development in Western countries (Gordon and Wong, 1986; McDonald and McMillen, 1990; Waddell et al., 1993; Ottensmann et al., 2008). In Beijing, however, the merit of such a study is further enhanced by the fact that urban centres are different from those in theories. How does a traditional centre such as the Tian’anmen Square (or the Forbidden City in general), or a new centre such as the Olympic sports facilities, perform in models based on market operations? The fact that these centres are neither employment nor service centres, but centres of perception and/or reputation adds new excitement to the established theories and methodologies in the study of urban spatial structure and housing price.

The paper is organised into six sections. Following this introductory section, section 2 provides a literature review on the conceptualisation of housing prices and the spatial restructuring of cities in post-1978 China. Section 3 introduces the data and methods. Section 4 analyses the emerging multiple urban centres defined in the city’s latest Master Plan. Section 5 presents the empirical results. Conclusions are drawn in section 6.

2. Conceptualising Housing Price and Urban Spatial Transformation in China

2.1 Housing Price Distribution and Urban Spatial Structure

In a perfect market economy, housing price is the equilibrium point where the willingness to pay meets the willingness to sell. Rosen (1974) used a ‘bid function’ and an ‘offer function’ associated with a bundle of attributes to conceptualise housing price and its determinants. This idea has become the foundation of the hedonic price model, which has been popularly used in property valuation (Mok et al., 1995; Huh and Kwak, 1997; Fik et al., 2003).

Three groups of attributes are the key variables that represent Rosen’s bundle of attributes in determining housing prices. They are the structural, neighbourhood and location characteristics of housing. Measurement examples include the age, structure, size and orientation of the housing unit, crime, income, ethnic mix of the neighbourhood and distances to urban centres and amenities (Lusht, 1997). Empirical research attesting these variables provides numerous support for the idea that the three groups of variables are relevant in modelling housing price, although their effects differ according to given situations.

The location variables provide a direct link between housing price and the spatial structure of the city. This connection can be summarised by the distance-decay function, in which an inverse relationship between housing price and distance to the urban centre is portrayed (Alonso, 1964). In the established urban economics model, urban centres are employment and/or service centres which housing buyers commute to and from every day. Given a fixed level of income and utility, therefore, there is a trade-off between commuting and housing in costs, thus a monocentric distance-decay function. Empirical results from the past half-century confirm that such a function exists in cities in the US, Australia, Canada, England, Scotland and Japan (Han, 2004, pp. 1261–122).

As polycentricism has emerged and become one of the defining characteristics of today’s urban landscape (Kloosterman and Musterd, 2001), the power of distance in explaining housing price variations using a model with one urban centre has declined. This began happening in as early as the 1960s, when Yeates and Garner (1971) claimed that only 10 per cent of housing price variation, in sharp contrast with the 75 per cent in the early 20th century, was explained by the distance variable. This change was an outcome of the improvement of the transport system, such as the rapid development of the highway system, the wide use of automobiles and the obsolescence of mass transit systems in US cities (Han, 2004, p. 1262). Anas et al. (1998, pp. 1441–1442) found considerable success in land value and density models by incorporating multiple centres in US cities. In this paper, we compare single centre with multiple centre hedonic housing price models in assessing the emerging polycentric urban structure in Beijing.

2.2 Urban Spatial Transformation in China

The past 30 years have witnessed the profound spatial transformation of cities in China, from value-flat to market-oriented. Before 1978, Chinese cities were basically sites for administration and industrial production. The urban spatial structure was determined by functional compatibility of land use according to the planning principles learned from the former Soviet Union, combined with the principle of least-cost industrialisation used by indigenous leaders. Mixed land use between employment and housing in the form of Danwei (i.e. work-unit) compounds was a main characteristic of the urban landscape (Bjorklund, 1986). Shops and services were provided as outlets to deliver consumer goods in the planned economy, leading to minimal variations in physical and social spaces (Pannell and Ma, 1983; French and Hamilton, 1979). In other words, the metropolitan space in pre-reform China is relatively flat in terms of the mixture of land use and minimal variations in social space.

Within the flat metropolitan space, there were symbolic buildings, squares and governmental offices located in the central cities. Examples include the Tian’anmen Square in Beijing, People’s Square in Shanghai and People’s Congress Hall in Chongqing. Although the political and administrative functions of these centres were associated with a certain level of concentration of jobs, they were by no means comparable to the Western-style central business district (CBD). Because commerce and producer services are treated as unnecessary in the Marxist ideology, the Western-style CBD accommodating businesses and services hardly existed in Chinese cities before 1978 (Qin et al., 2003).

Since 1978, Chinese cities have increasingly assumed the role of economic growth engines and centres of services. This new role of cities has led to rapid changes in economic structures. In Beijing, for example, the tertiary sector has contributed to an increasing proportion of GDP, from 23.7 per cent in 1978 to 72.1 per cent in 2007 (Lin, 2003). A spatial ramification of this new urban economy was the emerging business clusters. CBDs were established in many cities—some as a consequence of the already strong business functions in the centre of the city, while others were results of poor planning in anticipation of the forthcoming service economies (Lin, 2003; Economic Daily, 21 October, 2003). A CBD fever was reported in the media (21st-century Economic Report, 2 April 2003), which was a testament to the economic restructuring but also the desire for revenue income of local government and developers. High-level service centres emerged, accommodating shops and producer services in the city core and major suburban centres (for example, see Wang and Jones, 2002, on Beijing; Han and Qin, 2009, on Shanghai). The Danwei compound was no longer significant, as Danwei itself relinquished most of its planning functions that were associated with the fading planned economy. Many Danwei compounds were redeveloped by commercial and housing projects.

Beyond the inner city, suburbanisation of population and industries were obvious happenings in many Chinese cities. Empirical research using population census data of 1964, 1982 and 1990 confirmed that, during the 1980s, a number of Chinese cities (for example, Beijing, Shanghai, Guangzhou, Dalian, Shenyang and Hangzhou) showed signs of suburbanisation in the 1980s (for example, see Wang and Zhou, 1999; Ning and Deng, 1996). In Beijing, the inner-city population declined by 80 000 in eight years from 1982 to 1990; the population density in the inner city dropped by 1000 people per square km during the same period (Zhou and Meng, 2000). In Guangzhou, the proportion of manufacturing firms in the inner city decreased from 38.37 per cent to 27.19 per cent in the period 1980–89; in nearby suburbs, manufacturing firms increased from 10.29 per cent to 15.29 per cent in the same period (Chen and Cai, 1996).

With the process of inner-city redevelopment and suburbanisation, there also occurred the differentiation of space in terms of economic functions, social diversification and urban built form. Special development zones, in the form of industrial and business parks, were established in the fringe areas (Deng and Huang, 2004), whilst in the inner city, producer service firms found offices for their business operations (Han and Qin, 2009). The social space of large Chinese cities witnessed the growth of immigrants’ villages, often associated with the existing ‘villages’ in the city (Gu and Shen, 2003), whilst high-end housing clusters emerged in prestigious locations (Feng, 2003). The urban built form was also diversified to include high-rise and high-density housing as well as low-density villas. These changes have fundamentally reshaped the flat metropolitan space in China.

2.3 The Land and Housing Market

One of the major forces that has shaped the inner-city redevelopment and suburbanisation in Chinese cities is the emerging land and housing market (Zhou and Meng, 2000; Han, 2000; Ning and Deng, 1996). Before 1987, urban land was allocated administratively rather than through market mechanisms. The Amendment of Constitution in 1988 established the legal foundation for a land market, making urban land use rights tradeable and transferable (Dowall, 1993). Urban land use reform facilitated the relocation of inner-city industries, because the factories found it profitable to sell their land and move to the suburbs. With the abolition of the subsidised housing allocation system in 1998, urban residents were directed to the commercial housing market for their housing needs. A housing geography has emerged as a result of the socioeconomic stratification of the population, the production of different housing types (for example, low-end social housing vs high-end villas) and the location of housing projects (Han, 2004).

Interestingly, the Chinese government acted as both facilitator and player in the emerging land and housing market. Han and Wang (2003) demonstrated the local government’s role in forging redevelopment projects involving state-owned-enterprises in Chongqing. Numerous writings report the role of government as developers in building industrial parks (for example, see Pereira, 2003, for the Suzhou–Singapore industrial park project). To many local governments, land was seen as a major source of revenue in economic development. The practice of this idea led to acceleration in land development so that the local government benefited in the sales. Some of the land parcels were devoted to large office and commercial projects, thus equalising the distribution of these functions and developments throughout the metropolitan space. Local government competition in land and housing markets represents a force that spread out new developments including CBDs and other sub-centres in Chinese cities.

2.4 Modelling Beijing’s Urban Spatial Structure

Population redistribution as a result of inner-city redevelopment and suburbanisation in Beijing led to the development of a monocentric city structure in the 1980s and an increased polycentricity in the 1990s. Wang and Zhou (1999) used data from the third (1982) and the fourth (1990) national population census to model the population densities in Beijing. The results indicated that a negative exponential model based on a monocentric urban form using Tian’anmen as the centre of the city explained 67.1 per cent of the population density variation in 1982, but only 64.9 per cent in 1990. The density gradient dropped over the years, indicating the progression of suburbanisation. Using data from the fifth national population census (2000), Feng (2003) found that the polycentric model using Tian’anmen, Huayuan Lu, Bali Zhuang, Tuanjie Hu and Heping Jie as multiple centres yielded stronger power in explaining population distribution than that derived from the monocentric model. Wu (1998) examined land use changes in Guangzhou and suggested that a polycentric structure had emerged with regard to land use pattern.

Housing price variation in the metropolitan space provides another measurement to study the spatial structure of cities. In a comparison between Beijing and Jakarta, Han (2004) reports great variations in housing price distribution along the various directions linking the suburbs to the city centre (i.e. Tian’anmen). These price variations were partially shaped by location factors especially in relation to the distance between the housing project and the existing and emerging urban centres. Using a linear model formulation and 226 housing apartment samples, Yang (2001) found that a 1 km increase in distance to an urban centre (Tian’anmen) reduced housing prices by 237.8 yuan per square metre. In 2008, Yu et al. used a semi-log model form to analyse 885 samples of housing projects collected from the website of a real estate firm,¹ reporting that housing prices decreased by 4.4 per cent when the distance to the urban centre increased by 1 km. To date, there has been no effort to model the changing impact of the emerging centres on housing price variation and, in turn, to determine whether or not a polycentric urban spatial structure has evolved.

3. Methods and Data

3.1 Hedonic Regression Model

A hedonic price theory was developed to disentangle the bundle of housing characteristics that contribute to housing prices (Rosen, 1974). Using this theoretical basis, hedonic regression analysis estimates the implicit prices of characteristics of a house based on housing sales transactions, although only the final sales price is actually observed (Malpezzi, 2003). The coefficients of the hedonic regression are regarded as the hedonic (implicit) prices of the characteristics.

In this study, the dependent variable, housing transaction price, and continuous independent variables were transformed into their logarithmic values. In the existing literature, researchers have built hedonic models in linear, Box–Cox and log forms. This study has no intention to assess the performance of these forms, but chooses to use the log form as it works better than others in terms of its goodness-of-fit. Moreover, the log form model allows the estimated coefficients to be interpreted as the elasticity between the independent variable and the housing transaction price.

The robustness of the models was examined by looking at their sensitivity to the omission of other significant explanatory variables. This procedure was designed to test the stability of such coefficients and to explore possible issues of multicollinearity. In modelling spatial data, multicollinearity tends to be a potential problem as independent variables are likely to be geographically dependent on each other. Therefore the variance inflation factor (VIF) was used to test the multicollinearity issue. Results show that multicollinearity did exist in the models but it was not a severe problem. The largest VIF values are about 5.

Another potential problem is heteroscedasticity for which residual analysis was undertaken. Results from the White test reveal heteroscedasticity problems in modelling all the three years. This suggests that ordinary least squares (OLS) could be statistically inefficient, or even misleading. The generalised least squares (GLS) regression method was therefore used to cope with the heteroscedasticity problem. In statistics, the GLS estimate consistently provides the best linear unbiased estimator (BLUE) for the variables.

The specification of the hedonic model is shown in equation (1)

\ln (pric e_{i}) = β_{0} + β_{1} \ln X_{1, i} + β_{2} \ln X_{2, i} + β_{3} \ln X_{3, i} + ε_{i}

(1)

where, $pric e_{i}$ = transaction price for house i; $X_{1, i}$ = a vector of structural attributes for house i; $X_{2, i}$ = a vector of neighbourhood attributes for house i; $X_{3, i}$ = a vector of location attributes for house i; $β_{0}$ , $β_{1}$ , $β_{2}$ , $β_{3}$ = empirically estimated coefficients; and $ε_{i}$ = an error term assumed to be independent across observations and identically distributed from a normal distribution with mean zero and constant variance.

3.2 Housing Prices Database

In order to identify the changes in the impacts of (sub-)centres on housing price, the scope of this study included housing transaction prices in Beijing in 2001, 2003 and 2005. The dependent variable (housing price) was obtained from a database of house transactions maintained by the Beijing Construction Committee. The whole database contains 32 620 records for 2001, 104 339 records for 2003, and 176 327 records for 2005. We obtained permission to extract 2000 transaction records randomly from each of the datasets (2001, 2003 and 2005). Data items include the transaction price, name and street address of the properties and important although limited information about the structural characteristics (for example, area and level of the property, but not number of bathrooms). No neighbourhood characteristics are reported in the sample dataset. As geographically stratified sampling was not possible, the data cannot be organised according to the proximity of properties to subway stations.²

Not all the records contained complete information. Some of them had no address information and thus could not be geo-coded. The Beijing GIS database was collected from Tucheng Company, which provides professional services in GIS mapping. The transaction records of affordable housing heavily subsidised by the government are not included, because their prices are far below market price. After data clearance, there remain 1013 housing transaction records for 2001, 1314 records for 2003 and 1456 records for 2005. During the study period, the mean value of the unit housing price increased from 5770 yuan in 2001 to 6,095 in 2003, and further to 6,398 in 2005. The majority of the samples are within the Fifth Ring Road except a notable cluster in the Tongzhou District, in eastern Beijing (Figure 1).

Figure 1.

Spatial distribution of the housing samples in Beijing, 2001–2005.

3.3 Description of Attributes

Table 1 reports the summary statistics of transaction prices, structural attributes, neighbourhood attributes and location attributes.

Table 1.

Descriptive statistics of variables

	Minimum	Maximum	Mean	S.D.
2001 (n = 1013)
Transaction price (RMB yuan)	99 951	7 128 451	803 364	775 926
Area (square metres)	40	553	119	45
Level	1	35	9	6
Distance to the nearest park (PARK) (metres)	277	18 089	2 001	1 409
Distance to the nearest first-tier schools (SCHOOL) (metres)	44	23 095	2 598	2 208
Distance to the nearest subway station (SUBWAY) (metres)	101	18 768	1 677	1 754
Distance to the nearest highway (HIGHWAY) (metres)	77	3 320	985	745
Distance to Tian’anmen (TIANANMEN) (metres)	1 929	37 936	12 449	6 328
Distance to Jianguomen CBD (CBD) (metres)	425	40 351	13 390	5 585
Distance to Zhongguancun (ZHONGGUANC) (metres)	337	35 554	12 586	9 857
Distance to Olympic Centre (OLYMPIC) (metres)	1 195	30 309	11 411	7 288
2003 (n = 1314)
Transaction price (RMB yuan)	132 967	6 224 378	649 733	513 217
Area (square metres)	22	364	107	43
Level	1	31	9	6
Distance to the nearest park (PARK) (metres)	253	18 369	22 44	2 096
Distance to the nearest first-tier schools (SCHOOL) (metres)	102	22 980	3 411	3 216
Distance to the nearest subway station (SUBWAY) (metres)	53	29 816	2 526	3 349
Distance to the nearest highway (HIGHWAY) (metres)	46	9 593	1 186	1 059
Distance to Tian’anmen (TIANANMEN) (metres)	1 874	42 976	12 283	6 770
Distance to Jianguomen CBD (CBD) (metres)	435	40185	12941	6760
Distance to Zhongguancun (ZHONGGUANC) (metres)	302	43849	14546	8487
Distance to Olympic Centre (OLYMPIC) (metres)	1 586	36 743	13 321	6 394
2005 (n = 1456)
Transaction price (RMB yuan)	106 456	5 285 016	734 743	522 841
Area (square metres)	30	377	110	39
Level	1	29	8	6
Distance to the nearest park (PARK) (metres)	153	36 674	2 724	4 208
Distance to the nearest first-tier schools (SCHOOL) (metres)	43	35 125	4 629	4 428
Distance to the nearest subway station (SUBWAY) (metres)	103	48 505	3 749	5 888
Distance to the nearest highway (HIGHWAY) (metres)	72	32 346	1 630	3 158
Distance to Tian’anmen (TIANANMEN) (metres)	1 923	67 662	13 108	8 528
Distance to Jianguomen CBD (CBD) (metres)	247	62 710	12 797	8 879
Distance to Zhongguancun (ZHONGGUANC) (metres)	149	72 620	17 099	9 186
Distance to Olympic Centre (OLYMPIC) (metres)	1 456	64 603	14 794	8 327

Structural attributes

In the hedonic model, these are extracted from the sales transaction database as well, and include area (gross construction area) and level of the unit. Shown in Table 1, the mean value of house size decreased from 119 square metres in 2001 to 107 square metres in 2003, returning to 110 in 2005; the sample units were found on various levels ranging from ground level to level 35 in high-rise apartment buildings, with a mean level value at 9 or 8. On the basis of existing research (Jim and Chen, 2007), area and level are expected to have positive effects on housing prices. As all the housing records in the database are pre-sale transactions, age and construction condition which are commonly used in other empirical studies are irrelevant for hedonic regression in this study.

Neighbourhood attributes

In this study, these include distance to the nearest park (DPK) and distance to the nearest first-tier school (DSC).³ The parks were extracted from the GIS database provided by Tucheng Inc. A green land tract with an area of more than 10 hectares is defined as a park for this study. There are 253 such tracts, containing all the parks registered in the Beijing Municipal Administration Centre of Parks. The shortest distance between a housing sample and its nearest park is taken as DPK. The distance is captured with the aid of ArcView 3.2 Extensions developed by Jeness Enterprises (Jeness, 2004). DPK is expected to have a negative impact on housing prices (Kong et al., 2007).

A list of the first-tier schools was compiled by the authors, using information posted on the official website of Beijing Municipal Commission of Education and some house-searching websites.⁴ The list contains 50 top primary schools and middle schools in Beijing. Distance to the nearest first-tier school (DSC) is measured in the same way as DPK. Although there is, to the authors’ knowledge, no study on the impact of good schools on housing prices in China, DSC is expected to have a negative effect on housing price based on empirical studies in other countries (for instance, see Downes and Zabel, 2002) and the fact that Chinese culture pays great attention to education.

Location attributes

These include distance to the nearest subway station (DSB), distance to the nearest highway (DHW) and distances to the four urban centres—i.e. Tian’anmen (DTAM), Jianguomen CBD (DCBD), Zhongguancun (DZGC) and the Olympic Centre (DOPC). Subway stations include the 102 stops on Lines 1, 2, 5 and 13, and on the Batong Line (Baliqiao–Tongzhou). Although Lines 13, 5 and the Batong Line were built in 2003, 2004 and 2007 respectively, their route plans were announced before 2001. Here it is expected that DSB had a negative relationship with housing prices in 2001, 2003 and 2005. Highways include all the ring roads and expressways in the urban area of Beijing. Houses with a shorter distance to a highway are expected to have higher prices. DSB and DHW are measured in the same way as DPK and DSC. As there are no standard definitions for the ‘exact’ central points of Tian’anmen, the CBD, Zhongguancun and the Olympic Centre, this study takes the most symbolic buildings/point as the central points. The centroid of Tian’anmen Square is defined for measuring DTAM; the China World Trade Centre is used to measure DCBD; the Administration Centre of Zhongguancun High-tech Park is used to measure DZGC; and the Olympic Centre Stadium is used to measure DOPC. The distribution of the parks, schools, subway stations, highways and four urban centres is shown in Figure 2.

Figure 2.

Spatial distribution of urban facilities and centres in Beijing.

4. The Emerging Urban Centres in Beijing

Multiple centres are main outcomes of the spatial restructuring process in Beijing. Among the three emerging centres, the CBD is at the Third Ring Road, whilst the other two centres are even farther from the Tian’anmen core.

The CBD concept appeared in the 1993 Master Plan. Nevertheless, little change was made until the year 2000 when the municipal government made the decision to develop a CBD in order to attract foreign direct investment and transnational corporations. In 2001, the CBD Administrative Commission was established and began to formulate a development plan. The initial site was a 4 square km block. Due to overwhelming responses from the real estate market, the municipal government decided to extend the CBD eastwards by an additional 3 square km.

By September 2009, the total number of enterprises in the CBD reached 15 000, including 130 of the world’s top 500 enterprises. In Beijing, 60 per cent of the foreign bank branches and 80 per cent of the foreign automobile finance enterprises were located in the CBD.⁵ High-end producer services in the CBD contained most of the leading international enterprises, such as McKinsey, Price Waterhouse Coopers, Gallup, Dow Jones, Ogilvy & Mather and CB Richard Ellis. Media organisations such as China Central TV Station, Beijing TV Station, Hong Kong Phoenix Station, The People’s Daily newspaper, Beijing Youth newspaper and Time Magazine were also clustered in the CBD, forming a prime office centre.

The Zhongguancun area is a concentration of Beijing’s high-tech and IT industries. Located at the north-west corner between the Third Ring Road and the Fourth Ring Road, its formation began in the 1980s when researchers and academics in the research institutions and universities in the area opened their own research, development and consulting businesses. Many spin-offs of the institutions also evolved in the area. For example, Lenovo was established in 1984. These new establishments brought concentrations of electronics, software and other IT activities.

In 1988, the central government approved a proposal from the Beijing municipal government to set up a New-tech Experimental Zone, which later evolved into Zhongguancun Science and Technology Park. The core is a 10 square km stretch along Zhongguancun Road, which was subsequently defined as the Zhongguancun Centre in the 2005 Master Plan. The area includes not only universities and research institutions, but also indigenous and foreign enterprises in the IT industry. Examples of the former include Lenovo (based in the Chinese Academy of Sciences), Tsinghua Ziguang (Tsinghua University) and the Fangzheng Group (Peking University); foreign examples include IBM, Hewlett-Packard, Dell, Microsoft, Lucent, Motorola, Intel, General Motor, Panasonic, Nokia and SUN. The Zhongguancun Centre has become not only the ‘silicon valley’ of China, but also a major employment centre, attracting thousands of IT workers to Beijing.

The Olympic Centre represents a very different node from the CBD and Zhongguancun Centre. Instead of employment centres, the Olympic Centre is a concentration of sports facilities and their associated sport and cultural industries. The formation of this centre originated from the 11th Asian Games held in Beijing in September 1990. A cluster of sports facilities with a full range of amenities were constructed, which became a popular location in Beijing’s housing market instantly. After winning the bid for the 2008 Olympic Games in July 2001, billions of dollars of investment were poured into the area for the construction of new buildings, parks and infrastructure. Several five-star hotels were established in the area; a forest park of 6.8 million square metres was added to the north of the area as well; subway lines 8, 10 and 13 made the area the most accessible point in Beijing. With more and more residents getting richer, the sport and cultural facilities available in the centre were valued. Property values have been lifted by market demand accordingly since 2001 (SouFun, 2010).

5. Empirical Results

5.1 Monocentric vs Polycentric Models

In order to compare the performance of monocentric models and polycentric models, multiple regression models are calibrated using transaction data for 2001, 2003 and 2005. All attribute variables are entered. The OLS method was used first. However, results from the White tests on the residuals indicated a severe heteroscedasticity problem in all the regressions. Thus the GLS method was employed. The p-values of the White test are reported in Tables 2, 3 and 4. The VIF values of the variables are calculated in order to investigate the presence of multicollinearity. The coefficients of the variables and adjusted R ² values are reported in the tables as well.

Table 2.

GLS regression statistics, 2001 (n = 1013)

Independent variables	Model 1 Tian’anmen	Model 2 CBD	Model 3 Zhongguancun	Model 4 Olympic Centre	Model 5 Multicentres	VIF (Model 5)
CONSTANT	12.899***	12.115***	10.890***	11.385***	14.029***
	(132.28)	(67.50)	(37.39)	(94.54)	(64.34)
AREA	1.305***	1.348***	1.368***	1.356***	1.283***	1.102
	(349.05)	(56.70)	(43.63)	(87.95)	(67.74)
LEVEL	0.064***	0.089***	0.121***	0.107***	0.033***	1.251
	(16.08)	(10.07)	(15.27)	(12.55)	(4.93)
PARK	−0.082***	−0.136***	−0.137***	−0.088***	−0.082***	1.347
	(−9.90)	(−10.06)	(8.68)	(−7.61)	(−6.97)
SCHOOL	0.044***	−0.030**	−0.025**	−0.078***	0.048***	1.841
	(27.74)	(−3.672)	(−2.67)	(−8.93)	(5.10)
SUBWAY	−0.070***	−0.079***	−0.139***	−0.099***	−0.039***	1.409
	(−11.39)	(−11.29)	(−11.66)	(11.67)	(−4.18)
HIGHWAY	−0.081***	−0.159***	−0.153***	−0.142***	−0.014*	1.794
	(−20.02)	(−13.95)	(−11.40)	(−14.11)	(−1.82)
TIANANMEN	−0.488***				−0.226***	5.147
	(−44.38)				(−9.82)
CBD		−0.263***			−0.251***	4.013
		(−14.12)			(−12.72)
ZHONGGUANC			−0.114***		−0.169***	2.425
			(−9.16)		(−16.77)
OLYMPIC				−0.194***	−0.022	2.736
				(−13.86)	(−1.32)
Adjusted R ²	0.98	0.96	0.90	0.93	0.99
White test(p-value)	0.98	0.55	0.71	0.54	0.91

Notes: *** denotes significant at 0.001; ** denotes significant at 0.01; * denotes significant at 0.05. T-values are in parentheses.

Table 3.

GLS regression statistics, 2003 (n = 1314)

Independent variables	Model 1 Tian’anmen	Model 2 CBD	Model 3 Zhongguancun	Model 4 Olympic Centre	Model 5 Multicentres	VIF
CONSTANT	12.126***	11.881***	11.807***	12.880***	14.550***
	(87.77)	(109.78)	(75.63)	(151.38)	(109.78)
AREA	1.209***	1.158***	1.103***	1.113***	1.100***	1.146
	(76.85)	(90.25)	(83.35)	(96.81)	(112.25)
LEVEL	0.064***	0.057***	0.087***	0.073***	0.029***	1.154
	(15.59)	(8.74)	(13.57)	(9.90)	(6.09)
PARK	−0.023*	−0.058***	−0.037**	−0.012*	0.016**	1.339
	(−2.47)	(−5.47)	(−3.98)	(−2.41)	(2.62)
SCHOOL	−0.016*	−0.101***	−0.086***	−0.133***	−0.041***	1.954
	(−2.21)	(−14.55)	(−9.16)	(−23.54)	(−5.29)
SUBWAY	−0.077***	−0.058***	−0.128***	−0.072***	−0.011	1.620
	(−13.72)	(−15.24)	(−15.05)	(−15.12)	(−1.47)
HIGHWAY	−0.050***	−0.057***	−0.083***	−0.075***	−0.011	1.297
	(−6.27)	(−8.84)	(−11.71)	(−11.53)	(−1.41)
TIANANMENE	−0.370***				−0.091***	4.136
	(−30.83)				(6.28)
CBD		−0.227***			−0.253***	3.651
		(−76.58)			(−18.71)
ZHONGGUANC			−0.152***		−0.186***	2.275
			(−15.16)		(20.52)
OLYMPIC				−0.297***	−0.125***	2.102
				(−42.62)	(−9.07)
Adjusted R ²	0.92	0.94	0.91	0.92	0.99
White test(p-value)	0.67	0.77	0.73	0.16	0.62

Notes: *** denotes significant at 0.001; ** denotes significant at 0.01; * denotes significant at 0.05. T-values are in parentheses.

Table 4.

GLS regression statistics, 2005 (n = 1465)

Independent variables	Model 1 Tian’anmen	Model 2 CBD	Model 3 Zhongguancun	Model 4 Olympic Centre	Model 5 Multicentres	VIF
CONSTANT	12.195***	11.791***	11.719***	12.413***	13.871***
	(104.58)	(93.70)	(103.26)	(92.31)	(142.61)
AREA	1.140***	1.134***	1.148***	1.135***	1.122***	1.047
	(85.93)	(68.54)	(177.65)	(82.20)	(121.55)
LEVEL	0.028***	0.030***	0.055***	0.042***	0.020***	1.100
	(5.28)	(4.72)	(18.48)	(116.36)	(4.76)
PARK	−0.051***	−0.080***	−0.075***	−0.053***	−0.019***	1.37
	(−11.17)	(−14.60)	(−24.42)	(−10.65)	(−3.58)	4
SCHOOL	−0.026**	−0.084***	−0.059***	−0.091***	−0.039***	1.78
	(−3.10)	(−11.36)	(−10.31)	(−12.66)	(−7.27)	8
SUBWAY	−0.078***	−0.061***	−0.079***	−0.035***	0.005	1.79
	(−13.77)	(−13.18)	(−15.46)	(−6.97)	(1.03)	6
HIGHWAY	−0.061***	−0.092***	−0.094***	−0.086***	−0.024**	1.55
	(−10.80)	(−12.01)	(−28.13)	(−11.47)	(−3.97)	0
TIANANMENE	−0.276***				−0.029*	4.980
	(−19.60)				(−1.99)
CBD		−0.149***			−0.184***	3.847
		(−22.20)			(−14.15)
ZHONGGUANC			−0.157***		−0.188***	2.377
			(−10.69)		(−14.74)
OLYMPIC				−0.256***	−0.148***	2.173
				(−19.06)	(−14.16)
Adjusted R ²	0.88	0.91	0.95	0.93	0.99
White test (p-value)	0.84	0.75	0.77	0.74	0.79

Notes: *** denotes significant at 0.001; ** denotes significant at 0.01; * denotes significant at 0.05. T-values are in parentheses.

All the polycentric models (model 5) that include distances to multiple centres perform better than all the monocentric models (models 1, 2, 3 and 4). In the GLS regression results, the Adjusted R ² values show that the polycentric models offer greater explanatory power to housing price variances. As shown in Table 2, the polycentric model (model 5) explains 99 per cent of the housing price variances in Beijing, while the most powerful monocentric model (model 1, Tian’anmen-centred) explains 98 per cent of the variance in the 2001 dataset. For the 2003 dataset, the polycentric model still explains 99 per cent of the variance, while the most powerful monocentric model (model 2, CBD-centred) explains 94 per cent (Table 3). Similarly, the polycentric model (99 per cent) is more powerful than the best monocentric model (95 per cent in model 4, Zhongguancun-centred) for the 2005 dataset (Table 4). In all the three datasets, the explanatory power of the polycentric models is higher than that of the most powerful monocentric models at the levels of 1.0 per cent, 5.3 per cent and 4.2 per cent respectively.

It is of note that the Tian’anmen-centred monocentric models often used in the existing literature on Beijing show a significant change in their explanatory powers over the years, largely due to the increasing diversity of the urban spatial structure in Beijing. The explanatory power of Tian’anmen-centred models has decreased dramatically over the years, from 98 per cent to 92 per cent and 88 per cent. It suggests that the traditional Tian’anmen centre had weakened its explanatory power in the hedonic models. There have been numerous sub-centres evolved in various locations in Beijing (for example, Shangdi, Wangjing, Dahongmen, Wukesong, Xizhimen). These sub-centres may have exerted impacts on the housing values nearby, thus distorting the explanatory powers of the monocentric models. It is also interesting to note that, despite the waning explanatory power of Tian’anmen and the CBD, the explanatory power of Zhongguancun has increased. A possible explanation is that the IT industries in Zhongguancun developed well over the years and thus created more employment opportunities. As housing price is a trade-off with commuting cost, the increase in jobs in a centre suggests a reduction of commuting for many house owners and thus an increase in housing prices.

The polycentric urban models also reveal changes in the preferences of house-buyers for other structural, neighbourhood and location attributes—namely, apartment size, storey, proximity to the nearest park, first-tier school, subway station and highway (see Figure 3). The AREA variable is always a decisive factor in determining housing price. In 2001, 2003 and 2005, with a 1 per cent increase in the gross construction area of a residence, the transaction price will increase 1.283 per cent, 1.100 per cent and 1.122 per cent respectively.

Figure 3.

Changes in house-buyers’ preferences in Beijing, measured by regression coefficients in multicentre models, 2001–2005.

5.2 Significance of the Urban Centres

The R ² values in the monocentric models (Tables 2 –4) show that the importance to housing price of individual centres did not remain static over time (see Figure 4). The importance of Tian’anmen centre declined, whilst that of the Zhongguancun centre increased. In 2001, the monocentric models using the Tian’anmen centre were the most powerful ones in explaining the housing price variations among the four single-centred models. In 2003, the most powerful monocentric model was the one using the CBD. In 2005, the most powerful monocentric model was the one using Zhongguancun.

Figure 4.

Changes in impact of the centres on housing price in Beijing, measured by regression coefficients in multicentre models, 2001–2005.

The impact of Tian’anmen on housing prices decreased remarkably from 0.226 in 2001 to 0.091 in 2003 and then 0.029 in 2005. This suggests that, in 2001, the price of a house locating 1 per cent nearer to Tian’anmen would increase by 0.226 per cent, while the increase shrank to 0.091 per cent in 2003 and 0.029 per cent in 2005 with the same location change. The impact of the CBD on housing prices increased a little from 0.251 to 0.253 from 2001 to 2003, dropping to 0.184 in 2005. The impact of Zhongguancun on housing prices increased from 0.169 in 2001 to 0.186 in 2003, and went on increasing to 0.188 in 2005. Meanwhile, the impact of the Olympic Centre changed from being statistically insignificant in 2001 to significant in 2003 and more so in 2005. In 2005, locating 1 per cent closer to Zhongguancun would have a 0.188 per cent increase in housing price, with a 0.184 per cent increase when locating closer to the CBD, a 0.148 per cent increase closer to the Olympic Centre and a 0.029 per cent increase closer to Tian’anmen.

The changes in the explanatory power and coefficients of the four centres may be attributable to the transition and economic transformation in Beijing. With the introduction of market mechanisms into what was once a planned economy, the impact of a symbolic centre such as Tian’anmen has weakened gradually. Meanwhile, producer services, high-tech firms and cultural and recreational activities have gained more weight in the city’s economy. Thus the accommodation of producer service functions by the CBD, high-tech functions by Zhongguancun and cultural and sports activities by the Olympic Centre have resulted in their transformation into powerful influences on the housing price variations across Beijing.

6. Summary and Conclusion

China’s urban transformation since 1978 has been shaped by multiple forces and has produced interesting spatial outcomes. This paper examines one of the spatial outcomes in the context of the concentration and dispersion of urban activities focusing on the formation of a polycentric urban structure. Based on a large sample of housing transaction records, hedonic models are calibrated to assess the power of the multiple centres in explaining housing price distributions. These multiple centres are defined in the 2005 Beijing Master Plan. They are not necessarily the traditional service centres or employment clusters, but a mixture of centres with (Zhongguancun and the CBD) and without (Tian’anmen and the Olympic Centre) employment concentrations. Beijing’s urban centres are large geographical areas that accommodate symbolic structures, business functions as well as services, and draw activities to them on city, regional, national and international scales.

Does this unique combination of urban centres work in concert with a polycentric hedonic housing price model? The empirical results show that the models using multiple urban centres—i.e. Tian’anmen, the CBD, Zhongguancun and the Olympic Centre—produce better results than those using any one of these centres alone. These outcomes confirm that Beijing is moving from a monocentric urban form to a polycentric one and that the housing market is responding to some extent to the ‘polycentric development’ idea in the 2005 Master Plan. The emerging multiple urban centres have become key factors in influencing housing price variations across the city.

Although the empirical data include only snapshots of the changes in a five-year period in Beijing—unique as China’s capital city—the changing importance of the centres is revealing. During the study period, the impact of the traditional Tian’anmen centre on housing prices weakened whilst that of the newly emerged centres, such as the Olympic Centre, increased. This finding links the spatial dynamics of housing prices to the city’s transition and functional development. In the context of Beijing’s transition from a planned economy to a (quasi-) market economy and from a manufacturing base to a cultural centre with a better quality of life, multiple forces such as rising mobility and decreasing accessibility, as well as the decentralisation of jobs, services and population, have shaped the spatial restructuring process. One of the outcomes is that Tian’anmen has become less dominant in Beijing’s spatial organisation, while the Olympic Centre has emerged rapidly. This suggests that an urban centre with emerging functions exerts increasing influence on the spatial pattern of housing prices, while an old centre with traditional functions loses its influence. The overall trend of this development is the increasing polycentricity in Beijing. Recognition of this trend towards polycentric development has important implications for public policy; it needs to be addressed in future research by planning practitioners and academics.

Footnotes

Acknowledgements

The authors wish to thank the three anonymous referees for their constructive comments. Irene Canmon Han helped to proofread the manuscript. All the remaining errors are the authors’ responsibility.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC 41001103), the Australian Research Council (ARC DP1094801) and the Project of “985” in China (SPAP, RUC).

Notes

References

Alonso

(1964) Location and Land Use: Toward a General Theory of Land Rent. Cambridge, MA: Harvard University Press.

Anas

Arnott

Small

K. A.

(1998) Urban spatial structure, Journal of Economic Literature, 36, pp. 1426–1464.

Beijing Almanac Editor Committee (2002) Beijing nianjian [Beijing Almanac]. Beijing: Beijing Almanac Press [in Chinese].

Bjorklund

E. M.

(1986) The danwei: socio-spatial characteristics of work units in China’s urban society, Economic Geography, 62, pp. 19–29.

Chen

Cai

(1996) Guangzhou jiaoquhua de jincheng yu dinglijizhi [The process and dynamic mechanism of the suburbanization of Guangzhou], Redai Dili [Tropical Geography], 16, pp. 122–129 [in Chinese].

Deng

Huang

(2004) Uneven land reform and urban sprawl in China: the case of Beijing, Progress in Planning, 61, pp. 211–236.

Dowall

D. E.

(1993) Establishing urban land markets in the People’s Republic of China, Journal of the American Planning Association, 59, pp. 182–192.

Downes

Zabel

(2002) The impact of school characteristics on housing price: Chicago 1987–1991, Journal of Urban Economics, 52, pp. 1–25.

Feng

(2003) Zhngguo chengshi neibu kongjian chonggou [The internal urban spatial restricting in China]. PhD thesis, Peking University [in Chinese].

10.

Fik

T. J.

Ling

D. C.

Mulligan

G. F.

(2003) Modeling spatial variation in housing prices: a variable interaction approach, Real Estate Economics, 31(4), pp. 623–646.

11.

French

Hamilton

(1979) The Socialist City. Chichester: Wiley.

12.

Gaubatz

(1999) China’s urban transformation: patterns and processes of morphological change in Beijing, Shanghai and Guangzhou, Urban Studies, 36, pp. 1495–1521.

13.

Gordon

Wong

H. L.

(1986) The distribution of population and employment in a polycentric city: the case of Los Angeles, Environment and Planning A, 19, pp. 161–173.

14.

Shen

(2003) Transformation of urban socio-spatial structure in socialist market economies: the case of Beijing, Habitat International, 27, pp. 107–122.

15.

Han

S. S.

(2000) Shanghai: between state and market in urban transformation, Urban Studies, 37(11), pp. 2091–2112.

16.

Han

S. S.

(2004) Spatial structure of residential property-value distribution in Beijing and Jakarta, Environment and Planning A, 36, pp. 1259–1283.

17.

Han

S. S.

Qin

(2009) The spatial distribution of producer services in Shanghai, Urban Studies, 46, pp. 877–896.

18.

Han

S. S.

Wang

(2003) The institutional structure of a property market in inland China: Chongqing, Urban Studies, 40, pp. 91–112.

19.

Heikkila

Gordon

Kim

J. I.

Peiser

R. B.

. (1989) Whatever happened to the CBD–distance gradient: land values in a polycentric city, Environment and Planning A, 21, pp. 221–232.

20.

Huh

Kwak

S. J.

(1997) The choice of functional form and variables in the hedonic price model in Seoul, Urban Studies, 34(7), pp. 989–998.

21.

Jenness

(2004) Nearest features extension for ArcView 3.x, v. 3.8a. Jenness Enterprises (http://www.jennessent.com/arcview/arcview_extensions.htm).

22.

Jim

C. Y.

Chen

W. Y.

(2007) Consumption preferences and environmental externalities: a hedonic analysis of the housing market in Guangzhou, Geoforum, 38, pp. 414–431.

23.

Kloosterman

R. C.

Musterd

(2001) The polycentric urban region: towards a research agenda, Urban Studies, 38, pp. 623–633.

24.

Kong

F. H.

Yin

Nakagoshi

(2007) Using GIS and landscape metrics in the hedonic price modeling of the amenity value of urban green space: a case study in Jinan City, China, Landscape and Urban Planning, 79, pp. 240–252.

25.

Lin

Y. Q.

(2003) Zhongguo xuyao zhemeduo CBD ma? [Does China need so many CBDs?], Financial Digest, 2003, pp. 48–50 [in Chinese].

26.

Lusht

K. M.

(1997) Real Estate Valuation: Principles and Applications. Chicago, IL: Irwin Professional Publishing.

27.

Malpezzi

(2003) Hedonic pricing models: a selective and applied review, in: O’Sullivan

Gibb

(Eds) Housing Economics and Public Policy, pp. 67–74. Oxford: Blackwell Science.

28.

McDonald

J. F.

McMillen

D. P.

(1990) Employment subcenters and land values in a polycentric urban area: the case of Chicago, Environment and Planning A, 22, pp. 1561–1574.

29.

Mok

H. M. K.

Chan

P. P. K.

Cho

Y. S.

(1995) A hedonic price model for private properties in Hong Kong, Journal of Real Estate Finance and Economics, 10, pp. 37–48.

30.

Ning

Y. M.

Deng

Y. C.

(1996) Shanghai jiaoquhua yanjiu [The study on suburbanisation in Shanghai] in: Li

S. M.

(Ed.) The Prospect of Regional Economic Development in China. Taipei: National Taiwan University Press [in Chinese].

31.

Ottensmann

J. H.

Payton

Man

(2008) Urban location and housing prices within a hedonic model, The Journal of Regional Analysis and Policy, 38, pp. 19–35.

32.

Pannell

C. W.

L. J. C.

(1983) China: The Geography of Development and Modernization. New York: Halsted Press.

33.

Pereira

A. A.

(2003) State Collaboration and Development Strategies in China: The Case of the China–Singapore Suzhou Industrial Park (1992–2002). New York: Routledge Curzon.

34.

Qin

Zhao

Zhou

(2003) Chongqing CBD de jieding [The definition of CBD in Chongqing], Urban Planning Review, 2003, pp. 84–89 [in Chinese].

35.

Rosen

(1974) Hedonic prices and implicit markets: product differentiation in pure competition, Journal of Political Economy, 82, pp. 34–55.

36.

SouFun (2010) Beijing fangdichan shichang fenxi baogao [Analysis report of real estate market in Beijing]. Unpublished report [in Chinese].

37.

Waddell

Berry

B. J. L.

Hoch

(1993) Housing price gradients: the intersection of space and built form, Geographical Analysis, 25, pp. 5–19.

38.

Wang

Zhou

Y. X.

(1999) Modelling urban population densities in Beijing 1982–90: suburbanisation and its causes, Urban Studies, 36, pp. 271–287.

39.

Wang

Jones

(2002) Retail structure of Beijing, Environment and Planning A, 34, pp. 1785–1808.

40.

(1998) Polycentric urban development and land-use change in a transitional economy: the case of Guangzhou, Environment and Planning A, 30(6), pp. 1077–1100.

41.

Yang

(2001) An application of the hedonic price model with uncertain attribute: the case of the People’s Republic of China, Property Management, 19, pp. 50–63.

42.

Yeates

M. H.

Garner

B. J.

(1971) The North American City. New York: Harper and Row.

43.

Zhen

S. Q.

Liu

H. Y.

(2008) Zhuzhai jiage tidu de kongjianchayi he yingxiang yinsu yanjiu: yi Beijing weili [The spatial variation and affecting factors of the housing price gradients: the case of Beijing], Economic Geography, 28, pp. 406–410 [in Chinese].

44.

Zhongguancun Administration Committee (2009) Zhongguancun kejiyuan nianjian [Almanac of Zhongguncun Science Park]. Beijing: Jinghua Press [in Chinese].

45.

Zhou

Meng

(2000) Beijing jiaoquhua jiqi zhengce yingxiang [The Suburbanisation in Beijing and Policy Implication]. Beijing: Science Press [in Chinese].

Emerging Polycentricity in Beijing: Evidence from Housing Price Variations,2001–05

Abstract

1. Introduction

2. Conceptualising Housing Price and Urban Spatial Transformation in China

2.1 Housing Price Distribution and Urban Spatial Structure

2.2 Urban Spatial Transformation in China

2.3 The Land and Housing Market

2.4 Modelling Beijing’s Urban Spatial Structure

3. Methods and Data

3.1 Hedonic Regression Model

3.2 Housing Prices Database

3.3 Description of Attributes

Structural attributes

Neighbourhood attributes

Location attributes

4. The Emerging Urban Centres in Beijing

5. Empirical Results

5.1 Monocentric vs Polycentric Models

5.2 Significance of the Urban Centres

6. Summary and Conclusion

Footnotes

Acknowledgements

Funding

Notes

References