Spatial mismatch in Beijing, China: Implications of job accessibility for Chinese low-wage workers

Habitat International 44 (2014) 202e210

Contents lists available at ScienceDirect

Habitat International journal homepage: www.elsevier.com/locate/habitatint

Spatial mismatch in Beijing, China: Implications of job accessibility for Chinese low-wage workers Yingling Fan a, *, Ryan Allen a, Tieshan Sun b a b

Hubert H. Humphrey School of Public Affairs, University of Minnesota, 301 19th Avenue South, Minneapolis, MN 55455, USA School of Government, Peking University, The Leo KoGuan Building, Beijing 100871, China

a r t i c l e i n f o

a b s t r a c t

Article history: Available online

The spatial mismatch literature has historically been U.S.-centric. This paper offers a theory of how spatial mismatch may have become a growing problem in China. The research uses Beijing, China as a case study to empirically examine the magnitude and geography of spatial mismatch across low-wage workforce segments. It finds a significant jobs-housing mismatch among low-wage workers in Beijing, particularly for blue-collar workers and workers without local hukou (registered permanent residence). The degree of spatial mismatch increases after accounting for worker access to transit. The results indicate that spatial mismatch in Beijing is more due to a greater dispersion of workforce residences than job locations and a central city-focused public transit system incapable of serving the dispersed lowwage workforce. The research findings suggest that Chinese cities should be more strategic in where they build affordable housing and where they make future transit investments. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Jobs-housing mismatch China Job access Transit Migrant worker Affordable housing

Introduction In 1968, Kain formulated the original spatial mismatch hypothesis that racial discrimination in the housing market, exacerbated by job suburbanization and limited transportation options, resulted in poor job accessibility and employment outcomes among African-Americans in U.S. cities (Kain, 1968). Over time, the spatial mismatch literature has evolved to look beyond African-Americans and to include other disadvantaged groups in the U.S. such as Latinos, low-income single mothers, welfare recipients, and immigrants (Fan, 2012). Despite this evolution, U.S. researchers have consistently dominated the field and much of the literature has retained its focus on the specific context of U.S. cities. The U.S. focus of the spatial mismatch literature is not surprising. U.S. cities are generally more spread out, have lower job and population densities, and more segregated land use patterns than cities in other countries (Korsu & Wenglenski, 2010). Marked by stronger car dependency, U.S. cities also have less developed public transit networks than cities in many other countries (Button, 1998; Kenworthy & Hu, 2002; Kenworthy & Laube, 1999). In U.S. cities (except the very few that have well-developed public transportation systems, such as New York City), an individual without

* Corresponding author. Tel.: þ1 (612) 626 2930. E-mail address: [email protected] (Y. Fan). http://dx.doi.org/10.1016/j.habitatint.2014.06.002 0197-3975/© 2014 Elsevier Ltd. All rights reserved.

access to a car almost necessarily faces poor job accessibility. U.S. cities are also notable for their relatively high levels of economic and racial segregation, providing only limited residential location choices for low-income minorities (Massey & Denton, 1993). Compared to U.S. cities, Chinese cities are presumed to be more resistant to spatial mismatch and job accessibility problems. Some suggest that the supposed hallmarks of Chinese cities, including higher densities with closer proximity between jobs and housing, extensive public transit networks, and lower levels of segregation (Kenworthy & Hu, 2002; Knaap & Zhao, 2009), reduce the risk of uneven access to job opportunities between population groups with different socio-economic status. However, the presumed differences in spatial mismatch between U.S. and Chinese cities may have been exaggerated. Many Chinese cities have experienced urban sprawl affecting the location of housing and, to a lesser extent, jobs over the past two decades (Jiang, Liu, Yuan, & Zhang, 2007; Li, 2010). Despite these recent changes to the location of residences and jobs, limited research has focused on the issue of spatial mismatch in China. Even less research looks into how the issue may have affected China's socio-economically disadvantaged groups in particular (Wang, Song, & Xu, 2011; Zhou, Wu, & Cheng, 2013). With these shortcomings in the literature in mind, this article offers theoretical explanations and empirical examinations of spatial mismatch in China. More specifically, we discuss underlying factors that may have resulted in poor job accessibility of

Y. Fan et al. / Habitat International 44 (2014) 202e210

disadvantaged workers in China. We then empirically examine the pattern and magnitude of spatial mismatch in Beijing, China, while accounting for worker access to transit. Our research findings indicate a significant jobs-housing mismatch among low-wage workers in Beijing, particularly for those workers without local hukou (registered permanent residence). The mismatch is largely due to a greater dispersion of residences than job locations and a central city-focused public transit system incapable of serving the dispersed low-wage workforce. Spatial mismatch in China Relatively few studies have focused on the presence of spatial mismatch in China, but the emerging research indicates that spatial mismatch is a growing problem. Some of the first studies to shed light on the prevalence of spatial mismatch in Chinese cities focused on the relationship between residential location and commuting time. Cervero and Day (2008) indicated an increase in commute time for Shanghai residents who had relocated to a suburban neighborhood. Similarly, Wang and Chai (2009) found that Beijing residents who lived in danwei-supplied housing had shorter commute times than those who lived in market housing. Li (2010) found that both residences and jobs in the city of Guangzhou were suburbanizing at a rapid rate, resulting in increases in commuting distances and times. Wang et al. (2011) argued that the spatial imbalance between employment and population has increased significantly over time, and has become especially evident in the inner city where the loss of residents continues and the concentration of employment persists. One cause of spatial mismatch in China is the erosion of the danwei system. Under the leadership of Mao Zedong and into the early years of the reform era in China, much of the housing available in urban areas was allocated by danwei, which represented the main spatial and social organizing principle of urban life. Each danwei was designed to be autonomous, incorporating employment, housing, basic social services, and other aspects of social and cultural life in China into a single compound (Bray, 2005). As such, danwei systematically coupled housing and employment for the vast majority of urban residents in China (Wang et al., 2011), making spatial mismatch virtually impossible. Beginning in the early 1990s, reforms in the system of urban housing pushed danwei to support a fledgling real estate market by selling housing to workers, spinning off the real estate development functions to development companies, and eventually abolishing the allocation of housing through danwei altogether by 1998 (Wang & Chai, 2009; Wang et al., 2011). As a result, increasing numbers of urban residents buy housing on the private market. As Chinese residents increasingly turn to private market housing, changes in land policy and municipal-level fiscal behaviors have dictated that most housing (including most affordable housing) has been built in suburban locations, while the most expensive private market housing is closer to the center of cities (Zhou et al., 2013). As China's land reform shifted land leasing and land management responsibilities to local governments from the central government, local governments also inherited the responsibility to fund more of their municipal services (Zhou & Logan, 1996). This shift in rights and responsibilities has increased the incentive for local governments to use revenue generated from property development to augment their municipal budgets (Zhang & Fang, 2004). As a result, municipal governments often encourage the “highest and best use” of urban land, which frequently translates into activities such as luxury apartments and commercial uses (Zhou & Ma, 2000). For example, initial municipally-led redevelopment programs in China focused on “upgrading” inner city housing and resulted in massive displacement of residents who could not afford

203

the luxury housing that typically replaced the older housing (Zhang & Fang, 2004; Zhou & Ma, 2000) and resettled on the periphery of urban areas as a result (Xu, Chan, & Yung, 2014). These kinds of redevelopment programs have led to an inflated housing market in China that is increasingly unaffordable for many residents. In fact, according to the China Statistical Yearbook of 2011, the annual average sale price of residential apartments more than doubled between 1998 and 2010, rising from 1854 to 4,725RMB/m2. Changes to the urban housing market have coincided with the arrival of a massive wave of rural migrants. China's household registration system (hukou) classifies individuals according to their places of presumed regular residence (suozaidi). Migrants who live outside their officially-registered areas are often denied stateprovided education, housing, social security, and economic opportunities (Chan & Zhang, 1999; Fan, 2002). Defined as temporary residents (Zanzhu renkou) who left their officially-registered areas for six months or more, there were 221.4 million migrants in China in 2010 (approximately 17% of the total population) (National Bureau of Statistics of China, 2011). Migrants are permitted to work in cities on the basis of temporary residence permits, but have much less access to government subsidies and, in several respects, they occupy a social and economic status similar to that of unauthorized immigrants in the U.S. (Fan, 2001). Because they are ineligible for state-subsidized housing and are legally barred from or cannot afford most other types of housing in urban areas (Wu, 2002), rural migrants frequently rent housing located in “villages in city” (ChengZhongCun in Chinese) (Song, Zenou, & Ding, 2008). Most commonly located on the periphery of cities around rural and urban transitional spaces, these urbanizing villages are comprised of hundreds of thousands of housing units developed by native villagers who have taken advantage of their land ownership to develop low-cost, poor quality housing units for migrant workers (Song et al., 2008). Although it is probable that migrant workers experience spatial mismatch to a high degree as they concentrate in the periphery of urban areas, to date none of the studies on spatial mismatch in China has assessed the extent to which spatial mismatch affects migrant workers differently from workers with local hukou. In contrast to the U.S. spatial mismatch research that has historically focused on low-income minority residents, little of the existing literature on spatial mismatch in China focuses on disadvantaged or impoverished residents. One exception is research by Zhou et al. (2013) that investigates the growing spatial mismatch problem for residents in low-income neighborhoods in Guangzhou. Zhou et al. (2013) define their population of interest for assessing spatial mismatch by the respondent's housing, classifying subsidized-rental housing residents as disadvantaged. They find increased incidence of spatial mismatch for residents driven by a variety of mechanisms that differ between low-income and relatively more affluent residents (Zhou et al., 2013). Xu et al. (2014) offer another perspective on spatial mismatch in China for disadvantaged residents, suggesting that low income residents displaced from inner city neighborhoods lack the means to acquire relatively expensive housing close to employment centers. As such, they reason that economically disadvantaged residents have a mobility problem rather than a segregation problem that causes spatial mismatch. In general, the existing research on spatial mismatch for disadvantaged populations in China focuses on economic factors, such as limited incomes as revealed by the housing where residents live, rather than social factors, such as migrant status. The likelihood that spatial mismatch is an increasing problem in China is not necessarily mitigated by public transit systems. Public transit systems in Chinese cities provide frequent services and extensive network coverage, but increased road congestion has made long-distance trips difficult to achieve by transit within

204

Y. Fan et al. / Habitat International 44 (2014) 202e210

a reasonable amount of time (Pendakur, 1992; Shen, 1997). In large cities such as Beijing and Shanghai, buses are running at an average speed of around 10 km/h, which is slower than the average speed of bicycles (Shen, 1997). These land use and transportation trends, coupled with two other important phenomena in Chinese citiesdhukou status-based discrimination and a shortage of affordable housingdmay have already put disadvantaged groups at a higher risk for poor job accessibility. Indeed, these mechanisms along with the fact that employment centers remain concentrated closer to central cities are included in the major set of findings of Xu et al. (2014) in their study of spatial mismatch in Shanghai, China. Study area and data sources Beijing, China is selected as the case study area to illustrate the geography and magnitude of spatial mismatch in Chinese cities largely because of data availability issues. Data on job distribution and residential segregation have been largely inaccessible in China for entities without government ties. The inaccessibility is largely attributed to government concerns over the political sensitiveness of the data, a tradition of data hoarding practices, and a lack of confidence in data reliability and accuracy (Zhou & Wang, 2012). With data access challenges, Beijing is the only city for which we were able to obtain relevant employment and population data. We mainly relied upon personal connections and extensive manual efforts of digitizing paper records to collect data from Beijing. Beijing is one of the two biggest cities in China. As the capital of the country, it is unique in its development history as well as tax and subsidy system. For example, the public transit system of the city is heavily subsidized by the government, which consequently reduces transit pricing and influences commuting costs and residential location choices. Although Beijing is admittedly not representative of Chinese Cities in scale or political status, we expect that Beijing could be representative of Chinese cities in a number of ways, including the kind of planning and land reform practices used in the city, the kind of socio-spatial polarization experienced during post-reform era, and the increasing urban sprawl observed (Deng & Huang, 2004; Gu & Shen, 2003; Song, Ding, & Knaap, 2006). Fig. 1 shows a map of the Beiiing region and boundaries of the region's 240 jiedaos. Jiedao (subdistrict in Chinese) has been the basic administrative unit in Chinese cities for decades and is the lowest geographic level reported in publicly accessible

government statistical reports (Gu, Wang, & Liu, 2005). As depicted in Fig. 1, the two city centers in Beijing include the old center which is the site of the Forbidden City and bounded by the Second Ring Road, and the new center on the eastern side of the old city, between the Second and Third Ring Roads (Zhou, 1998). The new center was designated as the official Central Business District (CBD) of Beijing in the 1992 Beijing Master Plan. It is close to the foreign-embassy district, has easy access to the international airport, and contains some of Beijing's most luxurious hotels and apartment buildings as well as most upscale shopping centers. We define the geometric center of the Forbidden City as the exact location of the old city center, and the geometric center of the official CBD boundary in the 1992 Beijing Master Plan as the exact location of the new city center. Analysis of spatial mismatch often requires locational data on jobs and workforce residences. In this paper, locational data on jobs comes from the 2001 National Census of Basic Units, which was obtained through personal connections. The data include information on the location, employment size, industry sector of each business in Beijing, and are aggregated to the jiedao level for the analyses in this paper. Data on workforce residences come from the 2000 Population Census, which is available in printed copies of the 2001 China Statistical Yearbook. Analysis and findings Identifying disadvantaged workers In the existing spatial mismatch literature, race, income, and/or wage are widely accepted indicators of socio-economic status among U.S. workers. In contrast, race may not be relevant to identify disadvantaged workers in China since China's race and ethnic composition is largely homogenous with 92% of the population being Han Chinese. It is also practically impossible to use individual income or wage data to indicate locations of disadvantaged workers in China because the National Bureau of Statistics in China does not publish population census data on workers' wage or income levels in a way that makes it possible to link workers' income/wage data with geographic identifiers that are lower than city. Past attempts at assessing spatial mismatch in China have identified disadvantaged populations largely through housing choice, with residents living in subsidized housing the focal point of some research (Xu et al., 2014; Zhou et al., 2013). The problem with this approach is two-fold. First, rural migrants are generally

Fig. 1. Study area map and characteristics.

Y. Fan et al. / Habitat International 44 (2014) 202e210

ineligible for subsidized housing so this approach does not capture their experiences. Second, the large amount of displacement of low-income residents from central city neighborhoods in recent times has likely created more demand for subsidized housing than the existing supply. Therefore, using residence in subsidized housing to classify disadvantaged residents may fail to identify a substantial proportion of residents that are disadvantaged because of their lack of income, inability to access subsidized housing, or both. To address these data issues, we use industry sector-based worker distribution data to classify workers into low- and highwage categories. All Beijing workers in sectors that have an average yearly wage lower than 17,000 RMBs are considered lowwage workers (the average wage for all workers in Beijing was 16,350 RMBs). These low-wage workers are further subdivided by sector type into blue-collar and pink-collar workers, and by hukou status into those who are officially registered in Beijing (local) and those who are not (migrant). In this research, the sector-based lowwage workers are considered as disadvantaged workers who are most vulnerable to the adverse effects of spatial mismatch. Highwage workers, regardless of their sector type and hukou status are not considered as disadvantaged workers because they tend to have high levels of economic mobility and may even be able to gain a local hukou through the government's “blue -stamp” program (Fan, 2001). Table 1 illustrates the sector-based classification approach used in this analysis. The China Statistical Yearbook of 2001 used the 1994 Industry Classification Scheme (ICS1994: GB/T 4754d1994) to classify industry sectors. As shown in Table 1, there are a total of 16 broad industry categories defined in the ICS1994 codes. Among the sectors with low wage levels, we further identify blue- and pinkcollar sectors. Blue-collar sectors typically provide jobs that perform manual labor and that involve manufacturing, mining, building and construction, mechanical work, maintenance, repair and operations maintenance, or technical installations. Pink collar sectors tend to provide personal-service oriented clerical jobs, such as customer interaction, entertainment, retail and outside sales, and the like (England, 1993). Of course, using an industrial classification scheme to distinguish low and high-wage workers has limitations. Given the potential heterogeneity of wages within a particular industry (Bibb & Form, 1977), the distinctions we draw between low and high-wage earners should be viewed cautiously. Fig. 2 shows percentage distributions of Beijing workers by the sector-based worker classification. As expected, workers with local hukou have a much higher percentage of employment in high-wage

205

Fig. 2. Percentage distributions of workers by sector-based wage levels and sector type.

sectors (35%) than those without local hukou (10%). It is also worth noting that almost half of migrant workers (48%) work in low-wage pink-collar sectors. Examining the geography of spatial mismatch Traditionally, researchers examine the pattern of spatial mismatch using straightforward comparisons between where disadvantaged workers live and where they work. Similar to this literature, we limit our analysis to low-wage blue- and pink-collar workers. We reason that high-wage workers have the ability to choose the location of their residences without affordability being the primary consideration (Xu et al., 2014). Therefore, any spatial mismatch that occurs for high-wage workers is probably the result of a choice about where to live relative to employment opportunities. Further, high-wage workers are more likely to be able to afford a commute in comparison to disadvantaged workers. In addition to comparing spatial distribution patterns of actual jobs to spatial distribution patterns of disadvantaged worker residences, we improve on the typical methodology used in the spatial mismatch literature by also comparing patterns of transit accessible jobs to patterns of disadvantaged worker residences. We believe that the spatial imbalance between transit accessible jobs and worker residences is a more accurate indicator of spatial mismatch for disadvantaged workers in particular since it accounts for the availability of transit. Job accessibility of disadvantaged workers can be largely independent of the geography of

Table 1 Sector-based worker classifications. Sector description

A Farming, Forestry, Animal Husbandry and Fishery B Mining and Quarrying C Manufacturing D Production and Supply of Electricity Gas and Water E Construction F Geological Prospecting and Water Conservancy G Transport, Storage, Post & Telecommunication Services H Wholesale and Retail Trade & Catering Services I Finance and Insurance J Real Estate Trade K Social Services L Health Care, Sports & Social Welfare M Education, Culture and Arts, Radio, Film and Television N Scientific Research and Polytechnic Services O Government Agencies, Party Agencies and Social Organizations P Others

Average annual wage in Beijing, 2000 (RMB)

Low-wage

11,303 10,821 14,423 19,807 12,899 16,607 18,558 14,396 23,789 18,836 16,549 21,717 18,265 20,186 18,106 19,409

✓ ✓ ✓

Blue- collar

High-wage Pink- collar

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

206

Y. Fan et al. / Habitat International 44 (2014) 202e210

spatial mismatch between actual jobs and worker residences. For example, a balanced jobs and worker distribution in a sprawling, low-density metropolitan region can lead to poor job accessibility for disadvantaged workers, especially if the region has congested travel conditions and a transit system that fails to serve a wide range of locations. Conversely, job-housing imbalance does not necessarily produce poor job accessibility if a region has freeflowing travel conditions and a high-quality public transit system that allow disadvantaged workers to reach jobs far from their homes in a reasonable amount of time and at a reasonable cost. Thus, examining spatial patterns of transit accessible jobs and comparing the patterns to those of worker residences may provide more meaningful insights into the implications of spatial mismatch on job accessibility than examining spatial patterns of actual job locations. This analysis applies a cumulative opportunity approach to calculate the amount of transit accessible jobs at the jiedao level. That is, the calculation counts the number of jobs reachable by public transit from each jiedao centroid within a predetermined transit travel time. In this paper, our calculation involves the following assumptions and approximations. First, if a jiedao has its centroid within a half mile from any transit stops, the jiedao is considered as transit-served. Second, the transit travel time threshold for workers to reach a job is 60 min. The 60-min travel time includes bus/rail riding time, walking time from the origin block centroid to the starting transit stop and from the ending transit stop to the destination block centroid, and waiting and walking time associated with up to two transfers. Third, walking time is computed using an average walking speed of 3.4 miles/hr (1.5 m/sec). Fourth, a maximum of two transfers is allowed during the 60-min travel time. Additional transfers are not allowed due to limited geocomputing capacity and the fact that in 2000e2001 Beijing had a limited subway system with only two operating lines (an east-west line and a loop line). Each transfer is assumed to be associated with transfer-related walking time and 5-min transferrelated waiting time. Fifth, average peak-hour transit speeds are used for calculating bus/rail riding time. Specifically, urban bus services have an average peak-hour speed of 12 km/h, while suburban buses and subways have average peak-hour speeds of 15 and 35 km/h, respectively (Sun, Wang, & Wang, 2003). Maps in Fig. 3 show distribution patterns of worker residences, jobs, and transit accessible jobs for low-wage workers in Beijing, China. Distributions of transit accessible jobs follow concentric patterns almost perfectly for both low-wage blue and pink-collar jobs, despite non-concentric distributions of actual low-wage jobs and low-wage worker residences. This indicates that Beijing's transit system is heavily focused on the central cities and not oriented towards actual low-wage job or low-wage worker residential distributions. When compared across workforce segments, lowwage blue-collar worker residences and jobs are more spatially dispersed in the Beijing region than low-wage pink-collar worker residences and jobs. Regardless of blue- or pink-collar, local lowwage workers (workers with local hukou) have more dispersed residential patterns than migrant low-wage workers (workers without local hukou). When comparing job distribution patterns to worker residence distribution patterns, the distribution of jobs appears to have a better match to the worker residence distribution than the distribution of transit accessible jobs. Despite the better match, there are important differences between job distribution patterns and distribution patterns of local and migrant worker residences. Neither local, blue-collar workers nor migrant, blue-collar workers have a residential pattern well-matched with blue-collar jobs. Local, bluecollar workers are widely scattered across the Beijing region with no evident concentration. Migrant, blue-collar worker residences

appear in concentrated patches near the urban-rural fringe (in areas between the 4th and 5th ring roads). In contrast, blue-collar jobs tend to cluster at different urban-rural fringe areas (areas near the 5th and 6th ring roads). Likewise, neither local, pink-collar worker residences nor migrant, pink-collar worker residences have a residential pattern well-matched with pink-collar jobs. Both local and migrant pink-collar worker residences are less centralized and less concentrated than pink-collar jobs. Notably, migrant, pinkcollar workers have a residential pattern similar to migrant, bluecollar workersdthey tend to cluster in concentrated patches in areas between the fourth and fifth ring roads. To further understand how low-wage jobs and low-wage worker residences in Beijing are oriented or disoriented toward each other, Fig. 4 illustrates percentages of low-wage jobs and low-wage worker residences that were located within every 4-km intervals from city centers. Fig. 4 has two sub-graphs: one describing blue-collar jobs/workers and the other describing pinkcollar jobs/workers. If there is a perfect match between low-wage jobs and low-wage workers in the Beijing region, the workers line and the jobs line in the sub-graphs would coincide at every interval. Any gap between the jobs line and the workers line in a sub-graph would suggest spatial mismatch between jobs and workers. Note that spatial mismatch can occur between jobs and workers located at a given distance of city centers. Although a perfect spatial match between jobs and workers would result in coincided jobs and workers lines in Fig. 4, coincided lines do not necessarily mean a perfect match. To give a specific example, if all jobs located in the 4e8 km ring are at north (or east) while all housing within the same ring is located at south (or west), then there is a spatial mismatch even if the number of housing and jobs within the ring is equal. Together, Figs. 3 and 4 offer a complementary view of the geography of spatial mismatch in the Beijing region. As shown in Fig. 4, a jobs-housing mismatch exists among lowwage workers in Beijing, which is consistent with findings in Fig. 3. Specifically, the gaps between the workers and jobs line in the bluecollar distribution graph of Fig. 4 are evident between the 0e4, 4e8, 8e12, and 12e16 intervals, and the gaps between the workers and jobs lines in the pink-collar distribution graph are evident between the 0e4 and 12e16 intervals. The differences in the geography of spatial mismatch between blue- and pink-collar workers are due to the distinctive job and residential location patterns between workforce segments. As shown in Fig. 4, jobshousing mismatches among pink-collar workers in Beijing are largely due to greater percentages of jobs than workers in central areas within 4 km from city centers and greater percentages of workers than jobs in fringe areas beyond 12 km from city centers. Jobs-housing mismatch among blue-collar workers is largely due to greater percentages of jobs than workers in areas within 12 km from city centers and greater percentages of workers than jobs in fringe areas between 12 and 16 km as well as beyond 28 km from city centers. When looking at migrant workers and local workers in Beijing separately, it is important to note that the gaps between the migrant workers line and the jobs line are much more evident than those between the local workers line and the jobs linedindicating greater levels of spatial mismatch among migrant workers than local workers. This is largely because of migrant workers' limited presence in central areas and their large presence at 12e16 km interval from city centers. To be specific, 5% of migrant, low-wage blue-collar workers and 13% of migrant, low-wage pink-collar workers in Beijing lived between 0 and 4 km from city centers, compared to 9% of local, low-wage blue-collar workers and 19% of local, low-wage pink-collar workers. Twenty-five percent of migrant, low-wage blue-collar workers and 21% of migrant, low-

Y. Fan et al. / Habitat International 44 (2014) 202e210

207

Fig. 3. Distributions of low-wage worker residences, jobs, and transit accessible jobs. Note: The 5th and 6th ring roads are shown in this figure for geographic reference only. The 5th ring road was completed in 2003 and 6th ring road in 2009dboth after the data years (2000e2001) used in the paper.

208

Y. Fan et al. / Habitat International 44 (2014) 202e210

Fig. 4. Percentage distributions of worker residences, jobs, and transit accessible jobs by distance to the nearest city center in Beijing.

wage pink-collar workers in Beijing lived in urban fringe areas between 12 and 16 km from city centers, compared to 10% of local, low-wage blue-collar workers and 12% of local, low-wage pinkcollar workers. This unique residential concentration of migrant workers at the 12e16 km interval from city center shown in Fig. 4 is consistent with the finding in Fig. 3 that migrant workers appear to concentrate between the 4th and 5th ring roads. Another consistent finding between Figs. 3 and 4 is that the level of spatial mismatch appears to be lower between actual jobs and worker residences than between transit accessible jobs and worker residences. As shown in Fig. 4, the shape of the job distribution line follows the shapes of the worker distribution lines (especially the local worker distribution line) more closely than the shape of the line representing average job accessibility by transit.

We measure the magnitude of spatial mismatch using three indices of dissimilarity: the traditional Dissimilarity Index (D), the General Dissimilarity Index (GD), and the dissimilarity index based upon job access by transit (DTransit). D and GD were originally developed to measure the extent of housing segregation in space between members of different racial and ethnic groups within a given metropolitan area (Duncan & Duncan, 1955). The indices have been widely adopted by spatial mismatch researchers to describe the imbalance between residential and employment distribution (Horner & Marion, 2009; Martin, 2004; Stoll, 2006). The DTransit index is original, specifically developed in this paper for measuring spatial mismatch of disadvantaged workers in the Chinese context. The formulations of the three dissimilarity indices are as follows:

(1)

where, wi and ei represent respectively the number of workers and the number of jobs (employment) located in areal unit i of the study region; Pn Pn i¼1 wi and i¼1 ei represent worker and employment totals in the study region.

The General Dissimilarity IndexðGDÞ:  n   1 X P cwi  P cei  GD ¼ * n  n cw 2 ce  i¼1

i¼1

i

i¼1

i

n X

wj

d2 j¼1 ij

þ1

; cei ¼

n X

ej

j¼1

d2ij þ 1

(2)

(3)

where, cwi and cei represent respectively the composite weighted counts of workers and jobs (employment) located in areal unit i and its interacting units. As illustrated in Equation (3), all areal units in the study region are considered as interacting with each other based upon a distance decay function (1/d2þ1). dij represents the aerial distance between the centroid of areal unit i and the centroid of areal unit j. The Dissimilarity Index based upon Job Access by Transit (DTransit):

 n   1 X P wi  P cai  DTransit ¼ * n  2 i¼1  ni¼1 wi i¼1 cai

Quantifying the magnitude of spatial mismatch

 n   1 X P wi  P ei  The Dissimilarity IndexðDÞ: D ¼ * n n   2 i¼1 i¼1 wi i¼1 ei

cwi ¼

cai ¼

n X

  ej f tij

(4)

(5)

j¼1

  f tij ¼



1; if tij  60 mins 0; if tij > 60 mins

(6)

where, cai represents the composite counts of jobs (employment) accessible within 60 min of transit travel from areal unit i. As illustrated in Equation (5), whether jobs are considered as accessible is dependent upon based upon a travel time threshold function f(tij). tij represents the travel time by transit between the centroid of areal unit i and the centroid of areal unit j. All indices range between 0 (perfect balance) and 1 (perfect imbalance). Out of the three indices, DTransit has considerable advantages when measuring spatial mismatch in the context of Chinese cities. D treats areal unit boundaries as actual boundaries separating groups across unit, and thus boundaries inhibit interactions among groups. To take an extreme example, suppose that all jobs are located in a particular areal unit of the study region while all workers reside in a different areal unit. Whether these two areal units are 1 km apart or 20 km apart will not influence the D index (Stoll, 2006) and in both instances D will be equal to 1. GD is a more advanced index than D. It captures the potential spatial interaction between areal units by using distance-decayed composite counts and assuming that the spatial interaction between two areal units decreases in a continuous fashion as distance increases (Wong, 2005). This assumption on continuous spatial interactions is

Y. Fan et al. / Habitat International 44 (2014) 202e210

problematic for the analyses in this paper for two reasons. First, the majority of disadvantaged workers in China probably depend on public transit systems for their work commutes. For these workers, only jobs accessible via transit are relevant since all other jobs are, by definition, inaccessible. The assumption of continuous spatial interactions in GD means that all areal units in the region have some level of spatial interactions, which is unrealistic for disadvantaged workers in China given the majority of them do not have access to private vehicles. Second, the ability of workers in areal unit i to interact with jobs in areal unit j strongly depends on the cost of travel. The quality of transit services is rarely universal across locations. When workers have to rely on public transit to travel, time needed to travel between areal units i and j using transit is a much better indicator of the cost of travel than the linear distance between i and j. Fig. 5 illustrates values of these three indices by workforce segment. According to the DTransit values, jobs-housing mismatch in Beijing is much more pronounced among low-wage blue-collar workers (DTransit values range between 0.56 and 0.61) than lowwage pink-collar workers (DTransit values range between 0.33 and 0.43). When separating low-wage workers into local vs. migrant workers, the levels of jobs-housing mismatch are higher among migrant workers than local workersd0.61 compared to 0.56 among blue-collar workers and 0.43 compared to 0.33 among pink-collar workers. Values of dissimilarity indices change significantly by measurement methods. The values of D (the traditional dissimilarity index) and GD are considerably lower than DTransit for blue and pink-collar workers, regardless of hukou status. These findings highlight the importance of comparing methods and selecting methods based upon the research contexts when it comes to measuring the magnitude of spatial mismatch. Discussion and conclusions This research, although exploratory in nature and admittedly using decade-old data, is one of the few studies to date that empirically examines the spatial mismatch issue among low-wage workers in China. We find that low-wage workers in Beijing are subject to significant jobs-housing mismatch. The mismatch in general is due to a greater dispersion of workforce residences than job locations and a central city-focused public transit system incapable of effectively serving the dispersed low-wage workforce. Among low-wage workers, the pattern and magnitude of spatial mismatch are different for blue- and pink-collar workers and for local and migrant workers. There are greater levels of spatial mismatch among blue-collar workers than pink-collar workers. This is largely because pink-collar jobs and workers are relatively

209

more centralized than blue-collar jobs and workers. There are also greater levels of spatial mismatch among migrant workers than local workers. The high level of spatial mismatch among migrant workers is largely because of migrant workers' limited presence in central areas and their strikingly large presence at specific urban fringe areas (i.e., areas in between the 4th and 5th ring roadsdabout the 12e16 km interval from city center). This finding makes sense when considering the ongoing trend of luxury housing construction near the city center of Beijing (Zhang & Fang, 2004) and the large number of migrant workers living on the periphery of cities. Findings from the Beijing case study have important policy implications. First, the incorporation of job accessibility by transit in illustrating the spatial mismatch geography and in examining the spatial mismatch magnitude shows that Beijing has a public transit system that offers highly centralized transit services that are not oriented towards actual low-wage job or worker distribution. This indicates that the traditional transit improvements with a central city focus in Beijing have played a very limited role in improving job accessibility and mitigating jobs-housing mismatch among disadvantaged workers in the region. Second, our findings suggest a pessimistic outlook of Beijing's current affordable housing program. According to project-level affordable housing data published by the Construction Commission of Beijing Municipality, as of 2010, over 70% of affordable housing units in Beijing were located beyond 12 km from city centers, resulting in the residents living far away from likely job opportunities as well as good transit services. Looking forward, Chinese cities should carefully consider the location of the new affordable housing projects to ensure that these projects not only provide affordable housing but also improve low-income people's transportation mobility and access to job opportunities. In addition, according to China's affordable housing policy, only households with local hukou (registered permanent residence) have access to government supplied affordable housing. Efforts are needed to reform the affordable housing program so that it benefits the local poor as well as the large migrant population who suffer from greater levels of jobs-housing mismatch. It is worth noting that affordable housing production alone is unlikely to be sufficient for mitigating spatial mismatch. In Chinese cities, physical and market conditions may not allow affordable housing projects to be built in close proximity to urban centers since land in and near urban centers is often too expensive or has been developed to its full capacity. Further, the current incentives in place for local government officials to maximize their tax base by allowing real estate developers to construct luxury housing and commercial properties on land closer to the city center makes it

Fig. 5. Values of dissimilarity indices by workforce segments.

210

Y. Fan et al. / Habitat International 44 (2014) 202e210

unlikely that centralized affordable housing projects will be built without concerted action on the part of China's central government. Chinese local governments need more innovative and integrated programs to address the spatial mismatch issue. For example, local governments may initiate job creation programs that focus on economic and community development in urban fringe areas, especially in the 12e16 km band area from city centers where lowwage migrant workers are concentrated. Finally, the spatial mismatch phenomenon in Chinese cities merits additional research and discussion. Although this research helps to better understand the geography and magnitude of the phenomenon, many questions remain when it comes to spatial mismatch in China. For example, researchers may conduct more indepth studies that look at the root causes of the spatial mismatch issue in Chinese cities. What are the critical factors and processes that led to increasing spatial mismatch in Chinese cities? Are there roadblocks in China's city governance and public finance systems that make certain spatial mismatch mitigation strategies politically or fiscally infeasible? In addition, how do the magnitude and geography of spatial mismatch in China compare to those typically found in the U.S.? To what extent are U.S. policy solutions for addressing spatial mismatch relevant for Chinese cities? Answering these questions will help urban planners and policy makers in China devise solutions to the spatial mismatch problems currently experienced in Beijing and potentially other Chinese cities. Acknowledgment The authors thank the Peking University-Lincoln Institute Center for Urban Development and Land Policy for providing funding assistance that made this research possible. References Bibb, R., & Form, W. H. (1977). The effects of industrial, occupational, and sex stratification on wages in blue-collar markets. Social Forces, 55(4), 974e996. Bray, D. (2005). Social space and governance in urban China: The danwei system from origins to reform. Stanford University Press. Button, K. (1998). The good, the bad and the forgettabledor lessons the US can learn from European transport policy. Journal of Transport Geography, 6(4), 285e294. Cervero, R., & Day, J. (2008). Suburbanization and transit-oriented development in China. Transport Policy, 15(5), 315e323. Chan, K. W., & Zhang, L. (1999). The hukou system and rural-urban migration in China: processes and changes. The China Quarterly, 160, 818e855. Deng, F. F., & Huang, Y. (2004). Uneven land reform and urban sprawl in China: the case of Beijing. Progress in Planning, 61(3), 211e236. Duncan, O. D., & Duncan, B. (1955). A methodological analysis of segregation indexes. American Sociological Review, 210e217. England, K. V. (1993). Suburban pink collar ghettos: the spatial entrapment of women? Annals of the Association of American Geographers, 83(2), 225e242. Fan, C. C. (2001). Migration and labor-market returns in urban China: results from a recent survey in Guangzhou. Environment and Planning A, 33(3), 479e508. Fan, C. C. (2002). The elite, the natives, and the outsiders: migration and labor market segmentation in urban China. Annals of the Association of American Geographers, 92(1), 103e124. Fan, Y. (2012). The planners' war against spatial mismatch lessons learned and ways forward. Journal of Planning Literature, 27(2), 153e169. Gu, C., & Shen, J. (2003). Transformation of urban socio-spatial structure in socialist market economies: the case of Beijing. Habitat International, 27(1), 107e122. Gu, C., Wang, F., & Liu, G. (2005). The structure of social space in Beijing in 1998: a socialist city in transition. Urban Geography, 26(2), 167e192.

Horner, M. W., & Marion, B. M. (2009). A spatial dissimilarity-based index of the jobsdhousing balance: conceptual framework and empirical tests. Urban Studies, 46(3), 499e517. Jiang, F., Liu, S., Yuan, H., & Zhang, Q. (2007). Measuring urban sprawl in Beijing with geo-spatial indices. Journal of Geographical Sciences, 17(4), 469e478. Kain, J. F. (1968). Housing segregation, negro employment, and metropolitan decentralization. The Quarterly Journal of Economics, 82(2), 175e197. Kenworthy, J., & Hu, G. (2002). Transport and urban form in Chinese cities: an international comparative and policy perspective with implications for sustainable urban transport in China. disP-The Planning Review, 38(151), 4e14. Kenworthy, J. R., & Laube, F. B. (1999). Patterns of automobile dependence in cities: an international overview of key physical and economic dimensions with some implications for urban policy. Transportation Research Part A: Policy and Practice, 33(7), 691e723. Knaap, G., & Zhao, X. (2009). Smart growth and urbanization in China: can an American tonic treat the growing pains of Asia? In Y. Song, & C. Ding (Eds.), Smart urban growth for China (pp. 11e24) Cambridge, Massachusetts: Lincoln Institute of Land Policy. Korsu, E., & Wenglenski, S. (2010). Job accessibility, residential segregation and risk of long-term unemployment in the Paris region. Urban Studies, 47(11), 2279e2324. Li, S. M. (2010). Evolving residential and employment locations and patterns of commuting under hyper growth: the case of Guangzhou, China. Urban Studies, 47(8), 1643e1661. Martin, R. W. (2004). Spatial mismatch and the structure of American metropolitan areas, 1970e2000. Journal of Regional Science, 44(3), 467e488. Massey, D. S. (1993). American apartheid: Segregation and the making of the underclass. Harvard University Press. National Bureau of Statistics of China. (2011). The sixth national population census data gazette (no. 1). Retrieved 8/1/2011, from http://www.stats.gov.cn/tjgb/ rkpcgb/qgrkpcgb/t20110428_402722232.htm. Pendakur, V. S. (1992). Urban transportation in China: trends and issues. Transportation Research Record, 1372, 3e10. Shen, Q. (1997). Urban transportation in Shanghai, China: problems and planning implications. International Journal of Urban and Regional Research, 21(4), 589e606. Song, Y., Ding, C., & Knaap, G. (2006). Envisioning Beijing 2020 through sketches of urban scenarios. Habitat International, 30(4), 1018e1034. Song, Y., Zenou, Y., & Ding, C. (2008). Let's not throw the baby out with the bath water: the role of urban villages in housing rural migrants in China. Urban Studies, 45(2), 313e330. Stoll, M. A. (2006). Job sprawl, spatial mismatch, and black employment disadvantage. Journal of Policy Analysis and Management, 25(4), 827e854. Sun, F., Wang, Z., & Wang, J. (2003). Statistical analysis on the average speed of public buses in Beijing. Automotive Engineering, 25(3), 219e242 (in Chinese). Wang, D., & Chai, Y. (2009). The jobsehousing relationship and commuting in Beijing, China: the legacy of Danwei. Journal of Transport Geography, 17(1), 30e38. Wang, E., Song, J., & Xu, T. (2011). From “spatial bond” to “spatial mismatch”: an assessment of changing jobsehousing relationship in Beijing. Habitat International, 35(2), 398e409. Wong, D. W. (2005). Formulating a general spatial segregation measure. The Professional Geographer, 57(2), 285e294. Wu, W. (2002). Migrant housing in urban China: choices and constraints. Urban Affairs Review, 38(1), 90e119. Xu, Y., Chan, E., & Yung, E. (2014). Analysis of the mechanisms contributing to spatial mismatch in transitional Chinese cities. Journal of Urban Planning and Development, 140(2), 04013011. Zhang, Y., & Fang, K. (2004). Is history repeating itself? From urban renewal in the United States to inner-city redevelopment in China. Journal of Planning Education and Research, 23(3), 286e298. Zhou, Y. (1998). Beijing and the development of dual central business districts. Geographical Review, 88(3), 429e436. Zhou, M., & Logan, J. R. (1996). Market transition and the commodification of housing in urban China. International Journal of Urban and Regional Research, 20(3), 400e421. Zhou, Y., & Ma, L. J. (2000). Economic restructuring and suburbanization in China. Urban Geography, 21(3), 205e236. Zhou, J., & Wang, Y. (2012). Mobile-source GHG modeling institutions and capacities in China: findings based on structured interviews and on-line surveys. In The 92nd Transportation Research Board Annual Meeting. Washington D.C. Zhou, S., Wu, Z., & Cheng, L. (2013). The impact of spatial mismatch on residents in low-income housing neighbourhoods: a study of the Guangzhou Metropolis, China. Urban Studies, 50(9), 1817e1835.