Sources of migrant housing disadvantage in urban China

Environment and Planning A 2004, volume 36, pages 1285 ^ 1304 DOI:10.1068/a36193 Sources of migrant housing disadvantage in urban China Weiping Wu School of World Studies, Virginia Commonwealth University, Richmond, VA 23284-2021, USA; e-mail: wwu@vcu.edu Received 7 March 2003; in revised form 1 August 2003 Abstract. The increasing level of labor mobility in China challenges the current population-management structure. In particular, recent reforms in urban housing provision seem largely to overlook the needs of the migrant population. In this paper I examine the sources of migrant housing disadvantage in cities. Specifically, I analyze the institutional and socioeconomic factors underlying migrant housing choice and conditions, and how these factors influence migrants differently from the locals. Data are drawn primarily from citywide housing surveys and interviews conducted in Shanghai and Beijing. The findings show that migrants make housing decisions based on whether they intend to settle in the cities, and market-related factors such as income and education have a significant, positive impact on migrant housing conditions. But more importantly, the general disadvantage experienced by migrants has much of its root in the institutional restrictions associated with the hukou system that outweigh the combined effects of socioeconomic factors. Introduction The significant rise in labor mobility during the last two decades is a prominent feature of China's economic transition, reflecting a rapid process of industrialization and urbanization. Despite an impressive overall record of growth, income disparities between urban and rural areas, as well as regional imbalances, remain large. Primarily from rural to urban, much of the migratory flow takes place outside of the state plan and involves circular movements of rural labor in search of work to augment agricultural income. By the end of 2000 approximately 70 million rural migrants were working and living in urban areas according to official estimates. Nearly all of these people migrate without official change of household registration (hukou), although a small number manage to do so and achieve hukou change (permanent migration). This increasing level of mobility challenges the current population-management structure, as the hukou system still links residency with employment and social welfare. Most migrants continue to be regarded as temporary and have little chance of getting local hukou, no matter how long they have lived in the cities (often termed `temporary migrants', in this study I refer to them simply as `migrants'). Scholars agree that housing, an important element of urban amenities associated with hukou, remains difficult to attain for migrants (Chan and Zhang, 1999; Solinger, 1999; Wang and Murie, 2000; World Bank, 1997). Migrants are still largely excluded from the mainstream housing-distribution system. In theory they may purchase commodity housing in the cities, but bank mortgages are not available to them. In the secondary housing market, where older housing units change hands, a local hukou is often required. Subsidized public housing for low-income families is provided at below-market rents to qualified local urban residents only. Given these constraints, migrants make different housing choices and fare much worse in terms of housing conditions. The majority rent private housing or live in enterprise dormitories. They occupy far less space per person and overcrowding seems to be a feature of migrant housing. They also tend to live in less well-built dwellings that are less well built and which are less well equipped with kitchen/bathroom

1286 WWu facilities. It is not unusual to see a migrant family of three sharing a single rented room with no facilities, and using a corner of the room to set up a small cooking area with either a kerosene burner or a propane stove. Many of the prevailing theoretical studies on factors of migrant housing have evolved in a context in which private landownership, housing, and rental markets are functional, and thus these studies have largely focused on microanalyses. Chief among these individual factors are duration of residence, employment status, income level, and family status (Conway, 1985; Gilbert and Varley, 1990; Miraftab, 1997; Turner, 1968). There appears to be a direct relationship between housing choice (for example, renting versus ownership) and the economic status of migrants. Often it is only after migrants reach the stage of a secure job with reasonable income that they are able to become homeowners. Housing type and location also are directly linked to duration of residence in the city. Over time, migrants tend to move from rented rooms to squatter dwellings and then to houses. In addition, gender, household, and life-cycle factors are important in determining migrant housing choices and conditions (Miraftab, 1997). Other studies point to the importance of kinship and friendship ties in housing decisions (Banerjee, 1983; Conway, 1985). These inquiries, however, often fall short of considering the peculiar attributes associated with temporary and circulating migrants. Involving reciprocal flows, migrant circulation is characterized by movement from, and continuing ties with, a rural home base. Migrants who regard themselves as temporary members of the city often demonstrate different behaviors from permanent migrants (Goldstein, 1993; Nelson, 1976; Solinger, 1995). They tend to make different housing choices, invest little of their income to improve their living conditions, and demand fewer amenities and services. However, some evidence from Africa and Latin America suggests that, over time, urban ties surpass rural ties and temporary migrants eventually settle permanently in urban destinations (reviewed by Goldstein et al, 1991). What institutional and socioeconomic factors affect migrant housing choice and conditions in China? How do these influence migrants differently from the locals? These are the central questions of this paper. The focus of this paper is migrants who have moved to Shanghai and Beijing for employment purposes. The findings are primarily based on citywide migrant-housing surveys conducted in the two cities between 1998 and 2000, supplemented by in-depth interviews and results from official surveys. The first section of the paper is a discussion of the institutional barriers migrants face in urban destinations, and particularly in the housing system, followed by an outline of the research design. Then I empirically test the sources of migrant disadvantage in housing choice and conditions, and compare the effect of institutional and socioeconomic factors on migrant housing and between different population groups. The hukou system and urban housing The system of hukou has a profound impact on China's migrants. Implemented since the mid-1950s, the hukou system requires each citizen to register in one, and only one, place of regular residence. It entails two related parts: place of registration and type of registration (urban versus rural). This system has long been used to restrict migration, especially from rural to urban areas. The notion of temporary migrants is unique in China's contemporary context, as it does not denote a time frame but, rather, an official designationöhukou defines whether a migrant is permanent or temporary (also see Chan, 1996). The distinction between permanent and temporary migration is important, as permanent migration with official change of hukou has continued to be strictly controlled. Municipal authorities still use household registration as a basis

Sources of migrant housing disadvantage in urban China 1287 for providing urban services and maintaining infrastructure. To local urban residents and permanent migrants, urban amenities are either priced low or provided free. On the other hand, migrants without local hukou have very limited or no access to local schools, citywide welfare programs, state sector jobs, or the housing-distribution system. For instance, urban labor markets are highly segmented: the majority of migrants are restricted to jobs that are undesirable to the local population. Most migrant children cannot be enrolled in urban public schools, although a small number of them do so but at significantly higher expense than the charge made for local children. Housing had long been a form of social welfare to urban residents until recent reforms, and the rural ^ urban divide in housing provision dates back to the early socialist period. The dominant route, prior to 1999, was through a system of low-rent housing distributed either by work units or by municipal governments. This urban welfare housing system, however, did not apply to local residents with rural hukou, or to farmers in the countryside, who did not have access either to municipal or to work-unit public housing. Traditional family houses and private housing constructed on land allotted by production brigades were the norm for them, even in rural pockets within cities (Wu, 2002). Housing reforms gradually implemented during the last two decades have broadened housing choices for urban residents (see table 1). A new route is through market mechanisms, in which new commodity housing is developed and sold by real Table 1. Types of urban housing and their availability (source: based on Huang, 2003; Wu, 2002; and various government documents). Type of housing Commodity housing Economic and comfortable housing Muncipal public housing Work-unit public housing Low-rent housing Resettlement housing Private housing Rental housing Qualification Anyone can purchase at market price. However, only those with local urban hukou can qualify for bank mortgage loans. Local urban residents with low or medium income can purchase at subsidized prices. Sitting local urban tenants can purchase either ownership or use right, and can trade units. Other local urban residents can purchase either ownership or use right on the secondary housing market. Can be rented out with permit. Sitting local urban tenants can purchase ownership from their work units (mostly state owned) and transfer on the secondary housing market (except housing in institutions of higher education and some government agencies). Can be rented out with permit (and with work-unit approval in Beijing). For rental to local urban residents with lowest incomes, living on government allowances, and with per capita living area smaller than certain standards (for example 6 m 2 in Shanghai). Local residents being relocated from areas undergoing development can purchase at subsidized price (often lower quality housing in remote locations). Pre-1949 housing units passed on within family, and selfconstructed housing in rural areas. Local residents with rural hukou can construct new private housing on land allocated by their production brigades. Anyone can rent already-purchased commodity housing, already-purchased resettlement housing, and private housing.

1288 WWu estate companies. Both Shanghai and Beijing benefit from the national Comfortable Housing Project (anju gongcheng), which was launched in 1995 to create private sector housing (with government support) for low-income urban families to purchase (Li, 2000a). In the meantime the end of 1999 marked a turning point for China's housingdistribution system when the provision of all welfare housing (both through municipal and through work-unit distribution) was ended. Sitting tenants can choose to buy out the property right of their public housing. Individuals buying at market prices can enjoy full ownership rights, whereas those paying cost prices obtain a limited ownership share or only a use right (Wang and Murie, 2000; Zhang, 2000). Now urban residents have the option to trade these different types of housing on a secondary market. They also can rent directly from individual owners with rental permits, as well as rent private housing inherited by urban families or constructed by suburban farmers. More recently, low-rent public housing has become available to benefit urbanites with the lowest incomes and smallest living space (`Low-rent Housing Policy', China Daily 4 April 2003). Recent housing reforms, however, seem to overlook the need of the migrant population, despite its considerable size in many large cities. A local urban hukou continues to be an important qualification for accessing several type of urban housing, particularly those that are more affordable (see table 1). Migrants cannot acquire either the use right or ownership of muncipal or work-unit public housing directly because only sitting tenants (local urban residents) can do so. Both the Economic and Comfortable Housing and affordable rental units are also reserved for local urban residents only. On the secondary housing market, where older housing units are traded, participation generally requires a local hukou, although theoretically migrants can purchase housing there after completing a lengthy process of official approval (Beijing Evening News 27 June 2000). Commodity housing, the only property sector open for migrant ownership, is not affordable for most migrants. In-depth interviews with migrants reveal that the housing price considered by migrants as affordable averages around 160 000 yuan, a level much lower than what it would cost to buy a new commodity housing unitöabout 250 000 yuan in Shanghai and higher in Beijing. (1) Even local residents report difficulty in affording new commodity housing (Rosen and Ross, 2000). In addition, a local urban hukou is required to qualify for bank mortgages for new commodity housing (Bi, 2000). As a result of these restrictions, the rural ^ urban divide in housing continues even after rural migrants move to cities. The increasing mobility has added a new dimension (local versus nonlocal) to this divide, as migrants from urban origins do not enjoy the same access as local urban residents. Renting represents the best choice for migrants without local hukou. The rental sector involves a variety of housing types and has been active since at least the mid- 1980s when peasants began to trickle into Shanghai and Beijing. Urban residents who have purchased the ownership right of public housing can rent out their units (table 1). Shanghai has actually permitted the rental of public housing of which the sitting tenants have only a use right. In Beijing such housing is still theoretically prohibited from being rented out unless the approval of municipal housing authorities or relevant work units is given. But a large quantity of such housing has, in fact, been rented out. The bulk of rental housing is private housing in areas that used to be, or still are, agricultural within cities (Wu, 2001). In these areas local residents with rural hukou have been allocated ample land to build private living quarters, and tend to have (1) The price of a typical new commodity housing unit is based on a 1999 estimate (Shanghai Star 29 June 1999); migrant interviews were conducted in 1999 in Shanghai and in 2000 in Beijing.

Sources of migrant housing disadvantage in urban China 1289 much more housing space per person than urban residents living in the more built-up urban cores. In recent years accelerated urban development has led to the concentration of new functions (such as industrial parks, commercial subcenters, and technologydevelopment zones) in suburban areas. As a result, many rural residents are converted to urban status after their farmland is acquired for development. Because of their severed economic ties with rural villages, these residents tend to fair worse financially than the remaining rural residentsöso they have very strong incentives to rent out rooms for extra income. The growing demand of migrants for housing, and the lack of regulatory oversight, both contribute to the chaotic situation of the rental housing sector. As cities scramble to develop effective rental regulations, an increasing amount of deleterious building and rental activity continuesölargely in the form of unauthorized construction and the leasing of unsafe dwellings. This problem is particularly serious in urban ^ rural transitional areas where land is more readily available, the migrant population is more concentrated, and local residents have more incentive to rent out rooms. Even when regulations on rental housing take shape in some cities, concerns for adequate housing conditions and rental rights tend to be secondary. For instance, Beijing's regulations and Shanghai's early versions were promulgated merely for the purpose of maintaining public security, by using a rental-permit system. Another socialist institutionöthe work unitöalso plays a role in housing distribution, albeit with declining importance, and employees in state enterprises tend to have an edge over others. After 1999 state work units were no longer involved in housing provision and distribution, but many of them still offer housing subsidies to their employees. The most important form of subsidy is their matching of employee contributions to housing provident funds. Many state and collective enterprises also provide basic housing (in the form of dormitories) to migrant workers. Municipal authorities often prefer this arrangement as it provides a more controlled working and living environment in which enterprises can better manage matters related to temporary work and residence permits. A survey of 120 enterprises in four cities (Beijing, Wulhan, Suzhou, and Shenzhen) shows that, on average, about 75% of labor migrants employed by the enterprises live in institutionally provided dormitories (Knight et al, 1999; Wu, 2002). Research design The empirical data for this study derive primarily from new field research, including migrant-housing surveys, focused observation, and in-depth interviews, as secondary information on the subject is extremely limited. Studying two cities offers the quickest way of uncovering the impact of macrofactors on migrant housing patterns, particularly when the cities differ in their physical and policy environments. The choice of Beijing and Shanghai was made for a number of reasons. Ranked among China's largest metropolises, both have attracted similar numbers of migrants in recent years, in addition to a comparable local population base. Beijing, a northern city with harsh winters, is the political and cultural center of the country. Closely resembling a multinuclei model, the bulk of the city is organized around uniform work-unit compounds in a sprawling, somewhat homogeneous, urban form. Its housing and land systems remain heavily influenced by socialist planning. Located at the mouth of the Yangtze River, Shanghai is now a freewheeling southern city renowned for manufacturing, commercial, and trading activities. More or less following a concentric pattern, Shanghai has a history of residential differentiation and private housing ownership, dating back to the pre-1949 period. Shanghai has made more headway than Beijing in transforming its housing system.

1290 WWu Migrant-housing surveys The migrant-housing surveys provide in-depth housing information primarily concerning migrants, although a small number of locals also are included. A key rationale for the surveys is to produce several reference groups for comparisons. Two types of migrant are included in the sample: those with rural household registrations and those with urban ones. I also draw a small sample of permanent migrants who have moved to the cities with an official change of hukou. The last reference group consists of a small sample of local residents whose hukou has always been in the cities, drawn from the same neighborhoods as the migrants. Multistage stratified cluster sampling procedures were used to select respondents in the housing surveys. First, a number of districts were selected in three stratified geographical zones: central city, inner suburb, and outer suburb. Within each selected district in the central city and outer suburb, one neighborhood was picked at random. Because of the high migrant concentration in the inner suburb, two neighborhoods were picked in each selected district there. This resulted in a total of twenty-two neighborhoods as the areal sampling units in Shanghai (see figure 1) and eighteen in Beijing (see figure 2). Second, the total number of migrants in each of these Large river Central city Inner suburb Outer suburb District boundary Subdistrict boundary Figure 1. Surveyed neighborhoods in Shanghai.

Sources of migrant housing disadvantage in urban China 1291 neighborhoods was obtained from official sources. Population proportionate to size procedures were then used to determine the exact number of migrants to be surveyed in each neighborhood. In addition, a small number of migrants were selected from representative work units in different sectors. Selected migrants also needed to meet four criteria: aged 15 years or older; with nonlocal hukou; having migrated for employment reasons; and having stayed in the cities for over a month. These qualifications were intended to exclude those who were in the cities for cultural reasons (for example, training and attending conferences) and social reasons (for example, visiting families and friends, seeking medical treatment, tourism, and transient stay). The survey in Shanghai generated complete questionnaires for 1789 temporary migrants, 80 permanent migrants, and 137 local residents. It was conducted between December 1998 and March 1999 in 22 neighborhoods of Shanghai's 17 districts or counties (out of a total of 20) and 11 enterprises or institutions. The survey in Beijing was carried out between May and July 2000 in 18 neighborhoods of Beijing's 12 districts or counties (out of a total of 18) and 13 work units. The final outcome was complete questionnaires from 931 temporary migrants, 145 permanent migrants, and 154 local residents. After the surveys, a pool of longer term migrants (77 in Shanghai Central city Inner suburb Outer suburb District boundary Subdistrict boundary Figure 2. Surveyed neighborhoods in Beijing.

1292 WWu and 59 in Beijing), as well as a small number of locals, were selected and visited again for in-depth interviews. The basic demographic characteristics and geographical locations of the surveyed migrants closely resemble those of the larger floating population officially surveyed by both cities in 1997 (listed in parentheses below). For the Shanghai sample, the average age is 29.6 years (versus 29.9 in the 1997 city survey), 62.1% are male (versus 70.4), 68.5% are married (versus 67.4%), and 58.1% have attended junior high schools (versus 57.1%). For the Beijing sample, the average age is 28.6 years (versus 28.2 in the 1997 city survey), 61.0% are male (versus 66.9%), 61.7% are married (versus 57.0%), and 50.1% have attended junior high schools (versus 60.8%). In both cities, the housing surveys have a slight overrepresentation of female and married migrants. The geographical distribution across the central city, inner suburb, and outer suburb for the Shanghai sample is 37.4%: 44.6%: 18.0% (versus 41.5%: 40.3%: 18.2%) and for Beijing 20.8%: 62.4%: 16.8% (versus 16.0%: 62.9%: 21.1%). Although the stratified cluster sampling procedures did not yield random samples, a great deal of attention was paid to increasing the quality and representativeness of the data as much as possible. Geographical stratification reduced potential spatial skewness and, as described above, helped generate samples with distribution patterns comparable to those in official surveys. To capture migrants living in informal housing and work-related compounds (for example, construction sites and dormitories), a different set of sampling procedures was used for migrants in housing arrangements outside of residential neighborhoods such that the diversified nature of these housing arrangements was well reflected. Despite the relatively high quality of the survey data, there are limitations. Shanghai and Beijing may be different from other Chinese cities and, therefore, may not be representative. Even with repeated visits, migrants living in informal housing and public places are likely to be undercounted. Because of the small sample size of the locals in both cities, their attributes may not be fully representative of the larger local populations. Hence the comparisons between migrants and locals are illustrative in nature, and may not be conclusive. Supplementing these primary data sources, particularly for the locals, are results from the official 1997 Floating Population Surveys and from the 1995 1% Population Surveys in both cities. Qualitative methods, including in-depth interviews, observations, and site visits, helped to trace life-course events of migrants and community-level information. Measuring housing quality Most scholars agree that definitions of housing quality should encompass the interrelated nature of housing availability, affordability, and qualitative aspects of the residential environment (Lawrence, 1995). In this study, two critical measures of migrant-housing quality were used: housing choice or access, and housing conditions. Because the urban housing sector has been undergoing drastic reforms, housing price and rent levels are constantly changing. As a result, no reliable data are available for measuring housing affordability. Even though the average purchase price for commodity housing can be used as a proxy for affordability, the extremely small number of migrants in homeownership in cities will inevitably create bias. Therefore, housing affordability was not analyzed systematically in this study. The classification of housing type is complicated by the fluidity of China's housing system, as discussed earlier. The simple tenure choice between rental and ownership does not fully capture the complexity. In this research, as well as in several official surveys of migrants (Wang, 2001; Zhang, 1998; Zhou, 1996), key categories include the following: private housing (older housing units owned by individuals), commodity housing

Sources of migrant housing disadvantage in urban China 1293 (new housing units acquired through market mechanisms), renting private housing, renting public housing and living in dorms or worksheds provided by employers. Some migrants stay with urban residents, and a small number of migrants live in other types of accommodation (including hotels, self-built sheds, boats, and public places). Two indicators were used to measure housing conditions. The first is size, or per capita usable space, which includes living space and space for bathroom and kitchen facilities. This indicator has long been used within China to gauge progress in housing provision. The second indicator is a composite of seven qualitative aspects of housing, called the qualitative index. These seven aspects use a single scale: electricity (0 ˆ no, 1 ˆ yes), water (0 ˆ no, 1 ˆ yes), piped gas or propane (0 ˆ no, 1 ˆ yes), kitchen (0 ˆ no, 1 ˆ shared, 2 ˆ private), bathroom (0 ˆ public only, 1 ˆ shared, 2 ˆ private), type of structure (0 ˆ temporary, 1 ˆ permanent), and purpose of dwelling (0 ˆ residential and working or other purpose, 1 ˆ solely residential). Since the maximum composite score may be 9, the qualitative index is calculated as the sum of the seven aspects divided by 9. To understand the sources of migrant-housing disadvantage, two general groups of independent variables were used in the empirical analysis. Socioeconomic factors included age, education, gender, family status, household income, duration of residence in the city, and intention to stay (the last two are applicable to migrants only). The second group was institutional, including place of hukou, type of hukou, and type of work unit. Such division allows for direct tests of whether institutional factors are more important determinants of housing choice and conditions. Two sets of regression analysis were conducted to disentangle the effects of institutional and socioeconomic factors on housing choice and conditions. Housing choice is a categorical variable and therefore, logistic regressions were used to predict tenure choice (own versus rent) and rental sector (public versus private). Models were run for all population groups. Because the number of migrants in homeownership is very small, logistic regressions to predict housing tenure among migrants would not be reliable. Instead, the characteristics of homeowning migrants are shown descriptively and then compared with those of migrant renters. The effect of the same factors on housing conditions was analyzed through multivariate linear regressions, first of all population groups and then separately for migrants and locals. Because regression coefficients are unit dependent, my discussion focuses on standardized regression coefficients in assessing the extent to which each independent variable is influential (see also Fan, 2001). Empirical analysis of sources of migrant housing disadvantage Migrants make a very different range of housing choices from locals, as a result of their restricted access to the mainstream urban housing system. In both cities, homeownership is minimal (under 1%) for migrants. It is clear that the nonlocal ^ local divide is a more prominent barrier, resulting from the hukou system, than the rural ^ urban divide (see table 2, over). Renting represents the best choice for migrants, and more than half of them in both cities are renters. Private rental housing accommodates the largest number of migrants (table 3, over), especially in suburban areas that used to be, or still are, agricultural. In addition, when migrants find jobs in state and some collective enterprises, many of them also obtain access to the dormitory housing provided by those enterprises. Compared with the locals, migrants fare much worse in both indicators of housing conditions (table 2). Overcrowding seems to be a feature of migrant housing, with each person using one third of the space occupied by a typical local resident (7.8 versus 22:9 m 2 ). Migrants also tend to live in dwellings that are less well equipped with kitchen/bathroom facilities, are used for working or other purposes in addition to serving as residence, and are structurally less stable. For instance, about 69% of migrants have no access to bathroom facilities (either shared or private) inside their

1294 WWu Table 2. Housing ownership and conditions by place and type of hukou (frequency shown in parentheses). Temporary migrants (nonlocal) Locals a rate of per capita qualitative rate of per capita qualitative ownership usable index ownership usable index (%) area (m 2 ) (%) area (m 2 ) Urban 1.9 9.46 0.58 30.6 17.14 0.80 (363) (357) (362) (386) (381) (385) Rural 0.4 7.53 0.50 87.6 33.50 0.84 (2357) (2310) (2356) (129) (126) (129) a Locals include both local residents and permanent migrants. Table 3. Mean values for three population groups. Temporary migrants Permanent migrants Local residents Housing conditions Per capita usable space (m 2 ) 7.79 19.36 22.86 Qualitative index 0.51 0.78 0.83 Housing type Renting private housing 0.43 0.06 0.03 Renting public housing 0.14 0.29 0.34 Dormitory or workshed 0.33 0.08 0.01 Stay with local residents 0.04 0.01 0.01 Private housing 0.00 0.39 0.41 Commodity housing 0.01 0.03 0.06 Other housing a 0.05 0.13 0.14 Socioeconomic factors Age (years) 29.24 41.56 44.88 Gender (female ˆ 1) À 0.38 0.65 0.60 Education (years) 8.23 10.03 9.69 Marital status (reference: not married) À married with family in city 0.44 0.83 0.86 married without family in city 0.22 0.04 0.01 Household monthly income (yuan) 1275.01 1682.13 1969.36 Duration of residence in city (years) 4.03 13.46 na Plan to stay in city (plan to stay ˆ 1) À 0.41 0.88 na Institutional factors Type of hukou (rural ˆ 1) À 0.87 0.26 0.24 Type of work unit (reference: no employment) À state enterprise 0.20 0.37 0.26 collective enterprise 0.10 0.23 0.41 other types of employment 0.63 0.15 0.13 À Dummy variable. naðnot applicable. a `Other housing' choices include self-built shed, boat, hotel or inn, living on the street or in a hallway, and staying in hospital rooms for temporary migrants; and include relocation housing and housing sold by work units for local residents and permanent migrants.

Sources of migrant housing disadvantage in urban China 1295 dwellings, and 71% have no kitchen. (2) Housing conditions are worst for migrants living in the inner suburb, and the lack of kitchen and bathroom facilities is especially striking here: over 80% in both cities. This is a troublesome situation as the inner suburb is now the primary receiving area for migrants (Wang, 1995; Zhang, 1998). In addition, migrants in Beijing fare worse overall than those in Shanghai, both in size (7.1 versus 8:1 m 2 ) and in qualitative index (0.48 versus 0.53) of their accommodation. Another indirect indication of the housing problems experienced by migrants in cities is their own perception. Less than half of them are content with their housing size and facilities, but close to 90% feel happy about their commute distance (table 4). This is largely because migrants primarily choose housing based on its convenience to work or business (about two thirds of migrants in both cities responded so). The marked difference between the cities in the levels of satisfaction with housing expenses suggests that housing in Beijing may be much less affordable than in Shanghai. What is Table 4. Perception of housing situation in cities (cumulative percentages). Temporary migrants Local residents Shanghai Beijing all Housing size Very satisfied 9.1 3.4 6.6 13.3 Satisfied 39.0 44.1 41.2 43.3 Neutral 74.0 74.6 74.3 70.0 Dissatisfied 93.5 98.3 95.6 90.0 Very dissatisfied 100.0 100.0 100.0 100.0 Housing facilities Very satisfied 2.6 5.1 3.7 10.0 Satisfied 33.8 55.9 43.4 43.3 Neutral 75.3 84.7 79.4 70.0 Dissatisfied 98.7 100.0 99.3 96.7 Very dissatisfied 100.0 100.0 100.0 100.0 Commute distance Very satisfied 49.4 28.0 40.9 17.4 Satisfied 88.3 86.0 87.4 82.6 Neutral 96.1 96.0 96.1 82.6 Dissatisfied 100.0 100.0 100.0 100.0 Very dissatisfied 100.0 100.0 100.0 100.0 Housing expense Very satisfied 21.2 8.7 16.1 5.9 Satisfied 54.5 32.6 45.5 35.3 Neutral 77.3 60.9 70.5 64.7 Dissatisfied 95.5 97.8 96.4 94.1 Very dissatisfied 100.0 100.0 100.0 100.0 Compared to housing at home origin Much better 5.3 3.4 4.4 na Better 15.8 8.5 12.6 na About the same 21.1 15.3 18.5 na Worse 53.9 67.8 60.0 na Much worse 100.0 100.0 100.0 na Note: results in this table are based on in-depth interviews with a much smaller sample of migrantsð77 in Shanghai and 59 in Beijing, and 30 local residents. naðnot applicable. (2) The 1995 1% Population Surveys show that about 49% of local Shanghai residents have no bathrooms and 34% no kitchens. The numbers for Beijing residents are 35% and 13%, respectively.

1296 WWu most striking is that more than three quarters of migrants consider that their housing conditions in the cities are worse than those back home. Given the dismal conditions that migrants endure, nonetheless they express similar dissatisfaction with their current housing situations as the locals. In the analysis in the next two subsections I further investigate and empirically test the sources of migrant housing disadvantage. Housing choice With drastic reforms in the urban housing system, homeownership is increasingly attainable in Chinese cities. In Shanghai, for instance, about 60% of all families had bought their own homes by early 2001, either from their employers or from private developers (Far Eastern Economic Review 2001). Research shows that homeownership is a positive influence in people's financial standing and family lives. Given the transitional nature of China's housing system, both socioeconomic and institutional factors are important determinants of homeownership (Huang and Clark, 2002; Li, 2000b). For urban residents, job and work-unit ranks still play important roles in tenure choice, although lifecycle factors and income are becoming significant predictors. Even before the commencement of any major housing reform, the linkage between income and housing quality already existed (Logan and Bian, 1993). Institutional factors remain the foremost determinant of tenure choice when migrants are included in the analysis. The results of the logistic regression show that place of hukou (local versus nonlocal) is the single most important factor predicting homeownership (see table 5). Local residents and permanent migrants are much more likely than temporary migrants to own either commodity or private housing in the cities. This confirms my earlier theoretical analysis that the hukou system, segregating migrants from the local population, is much more important than socioeconomic factors in accounting for migrant disadvantage in housing choice. Affordable housing is particularly difficult for migrants to own in cities, as they cannot construct private housing (as they do back home) or purchase public housing units. In addition, work units still play some role in housing distribution, and employees of state enterprises are less likely to purchase commodity housing. On the other hand, a rural hukou and affiliation with collective enterprises predict a higher level of homeownership. This is primarily because local rural residents are much more likely to own private housing. As for those fortunate few migrants who do own their home; they are more likely to be older, better educated, with higher income, longer-term migrants (having lived in the destination city for five years or more), from urban origins, and with the intention to stay in the city (see table 6, over). These results confirm the general trend worldwide, that only those migrants who plan to stay in the city are willing to invest in more substantial housing. The critical difference lies in their commitment to the city because, as long as they intend to return to rural areas of origin, their behavior will be shaped differently (Goldstein, 1993; Nelson, 1976). In Beijing and Shanghai, longer term migrants who plan to stay also are more likely to have families living with them in the city, to make a better living, and to enjoy better housing conditions (table 6). When migrants have the chance of homeownership, they make their decisions based on whether they intend to settle in the cities. For example, a truck driver from Jiangxi, ``would like to buy a new housing unit primarily to get the `blue stamp' household registration [that may allow him to settle in Shanghai]'' (interview SHM51). (3) Thus, the system of granting only temporary urban residence permits to migrants is steering (3) In many large cities migrants may purchase, upon full payment at the time of sale, commodity housing of specified size and price and obtain `blue stamp' household registration that can become permanent after a specified number of years. This policy, however, favors only high-income migrants and is beyond the reach of most. Shanghai has recently discontinued the implementation of this policy.

Sources of migrant housing disadvantage in urban China 1297 Table 5. Logistic regressionöhousing choice for three population groups. Independent variable Tenure (own ˆ 1, rent ˆ 0) a Rent (private ˆ 1, public ˆ 0) b standardized coefficient odds ratio standardized coefficient odds ratio Socioeconomic Age (years) 0.190 1.032 0.039 1.007 Age squared 0.223 1.000 0.334 0.999 Gender (female ˆ 1) À 0.041 0.861 0.118 0.648** Education (years) 0.038 0.979 0.011 0.994 Marital status (reference: not married) À married with family in city 0.084 1.408 0.032 0.879 married without family in city 0.061 0.708 0.025 0.877 Household monthly income 0.064 1.000 0.075 1.000 (yuan) Institutional Place of hukou (local ˆ 1) À 1.378 405.623*** 0.394 0.107*** Type of hukou (rural ˆ 1) À 0.466 7.041*** 0.066 1.341 Type of work unit (reference: no employment) À state enterprise 0.201 0.311 ** 0.030 0.843 collective enterprise 0.126 1.943* 0.143 0.392** other types of employment 0.081 0.738 0.100 0.679 2 log likelihood of model 603.115 1629.416 Model w 2 806.961*** 359.344*** Degrees of freedom 12 12 Number of cases 1814 1576 Percentage correct 93.7 76.7 Significance levels: * <0:05; ** <0:01; *** <0:001. À Dummy variable. a `Owned' housing includes commodity housing and private housing. b `Rented' housing includes rented private housing and rented public housing. them away from considering homeownership in cities. Many of them have never given it a thought because they ``have not planned to stay in the city permanently'' (interview SHM74). The comments of a middle-aged nanny also reflect that many migrants never feel that they belong in the city: ``as the saying goes, farmers are always farmers. I do not dream of becoming a Shanghainese'' (interview SHM08). Most migrants, therefore, tend to invest their savings in building nicer houses back at home. The sentiment expressed by a shipyard worker in Shanghai is typical: ``eventually I will go home to Shandong and will remodel and decorate my house at home rather than [the] one [in Shanghai]'' (interview SHM75). Even in the rental sector, institutional factors appear to be important. Migrants are much less likely to rent public housing, and gravitate towards private rentals (see table 5). Again, the local ^ nonlocal distinction is the most significant predictor. Municipal public housing, in principle, can only be rented out to local urban residents in Beijing unless there are special approvals; whereas Shanghai has begun to allow tenants to sublease their units. In both cities, work-unit public housing is generally allowed to be subleased to employees of the same work units. As a result of these restrictions, migrants have limited access to the public housing units made available for rental by sitting local tenants. This further confirms that private rental is the only housing sector completely open and affordable to nonlocals.

Table 6. Mean values for temporary migrants. Housing tenure Type of hukou Intention to stay in city Length of residence owner renter urban rural yes no 5 5 years < 5 years Housing conditions Per capita usable space (m 2 21.81 8.65 9.46 7.53 8.65 7.20 8.31 7.57 Qualitative index 0.90 0.51 0.58 0.50 0.54 0.50 0.52 0.51 Socioeconomic factors Age (years) 34.06 29.77 30.12 29.11 29.82 28.84 32.94 27.56 Gender (female ˆ 1) À 0.65 0.40 0.39 0.38 0.35 0.40 0.32 0.41 Education (years) 10.00 8.04 10.47 7.88 8.39 8.11 8.21 8.24 Marital status (reference: not married) À married with family in city 0.76 0.63 0.38 0.45 0.50 0.40 0.59 0.38 married without family in city 0.00 0.14 0.20 0.22 0.18 0.24 0.24 0.21 Household monthly income (yuan) 3243.75 1408.59 1659.95 1215.11 1377.47 1201.58 1458.24 1190.32 Duration of residence in city (years) 7.43 4.17 3.72 4.08 4.58 3.65 na na Plan to stay in city (plane to stay ˆ 1) À 0.76 0.42 0.42 0.41 na na 0.49 0.37 Institutional factors Type of hukou (rural ˆ 1) À 0.59 0.89 na na 0.86 0.87 0.88 0.86 Type of work unit (reference: no employment) À state enterprise 0.18 0.07 0.29 0.18 0.19 0.20 0.21 0.19 collective enterprise 0.06 0.07 0.09 0.10 0.09 0.10 0.12 0.09 other types of employment 0.71 0.74 0.53 0.65 0.65 0.63 0.59 0.66 À Dummy variable. naðnot applicable. 1298 WWu

Sources of migrant housing disadvantage in urban China 1299 Table 7. Relationship between housing choice and conditions for temporary migrants. Housing choice Shanghai Beijing % per capita usable area (m 2 ) qualitative index % per capita usable area (m 2 ) qualitative index Renting private housing 49.0 8.8 0.52 32.0 7.9 0.49 Renting public housing 11.6 10.1 0.56 18.7 7.4 0.42 Dormitory or 28.8 5.8 0.54 41.6 5.7 0.46 workshed Staying with local 4.6 10.4 0.74 3.9 14.5 0.79 residents Private housing na na na na na na Commodity housing 0.7 19.6 0.88 0.4 21.3 0.98 Other a 5.4 6.8 0.39 3.3 5.4 0.37 Combined 100.0 8.1 0.53 100.0 7.1 0.48 naðnot applicable. a `Other' housing choices include self-built shed, boat, hotel/inn, living on the street or in a hallway, and staying in hospital rooms. Migrant housing choices differ to some degree between the two cities (table 7). Substantially more migrants are renting private housing in Shanghai than in Beijing, whereas more migrants live in dormitories in the capital city. The level of homeownership also is slightly higher in Shanghai. Much of the variation is, I believe, attributable to overall housing patterns in the two cities. Shanghai's private housing stock, particularly in the central city area, is relatively large thanks to its stronger presocialist legacy; whereas Beijing has more work-unit compounds built in the prereform socialist area and, therefore, a larger stock of public housing. The state sector is bigger in Beijing with more state-owned enterprises hiring migrant workers, many of whom are housed in dormitories. The capital city also has unusually high-priced commodity housing, because of the strong buying power of many government agencies and the high price level of land leases. This inevitably discourages homeownership for migrants, as well as for some local residents. (4) Housing conditions Recent reforms and increasing investment have brought marked improvements in housing conditions in urban China. For instance, per capita living space in Shanghai nearly doubled between 1990 and 2001 ( Shanghai Statistical Yearbook 2002). But migrants living in cities do not seem to benefit. Again, results from the regression analysis confirm that a nonlocal hukou places migrants at a significant disadvantage in housing conditions (see table 8, over). The effect of this institutional factor is so pronounced that it accounts for well over half of the variations both in housing size and in qualitative index for migrants. The impact of socioeconomic factors, by comparison, is much weaker, although education and income levels show sizeable influence. (4) The average commodity housing price in Beijing was 4771 yuan (US $576) per m 2 at the end of 2001 (People's Daily 31 January 2002). During the first half of 2001, commodity housing prices in Shanghai averaged 4173 yuan (US $504) per m 2 (``Property investment boom expected after the merger of domestic and nondomestic housing markets'', 21 October 2001, http://www.dtz.com.hk/ press room/pr prc101101 2.html).

Table 8. Regression on housing conditions (standardized coefficients). Independent variable Per capita usable area Qualitative index all groups temporary migrants locals a all groups temporary migrants Socioeconomic Age (years) 0.020 0.069** 0.043 0.003 0.000 0:015 Age squared 0.008 0.177 0.046 0.006 0.015 0.007 Gender (female ˆ 1) À 0.029 0.002 0.124** 0.098*** 0.149*** 0.055 Education (years) 0.078*** 0.102*** 0.074 0.111*** 0.121*** 0.151** Marital status (reference: not married) À married with family in city 0.016 0.053** 0.044 0.032 0.031 0.067 married without family in city 0.007 0.028 0.081* 0.033* 0.027 0.134** Household monthly income (yuan) 0.092*** 0.132*** 0.056 0.118*** 0.130*** 0.165*** Institutional Place of hukou (local ˆ 1) À 0.480*** 0.414*** Type of hukou (rural ˆ 1) À 0.095*** 0.054** 0.365*** 0.058** 0.099*** 0.117** Type of work unit (reference: no employment) À state enterprise 0.045 0.171*** 0.005 0.100 0.008 0.035 collective enterprise 0.117*** 0.013 0.138** 0.003 0.025 0.044 other types of employment 0.100*** 0.045 0.035 0.017 0.017 0.012 R 2 0.240 0.062 0.204 0.299 0.071 0.090 Significance levels: * <0:05; ** <0:01, *** < 0:001. À Dummy variable. Note: Regression coefficients are standardized. a Locals include both local residents and permanent migrants. locals 1300 WWu

Sources of migrant housing disadvantage in urban China 1301 The difference a local hukou can make also is borne out by a comparison of before-migration and after-migration housing conditions between temporary and permanent migrants. For permanent migrants, migration to the cities brings some improvement in housing conditions. But for temporary migrants, moving to the cities represents a big step backwards: housing space is reduced by close to three times, and the qualitative index deteriorates by half. Once in the cities, the housing conditions of permanent migrants with local hukou seem to be almost on par with those of local residents, both in size and in qualitative measures (see table 3). The housing surveys show that migrant-housing conditions are closely associated with housing choice (see table 7). The majority of migrants rent or live in dormitories, which tend to have worse overall housing conditions. When they stay with local residents, thus gaining indirect access to the urban housing-distribution system, their housing conditions improve markedly. For those migrants able to afford commodity housing, housing conditions are on a par with those of the locals. On the other hand, when access to urban housing is severely restricted, migrants have much poorer facilities while staying in self-built sheds or boats, living on the street or in a hallway, or staying in hospital rooms. These results confirm that there is a general disadvantage that applies to most migrants because of their restricted access to the urban housing-distribution system. For migrants only, income and education have significant positive impacts on both indicators of housing conditions (see table 8). Income could be expected to be influential because, even with restricted access to urban housing, migrants with higher incomes can afford to rent bigger places with better facilities. The effect of education on housing conditions is likely related to that of income. On the other hand, significant disadvantages in both indicators exist for migrants from rural origins. The distinction between rural and urban migrants is obvious for several important attributes. Compared with rural migrants, urban migrants tend to be much better educated and to earn a much higher income (see table 6); thus, the effect of hukou type may be reflecting that of education and income. Migrants working in state enterprises tend to suffer more overcrowding. This is largely because they are more likely to be living in dormitory-style housing (73% of them do so), which allows for much less personal space (table 7). Married migrants with their family and children in tow also suffer more overcrowding, indicating that urban life may be more difficult for larger families migrating together (World Bank, 1997). By and large, the division between local and nonlocal associated with the hukou system is the single most prominent predictor of housing conditions ömuch as it is with housing choice. Given the restricted access migrants experience in the cities, market-related factors such as income and education still have a significant, positive, impact on their housing conditions. In contrast, the effect of education and income is much weaker for the locals, especially on housing size. This suggests that market forces assert more control over housing sectors open to migrants (commodity housing and private rental). For local residents, on the other hand, additional institutional factors such as work-unit size and rank are sill influential (Huang and Clark, 2002; Logan et al, 1999). Also unlike migrants, a rural status actually indicates better housing conditions for the locals (table 8). This partially reflects the dual housing systems existing in Chinese cities: most rural residents have been allocated ample land on which to build private living quarters, often with private kitchens. As a result, local rural residents enjoy much more spacious housing and somewhat better facilities than do urban residents, in both cities.