The changing prevalence of housing overcrowding in post-reform China: The case of Shanghai, 2000–2010

Habitat International 42 (2014) 214e223

Contents lists available at ScienceDirect

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

The changing prevalence of housing overcrowding in post-reform China: The case of Shanghai, 2000e2010 Yina Zhang a, Jie Chen b, * a

School of Social Development and Public Policy, Fudan University, China School of Public Economics and Administration and Institute of Real Estate Research, Shanghai University of Finance and Economics (SHUFE); Phoenix Building 503, Wuchuan Road 111, Shanghai 200433, China b

a b s t r a c t Keywords: Housing overcrowding Housing poverty Subgroup decomposition Regression-based decomposition

Over the last two decades, China has experienced a drastic transformation of the housing system as well as rapid urbanization. By utilizing a pool of household-level micro data from three waves of national population census (2000, 2005 and 2010), this paper traces the evolution of housing overcrowding conditions in Shanghai since the marketization of the housing sector. We find that the overall incidence of housing overcrowding in Shanghai did not improve over the period from 2000 to 2010. The subgroup decomposition analysis shows that rural migrants consistently make up the majority of households living in overcrowded housing in Shanghai. The regression-based decomposition analysis further reveals that, even holding everything else equal, migrants are still much more likely to be subject to the risks of overcrowding than natives. We conclude this paper with discussions of policy implications. Ó 2013 Elsevier Ltd. All rights reserved.

Introduction Living in adequate housing conditions is widely accepted as one of the most important aspects of people’s lives and is often considered to be a key proxy indicator of people’s socio-economic status (UN-HABITAT, 2010). Until 1978, nearly all households in urban China lived in deprived housing conditions: the average living space per person in urban China was only 3.6 m2 (in terms of housing construction space, 6.7 m2), and 47.5% of urban households lived under extremely overcrowded housing conditions (living space per person less than 2 m2) (Hou, Ying, & Zhang, 1999). After the watershed termination of welfare housing provision in 1998, the market came to the central stage of housing provision in urban China (Wu, 2001). Over the period from 1998 to 2011, more than 9.3 billion m2 or approximately 100 million units of housing were built in urban China, leading the average housing construction space per person in urban China to improve from 18.7 m2 in 1998 to 32.7 m2 in 2011 (NBSC, 2011).

* Corresponding author. Tel.: þ(86) 21 65908835; fax: þ(86) 21 65104294. E-mail addresses: [email protected] (Y. Zhang), [email protected]. cn (J. Chen). 0197-3975/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.habitatint.2013.12.007

However, the success in expanding the urban housing stock through the market has been accompanied by a rapid increase in housing prices, making home purchases increasingly unaffordable for low-income households and particularly for young workers (Chen, Hao, & Stephens, 2010). The housing market has also experienced polarized housing consumption between different tenures and different socio-economic and demographic cohorts (Logan, Fang, & Zhang, 2010; Man, 2011). Meanwhile, since the beginning of 21st century, China has experienced rapid urbanization, and the urbanization rate currently grows at more than 1% per year (World Bank, 2012). Nonetheless, most rural to urban migrants are excluded from the formal housing market and is concentrated in so-called “urban villages” (Zheng, Long, Fan, & Gu, 2009). According to an official report, in 2011, only 37% of ruraleurban migrants were accommodated by the private rental market, and the rest mainly lived in overcrowded dorms or shanty sheds at workplaces (PFPC, 2012). Thus, although the shortage of the housing stock in urban China has been greatly alleviated over the last decade, tensions over distributional issues still plague the government. The literature has extensively studied the expanding inequality of housing consumption in the transition to markets in Eastern European countries and China (Huang & Jiang, 2009; Logan, Fang, & Zhang, 2009; Sato, 2006; Szelenyi, 1987). However, the

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

literature on the Chinese housing sector focuses mainly on the reform period or early post-reform period, but limited information is given about the situations in which the market dominates housing provision. This paper contributes to the literature by offering an updated examination of how the prevalence of housing deprivation has evolved in China over the last decade. In addition to describe a general national profile, we mainly focus on the dynamics of housing overcrowding in Shanghai by utilizing a pool of household-level micro data from three waves of national population census (2000, 2005 and 2010). The long period that the data covered allows us to trace the historical evolution of housing overcrowding in Shanghai from the early post-reform era to date. The full coverage and high creditability of census data assist us in producing reliable findings for the whole population. More importantly, the rich information of census data makes it possible to investigate household-level determinants of housing scarcity. The remaining sections of this paper are organized as follows: we first provide a brief literature review of the research on housing poverty and housing overcrowding, followed by explaining the methodology used in this paper. Later, we provide a brief introduction of housing deprivation in urban China and Shanghai and then examine the general profile of housing poverty in Shanghai with a series of housing poverty indicators. Further, we decompose poverty by subgroups, and multi-dimensional poverty indicators are also applied, followed by a multivariate analysis of the determinants of the likelihood of housing overcrowding. Finally, we conclude this paper with summary of the key findings and major policy implications. Literature review of housing poverty and overcrowding The concept of housing poverty The right to adequate housing has been long recognized as an important component of human rights in a number of international human rights instruments. Clarified in the 1991 General Comment No. 4 by the UN Committee on Economic, Social and Cultural Rights, the human right to adequate housing is derived from the right to an adequate standard of living and is believed to be of central importance for the enjoyment of all economic, social and cultural rights (UN-HCHR, 1991). According to Sen (1999), poverty is by nature multifaceted and the dimensions of poverty go far beyond inadequate income. To complement the limitation of “income poverty”, the concept of “housing poverty” was proposed in the UN-HABITAT’s (1996: 109) Global Report on Human Settlements 1996. The indicators of housing poverty are believed especially useful to measure the quality of household wellbeing in the absence of reliable poverty statistics based on people’s incomes and assets (UN-HABITAT, 1996). The literature has widely suggested that housing poverty may greatly contribute to aggravate the situations of income and wealth poverty (Matlack & Vigdor, 2008; Sato, 2006; Stephens & Steen, 2011; Zheng et al., 2009). Housing poverty is thus relevantly linked to the discussions of poverty because it can be seen as both a consequence and a source of poverty (Galster, 1987). As advocated passionately by the UN-HABITAT (2010), ensuring adequate housing is an effective means by which to alleviate poverty. Housing overcrowding: concepts and measurement Nonetheless, housing itself is also a multi-dimensional good. Housing is not just a question of four walls and a roof (UN-HABITAT,

215

2010). According to the UN-HCHR (2009) (the Office of the United Nations High Commissioner for Human Rights), major aspects of the right to adequate housing under ICESCR include habitability, accessibility to service, affordability and the security of tenure. The literature has long attempted to measure the quality of housing services (and then the extent of housing poverty) with a single composite indicator, but so far, no widely accepted consensus has been achieved (Fiadzo, Houston, & Godwin, 2013; Morris, Woods, & Jacobson, 1972). However, as agreed by many international organizations and government agencies worldwide, the most important element of housing poverty is the inadequate space of housing occupation (Blake, Kellerson, & Simic, 2007; UK Parliament, 2003; WHO, 2000: 5e14). Overcrowding reflects the inadequacy of the basic human need for shelter, and an insufficient amount space per person is fundamentally detrimental to people’s wellbeing (UNDP 2000). The overcrowding indicator is also one of key criteria used by UN-HABITAT (2010) to define an urban area as “urban slum”. The WHO (2000) confirms that the likelihood of disease transmission greatly increases in overcrowded environments. Overcrowding standards are either based on persons per room, space per person or both. For example, the UN-HABITAT (2010) defines overcrowding as more than three persons sharing the same room. The standard of overcrowding used in the UK Census is also room-based, which assumes that every household requires a minimum of two common rooms (excluding bathrooms). On the other hand, the indicator of living floor space per person is a key input to produce the Multidimensional Poverty Index (MPI) published by the United Nations Development Programme (UNDP, 2010). As recommended by the WHO (2000), overcrowding can probably best be measured in terms of the average living area per person in the place of residence. While the literature has yet to yield a single, widely accepted standard for a space-based overcrowding indicator (Blake et al., 2007), the WHO’s (2000) accepted standards for floor space are 7e9 m2 per person. Further, a recent US-HUD report proposes to measure the US standard of overcrowding with the criteria of 165 ft2/person (roughly 15 m2 per person) (Blake et al., 2007). In the UK Parliament (2003) Housing Bill, the space standard of overcrowding deems a minimum of 90 ft2 (roughly 8.1 m2) of floor area space for one person. Research methodology Overcrowding indicators This paper considers housing overcrowding to be the major dimension of housing poverty. Thus, in the following sections, the terms “housing overcrowding” and “housing poverty” are used interchangeably unless there are specific notations. For this reason, the analysis techniques in this paper are largely based on the anatomy of poverty studies. The Foster, Greer, and Thorbecke (1984) (FGT) class poverty index is a generalized class of poverty measures and is widely used in the literature. The FGT-class poverty index takes into account not only the headcount frequency (or ratio) of the “poor” but also the shortfall depth and inequality distribution of the “poor” (Foster et al., 1984). Major advantageous properties of FGT-class poverty indices include their “subgroup additive decomposition” (the overall level of poverty as a weighted average of subgroup poverty) and “subgroup consistency” (the overall level of poverty falls whenever poverty decreases within some subgroup and remains unchanged in the rest of the population) (Foster & Shorrocks, 1991).

216

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

The FGT-class overcrowding index can be denoted as follows:

Pðz; aÞ ¼

 H  1 X z  yi a N i¼1 z

(1)

where the parameter z is the agreed-upon overcrowding line (i.e., 8 m2/person), N is the total number of the population, H is the number of people being defined as overcrowded and yi is the individual’s housing space value; finally, the parameter a represents the degree of “aversion to poverty”, which measures the sensitivity of the overcrowding index to the depth of overcrowding. A larger a gives more weights for those with the severest overcrowding. It is easy to discern that the FGT-class overcrowding index is simply the overcrowding incidence ratio when a ¼ 0, the overcrowding intensity index when a ¼ 1 and the overcrowding severity index or overcrowding inequality index when a ¼ 2. The overcrowding incidence ratio implies overcrowding risk or exactly the fraction of the population that lives below the overcrowding line. The overcrowding intensity index captures the overcrowding depth or the average overcrowding gap, i.e., how far the deprived households’ housing conditions are below the overcrowding line on average. The overcrowding inequality index combines information on both overcrowding and inequality among those deprived as defined by the overcrowding line.

Decomposition of poverty by groups The FGT-class index is decomposable across K mutually exclusive groups (Foster et al., 1984)

Pðz; aÞ ¼

K X

qk Pk ðz; aÞ

(2)

k¼1

where qk is the population share of group k, and Pk(z,a) is the FGT index for group k.

Intuitional background: housing development in China and Shanghai Housing development in China and Shanghai Upon gaining power in 1949, the Chinese Communist government nationalized the ownership of land in cities and established public ownership over nearly all new housing stock in urban areas (Man, 2011). However, this state provision not only produced a serious housing shortage but also became a vital impediment to economic growth. Facing mounting pressures from housing shortages as well as budgetary burdens, the Chinese central government launched a nationwide housing reform in 1978 (Wu, 2001). In 1998, China’s employer-based welfare housing program was formally abolished, and the overwhelming majority of the public housing stock was quickly privatized in the early 21st century (Wang & Murie, 2011). The liberation of the housing sector fueled a property market boom and high rates of housing construction. The similar evolution of housing development was witnessed in Shanghai. In the history of Shanghai, housing conditions were always an acute issue (Chen et al., 2010). In 1949, the average living space of urban residents was just 3.9 m2 per person, and there were 1.15 million poor persons crowded in a 3.23 million m2 shantytown (Shanghai Statistics Yearbook, 1985). Nonetheless, investment in residential housing was scarce in Shanghai under the welfare housing system. A survey in 1985 showed that nearly half of the city’s 1.8 million households were living in overcrowded housing, 216,000 households had an average per person living space under 4 m2, and over 15,200 of those households had less than 2 m2 (MOST, 1995). The end of the welfare housing system in 1998 paved the way for the development of a full-fledged neoliberal housing market in Shanghai (Wei, Leung, & Luo, 2006). Together with the forces of globalization, the neoliberal development of the housing market stimulates extensive housing investment in Shanghai (Wu & Barnes, 2008). By the end of 2011, the total housing stock in Shanghai amounted to 550 million m2, which is 13.4 times greater than the housing stock in 1978 and 2.9 times greater than the housing stock in 1998 (Shanghai Statistics Yearbook, 2012).

Determinants of overcrowding

Measurement of housing overcrowding in China and Shanghai

Because the indicator of whether living in overcrowded conditions or not is a binary variable (valued at either 0 or 1), we apply the binary choice models. The basic form of the binary choice model can be succinctly written as follows:

The current paper uses housing space per person (in m2) to measure the amount of housing service consumed. Several previous studies have adopted the same strategy to analyze housing inequality in early post-reform urban China (Huang & Jiang, 2009; Li, 2009: 511e521). As argued by Li (2009), such a measure would facilitate the inter-temporal comparison of the state of housing wellbeing between households. Furthermore, China has experienced an unprecedented construction boom since the 1990s and the current housing stock in urban areas is predominated by relatively new dwellings; in 2010, more than 78% of dwellings were built after 1990, and 43% had been constructed since 2000 (NBSC, 2012). Thus, there is a high degree of quality standardization of the housing stock that will be compared in this study. For the case study of Shanghai, we deem those families with housing space per person less than 8 m2 to be families living under the housing poverty line or suffering from housing overcrowding. We refer to the WHO and UK standards of overcrowding to choose this criterion (cf. Section 2). This definition is also based on the minimum requirement for living floor space per person specified in The Administration Rules of Rental Housing in Shanghai (effective

Yi* ¼ bXi þ ε1

(3)

In Equation (3), Yi* represents the latent indictor for observation i’s binary outcome, Xi is a set of explanatory variables that will directly influence household i’s probability of living in overcrowded conditions, and εi represents the random residual term. The associated relationships between the binary variable Yi and the corresponding latent indicator Yi* will be as follows:

( Yi ¼

0; if Yi*  0 1; if Yi* > 0

(4)

The typical modeling method in such a situation is either the Logit Model (the random residual terms are assumed to have a logistic distribution) or the Probit Model (the random residual terms are normally distributed).

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

since October 1, 2011) (Shanghai Municipal Government, 2011). This legal document stipulates that the property owners of rental housing have the obligation to ensure that their tenants do not live in overcrowded housing: every tenant should have at least 5 m2 of living floor area space. Because 1 m2 of living floor area space corresponds to approximately 1.6 m2 of housing space (the latter takes account of common areas while the former does not), the standard of 8 m2 of housing space per person is thus not only a reasonable but also a legally based housing poverty line. Housing overcrowding in China and Shanghai: aggregate data According to the sixth national census (2010) (NBSC, 2012), the total construction space of the occupied housing stock in urban China doubled from 10.3 billion m2 in 2000 to 20.3 billion m2 in 2010. However, as the urban population increased by 44.3% throughout 2000 and 2010, the average housing space per person in urban China only achieved a 35.6% growth over the same period (22.36 m2 in 2000 and 30.33 m2 in 2010). The growth of the average housing space per person in the urban area of Shanghai was even lower, increasing by only 20.0% from 21.52 m2 in 2000 to 25.84 m2 in 2010 (cf. Table 1). Compared with housing conditions in 2000, the proportion of housing-challenged households (families with housing space less than 16 m2 per person) in urban China dropped heavily; the proportion plunged from 39.0% in 2000 to 23.5% in 2010. Nonetheless, the proportion of households experiencing housing overcrowding (families with housing space less than 8 m2 per person) remained high, decreasing only moderately from 11.6% in 2000 to 8.4% in 2010 (cf. Table 1). There are vast variations of the prevalence of overcrowding across regions in China. For example, among the four major cities that have province rankings, the overcrowding incidence in 2010 was only approximately 7% in the urban areas of Tianjin and Chongqing but almost double that in Beijing and Shanghai. Compared with Tianjin and Chongqing, Beijing and Shanghai also achieved a much lower alleviation of housing overcrowding over the period between 2000 and 2010. Particularly, the overcrowding prevalence in Shanghai dropped only 1% during this period. Thus, although the latest census data suggest that there has been a great improvement of average housing conditions among households in urban China under the market-oriented housing development, the distribution of housing improvement was quite uneven. Furthermore, the prevalence of housing overcrowding is still high by any reasonable standard, particularly in large cities where the migrant population grows fast, i.e., Shanghai and Beijing. In terms of housing space per person, the housing conditions in the rural areas of China are not very different from that in urban China (cf. Table 1). However, the prevalence of housing overcrowding is much lower in rural areas. For example, in 2010, only 3.5% of rural households were living in overcrowded housing (source: the sixth census database). Considering the essential difference in housing consumption between urban residents and rural residents, this paper focuses mainly on the situation of housing in urban areas. Nonetheless, the housing conditions in rural areas will be used as a comparison reference whenever needed.

217

over the period between 2000 and 2010. Our sample data consist of three sets of micro data that were randomly extracted from the Shanghai population census databases of 2000, 2005 and 2010, respectively. Specifically, we have 5559 households (15,682 individuals) for the year 2000, 27,939 households (77,677 individuals) for the year 2005 and 20,000 households (55,169 individuals) for the year 2010. The literature has long paid attention to the substantial gap in housing consumption between migrants and natives in China (Mobrand, 2006; Sato, 2006). Following the standard used in the literature (Huang & Jiang, 2009; Zheng et al., 2009), a household is defined as a migrant family if the head of the household does not have local Hukou (registration status). However, according to the nature of the head of the household’s Hukou, Chinese households can also be divided into urban residents (with non-agriculture Hukou) and rural residents (with agriculture Hukou). Therefore, we classify the whole resident population into four distinctive categories: urban natives, rural natives, urban migrants and rural migrants. The demographic structure and the distribution of housing space across the four subgroups among our sample data are shown in Table 2. This table shows that the population share of migrants in Shanghai has significantly increased since 2000, rising from 19.4% in 2000 to 27.5% in 2005 and further soaring to 39.6% in 2010. Meanwhile, rural migrants also make up a growing share of the migrant population: jumping from 64% in 2000 to 79% in 2005 and keeping constant at 80% in 2010. One can verify that the demographic distribution of our sample data is very close to the aggregate profile of the whole Shanghai population in the years analyzed (SSO, 2012). Table 2 suggests that a large inequality persisted in the housing conditions within the resident population in Shanghai over the period from 2000 to 2010. For the whole population, the Gini coefficient of housing space per person was 0.3982 in 2000, slightly rising to 0.4127 in 2005 and further increasing to 0.4326 in 2010. The finding that the inequality of housing consumption in Shanghai does not significantly enlarge under the market-deepening era is broadly consistent with the findings for the case of Guangzhou by Li (2012). At first glance, a persistently large gap in the housing conditions existed between natives and migrants. For example, in the year 2010, while the mean housing space per person among natives was as high as 33.8 m2, the mean housing space per person was only half this figure (17.3 m2) for migrants. However, if we compare only urban natives and urban migrants, we find that the difference in the average housing space per person was relatively small in any census year. In addition, we also find that the inequality of housing conditions was consistently greater among migrants than among natives. Within the four demographic subgroups, rural migrants had the largest Gini coefficient of housing space distribution (cf. Table 2). Limited by space, the current paper focuses mainly on the theme of overcrowding, and the topic of housing inequality will be explored in an accompanying paper. After showing a general picture of the housing conditions in Shanghai over the period from 2000 to 2010, we analyze the dynamics of three dimensions of housing overcrowding (incidence, intensity and inequality) in the following section. Housing overcrowding indicators and their decompositions

Data sources and sample description To investigate in detail the changing prevalence of housing overcrowding in post-reform China, we study the case of Shanghai and examine the dynamics of housing overcrowding in Shanghai

The housing overcrowding incidence ratio The incidence of housing overcrowding is sensitive to the criteria of overcrowding. However, the gap in the overcrowding incidence across subgroups remains largely unchanged within any

218

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

Fig. 1. The incidence ratio of living in overcrowded housing in Shanghai (2000, 2005 and 2010). Note: 0 stands for native group and 1 stands for migrant group.

reasonable range of overcrowding standards. See, for example, how the gap in the overcrowding incidence ratio between natives and migrants varies with the overcrowding line of each of the three census years (cf. Fig. 1). Thus, whereas our analysis of housing overcrowding is mainly based on the standard of 8 m2 per person, it has the capability to capture major characteristics of housing overcrowding pattern in Shanghai that are not specific to a certain overcrowding line. Table 2 reveals a simple fact: quite unexpectedly, the overall incidence of housing overcrowding in Shanghai did not improve over the period from 2000 to 2010 but instead generally became worse: the incidence rose sharply from 12.9% in 2000 to 17.2% in 2005 and slightly decreased to 15.7% in 2010.1 We also find the inverted U-shape of the evolution of overcrowding incidence also appears within the migrants but not for the case of natives (cf. Table 2). For urban natives, the overcrowding incidence consistently dropped from 11.5% in 2000 to 9.2% in 2005 and further decreased to 5.5% in 2010. The overcrowding incidence within rural natives was always marginal, approximately 1e2% in any of census year. Our decomposition analysis shows that in 2000, approximately 52.7% of overcrowding was experienced by urban natives, but this share dropped to 31.5% in 2005 and further declined to 18.5% in 2010. On the other hand, the share of rural migrants in the overcrowded population soared from 39.3% in 2000 to 61.2% in 2005 and further increased to 75.5% in 2010. This finding suggests that if the government has an aim to eliminate the incidence of aggregate

1 The overcrowding incidence for Shanghai in 2000 by our sample data (reported in Table 2) is lower than that of Census Bulletin (reported in Table 1): 12.9% vs. 15.6%. Except for the possible sample attrition bias, this result could be because our sample data contains all households, while the Census Bulletin concerns only households with a family status and ignores those with collective Hukou; the latter represents roughly 8% of the total population.

housing overcrowding, rural migrants should become the major target group. The housing overcrowding intensity index As Table 2 shows, the overcrowding intensity index for the whole sample rose from 0.0392 in 2000 to 0.0562 in 2005 and then remained stable at 0.0511 in 2010. The case of 2010 may be used to interpret in the economic implications of the overcrowding intensity index: to eliminate aggregate housing overcrowding completely in this year, the municipal government of Shanghai would need an average increase of 0.4 m2 (0.0511 times 8 m2) of housing space per resident. However, this implication is equivalent to require an additional 9.2 million m2 of housing space to be supplied in Shanghai (0.4 m2 times 23.02 million residents). If, however, the housing overcrowding line is set at 15 m2 per person, a level that still qualifies for applying for Economic Comfortable Housing in Shanghai (SHFG, 2009) or a level that is close to the US standard of overcrowding (Blake et al., 2007), the overcrowding intensity index is 0.1554 in 2010. This suggestion would require 2.331 m2 per resident or 53.66 million m2 of newly constructed housing space in total to eliminate the housing overcrowding in Shanghai. Such a number is about 3 times the average annual new housing supply in Shanghai, which suggests that there is still a great deal of room for further housing investment in Shanghai. The housing overcrowding inequality index Table 2 shows that the historical evolution of the housing overcrowding inequality index largely mirrors the development of overcrowding incidence or overcrowding intensity. It also indicates that the rankings of the housing overcrowding inequality index across subgroups are broadly consistent with the rankings of overcrowding incidence or overcrowding intensity. However, we find that the severity of housing

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

overcrowding is marginal among rural natives (far less than 0.1 in any census year) but about 6 times higher among rural migrants (0.06e0.07). This finding suggests that the distribution of housing is very uneven among housing-poor rural migrants. Thus, the government should employ different intervention policies to address housing overcrowding problems across subgroups. For example, a redistributive policy that helps to reduce the inequality of the subgroup that is associated with highest overcrowding inequality index may most significantly reduce aggregate overcrowding (Araar & Duclos, 2007). Limited by space, how intervention policy should be steered to reduce housing overcrowding or housing poverty will be addressed in an accompanying paper. Regression-based analysis of housing overcrowding in Shanghai, 2000e2010 This paper has so far adopted an analytic approach to study the evolution of aggregate profiles of housing overcrowding and gaps across socio-demographic groups. However, it will be more critical to know whether the housing differential between migrants and natives is mostly a consequence of institutional barriers or a market-based result of differences in human capital attainments across subgroups. To achieve this goal, we run multivariate regression models to discover what factors affect a household’s likelihood of living in overcrowded housing. Because the status of overcrowding is a binary outcome, the logit model is used. We run logit models on sample observations for each census year. Because the estimated coefficients of explanatory variables in the logit model do not carry intuitive implications, we choose not to report the coefficients themselves but instead to report the marginal effects of these explanatory variables in the logit models. These marginal results are shown in Table 3, together with the fitness statistics of logit models. The models in columns 2e4 are regressed mainly on the characteristics of the head of the household. To check how the household’s possibility of overcrowding may be affected by the characteristics of spouses, columns 6e8 report the results of models regressed on couple-level information. Through columns 2e4 in Table 2, we find that in any census year, a household’s likelihood of living in overcrowded housing increases with family size, cohabitating with a spouse and children but decreases sharply with the increase of the head of the household’s educational attainment. In 2005 and 2010 but not in 2000, the age of the head of the household also matters, suggesting a strong cohort effect. In 2010, having a middle-aged household head (aged 35e50) would lead to the lowest likelihood of living under overcrowded conditions. The impacts of the head of the household’s industry sector and occupation changed over time. In 2000, industry and occupation did not have any significant effects on overcrowding likelihood. Some marginal differentials across particular industries appeared in 2005, but the inter-industry gap disappeared again in 2010. However, significant differences across occupations were found in 2005, and they become further manifested in 2010. This confirms that occupation-based factors are becoming increasing important component of housing inequality in China, as reported in recent literature (Huang & Jiang, 2009; Li, 2012; Logan et al., 2010). A key concern in this paper is the impact of Hukou type on overcrowding likelihood. We find that even after controlling for the impacts of the head of the household’s demographic characteristics and human capital attainments, migrants are much more likely to suffer overcrowding than natives are. Column 2 in

219

Table 3 suggests that in 2000, while rural migrants had the least access to non-overcrowded housing, urban migrants were not much different from comparable urban natives in meeting basic housing needs. However, perhaps the most striking finding in this paper, as shown in columns 3e4 in Table 3, conditional on everything else being equal, urban migrants were as likely as rural migrants to be subject to the risks of overcrowding in 2005 and 2010. Nonetheless, we refrain from attributing the significant attainment-adjusted gap of overcrowding risks between migrants and natives to institutional discrimination in the housing market. The roughly equal average housing space per person between urban natives and urban migrants (cf. Table 2) denies the importance of this factor, if there is any. Previous studies in other Chinese cities have also provided evidence that Hukou is now playing a relatively small role in determining a household’s housing consumption (Huang & Jiang, 2009; Li, 2012). Columns 6e8 in Table 3 report the results of models regressed on couple-level information for each census year. Compared with the regressions based on the head of the household, major additional insights include: if the head of the household is native, marrying a migrant spouse significantly increases the overcrowding likelihood of this household; on the other hand, if the head of the household is a migrant, marrying a native spouse has large benefits in terms of reducing the overcrowding risk of this household. We also find that a spouse’s human capital (age, education attainment, the occupation and industry sector in which he/ she works) matters to the household’s overcrowding likelihood. This finding may not only have important policy implications but also are useful in guiding the model specifications in related research.

Summary and conclusions Over the last two decades, China has experienced a drastic transformation of the housing system as well as rapid urbanization. This change carries wide implications of investigating how these socio-economic transformations have alleviated or aggravated the prevalence of housing overcrowding in China. Using three waves of census data (2000, 2005 and 2010), this paper traces the evolution of housing overcrowding conditions in Shanghai since the marketization of the housing sector. Strikingly, we find that the aggregate incidence of overcrowding in 2010 was higher than that in 2000. In 2010, a significant fraction of households in Shanghai was living in excessively overcrowded housing (15.7% if using the overcrowding line at 8 m2 per person). We also show that there was a persistent and large gap of the prevalence of overcrowding across households with different types of Hukou. The decomposition analysis shows that approximately 70% of the total housing overcrowding in Shanghai is experienced by rural migrants. The results of household-level multivariate regressions suggest that the demographic attributes and human capital attainments of the head of the household are important determinants of the likelihood of household-level overcrowding. Particularly, a household’s risks of being entrenched in overcrowded housing drops substantially as the educational attainment of the head of the household increases. Over time, the occupation of the head of the household became increasingly important in determining the household’s chance of living in adequate housing. An extended analysis by employing couple-level information shows that a spouse’s socio-economic characteristics are also important factors in understanding the household’s housing wellbeing.

220

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

The findings in this paper rebuff the belief that the maturation of a free housing market would alleviate housing poverty automatically. Rather, Chinese policymakers should be urged to pay much greater attention to those disadvantaged groups in the housing market. As our evidence has shown, given the soaring housing prices and increasing expensive rental housing market in major cities, a large number of Chinese households, especially rural migrants, could not meet basic housing needs with their own resources. Fortunately, the Chinese government has formally recognized this acute problem and put the expansion of public housing at the top of its political agenda (Wang & Murie, 2011). The recently emphasized Public Rental Housing has included migrants as a part of its target population (Chen, Yang, & Wang, 2014). Nonetheless, how the government-led public housing supply should be strategically combined with the market mechanism to speed up the elimination of housing deprivation in China remains an issue that is open to debate (Man, 2011). Meanwhile, as suggested by Chen et al. (2010), means-tested housing subsidies may be more effective as well as cost-efficient in improving rural migrants’ housing consumption. Several studies have also investigated the major policy instruments that the Chinese government has adopted to alleviate housing affordability issues for low-income householdsdthe HPF (Housing Provident Fund) and Economic Housing (Chen et al., 2014; Man, 2011; Wang & Murie, 2011). These authors have typically found that these programs have significant deficiencies and need substantial improvements in the future. Future research is encouraged to examine these issues in detail.

Given this evidence, the rise of the incidence of overcrowding in Shanghai over the period from 2000 to 2010 seems to be an inevitable result of the drastic changes in the demographic structure. Particularly, the shares of relatively poor and lesseducated rural migrants in the total population rose from 12.4% in 2000 to 21.8% in 2005 and to 31.7% in 2010 (cf. Table 2). Nonetheless, it is the industry structure of China’s urban economy as well as China’s unique urbanization mode that lead to this specific demographic evolution (World Bank 2012). As analyzed in the previous literature (Chen et al., 2010; Logan et al., 2009; Sato, 2006; Szelenyi, 1987), the widening inequality of labor incomes in the market economy would trigger the decline in the average affordability of market housing among lowincome households under the marketization process of housing provision. Furthermore, the literature has suggested that China’s post-reform housing policy is dominated by a goal to stimulate local growth through enhancing the attractions of real estate investment (Chen, Guo, & Zhu, 2011). Thus, the local government does not have strong incentives to ensure decent housing for the population at a low cost (Man, 2011). Meanwhile, rural migrants have little political power to influence local public welfare policy under the Chinese political system (Mobrand, 2006; Zheng et al., 2009). Thus, the Chinese Hukou system and the institutional exclusion of rural migrants from the public housing welfare system have considerably exacerbated their housing situation (Wang & Murie, 2011). In Shanghai, for example, local low-income households with housing space under 8 m2 per person can access heavily-subsided Cheap Rental Housing, but migrants are completely excluded from this program (SHFG, 2009). On the other hand, we also suspect that the high overcrowding risk among migrants could also be a result of migrants’ self-selection. The increasing difficulty of accessing Hukou in large cities such as Shanghai induces more and more low-income migrant households to give up the plan to settle down in the host city permanently. Driven by the motivations to bring more money to their hometowns, low-income migrants may lower their housing consumption to the minimum possible level. Zheng et al. (2009) offered a similar explanation for the extremely meager housing consumption among rural migrants in Beijing.

Acknowledgment The research work is supported by the funding from NSFC (71173045), Ministry of Education (13JZD009), MOHURD (2012-12), SH-POPSS (2012ESH003), Fudan University 985-III Project (2012SHKXQN012). Appendix

Table 1 Housing conditions in urban China and four major cities, 2000e2010. Region

Whole area Families Persons (1000s) (1000s)

Urban area Room numbers Housing space Housing-poor Housing- Families Persons Room numbers Housing space Housing-poor Housingper person (m2) ratio (%) per family per person (m2) ratio (%) difficult (1000s) (1000s) per family difficult ratio (%) ratio (%)

Y2000 China 340,491 1,178,271 2.72 Beijing 4097 11,922 2.75 Tianjin 2977 9218 2.23 Chongqing 9142 29,484 2.48 Shanghai 5299 14,787 2.10 Y2010 China 401,934 1,239,981 3.12 Beijing 6680 16,340 2.33 Tianjin 3662 10,262 2.24 Chongqing 10,001 26,994 2.96 Shanghai 8253 20,593 2.04

22.77 21.03 19.09 26.67 24.00

9.1 14.9 11.3 6.0 15.6

36.7 38.4 43.5 41.0 26.2

131,298 408,814 2.39 3231 9079 2.30 2211 6504 1.94 3156 9241 2.23 4671 12,985 1.94

22.36 20.01 18.32 23.37 21.52

11.6 17.7 14.2 11.3 16.8

39.0 41.5 46.8 36.8 44.4

31.06 29.28 26.07 36.99 27.25

6.0 13.7 5.9 4.8 15.7

20.3 31.0 22.9 36.0 12.5

207,189 590,124 2.65 5803 13,966 2.09 2876 77,426 2.00 5086 13,646 2.53 7302 18,343 1.93

30.33 28.23 26.11 32.05 25.84

8.4 14.6 7.0 7.8 15.6

23.5 32.0 23.9 18.8 35.7

Data source: the Fifth National Census (2000) and Sixth National Census (2010) (NBSC, 2012). Note 1: In the Chinese statistical system, there are two types of households, one is family household and the other one is collective household (Jiti hu) where many people share one hukou (household registration status) collectively. The data here refers to only family households and thus the number of “persons” in the table is slightly less (ca. 10%) than the number of total persons reported in the census publications. Note 2: Housing-poor: families with housing space per person less than 8 m2; housing-difficult: families with housing space per person less than 16 m2.

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

221

Table 2 The distribution of housing space and housing overcrowding in Shanghai, 2000, 2005 and 2010. Housing space per person (m2)

Inequality

FTG Indicators of overcrowding

Subgroup overcrowding ‘share’

Mean

Meanjpoor

Mean gapjpoor

Gini

a¼0

a¼1

a¼2

a ¼ 0 (%)

a ¼ 1 (%)

a ¼ 2 (%)

100 59.10 21.47 7.00 12.43

24.90 20.72 43.21 21.74 14.93

5.57 5.74 5.82 5.80 5.31

2.43 2.26 2.18 2.20 2.69

0.3982 0.3482 0.2983 0.3245 0.4639

0.1292 0.1152 0.0151 0.1024 0.4078

0.0392 0.0326 0.0041 0.0282 0.1372

0.0162 0.0129 0.0016 0.0104 0.0607

52.69 2.51 5.55 39.25

49.16 2.25 5.03 43.56

46.90 2.15 4.48 46.47

100 59.74 12.73 5.72 21.81

24.04 24.40 45.62 21.49 11.12

5.39 5.65 5.69 5.36 5.26

2.61 2.35 2.31 2.64 2.74

0.4127 0.3545 0.2841 0.3951 0.3964

0.1716 0.0915 0.0091 0.1883 0.4813

0.0560 0.0269 0.0026 0.0621 0.1650

0.0242 0.0110 0.0010 0.0270 0.0734

31.45 1.10 6.28 61.17

28.29 1.05 6.35 64.31

26.63 0.96 6.38 66.03

100 52.74 7.71 7.83 31.73

27.23 31.47 49.40 30.48 13.98

5.40 5.48 4.85 5.58 5.37

2.60 2.52 3.15 2.42 2.63

0.4326 0.3570 0.3257 0.4174 0.4350

0.1572 0.0551 0.0156 0.1066 0.3738

0.0511 0.0174 0.0061 0.0322 0.1227

0.0235 0.0079 0.0034 0.0135 0.0567

100.00 18.48 0.76 5.31 75.45

100.00 17.94 0.93 4.94 76.20

100.00 17.68 1.11 4.50 76.71

Pop. share (%)

2000 All Urban natives Rural natives Urban migrants Rural migrants 2005 All Urban natives Rural natives Urban migrants Rural migrants 2010 All Urban natives Rural natives Urban migrants Rural migrants

Source: the fifth census database and the sixth census database.

Table 3 Marginal effects of demographic and socio-economic variables on the probability of living in overcrowded housing in Shanghai. Household head information

hpoor 2000

2005

2010

Female

0.008 (1.129)

0.006 (1.427)

0.013* (2.432)

Live with spouse Live with child(ren)

0.037*** (5.344)

0.029*** (7.395)

0.025*** (3.754)

0.004* (2.295)

Household size

2010

poor

hpoor

0.007 (1.089)

0.017* (2.469) 0.026*** (4.043)

0.053*** (11.923)

0.002 (0.398)

0.018*** (3.562)

0.004 (0.836)

Household size

0.016*** (5.024)

0.001 (0.177)

0.030*** (14.153)

0.025*** (11.572)

0.066* (2.440)

0.068 (1.574)

0.075*** (11.041)

0.045*** (3.536)

0.045 (0.612)

0.285 (1.480)

Urban migrants

0.003 (0.217)

0.260*** (11.824)

0.286*** (9.680)

Rural migrants

0.112*** (5.745)

0.283*** (22.752)

0.276*** (19.290)

Reference: Age 20 below Age 21e34 0.062 (1.361)

0.017* (2.213)

0.031** (3.198)

Age 35e50

0.035 (1.190)

0.035*** (4.251)

0.042*** (4.141)

Age 50e64

0.017 (0.471)

0.043*** (5.707)

0.031** (3.247)

Age 65 above

0.002 (0.045)

0.058*** (4.381)

0.003 (0.098)

Reference: middle school and less High school 0.005 (0.807)

0.017*** (4.109)

0.014** (3.254)

University

2005

poor Female-household 0.001 (0.157) head Live with child(ren) 0.015 (1.704)

Reference: Urban natives Rural natives 0.075*** (9.142) 0.078*** (14.553) 0.050*** (5.156)

3-year college

Couple information 2000

0.019*** (4.130)

Reference: both are urban natives 0.071* (2.055) Urban native þ migrant spouse Rural native þ native spouse Rural native þ migrant spouse Urban migrant þ native spouse

0.103*** (10.882)

0.022 (0.292)

0.060 (0.592)

Urban migrant þ migrant spouse Rural migrant þ native spouse Rural migrant þ migrant spouse HH < 40 and spouse  40

0.004 (0.210)

0.271*** (7.593)

0.007 (0.225)

0.140 (0.849)

0.171*** (5.377)

0.287*** (13.158)

0.345*** (14.315)

0.005 (0.393)

0.023 (1.938)

0.005 (0.293)

0.006 (0.769)

0.002 (0.187)

0.018** (3.083)

0.003 (0.661)

HH  40 and 0.008 (0.496) spouse < 40 0.041*** (5.072) 0.049*** (8.676) 0.044*** (7.721) HH  40 and 0.027** (2.882) spouse  40 Reference: both low educated 0.034*** (3.739) 0.074*** (15.748) 0.071*** (12.959) Low educated and 0.012 (1.188) spouse high educated

0.356*** (7.948)

0.029*** (3.368) 0.042*** (5.362) (continued on next page)

222

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223

Table 3 (continued ) Household head information

hpoor 2000

2005

2005

2010

poor

poor

hpoor

0.026** (3.029)

0.037*** (4.931)

0.033*** (3.805)

0.075*** (15.908)

High educated and spouse low educated

0.073 (1.956)

0.026 (0.751)

0.107 (1.812) 0.089 (1.642)

0.053 (1.052) 0.027 (0.652)

Both high-educated 0.040*** (4.272) 0.079*** (13.452) _Iinds_coup_1 (d) 0.017 (1.428) _Iinds_coup_2 (d) 0.025 (1.813) 0.029** (2.715)

Graduate

Reference: Agriculture etc. Mining and 0.037 (1.004) manufacture Construction 0.034 (0.706) Transportation 0.057 (0.998) and communication 0.073 (1.166) Wholesale, retail, food and hotel 0.044 (0.652) Finance, insurance and real estate Personal and 0.087 (1.002) private service 0.052 (1.009) R&D, medical and social service Governments 0.024 (0.466) and organizations Reference: Leaders and managers 0.004 (0.173) Professional and technical workers Staffs and 0.003 (0.142) clericals 0.043 (1.673) Sales and service employees 0.065 (0.990) Workers in farmingrelated sectors 0.024 (1.209) Production and manual workers Reference: city area Town area 0.029*** (4.023) Rural area 0.014 (1.432) N 3502 Pseudo R2 0.244

Couple information 2000 2010

0.083*** (11.313) 0.043** (3.184) 0.004 (0.453)

0.085 (1.723)

0.009 (0.262)

_Iinds_coup_3 (d)

0.025 (1.709)

0.012 (1.602)

0.090 (1.491)

0.035 (0.763)

_Ioccup_cou_1 (d)

0.003 (0.200)

0.050** (3.012)

0.129* (1.968)

0.030 (0.696)

_Ioccup_cou_2 (d)

0.002 (0.153)

0.030** (3.156)

0.141* (2.149)

0.051 (1.029)

_Ioccup_cou_3 (d)

0.001 (0.064)

0.038*** (3.936)

0.104 (1.559)

0.026 (0.523)

0.026 (1.625)

0.052* (2.379)

0.067*** (3.735)

0.087*** (3.807)

0.072*** (5.013)

0.093*** (4.908)

0.018 (0.520)

0.190* (2.049)

0.077*** (6.144)

0.110*** (6.431)

0.042*** (3.906)

0.024*** (6.023) 0.036*** (8.669) 13212 0.329

N Pseudo R2

2480 0.244

17791 0.298

7312 0.333

6887 0.376

Note: t statistics in parentheses; Marginal effects, for discrete change of dummy variable from 0 to 1. *p < 0.05; **p < 0.0; ***p < 0.001.

References Araar, A., & Duclos, J. -Y. (October 2007). Poverty and inequality components: a micro framework. Blake, K. S., Kellerson, R. L., & Simic, A. (2007). Measuring overcrowding in housing. Chen, J., Guo, F., & Zhu, A. (2011). The housing-led growth hypothesis revisited: evidence from the Chinese provincial panel data. Urban Studies, 48(10), 2049e 2067. Chen, J., Hao, Q., & Stephens, M. (2010). Assessing housing affordability in postreform China: a case study of Shanghai. Housing Studies, 25(6), 877e901. Chen, J., Yang, Z., & Wang, Y. P. (2014). The new Chinese model of public housing: a step forward or backward? Housing Studies. in press. Fiadzo, E. D., Houston, J. E., & Godwin, D. D. (2013). Estimating housing quality for poverty and development policy analysis: CWIQ in Ghana. Social Indicators Research, 53(2), 137e162. Foster, J. E., & Shorrocks, A. F. (1991). Subgroup consistent poverty indices. Econometrica, 59(3), 687e709. Foster, J., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures. Econometrica, 52(3), 761e766. Galster, G. C. (1987). Housing discrimination and urban poverty of AfricaneAmericans. Journal of Housing Research, 2(2), 87e120.

Hou, X., Ying, H., & Zhang, Y. (1999). To meet housing needs of all people: Major breakthroughs of urban housing system (in Chinese). Guangxi Normal University Press. Huang, Y., & Jiang, L. (2009). Housing inequality in transitional Beijing. International Journal of Urban and Regional Research, 33(4), 936e956. Li, S. (2009). Housing inequality in urban China: Guangzhou 1996 and 2005 (Vol. 3) (pp. 511e521). Espace populations sociétés. Li, S. (2012). Housing inequalities under market deepening: the case of Guangzhou, China. Environment and Planning A, 44(12), 2852e2866. Logan, J. R., Fang, Y., & Zhang, Z. (2009). Access to housing in urban China. International Journal of Urban and Regional Research, 33(4), 914e935. Logan, J. R., Fang, Y., & Zhang, Z. (2010). The winners in China’s urban housing reform. Housing Studies, 25(1), 101e117. Man, J. Y. (2011). In J. Y. Man (Ed.), China’s housing reform and outcomes. Cambridge, MA: Lincoln Institute of Land Policy. Matlack, J., & Vigdor, J. (2008). Do rising tides lift all prices? Income inequality and housing affordability. Journal of Housing Economics, 17(3), 212e224. Mobrand, E. (2006). Politics of cityward migration: an overview of China in comparative perspective. Habitat International, 30(2), 261e274. Morris, E. W., Woods, M. E., & Jacobson, A. L. (1972). The measurement of housing quality. Land Economics, 48(4), 383e387.

Y. Zhang, J. Chen / Habitat International 42 (2014) 214e223 MOST. (1995). Housing settlement project in Shanghai. Shanghai: MOST. NBSC. (2011). Series reports on China’s economic and social development achievements during the 11th five-year planning period (2006e2010) (Vol. 9). National Bureau of Statistics of China (NBSC). Available at: http://www.stats.gov.cn/tjfx/ztfx/ sywcj/t20110307_402708357.htm Accessed 01.03.13. NBSC. (2012). Tabulation on the 2010 people census of the People’s Republic of China. The Official Website of National Bureau of Statistics of China (NBSC). Available at: http://www.stats.gov.cn/tjsj/pcsj/rkpc/6rp/indexch.htm Accessed 01.03.13. PFPC. (2012). In PFPC (Ed.), Report on China’s migrant population development 2012 national population and family planning commission of P.R. China. Beijing: China Population Press. Sato, H. (2006). Housing inequality and housing poverty in urban China in the late 1990s. China Economic Review, 17(1), 37e50. Sen, A. (1999). Poverty as capability deprivation. In Development as freedom (pp. 87e319). New York, NY: Random House, Inc. Shanghai Municipal Government. (2011). The administration rules of rental housing in Shanghai. The Law Database of the Shanghai Municipal Government. Available at: http://www.shanghai.gov.cn/shanghai/node2314/node2319/node12344/ U26ai28161.html Accessed 01.03.13. SHFG. (2009). The regulation policy on cheap rental housing scheme in Shanghai (2009). SHFG (Shanghai Municipal Bureau of Housing Security and Building Management). Available at: http://www.shfg.gov.cn/fgdoc/lzzfzc/200911/ t20091111_331721.html Accessed 01.11.09. SSO. (2012). The handbook of key survey results of the sixth population census in Shanghai. Shanghai: SSO. Stephens, M., & Steen, G. van (2011). “Housing poverty” and income poverty in England and The Netherlands. Housing Studies, 26(7e8), 1035e1057. Szelenyi, I. (1987). Housing inequalities and occupational segregation in state socialist cities. International Journal of Urban and Regional Research, 11(1), 1e9. UK Parliament. (2003). Housing (overcrowding) bill. United Kingdom: House of Commons.

223

UNDP. (2000). Human development report 2000: Human rights and human development United Nations Development Programme. New York, NY: Oxford University Press. UNDP. (2010). Human development report 2010: The real wealth of nations: Pathways to human development United Nations Development Programme. New York: Palgrave Macmillan. UN-HABITAT. (1996). An urbanizing world: Global report on human settlements. UNHABITAT. UN-HABITAT. (2010). Housing as a strategy for poverty reduction in Ghana United Nations human settlements programme. Nairobi, Kenya: UN-HABITAT. UN-HCHR. (1991). CESCR general comment 4: The right to adequate housing (Art.11(1)). UN-HCHR. UN-HCHR. (2009). The right to adequate housing. UN-HCHR. Wang, Y. P., & Murie, A. (2011). The new affordable and social housing provision system in China: implications for comparative housing studies. International Journal of Housing Policy, 11(3), 237e254. Wei, Y. D., Leung, C. K., & Luo, J. (2006). Globalizing Shanghai: foreign investment and urban restructuring. Habitat International, 30(2), 231e244. WHO. (2000). Overcrowding. WHO. Available at: http://www.who.int/ceh/ indicators/overcrowding.pdf Accessed 01.06.13. World Bank. (2012). China 2030: Building a modern, harmonious, and creative highincome society. Washington, DC: World Bank. Wu, F. (2001). China’s recent urban development in the process of land and housing marketisation and economic globalisation. Habitat International, 25(3), 273e289. Wu, J., & Barnes, T. (2008). Local planning and global implementation: foreign investment and urban development of Pudong, Shanghai. Habitat International, 32(3), 364e374. Zheng, S., Long, F., Fan, C. C., & Gu, Y. (2009). Urban villages in China: a 2008 survey of migrant settlements in Beijing. Eurasian Geography and Economics, 50(4), 425e446.