- 1 China’s hukou or household registration system comprises a dual classification, firstly the differ (...)
1China is undergoing unprecedented urban transformation. Official statistics put the total urban population at the end of 2006 at 577.06 million, with an urbanization rate of 43.9 percent. The corresponding figures for 1980, when the reform was about to begin, were 191.40 million and 19.39 percent. In other words, some 386 million people were added to China’s urban areas over a period of 26 years. While the early reform years saw the revival of market towns and rapid growth of small cities in association with rapid development of rural or township and village enterprises, in more recent times especially since the mid 1990s growth has been concentrated in the country’s largest metropolitan centres [Li, 2008 ; Gu et al., 2008]. Between 2000 and 2004, the average population growth rate of super-large cities (cities with non-agricultural population1 exceeding 2 million) reached 10.77 percent per annum. On the other hand, the population of small cities (cities with non-agricultural population less than 200,000) actually decreased by an average of 2.59 percent [Li, 2008, p. 362].
2Yet, despite the immense population pressure, the housing conditions for the average urban dweller in China have registered remarkable improvements over the past two decades. According to the Statistical Yearbook of China, the per capita housing space consumed in urban areas increased steadily from 3.6m2 in 1978 to 10.0 m2 in 1985, 16.3 m2 in 1995, 20.8 m2 in 2001, and 26.1 m2 in 2005 [Li and Li, 2008, p. 384]. Admittedly, the official statistics refer to the situation of people with the proper hukou, and it is generally believed that migrants’ housing conditions are generally worse than those of other urban dwellers [Wang et al., forthcoming]. Nonetheless, the progress made is still immense. Census data also show that other aspects of housing conditions such as the availability of separate toilet and kitchen inside the dwelling unit as well as home furnishing also manifest substantial gains. Moreover, the urban homeownership rate jumped from 24 percent in 1990 to 72 percent in 2000 and further to 82 percent in 2005 [Li and Li, 2008, p. 386].
3Underlying the huge increase in per capita housing space amidst phenomenal population growth are massive investments in housing construction. The 2000 Population Census shows that 47.6 percent of the urban housing stock was less than 10 years old. In Shanghai, the largest city of the country, the figure was 51.3 percent [Li and Li, 2008, p. 387] The pace of housing construction accelerated even further in the new millennium. New housing starts by real estate developers increased from 188.0 million m2 in 1999 to 551.6 m2 in 2005, and the total number of commodity housing completions between 1999 and 2005 reached 19.88 million units [Li and Li, 2008, p. 399].
4Surely, the improvements are unlikely to be felt evenly among all segments of the population. Some would have benefited more while others less. For certain groups the housing conditions could be even worse than before. While there existed housing and other inequalities under the former socialist redistributive economy, the nature of housing inequality in China today is much more complicated, with market forces playing an increasingly influential role. The gradualist and incremental nature of China’s reform [Zhu, 1999], particularly the continuing domination of the Chinese Communist Party (CCP or Party) and the state over economy and society, however, implies that the nature of housing inequality in China is unlikely to be the same as that in Western countries.
5In China under Mao the major social cleavage was the rural-urban divide, which was underpinned by both the hukou or household registration system and a price system variously known as scissors’ difference, which favoured urban-industrial goods over primary commodities [Chan, 1992]. Only people with non-agricultural hukou would be provided with grain rations and assigned to a danwei or work unit [Lu and Perry, 1992]. The latter, which could be a state or Party organization or a state enterprise, was much more than an employer. It also acted as the single most important institution performing redistribution functions as well as exercising social and political control. The great majority of urban dwellers were danwei members, and lived in danwei-provided housing within the danwei compounds.
6While most danwei dwellings were of similar size and quality, the lopsided emphasis on production over consumption resulted in perennial severe housing shortage [Chan, 1992] Not every danwei member would be able to secure a danwei flat. Priority in the allocation of danwei housing was based on occupational rank, seniority, Party affiliation and rank in the Party, and recognized contribution to the state. Also, not every danwei was equally endowed with resources. Enterprises not in industries top on the development agenda were deprived of resources. Housing conditions in such danwei tended to be much worse than those that were emphasized [Bian and Logan, 1996]. But it was those people without danwei housing who had to stay in dilapidated dwellings inherited from pre-1949 times, the great majority of which came under the control of the municipal housing bureau upon the socialist transformation of the 1950s, endured the worst housing conditions [Li and Li, 2008].
7Accompanying the economic reform was reform in urban housing provision. Initially the reform focussed on the supply side, with the establishment of real estate companies to replace individual danwei as the main developers of new housing. Yet most commodity housing units built by the real estate companies were bought by the individual danwei for subsequent allocation as a welfare item to their workers [Li, 2000]. With the promulgation of the “Views on Nationwide Implementation of Urban Housing Reform” in 1991 and the “Decision on Deepening Urban Housing Reform” in 1994, the emphasis of the reform gradually shifted to the demand side, with homeownership promotion being a major objective. Danwei were encouraged to sell their housing stock to sitting tenants at highly discounted price and provide substantial assistance to those currently without danwei housing to purchase in the open market. Housing thus sold is termed reform housing. Established criteria such as occupational rank and seniority in the workplace were the major considerations in prioritizing the sale of reform housing and the price charged [Wang and Murie, 1999 ; Li and Li, 2008].
8In 1998, the State Council called for the ending of the welfare allocation of housing. That is to say, danwei would cease to provide subsidized housing to their workers. From then onwards, most households have to satisfy their housing needs through buying or renting in the market. Under such circumstances, wealth and income, which have become increasingly polarized under the reform, are likely to become the most important determinants of housing consumption. Although municipal governments have continued and even expanded the “comfortable housing scheme (anju gongcheng)”, under which selected housing development projects would be put on sale to low and middle income households at prices some 30 percent below prevailing market levels, such housing remained quite expensive and beyond the reach of the target population. The last two years of the twentieth century saw workers in state-owned enterprises and state and Party organizations stampeding to buy reform housing, a phenomenon variously known as “catching the last train” [Li and Yi, 2007]. Of course, there were those retirees and low income workers, as well as those who were forced to be permanently “off duty” or xiagang because of plant closures or firm restructuring in conjunction with the enterprise reform of the mid and late 1990s. Expectedly, these people could not afford to buy even with big discounts.
9In the early 2000s, one city after another, reform housing was given full property right and allowed to enter the market. With the huge difference between the market and subsidized price, this is tantamount to the conferment of windfall profits to the purchasers of reform housing, who are now able to trade their reform housing unit for spacious commodity housing in a well landscaped and well maintained segregated gated housing estate (Huang, 2005). This, together with phenomenal increases in housing price, has many times magnified the wealth inequality in urban China, which now increasingly manifests in terms of housing tenure and the value of the residential property the individual owns [Li, 2007]. It may be argued that people who did well in the former socialist planned economy are doing even much better in the present socialist market economy. Those who were left behind in the latest round of housing reform have to live with their dilapidating dwellings or wait for resettlement to remote suburbs upon redevelopment of their existing premises.
10The discussion so far has been concerned with people with the proper hukou. The reform has brought hundreds of millions of migrants from the country’s vast rural hinterlands to the major metropolitan centres. A two-class society comprising those with and without the proper hukou today characterizes all major cities in China [Chan, 2009]. The great majority of the migrant population suffer from discrimination in both the job and housing markets. In particular, subsidized housing is largely out of reach to the migrant population. And, given the high price, formal market housing is unaffordable to a large proportion of migrants. Many have to be contented with substandard housing with dubious legal status in the urban villages on the former urban-rural fringe [Jie and Taubmann, 2002].
11The highly complicated and varied housing consumption scene resulting from the different phases of economic and housing reform described above points towards not only heightened housing inequality but also both continuities and changes in the nature of housing inequality. Based on data derived from household surveys conducted in Guangzhou in 1996 and 2005, the present paper attempts to provide quantitative indicators measuring the extent of such changes. It may be noted that in the earlier survey year a large proportion of commodity housing built was first sold to individual danwei, which then rented or sold the housing to their workers. In the latter year the market dominated both the supply and demand sides. Thus, comparisons based on the two surveys allow us to gauge the difference in the extent and nature of housing inequality between a system which is primarily market based and one in which danwei and other socialist redistributive institutions still dominated the housing provision regime. In the following section I first discuss in greater detail how the two surveys were conducted and the comparability of the two datasets. I then present the empirical analysis, which is divided into two parts. The first is an overall assessment of housing inequality in the two survey years, and the second is a quantitative comparison of the nature of housing inequality between the two years.
12Both the 1996 and 2005 Guangzhou surveys were conducted through household interviews with the assistance of the Centre of Urban and Regional Studies of Sun Yat-sen University, Guangzhou. Strictly speaking, the two surveys were targeting at two different types of housing and hence populations. More specifically, the 1996 survey only covered newly constructed commodity housing, that is, housing put up by developers for sale at market price. But the 1996 survey was carried out prior to the ending of the welfare allocation of housing. While some work units continued to build housing for their workers, in Guangzhou supply-side commodification was well under way. The bulk of new housing supply was provided by real estate developers, although the main buyers remained the individual danwei. In the sample about 25 percent of the households purchased or rented directly in the open market [Li, 2000]. The rest remained within the subsidized housing sector.
13Arguably, the 1996 sample still provided a snapshot of the general housing situation of Guangzhou at the time of survey, despite its specificity. Surely, caution has to be exercised in interpreting the survey findings. By definition, all respondents interviewed were recent movers. Moreover, they were either those who were privileged enough to be assigned new apartments by the respective danwei or by the municipal housing bureau, or those who were able to purchase or rent in the open market. All these suggest a built-in bias towards the better-to-do groups. As such, the degree of housing inequality revealed by the sample is likely to be an underestimate of the extent of housing inequality for the population as a whole. However, by introducing statistical controls in a multivariate setting, it is still possible to provide reasonable estimates of the effects of individual factors on housing consumption.
14The sampling frame was constructed by reference to the list of presale commodity housing projects issued by the Municipal Land and Housing Administration Bureau over the period January 1992 - August 1994. The target sample size was 1000 housing units (and hence 1000 households). A multi-level probability proportional to size (PPS) sampling strategy was adopted. In particular, the sample was designed so that the number of housing units surveyed in each of the original eight urban districts of Guangzhou was proportional to the total number of presale housing units in the given district. Within each urban district, a number of “streets” (or sub-districts) were selected as on a randomized basis, according to the size distribution of the sub-districts in respect to the number of housing units. This geographically based sampling strategy ensured adequate representation of different types of residential areas in different parts of the city. Included in the questionnaire were a wide range of questions on various aspects of housing consumption as well as on the socio-demographic attributes and employment characteristics of each member of the household.
15In comparison, the 2005 survey was a more general survey, with the target population being all households in Guangzhou. Again, a multi-level PPS sampling strategy was adopted. Street- or Sub-district-level tabulations provided by the Public Security Bureau on the size distribution of households (including those without the local hukou) over the original eight urban districts plus the newly incorporated district of Panyu were used to construct the sampling frame. The target sample size was 1200, to be distributed according to the PPS principle over the nine urban districts surveyed.
16Comparison with data given by the 2005 National One Percent Sample Population Survey shows reasonably closed correspondence between the Guangzhou sample and the mini census (Li, forthcoming). However, the Guangzhou survey under-sampled migrants without proper hukou. This is largely because only households residing in permanent housing were interviewed.
17Housing is a multi-dimensional good. Size of dwelling, type of structure in which the dwelling is located, internal design, decoration and furnishing, dwelling age and state of maintenance, estate and neighbourhood amenities, accessibility to employment and shopping facilities, etc. are all important dimensions affecting how much and what kind of housing service a dwelling provides. While it is not absolutely essential, the construction of a composite measure would facilitate the comparison of the state of housing well being between individuals or households, and hence the assessment of housing inequality.
18Commonly employed in the analysis of housing demand is the price or rent (annualized price) of the dwelling unit. Hedonic theory argues that the price of a dwelling, which is the outcome of the interplay between a myriad of supply and demand factors, can be seen as the summation of the values attached to individual housing attributes [Olsen, 1987 ; Straszheim, 1973]. Thus, price or value is commonly used as a measure of the composite good, housing service. However, the use of the price variable presumes the presence of a more or less competitive market, which is clearly not the case in China where the state continues to exert substantial influence on the land and housing market and where collusion between real estate developers and local government is prevalent [Zhu, 1999]. The price variable was particularly problematic in 1996 when large numbers of housing units were sold by danwei and the municipal housing bureau to sitting tenants at heavily discounted prices.
19In the present study dwelling floor area (in m2) is used to gauge the amount of housing service consumed. Admittedly this is not a perfect measure. But this is a variable readily susceptible to inter-temporal comparisons. The fact that the 1996 survey included only newly completed housing whereas the 2005 survey tried to proportionately include all types of housing could render the comparison somewhat problematic. However, Guangzhou, along with other major cities in the country, has experienced a housing construction boom unseen elsewhere in the world. The housing stock is very new, mostly built in the 1990s and beyond. The massive scale of housing construction in recent years has introduced a degree of standardization in the floor space measure for the two surveys.
20Between the two survey years the Chinese economy and society had undergone tremendous changes. Almost incessant double-digit growth over the period not only resulted in a general increase in income, but it also brought about much greater concentration of wealth towards business proprietors, managers of major state-owned enterprises, and people with access to valuable business and state information and with close connections to government at different levels. Furthermore, the enterprise reform of the 1990s and the growth of foreign-funded enterprises, which tended to reward the top managers and professionals remuneration packages mimicking those in their home countries, led to much enlarged income spreads. On the housing front, the 1998 reform ended the welfare allocation of housing. Since then, wealth and income have probably become the single most important factors underlying housing consumption. All these argue for a substantial increase in housing inequality.
21Figures 1a and 1b plot the distribution of dwelling floor space in the two samples. The mean floor space consumed increased slightly from 58.9 m2 for the 1996 sample to 62.3 m2 for the 2005 sample. This rather small increase has to be judged against the fact that the 1996 data were only concerned with newly completed commodity housing units, which were of better quality and generally larger than the pre-1949 dwellings as well as the danwei apartments built in earlier times. Also, the 1996 sample probably pertained to the better-to-do groups. It may be safe to conclude that the overall improvement in housing consumption was substantially larger than the difference in the mean value between the two samples. The corresponding standard deviations were 22.93 m2 and 25.81 m2, respectively. Thus, while in both years there existed quite substantial variations in the level of housing consumption, the difference between the haves and have-nots appeared not to be too extreme.
Figure 1a. Distribution of Dwelling Floor Space, 1996 Guangzhou Survey
Figure 1b. Distribution of Housing Floor Space, 2005 Guangzhou Survey
22The coefficient of variation (CV), defined as the standard deviation divided by the mean value, is a commonly employed measure of inequality. The CV increased slightly from 0.39 for the 1996 sample to 0.41 for the 2005 sample. The slight increase in the CV, again, has to be judged against the difference in the target populations of the two surveys. Above it was argued that the 1996 sample would probably bias the inequality estimate downward. If this is indeed the case, then it is safe to conclude that the increase in housing inequality was likely to be even smaller than what the difference in the CV suggests. Another commonly employed measure of inequality is the Gini coefficient, which ranges from 0 (absolute equality) to 1.0 (absolute inequality). Both samples yield a Gini coefficient of 0.21, again indicating stable and relatively mild housing inequality levels, irrespective of the massive changes that had taken place both in the housing and job markets between the two survey years.
23This is a rather surprising finding. In both samples migrants were under represented. The degree of housing inequality revealed would be higher if this sampling bias was corrected. But this would not materially affect the conclusion that the degree of housing inequality in Guangzhou remained more or less the same between 1996 and 2005. Perhaps it may be noted that neither the CV nor the Gini coefficient computed takes into consideration spatial variations and hence accessibility to jobs, schools, medical services, shopping and neighbourhood amenities, and the likes.
24In Guangzhou and other cities of the country, almost all good schools and hospitals as well as up-market shopping facilities and entertainment facilities, in addition to the bulk of high status administrative and managerial jobs, are located within or close to the inner core [Li and Li, forthcoming]. Previous studies reveal strong residential preference for central locations in the city, which implies steep rent gradients [Wang and Li, 2006]. Dwellings with the same floor area in central locations and those in distance suburbs yield quite different levels of utility to the household. To the extent that better-off households are able to reside in highly priced commodity housing in redeveloped inner-city neighbourhoods while large numbers of xiagang workers and retirees displaced by redevelopment programmes are forced to relocate to far-flung suburbs lacking amenities, the actual degree of housing equality in more recent years could be a lot more severe than was indicated by the CV or Gini coefficient reported above. Moreover, market-led redevelopment has turned many inner-city neighbourhoods and former work-unit compounds to homogeneous housing estates, both in terms of socioeconomic and housing composition. As a result, the social and geographical distance between the better-to-does and the less-well-offs has probably increased by a wide margin. This increase in residential differentiation, which could greatly heighten the sense of injustice especially among the deprived, is not captured by the overall inequality measures.
25The same level of housing inequality could arise from different underlying factors. This section analyses the determinants of housing consumption in the two survey years using multiple regression. Previous works on housing demand suggest that demographic characteristics such as household size, age of the household head, and stage in the family life cycle are fundamental in determining housing needs and preferences and hence the amount of housing consumed [Clark and Dieleman, 1996]. Of course, in a market setting, income is of paramount importance. Studies conducted in the United States show that the income elasticity of the demand for housing service is in the region of 0.5, that is, an 1 percent increase in income will result in 0.5 percent increase in housing consumption, other things being constant [Glaeser et al., 2008]. Education attainment, as a measure of human capital, is a major determinant of income ; in addition, education also structures residential preferences and the ability to make appropriate housing decisions under fast-changing market conditions. Occupational status is another constituent component in assessing socioeconomic status. Multinational firms, for example, often offer subsidized housing to expatriate workers and high ranking managerial and professional staff.
26In a socialist redistributive economy, rank in the work unit and in the CCP largely determines the ability to command redistributive powers [Szelenyi, 1983 ; Bian and Logan, 1996]. Advancement along the job ladder is closely tied to formal education credentials when the ability to generate profit is not a major concern. In the case of China, studies using data of the 1980s and 1990s have found all three variables, namely, education, occupational rank and CCP membership, to be significant factors affecting urban housing consumption [Bian and Logan, 1996 ; Huang and Clark, 2003 ; Li, 2000 ; Li and Li, 2006]. Another variable which is of importance is the kind of work unit to which an individual is affiliated. Priority sectors are better endowed with resources. Also, national-level state enterprises are much better resourced than provincial-level ones, which, in turn, are much better resourced than enterprises established by city and county authorities [Bian and Logan, 2003 ; Li and Li, 2008]. Of course, there is hukou status, which assumes particular importance today, with hundreds of millions of migrants of rural origin flocking to the country’s largest metropolitan centres.
- 2 The logarithmic scale gives better fit than the linear scale. Also with income also expressed in l (...)
27To quantitatively assess how the different factors discussed above would impact on housing consumption and the extent to which their respective effects might have changed over time, I estimated two multiple regression models ; the first was based on the 1996 sample and the second the 2005 survey. In the regression models the dependent variable is dwelling floor space (m2) in logarithmic scale.2 The independent variables are quantitative measures of the said demographic, socioeconomic and institutional factors. The regression results are summarized in the following. A full listing of the regression results can be obtained from the author.
28Let us first consider the demographic attributes. Age, which correlated closely with seniority in the workplace especially under low or non-existent job mobility in the former centrally planned economy, is highly significant and positive for the 1996 equation. However, age is not significant in the 2005 equation and the coefficient estimate is close to zero. Probably this is because danwei provision of housing was no longer common in the latter year. Household size is significant and has similar effects in both equations. However, the family type variable is not significant in either equation. Having children in the family was not accompanied by a larger dwelling in a housing regime when danwei still dominated housing provision. It remains an insignificant factor in a regime in which housing consumption is primarily market determined.
29Household income is probably the most important variable gauging socioeconomic status. The 1996 equation gives an income elasticity of demand for housing of 0.14, whereas the 2005 equation gives an elasticity of 0.19. Thus, income has become more important in determining housing consumption in China over time. While the 2005 estimate is still quite low, it has to be judged against the income measure employed, which refers to current income. A more appropriate income measure would be permanent or expected income, which would yield higher elasticity estimates [Olsen, 1987].
30Education attainment is another commonly employed indicator of socioeconomic status. In 1996, people with tertiary education had much higher level of housing consumption than did people without. But in 2005, education became insignificant. Very similar findings are revealed for occupational status, another indicator of socioeconomic status. More specifically, managerial and professional workers in 1996 enjoyed higher levels of housing consumption than others but not in 2005. Evidently, with market deepening the direct effects of education and occupational rank have diminished. Surely they will continue to affect housing consumption, but only through their effects on income.
31With respect to the institutional variables, first there is danwei type, which is found to be insignificant for both 1996 and 2005. A second institutional variable is source of housing provision. In both years, housing provided by the municipal housing bureau was the smallest. On the other hand, housing obtained in the market, which did not differ too much in size from housing provided by danwei in 1996, was 14.0 percent larger in 2005. In other words, people who are able to move from municipal and danwei housing to market housing are those who gained most under the latest round of urban housing reform.
32Previous works have established the importance of CCP membership in access to housing resource. In both years, CCP membership brought about more than 10 percent increase in housing consumption for an otherwise identical household. Considering that membership in the CCP probably also affects promotion prospects, its total effect on housing consumption in Guangzhou is quite substantial, despite the drive towards marketization and the neoliberal rhetoric of merit-based rewards in conjunction with globalization.
33Finally, there is hukou status. Because of the fact that the 1996 survey was targeted only at newly constructed commodity housing, the migrants in the sample were likely to be those who had successfully established themselves in the city. As such the hukou variable is only meaningful for the 2005 equation, which shows that local hukou is associated with 11.9 percent more housing space than non-local one. Discrimination in the housing market persists in a largely market based housing provision regime.
34The results of the statistical analysis using the 1996 and 2005 survey data reveal both quite surprising findings and findings in line with expectation. Regarding the overall level of housing inequality, the two surveys yield similar CVs and almost identical Gini coefficients. Both measures indicate relative mild levels of inequality in the two years. At least in the case of Guangzhou, which was relatively advanced in its transition to a housing regime based primarily on market provision even back in 1996, the newest phase of the reform apparently has not worsened significantly the degree of housing inequality. Most people have been able to benefit from the improvement in housing space brought about by rising incomes and the massive housing construction boom in recent years.
35To the policy maker the fact that the overall level of housing inequality has remained quite mild and stable would be a pleasing piece of information. However, as was pointed out above, the gross measures of housing inequality have not taken into account spatial variations in accessibility to job and urban amenities ; nor have they taken into consideration the social distancing effects caused by increasing spatial segregation, particularly at the neighbourhood level with the increasing prevalence of homogenous gated commodity housing estates. Moreover, it is not only the objective level of inequality that is of concern. Often, the subjective perception of increasing inequality arising from, for example, the difficulty of young income earners to enter homeownership because of the massive increase in property prices, or developers’ aggressive advertising campaigns promoting super luxurious villa housing and condominiums targeting at the nouveau riches, could result in widespread dissatisfaction and even social discontent. It should also be pointed out that the market institutions were more established in Guangzhou even before the 1998 housing reform because of the influence of Hong Kong. In cities where the state sectors had a much stronger dominance over the economy and society, and where danwei provision of housing was much more entrenched, the enterprise reform and the latest marketization drive in the housing realm could bring about vastly different effects on the housing distribution. It would be useful to conduct studies using data from other cities in future to provide a comparative perspective.
36With respect to the individual factors affecting housing consumption, life-cycle effects have remained relatively weak, although household size continues to be of importance. Income has assumed greater significance. Concomitant with the increase in the income effect, education in the latter year no longer yielded significant direct effect on dwelling floor space consumed. The same can be said for occupational rank. However, CCP membership has remained a major factor affecting housing space provision, even after holding occupational status, danwei type and other variables constant, and the magnitude of influence in 2005 was more or less the same as in the 1996. Probably because of sampling selectivity bias hukou status does not show the expected result in the 1996 equation. But in the 2005 equation non-local hukou is associated with significantly lower housing space, suggesting continuing discrimination in the housing market along hukou lines. On the whole, it can be concluded that while market considerations have played a much larger role in more recent years, institutions characterizing the socialist redistributive economy continue to exert their influence in structuring housing consumption and hence housing inequality in China.
The author would like to thank Miss Becky Pang for her assistance in data analysis. He is also indebted to the anonymous referees for valuable comments. This research is funded in part by Hong Kong Research Grants Council, Grant No. HKBU242907.