The intra-urban distribution of skilled migrants: a comparison between Shanghai and Nanjing

Transcription

1 The intra-urban distribution of skilled migrants: a comparison between Shanghai and Nanjing CAN CUI, STAN GEERTMAN, PIETER HOOIMEIJER Abstract With the maturing of market-oriented reforms over the last two decades, economic restructuring and labor mobility have accelerated, resulting in a remarkable increase in the number of skilled migrants to Chinese cities. Skilled migrants not only differ from traditional rural-to-urban migrants in terms of recruitment in the labor market, but also in their access to housing. Consequently, their distribution over residential areas in the city differs too. Yet, the economic structure, the spatial structure, the housing provision and the migration policy vary substantially from one city to another, which might result in different patterns in different cities. In this paper, a comparative analysis between Shanghai and Nanjing has been undertaken to investigate the driving force underlying the residential distribution of skilled migrants versus all migrants and all population at sub-district level, using interprovincial migration data from China s 2000 census. The results show that skilled migrants have better access to the cities compared with total migrants and even compared with total population, as they concentrate in the sub-districts with more professional job opportunities and larger share of (former) work unit housing or high priced housing in both Shanghai and Nanjing. Shanghai is more advanced in the process of housing commercialization, indicated by the stronger impart of market forces in the housing market on the concentration of skilled migrants. Keywords: skilled migrants; residential distribution; local context 1. Introduction Rapid economic development and disparities in regional developments bring about a massive domestic migration towards Chinese cities in the last three decades, which in return accelerates the process of industrialization and urbanization in the cities. At the early stage, a vigorous development of manufacturing sector and construction industry has attracted a large population of low-skilled labor, dominated by rural-to-urban migrants. As Chinese cities change from being centers of production to centers of consumption, consumerism has emerged. The employment structure in Chinese cities is gradually shifting away from basic manufacturing towards a more diverse range of positions, such as professionals, secretaries, managers, and service workers (Pigott, 2002). At the same time, as Chinese higher education is in the process of universalization, more and more people are well trained and more capable of engaging in skilled jobs. The diversified job opportunities in the urban labor market together with the growing population of qualified labor leads to a remarkable increase in the number of skilled migrants. Moreover, this number is expected to continuously increase with the deepening of economic restructuring. Another prominent feature of Chinese cities is that the urban spatial structure is experiencing striking reorganization during the course of urban renewal and sprawl, triggered by economic development (Song et al., 2010). Social stratification is intensified due to income distribution system reforms and the relaxation of urban household registration regulations in market-oriented economy. The combination of spatial reorganization and social stratification is creating residential differentiation (Li and Wu, 2008; Liu et al., 2008; Wu, 2002; Wu and Li, 2005). In this differentiated urban space, the spatial arrangement of opportunities is changing migrants access to the cities. In the central area, reemergence of

2 commercial activities is shaping a new central business district (CBD); contrarily, industrial establishments are relocated to the urban outskirt. Meanwhile, housing system also changed dramatically from a welfare-oriented towards a market-oriented system (Wang, 2000; Wang & Murie, 2000; Zhao, 2010), making it possible for residents to freely choose their residence location. From the perspective of spatial dimension, the redevelopment in the central city, and substantial commodity housing construction in new suburban areas have greatly reshaped the spatial distribution of the residents. The emerging of skilled migrant population and the changing spatial structure of cities arouse our interests in their residential distribution and the driving forces behind it. Skilled migrants have more resources, and can easily overcome institutional barriers; hence their residential distribution and the driving forces behind it are theoretically different from other migrants. In earlier study, residents in cities are classified into three categories by their hukou status (Chan et al., 1999; Fan, 2002; Sun and Fan, 2011), namely, urban natives, permanent migrants and temporary migrants. It results in a socioeconomic hierarchy in which permanent migrants, mainly composed of well-educated elites, are the most prestigious, urban natives are in the middle and temporary migrants are the least prestigious. In this paper, we propose a different way of classifying migrants. According to their occupation, migrants are divided into skilled and low-skilled workers. Skilled migrants are mainly administrative and professional staff enrolled in government, institutions, state-owned enterprises, and emerging service sector. This shift in classification can facilitate better understanding of the impact of occupation on migrants residential location; hence it will shed a light on the role of occupation in overcoming institutional barriers. In this paper, in order to better understand the residential distribution of skilled migrants in different local contexts, a comparative analysis between Shanghai and Nanjing will be undertaken to examine the impacts of location-specific factors on skilled migrants concentration in comparison with total migrants and total population. Based on the data from 2000 National population census, results are generated by adopting regression models at the sub-district level. The following content is organized in five parts. It begins with a literature review on migrants settlement choice. Description of data and Study area is followed. Then, a comparison between Shanghai and Nanjing in aspects of spatial structure, economic structure, housing segment and characteristics are provided. An empirical analysis is presented to unveil the residential distribution of skilled migrants and the driving forces behind it. The findings are summarized and discussed in the last section. 2. Literature Review Most of the existing literature treats all Chinese domestic migrants as one single undifferentiated mass (partly due to lack of disaggregated data) (Chan et al., 1999). In general, migrants settlements are found in urban periphery (Wu, 2008). The uneven residential distribution of migrants is attributed to factors, such as job opportunities, the availability of affordable housing, social networks (Wu and Li, 2005; Wu, 2008). Since migration in China is primarily employment-driven, same as most developing economies, the best predictor of migrant concentration is employment opportunities, particularly the number of state-owned establishments and industrial establishments most of which have been relocated in the suburbs (Wu, 2008). Housing availability also largely affects migrants residential pattern (Logan et al., 2009; Wang et al., 2012; Wu, 2004; Wu and Li, 2005). The disadvantage in access to subsidized public housing confines migrants tenure choice (Logan et al., 2009). Liu et al. (2012) state that migrants are much less likely to be homeowners, and they consume smaller and more poorly facilitated housing. Market-related factors, such as income and education, together with institutional factors, contribute to their severely disadvantaged position in the housing system. In addition, kinship also plays an important role in migrants residential choice. Migrants from particular (rural) areas are likely to congregate in parts of cities, named - 2 -

3 after their origin province, such as Zhejiang Village, Anhui Village in Beijing (Gu and Liu, 2008; Stephens, 2010). In earlier times, migrants were dominated by low-skilled rural-to-urban migrants. While with the shifting in migrant composition, it is getting problematic ignoring the variations within migrant population. Some scholars have realized the growing heterogeneity of migrants and subdivided them according to their hukou status. Chan et al. (1999) address the substantial differences between hukou and non-hukou migrants. Hukou migrants tend to originate from urban areas, have a highly skewed share of the college-educated, and are employed in skilled jobs. Logan et al. (2009) classified migrants into urban and rural migrants, and further classified each group into recent and established according to the length of residence. Established urban migrants (with urban hukou and have stayed in the city over 5 years) are found to have a relative advantaged position in the housing system even compared with urban natives. Focusing on the differentials between permanent migrants and temporary migrants, Sun and Fan (2011) conclude that permanent migrants, represented by young, highly educated individuals and those in prestigious occupations, increasingly assume the position of social and economic elites, whereas temporary migrants are the disadvantaged and disenfranchised. There are also some studies distinguishing people based on their occupation. The resources and restrictions living in the cities may differ a lot depending on their occupation, resulting in different access to the cities. Yeh et al. (1995) reveal residential differentiation in terms of occupation in Guangzhou. Results show that the cadres occupy a belt in the suburbs immediately adjacent to the old city area; the intellectuals reside mainly near universities, the factory workers are more widely scattered, with a few concentrated in clusters; finally, the farmers occupy the outer edge of the city. In Lo s study (1994), one more group, the emerging workers in the tertiary sectors, is identified and found clustering in the old city area. Bian and Liu (2005) report that households headed by managerial and professional elites attain more space and higher-quality housing. Similarly, Liu et al. (2012) demonstrate that housing tenure also varies significantly across groups with different occupations in Nanjing. People who work for government agencies and institutions and managers in state-owned and collective enterprises are more likely to own or rent subsidized public housing, and occupy larger and better quality housing. However, little empirical evidence has captured the differences within domestic migrant population in terms of their occupation. In international migration theories, skilled migrants, defined by generic occupational terms such as professional and manager, stared to be recognized and intensively studied since the mid-1960s (Koser and Salt, 1997). Most of these researches discuss the movement of highly skilled persons from developing countries to developed countries ( brain drain ) and focus on its volume, composition, impacts, and policy implications (Docquier and Rapoport, 2004; Findlay et al., 1996; Kofman, 2000; Vertovec, 2002). Regarding the determinants of residential distribution, Vertovec (2002) points out that schools and universities are a foremost source of recruitment for skilled migrant networks. Skilled migrants rely more on networks of alumni, colleagues or organizations and less on kin-based networks than low-skilled migrants. Concerning domestic skilled migrants, it is virtually impossible to find much discussion. To fill this gap, this paper aims to examine the residential pattern of skilled migrants in the differentiated urban space. The overview of the literature indicates that employment opportunities, housing availability, and urban spatial structure exert impact on migrants residential choice, which is the basis of our hypothesis in this paper. However, the way of how employment opportunities, housing availability, and urban spatial structure affect skilled migrants is expected to work differently, as their recruitment in the labour market and access to the housing market differs prominently from other migrants

4 3. Data & Study Area 3.1 Data The main findings of this paper are based on the dataset from 2000 National population census and economic census in Shanghai and Nanjing. The dataset from the census uses subdistrict as a sampling unit. In this paper, we specifically focus on interprovincial migrants, who moved to the surveyed sub-district from other provinces during the preceding five years (Nov. 1, Nov. 1, 2000). Unless specified otherwise, the migrants mentioned in the following part refer to interprovincial migrants. Intra-provincial migrants are not taken into our research, because in the database of the population census, intra-provincial migrants also include intra-urban movers. The following analysis is at sub-district level, which is the finest level of obtainable data and also important in the formation of spatial differentiation, for it is equivalent to the base-level organization of residential areas (Wu and Li, 2005). Employment data is drawn from the economic census, which is workplace-based. 3.2 Study area Shanghai and Nanjing are both within the metropolitan area of the Yangtze River Delta, which is one of the most developed and dynamic economic regions in China. These two cities are both developed comparing with most of other cities in China. Whereas, Shanghai and Nanjing differs in terms of administrative rank, population size, and the economic role in the Yangtze River Delta; there are also some differences in housing provision, hukou system and other migrant related policies. The comparison between Shanghai and Nanjing will provide insight into the driving forces in different local context underlying the spatial concentration of skilled migrants. Fig.1 Four rings of Shanghai and Nanjing Shanghai is a municipality directly under the central government, equivalent to a province at administrative level. According to the Fifth China Census (2000), the total population of Shanghai reached 16 million. It was divided into 20 county-level districts: 17 districts and 3 counties in Nanjing, the capital of Jiangsu province, is the second largest commercial center in the Yangtze River Delta, after Shanghai. In 2000, the total population was 6 million

5 Nanjing consisted of 15 county-level divisions, of which 10 are districts, and other 5 are counties. Considering the administrative boundaries, road rings, geographic features, and development phase, both Shanghai and Nanjing are divided into 4 rings: urban core, central areas, inner suburbs, outer suburbs (Fig.1). Concentrating on the urban built-up area, the first 3 rings are chosen as the study areas for both Shanghai and Nanjing. 4. Differences in Local Context Focusing on the first three rings makes it comparable between Shanghai and Nanjing, respectively covering an area of 1589 km 2 and 1403 km 2 (Table 1). Within the study area, there are 154 sub-districts in Shanghai and 83 sub-districts in Nanjing. The larger number of sub-districts in Shanghai indicates the smaller sub-district level division. As one of the most populous cities in the world, the population density in Shanghai is almost three times of that in Nanjing. Shanghai is also a more popular migration destination than Nanjing. Local residents in Shanghai take up 63.3% of the total population; while in Nanjing, 67.3% are local residents. Because of the higher urbanization level in Shanghai, the proportion of people holding agricultural hukou is only 24%. After this summary table, a detailed comparison between Shanghai and Nanjing in aspects of spatial structure, economic structure, and the housing market will be given in the following sections. Table 1 Study areas (first three rings) of Shanghai and Nanjing Shanghai Nanjing Area 1589 km km 2 The number of sub-districts Total Population 11,823,739 3,973,322 The number of local residents 7,483,225 (63.3%) 2,675,173 (67.3%) Agricultural hukou 24.0% 26.0% 4.1 Spatial structure Fig.2 Study area of Shanghai - 5 -

6 Even though every district has its own center, the urban core (the first ring) in Shanghai is around people s square on the west bank of Huangpu River. Prominent business areas include Lujiazui on the east bank of Huangpu River, and The Bund in the west bank of the Huangpu River. The city hall and major administration units are located around People s square and Nanshi District, which also serve as a commercial area. Other major commercial areas include Xujiahui in Xuhui district. Meanwhile, the urban core has long been the residential core. With the combined effect of redevelopment in central city (the first and second rings), new housing construction in the inner suburbs (the third ring), and the satellite town program after 1950s, the inner suburbs are experiencing rapid transition from rural to urban uses. As a result, inner suburban sub-districts gradually are accommodating a large population while the population in outer suburbs remains stable (Gao and Jiang, 2002). To change the mixed pattern of land use and industrial fragmentation in the central city, relocating factories in the central city to suburbs has been carried out by means of establishing a number of industrial zones (usually called economic and technology development zone, ETDZ for short), including Hongqiao ETDZ, Jinqiao Export Processing Zone (EPZ), Minhang ETDZ, Caohejing High-Tech Park, and Zhangjiang High-Tech Park. This process of industrial relocation, albeit slow and with mixed results, has freed up a significant amount of space in the urban core and led to an industrial concentration in the second ring (Wu, 2008). Pivot infrastructures and efficient traffic networks have definitely facilitated the population dispersion process and allowed the comprehensive development of the "big Shanghai" integrating east and west parts (Wu, 2004). After several decades of development and suburbanization, Shanghai already became a representative poly-centric city in China. Urban core Fig.3 Study area of Nanjing Comparatively, Nanjing is a compact city, with high population concentration within the urban core (the first ring), which was circled by the ancient city wall before. Administrative offices, higher education, as well as high-ranking residences were mainly concentrated in Gulou and Xuanwu districts (Fig. 3, also for the names mentioned in the following text). Even - 6 -

7 within the urban core, the north-south segregation was a prominent characteristic (Liu and Wu, 2006). The southern part of the urban core is a traditional old urban region where handicraft workers, small stores and ordinary urbanites concentrated. Since the 1980s, most of the shanty areas and some old urban neighborhoods were redeveloped into mixed areas of residences and commerce. After the implementation of housing reforms, the disconnection between employment and residence has begun to gradually change the traditional residential pattern in which the workers in the same enterprise or administrative institution live in the same community. After 1950, the spatial structure of Nanjing was influenced by the state-led urban development. A series of state-owned heavy industries, including electrical, mechanical, chemical and steel factories, were built in the second and third ring, extending from the Zhongyangmen to Maigaoqiao in the northern part, and from Yuhuatai to Banqiao in the south west. Since the 1990s, Nanjing has experienced industrial restructuring, from laborintensive to capital- and technology-intensive industries, and from traditional manufacturing sectors to modern service sectors. Nanjing New & High-Tech Industry Development Zone, Nanjing Economic and Technological Development Zone, and Jiangning Development Zone were successively established in 1991, 1992, driving urban expansion into the peripheral areas. Despite the efforts of developing the suburb areas, Xinjiekou still remains the commercial and business center of Nanjing. 4.2 Economic structure Although both situated in Yangtze River Delta, Shanghai and Nanjing varies a lot particularly in terms of economic structure. In the last two decades Shanghai has been one of the fastest developing cities in China even in the world, becoming the commercial and financial center of mainland China. Retrieving from the establishment database of 2001, the average proportion of service employments in a sub-district reached up to percent (table 2). The three largest service industries in Shanghai are financial services, retail, and real estate. Nevertheless, Shanghai is still one of the main industrial centers of China, with the production employments taking up percent averagely in all sub-districts. Table 2. Different kinds of employment in Shanghai and Nanjing Shanghai (154 sub-districts) Nanjing (83 sub-districts) range Mean Proportion range Mean Proportion Administrative employment % 0-1, % Professional employment ,364 3, % 0-14,739 3, %. Service employment ,239 17, % 0-46,198 8, % Production employment 2,821-73,288 18, % ,333 25, % Compared with other major (coastal) cities, Nanjing is considered a relatively conservative city (Luo & Wei, 2006) and less impacted by deindustrialization compared with Shanghai. In the 1950s, the central government of China invested heavily in Nanjing to build a series of state-owned heavy industries, as part of the national plan of rapid industrialization. The current industry of the city basically inherited the characteristics of the 1960s, with electronics, cars, petrochemical, iron and steel, and power as the "Five Pillar Industries". As shown in table 2, service industry started to gain prominence in 2000, with average percent service employments in all sub-districts. As a traditional heavy industry city, the production employments unsurprisingly mount up to percent of the total employments averagely in all sub-districts

8 4.3 Housing segment and characteristics Under socialism, housing was treated as a welfare provision allocated by work unit or government based on such criteria as occupation, administrative rank, job performance, political connections (Zhou and Logan, 1996). Starting in the late 1980s, the housing allocation system was abolished and a series of housing reforms were implemented, aiming at commercializing the housing market. Privatization of work unit housing took the form of transferring ownership to the sitting tenants at highly subsidized prices (Logan et al., 2010). That means those who already obtained better residential status from housing allocation system still enjoy privileged position during reform. A housing finance system was developed, and mortgage loans started to be offered to housing buyers by financial institutions. Profound changes have since taken place, both in housing provision and in housing consumption. However, it took time to put these reforms into practice. The housing market was going through a transition, in which the privatization of work unit housing was implemented, and the commercial housing began to burgeon. Institutions inherited from the socialist period and the emerging market mechanisms are intertwined to contribute to housing differentiation after the introduction of housing reform (Liu et al., 2012). Table 3. Differences of housing provision and characteristics between Shanghai and Nanjing Housing tenure structure Shanghai (154 sub-districts) Nanjing (83 sub-districts) range mean range mean Self-built % 18% % 25.6% Commercial purchase % 8.7% % 3.3% Economical purchase % 4.0% % 3.0% Former work unit housing purchase % 23.3% % 32.2% Rent work unit housing % 33.0% % 22.6% Rent commercial housing % 6.8% % 6.7% Housing characteristics Proportion of multi-functional housing 0-6.6% 2.1% % 2.5% Housing quality index Per capita dwelling space (m 2 ) Purchase price (1000 yuan) Renting price (yuan) In 2000 Population Census, there are six types of housing classified by the ways how the residents obtained their house. Four categories are the owner-occupied housing, and two types are the rental housing. The ownership rate in Shanghai is 54% in all sub-districts, a slight lower than 64.1% in Nanjing. Within the category of owner-occupied housing, although we already confine our study area to the urban areas of the cities, there is still average 18% selfbuilt housing in a sub-district of Shanghai. In Nanjing, this proportion is even higher than 25%. The self-built housing is a common housing type in rural areas, which is built by the villagers on the collective-owned land. The large share of self-built housing in Nanjing indicates the lower urbanization level in 2000 compared with Shanghai. The share of commercial purchased housing in Shanghai is almost three times of that in Nanjing, manifesting the advanced commercialization process of the housing market. The purchased former work unit housing accounts for 23.3% in Shanghai, lower than 32.3% in Nanjing. Probably because there is higher share of state-owned enterprises in Nanjing, allocating work unit housing to the employees; thus there is higher share of former work unit housing in the 1 Exchange rate (1999): 1 dollar = 8.28 yuan - 8 -

9 housing market. Rental work unit housing amounts to 33.0% in Shanghai, much higher than 22.6% in Nanjing. A possible explanation is that the rental price is pretty high in Shanghai, so even after reforms have been implemented for several years, the work unit continues to provide rental housing for their employees to compensate the high rental price in Shanghai. With respect to housing characteristics, Shanghai and Nanjing share several similarities, with almost the same proportion of multi-functional housing, approximate housing quality index, and the same amount of per capita dwelling space. The multi-functional housing means a house used as a residence as well as a work place, which is a typical housing type for selfemployed persons, such as small business runners. Housing quality index varies between 0 and 100, which is calculated based on the inner facilities of the house (kitchen, energy source, tap water, bath facilities and toilet). The average indices over 80 show that the houses in Shanghai and Nanjing are possessed of basic facilities. The evident difference between Shanghai and Nanjing is the purchase & renting price. It does not come as a surprise, because the higher housing price in Shanghai is associated with the advancement of economic development and the advanced phase of housing commercialization. However, the values of the housing price are much lower than what we expect. The housing price here is the average price in each sub-district, including private as well as public sector. In 2000, there are quite large share of public housing (over 50%). The privatized work unit housing and rental public housing are highly subsided. As a result, the average housing price is well below the market price (approximately 3300 yuan per square meter in Shanghai, 2500 yuan per square meter in the urban core of Nanjing). 5. Residential Distribution of Skilled Migrants There are six occupational categories in the Census: agricultural worker, production worker, sales & service worker, clerk, professional/technician, and administrator/manager. The image of migrants coming to cities to work in manufacturing industries is clearly outdated as early as Migrants occupation is shifting towards the service sector, with an average proportion over 40% both in Shanghai and Nanjing. Among these six categories, professional/technician and administrator/manager are defined as skilled jobs in this paper. Table 4. Occupation structure of migrants Shanghai (154 sub-districts) Nanjing (83 sub-districts) range Mean range Mean Administrator/manager 0-9.9% 2.1% % 3.2% Professional/technician % 6.5% % 10.1% Clerk % 5.6% % 3.9% Sales & service worker % 42.1% % 42.9% Production worker % 39.8% % 32.6% Agricultural worker % 3.9% % 7.2% Description Skilled Migrants Total migrants Skilled Migrants Total population Table 5. Descriptive statistics of skilled migrants Shanghai Minimum: 0.68% Maximum: 25% Mean: 8.63% Standard Deviation: Minimum: 0.08% Maximum: 2.08% Mean: 0.77% Standard Deviation: Statistical description Nanjing Minimum: 1.00% Maximum: 41.10% Mean: 13.34% Standard Deviation: Minimum: 0.02% Maximum: 2.04% Mean: 0.48% Standard Deviation:

10 Focusing on skilled migrants and using total migrants and total population as two comparison groups, table 5 shows the descriptive statistics. Among migrants in Nanjing, there are as many as 13.3% of them engaged in skilled jobs; while in Shanghai, there are only 8.6% skilled migrants, because Shanghai is a popular and traditional migration destination not only for skilled migrants but also for low-skilled migrants. High demand for labor in manufacturing sector and infrastructure construction has drawn a large flow of low-skilled migrants, thus the share of skilled migrants is not high. Because skilled migrants we are investigating are only those who migrate across provincial borders, as expected, the share in total population is quite small. To visualize the differentials of skilled migrants residential distribution in comparison with total migrants as well as total population, LISA maps are employed (Anselin et al., 2006). With different comparative population groups, the LISA maps show distinct patterns (Fig.4 & Fig.5). Compared with total migrants, skilled migrants are markedly clustered in the urban core of both Shanghai and Nanjing (Fig.4), illustrating their better access to the urban core. This result is in accord with the findings of most existing literature that migrant population, dominated by low-skilled migrants, are found in peripheral areas, spatially excluded from the core areas of the cities (Wu and Li, 2005; Wu, 2008). Subgrouping migrants into skilled and low-skilled groups reveals the heterogeneity among migrants in terms of residential distribution. (a) Shanghai: Moran s I = Pseudo P-value: 0.01 (99 permutations) (b) Nanjing: Moran s I = Pseudo P-value: 0.10 (99 permutations) Fig.4 LISA maps for the proportion of skilled migrants to total migrants in Shanghai and Nanjing When compared with total population, skilled migrants are clearly clustered in Shanghai, with Moran s I score of , demonstrating high spatial association among the neighboring sub-districts in terms of attracting or distracting skilled migrants (Fig.5 (a)). The highest concentration of skilled migrants appears in the southwest part, mainly Xuhui district, which is a historically upper end area of Shanghai with many education, research institutions and Caohejing high-tech park located there. This pattern is in line with Wu and Li s findings (Wu and Li, 2005) of the concentration of well-educated residents in the southwest of Shanghai. In Nanjing, insignificant Moran s I score of indicates that the distribution of skilled migrants does not differ much from that of total population (Fig.5 (b)). In other words, no sub-district attracts skilled migrants in particular compared with total population

11 (a) Shanghai: Moran s I = Pseudo P-value: 0.01 (99 permutations) (b) Nanjing: Moran s I = Pseudo P-value: 0.10 (99 permutations) Fig.5 LISA maps for the proportion of skilled migrants to total population in Shanghai and Nanjing 6. Explaining Residential Distribution of Skilled migrants The previous section provides the geographical patterns of skilled migrants residence. To go beyond the descriptive analysis and gain some understanding of the location-specific factors that determine the residential location choice of skilled migrants, regression models will be carried out. Two dependent variables are respectively the share of skilled migrants in total migrants and in total population in each sub-district. They are log transformed before taken into the models. Table 6. Explanatory variables Type Variables Description Employment Professional jobs Service jobs Production jobs Number of employment in Professional establishments Number of employment in Service sectors Number of employment in Production establishments Housing Location Tenure Housing condition House price Ring Work unit purchase Public rent Housing quality index Per capita space Multi-functional house Purchase price Renting price ring1 ring2 ring3 The proportion of purchased former work unit housing The proportion of public housing renting An index based on five aspects: with or without kitchen, energy source for cooking, with or without tap water, bath facilities and individual toilet per capita dwelling space (square meter) The proportion of houses as both residence and commercial places Average price for purchasing or building a house Average price for renting a house per month ring1: urban core ring2: central area ring3: inner suburb Based on the literature review, three groups of location-specific factors are hypothesized to

12 be able to explain the concentration of skilled migrants in comparison with total population and total migrants: employment opportunities, housing characteristics, and the location of the sub-district (Table 6). We assume that skilled migrants, resembling other migrants that searching for an employment is their original intention of coming to the cities, are likely to reside nearby workplace especially at the early stage of living in the cities. The number of different employment types within each sub-district is retrieved from economic census. Skilled migrants, as defined, are administrator/manager or professional/technician, thus they are expected to locate in the sub-district with more number of professional jobs (finance, high-technology, medical care, education & research, culture & art), and less number of service and production jobs. The existence of lingering housing dependency between households and institutions still impacts housing consumption (Huang and Clark 2002). Skilled migrants are hypothesized to live in better housing (better housing quality and larger space) and have better access to public housing sectors, thus the explanatory variables of work unit purchase, public rent, housing quality index, and per capita space are supposed positively affect the concentration of skilled migrants. An increase of multi-function housing in a sub-district, which is a typical housing type for self-employed persons, may decrease the possibility of a concentration of skilled migrants in that sub-district. Housing price, which is an indicator of housing quality and accessibility in a market-oriented system, is hypothesized to be positively related with the share of skilled migrants in a sub-district. Location variables will also be tested in the model. As seen from the Fig.4 and Fig.5, compared with total migrants, skilled migrants are clearly overrepresented within the first road ring both in Shanghai and Nanjing. Ln( Pi ) 1Ei 2H i 3Li P i : the percentage of skilled migrants to the total population and total migrants in the i th subdistrict E i : employment opportunities variables of the i th sub-district H i : housing attribute variables of the i th sub-district L i : location variables of the i th sub-district We started with standard linear regression-ordinary least square (OLS), which is a type of non-spatial regressions (equation 1). After running models using different combinations of explanatory variables, several variables have been dropped from the final models because of multi-collinearity problem or because they do not exert any impact in the models. Moreover, the larger set of explanatory variables will lead to low degrees of freedom, which makes statistical power goes down. The number of service jobs in a sub-district is highly correlated with the number of professional jobs (0.806 ** in Shanghai, ** in Nanjing); while production job opportunities do not exhibit a significant impact. Hence only the number of profession jobs will be taken into the final model. Within the group of housing variables, housing quality index and per capita space are discarded, respectively because housing quality index is highly associated with the share of purchased former work unit housing (0.575 ** in Shanghai, ** in Nanjing) and per capita space is strongly and negatively correlated with the share of public renting ( ** in Shanghai, ** in Nanjing). It is consistent with the previous findings (Logan et al., 2010) that the quality of purchased former work unit housing is desirable among other housing tenures and the public rent offers less space. Location variables are also excluded from the final models, as the variance between these three rings is largely reflected by employment and housing variables. Both in Shanghai and Nanjing, the number of professional job and the share of public rent is positively related with the first ring, whereas negatively related with the third ring. (1)

13 Table 7. The changing Moran s I Skilled migrant/ Total population Shanghai Skilled migrant/ Total migrants Skilled migrant/ Total population Nanjing Skilled migrant/ Total migrants Before OLS regression *** *** * * After OLS regression *** * ** Notes: *** significant at 0.01 level; ** significant at 0.05 level; * significant at 0.1 level. Taking remaining explanatory variables into the OLS model, the scores of Moran s I of the share of skilled migrants in total population and in total migrants have decreased remarkably (Table 7), demonstrating the combined influence of independent variables in explaining the variance in the concentration of skilled migrants, particularly for Shanghai. But, still the statistically significant Moran s I indicates strong spatial autocorrelation of the residuals (Anselin et al., 1996). It violates a regression assumption of independence of observations. Furthermore, spatial dimension of social and economic may truly be an important aspect of a modeling, as regional theory relies on notions of spatial interaction and diffusion effects, hierarchies of place and spatial spillovers (S4, 2006). Therefore, spatial lag model is a proper alternative taking spatial dependencies into account. Assuming that dependencies exist directly among the levels of the dependent variable, the concentration of skilled migrants in sub-district i is affected by the concentration in neighboring sub-districts 2. A lag term, which is a specification of the share of skilled migrants in neighboring sub-districts, is included in the regression, and its coefficient and p-value are interpreted as for the independent variables (equation 2). It has been proved that spatial lag model indeed yield improvement to the original OLS models. The results are presented in table 8, using standardized regression coefficients to assess the extent to which each independent variable is influential. Ln( Pi ) 1Ei 2H i 3Li 4 WijP j P j : the share of skilled migrants in a neighboring sub-district W ij : weight matrix (2) Explanatory variables Table 8. Spatial lag regression model Shanghai Skilled migrant/ Total population Skilled migrant/ Total migrants Skilled migrant/ Total population Nanjing Skilled migrant/ Total migrants (Constant) Professional jobs ** ** *** Work unit purchase *** ** *** Public rent *** ** *** Purchase price *** *** *** Renting price *** * Multi-functional house *** ** *** SPATIAL LAG *** *** ** *** Sum. Statistics: N R Log likelihood Notes: *** significant at 0.01 level; ** significant at 0.05 level; * significant at 0.1 level. 2 The sub-districts, whose centroids are within 5 km, are considered as neighboring sub-district of each other.

14 We will look into the results from three different angles. The first one is the relative importance of each explanatory variable. Another important angle is examining the relative position of skilled migrants in comparison with total population and total migrants. The typical features of skilled migrants distribution will be revealed only with the presence of comparison groups. The last one relies on the comparison between Shanghai and Nanjing as the local context differs. The results in the four columns show that the housing market is the strongest driving force of skilled migrants residential distribution, no matter in Shanghai or in Nanjing, no matter compared with total population or with total migrants. It differs from Wu s conclusion made for all migrants that employment opportunity is the better predictor compared to the housing availability for migrants concentration (Wu, 2008). It is understandable that the total migrants, with large share of low-skilled and temporary migrants, put the proximity to existing or potential employment as the priority when choosing residential location. As for skilled migrants, although employment opportunity does exert a significant impact, the effects of housing variables are much stronger, indicating that housing availability and housing characteristics are decisive for skilled migrants when choosing residential location. Skilled migrants have better access to the cities compared with not only total migrants but also total population. In Shanghai, professional job opportunity in a sub-district significantly contributes to the chance of attracting skilled migrants both in comparison with total population and total migrants. In Nanjing, the professional job opportunity predicts an increasing likelihood of the concentration of skilled migrants versus total migrants. Compared with total population (the first column), the results in the models for Shanghai and Nanjing show that skilled migrants concentrate in the areas either with high share of purchased former work unit house or with higher purchase and renting price. These results are in line with the hypothesis that skilled migrants are at an advantageous position in accessing better housing even compared with total population. When compared with total migrants, the advantageous position of skilled migrants in housing market is unsurprisingly more evident. Their advantageous position in public sector is manifested by the concentration in the sub-districts with higher share of former work unit housing and public renting. It is explained earlier that skilled migrants are managerial and professional staff in government, institutions, or stateowned enterprises, which conferred them an exclusive access to work unit housing before housing reforms. Later reforms to capitalize housing stock as a personal asset still favored managerial and professional staff. In the sector of commercial housing, the advantageous position of skilled migrants is reflected in the better access to the areas with higher purchase housing price. The models for Shanghai and Nanjing are possessed of a lot of similarities, whereas, the advantageous position of skilled migrants versus total population derives differently. In Shanghai, the concentration in the areas of higher purchase price suggests that skilled migrants are of high socioeconomic standing even compared with total population. While in Nanjing, skilled migrants are more likely within the institutional system, having good access to public sector of the housing market. Another visible difference between Shanghai and Nanjing is the opposite impact of the proportion of multi-functional house on skilled migrants concentration. It may be due to the different distributions of multi-functional housing. In Shanghai, high share of multi-functional house reduces the probability of attracting skilled migrants, because this kind of housing is found in the suburbs. However, in Nanjing, there is certain share of multi-functional housing in the urban core, where skilled migrants cluster. The most important distinction between Shanghai and Nanjing is the effects of spatial lag. As explained earlier, spatial lag measures the average influence on observations by their neighboring observations. The positive and highly significant effects of spatial lag in Shanghai suggest an agglomeration pattern of skilled migrants distribution at sub-district level compared with both total population and total migrants. On the contrary, segregation

15 presents in skilled migrants distribution in Nanjing, as indicated by the negative and highly significant impact of spatial lag. This difference in distribution pattern at sub-district level might be subjected to the different spatial structures of these two cities. The development of land use is continuous, and the administrative division at sub-district level is finer in Shanghai. Therefore, spillover effects are likely to occur. But in Nanjing, neighboring sub-districts are overall heterogeneous, resulting in pockets of skilled migrants. 7. Conclusion and Discussion Skilled migrant population, as an unrecognized group, is growing with the economic restructuring in Chinese cities in the past decades. Although they come into the cities as strangers like other migrants, their skills and preferable occupation may contribute to their superior position in the labor and housing market compared with total migrants or even compared with total population, which may results in a different residential distribution in cities. This study aims at uncovering the residential distribution of skilled migrants and revealing the location-specific factors behind it in comparison with total migrants and total population. Based on the 2000 population census data in Shanghai and Nanjing, spatial regression models are employed to explore the varied influence of location-specific factors on the concentration of skilled migrants at sub-district level. The advantageous position of skilled migrants in accessing to the cities shows up no matter compared with total migrants or compared with total population. It can be explained by their superior position in the labor and housing market. Their settlements are located in the areas with larger number of professional jobs and higher share of public housing or expensive housing. This conclusion is in contrast with previous findings, taking all migrants as a whole, that migrants are found in cheap, low quality and high density housing (Jie and Taubmann, 2008; Zhou and Cai, 2008; Liu et al., 2010). Yet, the results for skilled migrants are similar with Fan s findings that permanent migrants are actually the elite in the city, enjoying a higher position even than urban natives (Fan, 2002). In this paper, migrants are classified into skilled and low-skilled ones according to their occupation instead of their hukou status, but we arrive at similar conclusions, proving that being engaged in skilled labor makes it easier for migrants to overcome institutional barriers. Focusing on the comparative analysis between Shanghai and Nanjing, on the one hand, the similarities in the results proves that our spatial regression models are valid for these two cities. We may assume that running the models for other relative developed cities in China may generate similar results and conclusions. On the other hand, the differences in the results between Shanghai and Nanjing certainly reflect the different local context. When compared with total population, high prices of the commercial housing predict the concentration of skilled migrants in Shanghai, whereas, in Nanjing, skilled migrants have easier access to the areas with high share of public housing. The still opened public housing sector indicates that Nanjing lag behind Shanghai in housing commercialization. The different land use pattern and different scale of sub-districts between Shanghai and Nanjing may contribute to the opposite direction of spatial dependencies. In Shanghai, the concentration of skilled migrants has spillover effects, results in clusters of highly concentrated sub-districts. However, in Nanjing, the concentration of skilled migrants in one sub-district contrarily reduces the attractiveness of neighboring sub-district for skilled migrants, results in pockets of skilled migrants. In this paper, we have investigated the residential distribution of skilled migrants in 2000 and have explored its determinants using three groups of location-specific factors. With deepening of economic restructuring and housing marketization, the 2010 population census data is expected for further study on the spatial evolution of skilled migrants residential distribution. Limited by the area-based dataset, the analysis is also area-based; consequently, the conclusion cannot be drawn at individual level. Yet we can proceed on the individual level

16 analysis using field work data which are collected and organized at individual level in our future research. References Anselin, L., I. Syabri and Y. Kho (2006) GeoDa: an introduction to spatial data analysis. Geographical analysis 38.1, Anselin, L., A. K. Bera, R. Florax and M. J. Yoon (1996) Simple diagnostic tests for spatial dependence. Regional science and urban economics 26.1, Bian, Y. and Y. Liu (2005) Shehui fenceng, zhufang chanquan yu juzhu zhiliang dui zhongguo 'wupu' shuju de fenxi [Social stratification, housing property rights and living quality an analysis of Chinese census 2000]. Social Science Research 3, Chan, K.W., T. Liu and Y. Yang (1999) Hukou and non-hukou migrations in China: comparisons and contrasts. International Journal of Population Geography 5.6, Docquier, F. and H. Rapoport (2008) Skilled migration: the perspective of developing countries. In J. Baghwati and G. Hanson (eds.), Skilled migration: prospects, problems and policies, Russell Sage Foundation, New York. 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, Findlay, A.M., F.L.N. Li, A.J. Jowett and R. Skeldon (1996) Skilled international migration and the global city: a study of expatriates in Hongkong. Transactions of the Institute of British Geographers 21.1, Gao, X. and Q. Jiang (2002) Dui shanghai chengshi renkou fenbu biandong yu jiaoquhua de tantao [Changes in Shanghai s urban population distribution and suburbanization] Urban Population 26.1, Gu, C. and H. Liu (2008) Social polarization and segregation in Beijing. In J. R. Logan (ed.), The new Chinese city: globalization and market reform, Blackwell, Oxford. Huang, Y. and W.A.V. Clark (2002) Housing tenure choice in transitional urban China: a multilevel analysis. Urban Studies 39.1, Jie, F. and W. Taubmann (2008) Migrant enclaves in large Chinese cities. In J. R. Logan (ed.), The new Chinese city: globalization and market reform, Blackwell, Oxford. Kofman, E. (2000) The invisibility of skilled female migrants and gender relations in studies of skilled migration in Europe. International Journal of Population Geography 6.1, Koser, K. and J. Salt (1997) The geography of highly skilled international migration. International Journal of Population Geography 3.4, Li, Z. and F. Wu (2008) Tenure-based residential segregation in post-reform Chinese cities: a case study of Shanghai. Transactions of the Institute of British Geographers 33.3, Liu, Y., S. He, F. Wu and C. Webster (2010) Urban villages under China's rapid urbanization: unregulated assets and transitional neighbourhoods. Habitat International 34.2, Liu, Y., S. He and F. Wu (2008) Urban pauperization under China's social exclusion: a case study of Nanjing. Journal of urban affairs 30.1, Liu, Y., S. He and F. Wu (2012) Housing differentiation under market transition in Nanjing, China. The Professional Geographer 64.4, Liu, Y. and F. Wu (2006) Urban poverty neighbourhoods: typology and spatial concentration under China's market transition, a case study of Nanjing. Geoforum 37.4, Lo, C. P. (1994) Economic reforms and socialist city structure: a case study of Guangzhou, China. Urban Geography 15.2, Logan, J.R., Y. Fang and Z. Zhang (2009) Access to housing in urban China. International Journal of Urban and Regional Research 33.4, Logan, J.R., Y. Fang and Z. Zhang (2010) The winners in China's urban housing reform. Housing Studies 25.1, Luo, J. and Y.D. Wei (2006) Population distribution and spatial structure in transitional Chinese cities: a study of Nanjing. Eurasian Geography and Economics 47.5, Pigott, C.A. (2002) China in the world economy: the domestic policy challenges. Organization for Economic