The Inception of Housing Pathways in Urban China: The Declining Household Formation of Young Adults from 2011 to 2017

The homeownership rate of young adults has surged to an unprecedented level in urban China, despite rising housing prices and significant rural-urban migration. A trend analysis of nationally representative microdata shows that household formation is the missing link in the paradox and that many young adults aged 18-44 have failed to form independent households from 2011 to 2017, thereby delaying the start of their housing pathways. When factors such as socioeconomic and institutional attributes are controlled for, age differences in household formation decrease as expected. However, the age differences grow surprisingly larger over the study period, reflective of reform-induced changes in resource allocation. Further analysis demonstrates significant heterogeneity in headship status. While local young adults are squeezing into homeownership, migrants are overrepresented in the relatively stunted rental sector. Thus, while migration has brought newcomers to urban China and kept the headship rates from falling even further, institutional barriers have blocked migrants' housing pathways. Overall, the pace of change is breathtaking. There is a growing divergence in young adults' housing pathways, which depends on the timing of market entry, institutional attributes, housing prices, and personal income.

Published in International Journal of Urban Sciences The Inception of Housing Pathways in Urban China: The Declining Household Formation of Young Adults from 2011 to 2017 Zhou Yu Department of Family and Consumer Studies University of Utah ALFRED EMERY BLDG 254 225 S. 1400 E. Salt Lake City, UT 84112 Email: Tel: [e-mail] 801-585-0437 Michael Haan Department of Sociology Western University Email: [e-mail] Tel: [phone] x85110 Chengdong Yi School of Management Science and Engineering Central University of Finance and Economics Email: [e-mail] 1 ABSTRACT The homeownership rate of young adults has surged to an unprecedented level in urban China, despite rising housing prices and significant rural-urban migration. A trend analysis of nationally representative microdata shows that household formation is the missing link in the paradox and that many young adults aged 18–44 have failed to form independent households from 2011 to 2017, thereby delaying the start of their housing pathways. When factors such as socioeconomic and institutional attributes are controlled for, age differences in household formation decrease as expected. However, the age differences grow surprisingly larger over the study period, reflective of reform-induced changes in resource allocation. Further analysis demonstrates significant heterogeneity in headship status. While local young adults are squeezing into homeownership, migrants are overrepresented in the relatively stunted rental sector. Thus, while migration has brought newcomers to urban China and kept the headship rates from falling even further, institutional barriers have blocked migrants’ housing pathways. Overall, the pace of change is breathtaking. There is a growing divergence in young adults’ housing pathways, which depends on the timing of market entry, institutional attributes, housing prices, and personal income. Keywords: Household formation; headship rates; young adults; migration; institutional barriers; housing pathways Highlight: 1. It is a great puzzle that young adults have a high rate of homeownership in urban China, despite rapidly rising housing prices and large rural-urban migration. 2. This trend analysis shows a declining rate of young adults’ household formation and helps solve this puzzle. 3. It is getting harder for young adults to start their housing pathways than before, while there is a growing divergence in young adults’ housing pathways. 4. Local young adults and migrants are squeezing into homeownership and rental housing respectively. 5. Institutional barriers have blocked migrants’ housing pathways and lowered the expected demand for urban housing. 2 Introduction Within the last decade, urban China has witnessed both rapid growth and significant changes in allocations within the housing sector (Chen, Guo, and Wu 2011; Gan 2014; Li 2017; Chen and Wen 2017). Particularly notable are the rapidly increasing homeownership rates of young adults below the age of 45 which, having started at about 20% in 1990, have jumped to over 80% as of 2017 (Clark, Huang, and Yi 2019; Li 2017). This marks a major success for the country, as these rates are amongst the highest in the world (Clark, Huang, and Yi 2019; Li 2017). This is a surprising finding, however, given that urban China has experienced large increases in not only the entry costs to homeownership, but also rural-urban migration (Or 2017), both of which should have dampened the homeownership rates of young adults. A growing body of literature documents the central role of household formation in a variety of outcomes, such as land development and public utility consumption (e.g., Haurin and Rosenthal 2007; Furlong 2016). Generally, households are regarded not only as agents of social reproduction and urban transformation (Buzar, Ogden, and Hall 2005; Douglass 2014), but also as a sign of residential independence and a vital link between population and housing (Huang, He, and Gan, 2021; Paciorek 2016; Goldscheider and Da Vanzo 1989). Many researchers within this area use the term headship rate, or the number of households per person, to measure household formation and housing demands (e.g., Carliner 2003; Skaburskis 1994; Smith et al. 1984; Santi 1990), wherein higher headship rates equate to more households being formed within a given population. Similarly, Kendig (1990) puts forward a life course approach to understanding housing attainment, accentuating the role of household formation as a preparatory move toward homeownership and better housing outcomes in the 3 subsequent life course journey. Young adulthood is often an eventful period within the life course (Elder 1975; Riley 1987), as, during this time, young people make several steps along their housing pathway1 through the process of leaving the family home, becoming residentially independent, and achieving homeownership (Clark, Deurloo, and Dieleman 2003; Sweet 1990). Consequently, the headship rates of young adults increase rapidly until middle age when household dissolution becomes more common. During this initial timeframe, however, young adults, as new entrants to the housing market, are particularly sensitive to changing market conditions (Lee and Painter 2013; Forrest and Yip 2012). Using the headship housing pathway milestone, we can compare young adults within the same age groups to gauge their relative positions in their housing pathway trajectories. This is an important area of research, as the initial debt accumulated early on often has a negative effect on housing pathways in later life (Clark, Deurloo, and Dieleman 2003; Pitkin 1990). In many countries, young adults face growing difficulties in accessing housing, as they are slower than previous generations to leave the family home and live independently (McKee 2012; Mackie 2016; Cobb-Clark 2008). This delayed independence, in turn, has serious consequences on marriage, social mobility, and subsequent success in the housing market. It is surprising then that few studies have examined the propensity of young adults to form independent households, especially within the context of urban China. Often, this 1 The housing pathway is defined in Clapham’s (2005) work, and indicates an interest in “the patterns of interactions concerning house and home, over time and different localities,” which aims to build upon the idea of the housing career (p. 27). See Clapham 2005, Hirayama 2010, and Hochstenbach and Boterman 2015 for a larger discussion on the housing pathway. 4 demographic does not experience this transition and, as a result, is excluded from homeownership statistics. Part of this is due to the fact that young adults in urban China face unique challenges. For instance, the one-child policy, implemented in the 1980s, has delayed marriages, sharply reduced family sizes, and led to a substantial decline in the size of successive birth cohorts (Banister 1987; Peng 2011; Jiang, Feldman, and Li 2014, ). Young adults are also burdened by the responsibility of caring for their parents who, although being more numerous, accumulated little wealth in the socialist era (Yu 2020). Additionally, the economic reform has substantially altered the life course trajectories and housing pathways of this demographic. Regardless of their higher levels of education and income than previous generations, young adults have not only directly bore the brunt of escalating housing prices, but also have less access to the socialist redistributive system (Xie and Jin 2015; Li 2017; Deng, Hoekstra, and Elsinga 2016). Nonetheless, the literature remains unclear as to whether declining household formation has artificially elevated the homeownership rate (Yu and Myers 2010; Haurin and Rosenthal 2007). To add to the growing literature surrounding housing pathways of young adults, this study examines changes in the headship rates of those aged 18 to 44 in urban China from 2011 to 2017. This period is of particular interest as it is punctuated by rapid socioeconomic changes and temporal dynamics in household formation, alongside a roughly five percentage point increase in homeownership (Huang, He, and Gan, 2021). Our choice of age range coincides with the period in which most people will have made their residential decisions. It is also the range where there has been a tremendous surge in homeownership rates (Clark, Huang, and Yi 2019; Li 2017). By leveraging the 2011 and 2017 China Household Finance Survey (CHFS), 5 which is a nation-wide, individual-level dataset, we use trend analysis to examine how individual age groups have changed their headship rates over the study period while controlling for influencing factors at both the macro and individual level. To our knowledge, our study is among the first to conduct such an analysis. We hypothesize that: (H1) As the economic reform deepens, market abilities, rather than institutional and territorial affiliations, will explain the bulk of the age differences seen in household formation and their changes over time; (H2) the headship rates of younger age groups will increase over time while age differences decline. This is because young adults today, who have higher levels of income and education than those of the past, should have benefited more from the market allocation of resources and shown greater preferences for privacy (the opposite of this is the depreciation hypothesis, wherein the relative importance of income and education for household formation declines as a result of growing competition in the housing sector); and (H3) the floating population, or migrants who are registered outside the current city, will have lower headship rates than local residents due to the presence of fewer resources and overall lower headship rates in rural China where the vast majority of migrants reside. In the following section, we will review the literature, discuss the methods for our analysis, and summarize major data trends, while the conclusion will cover our major findings. Literature review Urbanization and economic reform Spanning more than three decades, China has witnessed rapid urbanization and significant shifts in terms of rural-urban migration (Fan 2007; Skeldon 2011; Friedmann 2005; 6 Day and Ma 1994). As shown in Figure 1, the percent of the Chinese population living in urban areas has increased from 36% in 2000 to 59% in 2017 and the growth was largely uninterrupted (National Bureau of Statistics of China 2020). Similarly, the number of migrants, or rather those who do not have permanent residency status in cities, has also increased within this timeframe, from 121 million in 2000 to 245 million in 2017 (Wu et al. 2019 ; Liang, Li, and Ma 2014; National Bureau of Statistics of China 2020). More troublesome is that the overwhelming majority of these migrants are young adults from the countryside, who are often suffering from housing deprivation in urban China (Fang and Zhang 2016; Wu and Zhang 2018; Chan 2018; Zhou 2010). (Figure 1) Within the 2010s alone, the urban housing sector has grown tremendously due to both the privatization of the public housing stock and the mass construction of commercial housing (Mak, Choy, and Ho 2007; Walder and He 2014). While urban homeownership rates are among the highest in the world, household wealth tends to be heavily concentrated in private homes. Thus, homeownership has become not only a major means of wealth accumulation, but also a source of inequality (Wu, Bian, and Zhang 2018; Li, Li, and Ouyang 2017; Ren and Hu 2016; Xie and Jin 2015). As part of the economic reform, urban China has been transformed from a state-oriented economy, in which political loyalty is prized and economic resources are allocated under a centrally controlled system, to a market economy, in which economic productivity is rewarded based on market principles (Polanyi 1957; Szelenyi and Kostello 1996; Nee and Matthews 1996). However, research on housing allocation in China has shown some mixed results. For 7 instance, some findings indicate that market abilities seem to have superseded institutional affiliations in housing allocation (Song and Xie 2014 ; Jiang 2006) while others have pointed to the persistent effect of hukou on housing outcomes, or rather the various housing challenges faced by rural migrants in urban China (Fang and Zhang 2016; Wu and Zhang 2018; Huang et al. 2017). The influencing factors of household formation The literature on household formation has been largely developed in industrialized countries to explain the dramatic household growth seen during the post-World War II era (e.g., Smith et al. 1984; Kobrin 1973; Miron 1989). One pervasive trend found is the move toward separate living in smaller and less complex households. In response to not only this trend, but also as means of illuminating the growing differences between demographic groups in industrialized countries (Burch 1995), a number of frameworks have been developed. Among these, four stand-out explanations include culture/taste, economics, family structure/demographics, and the life course. First, the culture/taste hypothesis presumes that there exists a relatively general preference for ‘privacy’ or ‘independence.’ As housing preferences tend to vary between different population groups, so too do headship rates and, as a result of changing norms and values, residential patterns (Goldscheider and DaVanzo 1986; Pampel 1983; Mulder 2013). In addition to this, educational attainment is often positively associated with not only household formation (Yu 2017; Yu and Myers 2010), but also secularization in almost all societies (Albrecht and Heaton 1984). However, it is important to note that the taste hypothesis is difficult to measure and test and, as a result, has traditionally been used as a last resort to 8 explain the residual differences in household formation between groups (Pampel 1983; Cooper and Luengo-Prado 2018). Second, economists view household status as a composite good, regarding households as both a consumer and a producer of goods and services (Goodman 1986; Carliner 1975; Becker 1993). As such, people are hypothesized to make optimal grouping decisions based on the demand for market goods and labor market conditions. Further, income and housing costs are the primary influencers of household formation, while other socioeconomic and demographic factors mediate the process (Ermisch 1981; Haurin, Hendershott, and Kim 1993). Within the context of the post-WWII era, rising real income is thought to be linked to higher levels of residential independence, as it allows people to afford separate living spaces and “consume privacy” (Michael, Fuchs, and Scott 1980). Although increased access to housing vouchers works to elevate the headship rates of young adults, especially on the demand of rental housing (Borsch-Supan 1986), the Great Recession of 2008 has had a compressing impact, thereby garnering greater policy attention, especially in the U.S. (e.g., Kiefer, Atreya, and Yanamandra 2018 ; Goodman, Pendall, and Zhu 2015; Fisher and Woodwell 2015). In the subsequent recovery period, young adults seem to have put more weight on economic conditions and become more sensitive to employment and rental costs in their household formation decisions than previous generations (Newman, Holupka, and Ross 2018; Cooper and Luengo-Prado 2018; Lee and Painter 2013; Wiemers 2014). However, while the economic approach has intuitive appeal, income and housing cost alone cannot fully explain the observed patterns of historical changes in living arrangements. In this regard, there are a few things worth mentioning. For one, rising income has led to a 9 decline in household formation (Aguila, Park, and Vega 2020). Further, there are large and persistent differences in headship rates between not only racial/ethnic groups, but also immigrants and the native-born that cannot be fully explained away by income and housing costs (Yu and Myers 2010; Haurin and Rosenthal 2007; Phua, Kaufman, and Park 2001; Lee and Edmonston 2014). Additionally, another concern is the “stable preferences” assumption. What this neglects to acknowledge is that not only is the concept of ownership contested and evolving (Wang et al. 2020; Li 2017), but the rules surrounding resource allocation are often murky (Song 2010; Logan and Bian 1993) and social norms are fluid and changing (Nee and Matthews 1996; Chen, Chen, and He 2019). Thus, when public rental housing was converted to homeownership through housing reform and renters became owners overnight, the preferences could hardly be stable, nor could the transformation of ownership status be a rational decision from the individual’s perspective. The third explanation, family structure or demographics, has contributed to understandings of household formation, especially the long-term trends toward smaller, less complex households (Kobrin 1976; Sweet 1984). Until the 1980s, population structure, such as marriage patterns, kinship, and fertility, were conducive to independent living (Burch and Matthews 1987 ; Waite, Goldscheider, and Witsberger 1986). However, while public transfers and social securities have made independent living more feasible (Berk and Berk 1983), the decline of kin-based social support has given divorced or widowed women fewer living arrangements options (Burch 1995). Nonetheless, the formally married have increased their headship rates over time, while the unmarried have changed little in the U.S. (Carliner 1975). 10 Other demographic factors, such as Second Demographic Transition, which is characterized by delayed marriages and declining fertility rates, have led to a divergence of headship rates between both countries and demographic groups (Lesthaeghe and Moors 2000; Sobotka 2008; Kuijsten 1995). The fourth and final factor, the life course approach, has also been used to explain longterm trends in household formation. Highlighted here are the variable ways through which individuals gradually improve their housing status until later in life (Kendig 1990). In this way, household formation is an important life transition and a converging point between the life course and the housing pathway (Mulder and Hooimeijer 1999; Clark 2012). Thus, geographers and housing demographers have emphasized not only the temporal dimension of household formation (Myers 1999; Pitkin 1990; Clark, Deurloo, and Dieleman 2003), but also housing adjustment inertia (Pitkin, Masnick, and Brennan 1990; Börsch-Supan 1989). While researchers disagree about the relative importance of these four sets of explanations, they acknowledge their interconnected and, in many cases, endogenous nature. Household formation in China China has a unique political structure, economic system, and cultural traditions. Therefore, the patterns of household formation are likely different from those of western countries, which have been the primary focus of the existing literature. However, rapid globalization in the past forty years has led countries to have substantial economic and cultural exchanges. Such exchanges, coupled with China's rapid urbanization and economic growth, may lead to similarities in the patterns of household formation. Our general expectation is that, as a result of economic reform, the Chinese patterns are similar to those identified in western literature. 11 Existing research in China shows a rapid growth in the number of domestic households, as well as a trend of simplification in household structure, similar to what is seen in the postWWII period in the West (Hu and Peng 2015). Additionally, these studies attest that growing gender imbalances and institutional attributes, unique to China, have distinct effects on household formation (Yu 2017). However, these tend to be based on summary data and analyses of the initial period of China’s housing reform. As such, little is known about recent changes at the individual level. This lack of research is largely due to the paucity of publicly available, nationally representative datasets. This is a gap that our current study aims to fill. Data and methods This study relies on the 2011 and 2017 China Household Finance Survey (CHFS) microdata. This is particularly useful, given the relatively long duration, the nationally representative, comprehensive, and biennial nature of the data (Gan et al. 2013). In terms of collection, the same team has carried out the survey using largely consistent questionnaires and sampling methodology over time, thereby permitting a trend study. Additionally, we include official data to control for housing prices (National Bureau of Statistics of China 2020). The dependent variable of interest is headship status, which indicates whether someone is a householder (or household head) within a population group. For each household, the person who was the most knowledgeable about the household was interviewed and designated as the householder. Importantly, there is exactly one householder per household (or occupied housing unit) at any time. The number of householders equals the number of households in any given location. Further, the unit of analysis is the individual. For the purposes of this analysis, only urban 12 residents, who can be either local hukou holders or migrants, are included. In addition, we use the 2013 and 2015 data in the robustness check to ensure that our results are stable over time and can stand up to alternative specifications. The decision to use household headship rate as a measure of housing attainment allows for a straightforward analysis. However, the householder status is not necessarily randomly assigned in married-couple households. Rather, research shows that men are much more likely to be designated as the householder within married couples (Kent 1992; Myers 1992, ). To mitigate this potential concern, we split the headship status of married couples evenly and separately identify the never married from the formerly married in the multivariate analysis. Moreover, a potential bias of the trend analysis is that membership within the sample differs. The age group comparison does not corrupt our understanding of the trend, however, as long as the sampling strategy has remained consistent, as is the case for the CHFS data. Another tool to guard against potential bias is to use multivariate analysis to control for the covariates, which will be carried out in this study. Important to note is that this trend analysis is premised on the assumption that, while young adults today are quite different from those of decades past, they have stable housing preferences in the short run (Ham 2012). In other words, young adults in urban China should have similar headship rates by age group, especially after adjusting for their socioeconomic and demographic differences. Any residual differences in headship reflect changes in the entry point of an individual’s housing pathway. With this in mind, our analysis undergoes a number of key steps. First, we present descriptive statistics for a large sample in both 2011 and 2017. Then, we focus on young adults 13 aged 18-44. Following the descriptive analysis, two sets of multivariate analysis are carried out: binomial and multinomial logistic regressions. Based on the random utility assumption, we employ discrete choice modeling (McFadden 1978). Finally, we conduct robustness checks using data from 2013 and 2015. While the binomial logistic regression result is relatively straightforward to interpret, the advantage of the multinomial logistic regression is its ability to capture the complexity of headship status in urban China. This complexity reflects the country’s socialist legacy, the growing housing market, successive government interventions in the housing market, and reform measures implemented over time. The regression model used in this analysis is specified as follows: Equation (1) H = AG + Yr + AG * Yr + HC + ECON + DEMOG + INSTITUTION where, H = the headship status; AG = age group (AG1: age 18-34 and AG2: age 35-44); Yr = changes over the study period from 2011 to 2017 by the reference group (AG2); AG * Yr = the period effect on AG1 relative to that of the reference group2 (AG2); HC = individual’s human capital level; ECON = individual’s economic characteristics including personal income (logged and adjusted for inflation), employment status, employment sector, different tiers of regions (a proxy for the level of economic development), and housing prices at the provincial and city level (logged and adjusted for inflation); DEMOG = individual’s demographic factors including marital status and sex; and INSTITUTION = individual’s migrant status and access to the housing provident fund. 2 Although it would have been possible to run more interactions, since age progression is our main focus we include only one. 14 In the first set of multivariate analysis, we follow previous studies (e.g., Mulder and Smits 2013; Santi 1990) and treat headship status as a joint decision manifested in two unranked categories: householder or non-householder (base). In the second set, we also follow previous studies (e.g., Leppel 1986; Yu and Myers 2010; Haurin and Rosenthal 2007), but use multinomial logistic regression in the analysis, separating headship status into five unranked types (non-householder/non-head = 0, householder of a renter household = 1, householder of a self-build unit = 2, householder of a special-access unit = 3, and householder of a commercial unit = 4)3. The first type of householder status, nonhouseholders, are those who may live with family members, friends, or roommates and acts as the reference category. The second type, renter householders, makes up a smaller portion of the sample because, while most of the housing stock in cities were rentals before the 1990s, this share is rather small in urban China today. The third type, self-build units, are primarily bought for consumption and are often the least preferred option for homeownership, since they are non-tradable in the housing market and often of low build quality (Logan, Fang, and Zhang 2009). The fourth type is householders of special-access housing, wherein individuals have acquired their housing units outside of the regular housing market. While special-access housing units are generally more desirable than self-build units, their tradability is often murky and their existence reflects the transitional nature of the Chinese economy. The fifth and final type of householder is that of commercial housing, which is the most desirable type of headship in urban China (Walder and He 2014; Li 2012). Compared to special-access and 3 In case an individual owns multiple housing units, we use the primary residence. 15 self-build housing units, these tend to be bigger, newer, much more expensive, and are a key component of wealth accumulation as they can be freely traded in the housing market. This second set of analyses are particularly important, as Mulder (2013) suggests that housing tenure and type are the two central dimensions of an individual’s housing aspiration. This is due, in part, to their association with various housing needs and the timing at which an individual leaves the parental home, gets married, and becomes parents. It is perhaps even more important to distinguish the different types of housing units in urban China, given the country’s adoption of many market-reform measures over the years in conjunction with the strong imprint of socialist legacy. The five types of householder status, separately categorized in the dependent variable, are very different in terms of ownership, build quality, desirability, and geographic locations (Logan, Fang, and Zhang 2009; Wu 2004). As such, we examine the coefficients of key variables that influence an individual’s householder status, measuring the attractiveness of each alternative status relative to being a non-householder. Covariates The three key variables of interest are: (1) headship differences between the 18-34 (AG1) and 35-44 (AG2) age groups, (2) the Year variable, or how the headship of the reference group (AG2) has changed from 2011 to 2017, and (3) the period effect, or how the headship of AG1 has changed over the study period relative to the changes witnessed for the reference group (AG2). If the reported odds ratio or relative risk ratio is above 1, the independent variable is deemed to be positively associated with the dependent variable. Aside from these, there are additional noteworthy measures included in the analysis. For instance, age is a critical parameter in household formation, as it is directly linked to the 16 housing pathway and reform measures implemented over time in the Chinese emerging housing market. Research has shown that housing consumption and household formation have followed a life cycle pattern in the short run (Myers 1999; Paciorek 2016; Cooper and Luengo-Prado 2018). That is, the same age groups tend to have similar rates of headship. Thus, we will make the same life cycle assumption and examine how headship rates have changed between age groups over time after controlling for the covariates. However, the variations in household formation could be a result of socioeconomic differences, demographic factors, and the unique institutional setting in urban China. As such, it is important to control for these confounding factors in multivariate frameworks so as to uncover age differences in household formation and changes over time. Moreover, independent variables of interest include human capital, as well as economic, demographic, and institutional attributes. Based on previous research, those who are highly educated, have higher levels of personal income, were formerly married, and are currently employed tend to have higher headship rates. For our analysis, we also control for the geographic location of the observation. Those in the tier-1 regions, which include Beijing, Shanghai, and Guangdong, should have higher headship rates as, not only do these residents tend to be more individualistic and independent minded (Yu 2017), but these regions have better public services and facilities, which are conducive to household formation (Wu et al. 2019). As well, the cost of housing is not only high, but increasing quickly in these areas (Deng, Gyourko, and Wu 2012). For example, from 2011 to 2017, housing prices rose by about 100% and 20% in tier-1 and tier-2 regions, respectively, but increased little elsewhere. Institutional attributes, including migrant status and access to the housing provident fund, 17 are rather unique to China. That is, migrants tend to have very limited access to public services, medical care, and subsidized housing and, in comparison, local residents are expected to have higher headship rates (Wu and Zhang 2018; Chan 2018; Cheng and Selden 1994). At the same end, having access to the housing provident fund—a special perk for urban residents who work for state owned enterprises and large private companies—should also be positively associated with household formation (Tang and Coulson 2017; Xu 2017). In what follows, we summarize the findings of our analysis based on the methods outlined above. Analysis Descriptive findings We first present summary statistics for both 2011 and 2017. Between these two observation periods, the overall headship rates for those 18 and older in urban China increased by 2.2 percentage points, as indicated in Table 1 below. Coupled with steady population growth as a result of rural-urban migration and, to a lesser extent, natural growth, there are a large number of households in urban China over the six-year period. Not surprisingly, this has led to a significant building boom as well as dramatic increases in housing prices, particularly in top tier cities (Glaeser et al. 2017; Tsai and Chiang 2019; Scutt 2017). Further, headship rates appear to increase, hitting a peak at the 45-54 age group in 2011 and the 65+ age group in 2017. Although the unadjusted age specific headship rates are below those of industrialized countries, the pattern is largely consistent. (Table 1) To outline these differences in headship rates by age group, we include Figure 2 below. 18 While those aged 18-34 saw a 7.4-point decrease in headship rates from 2011 to 2017, the 65+ group had a 4.3-point increase. Thus, it can be argued that younger age groups saw bigger drops4. Due to this trend, the remainder of the analysis will focus on young adults between the ages of 18 and 44. (Figure 2) Table 2 reports the summary statistics of young adults aged 18 to 44 in 2011 and 2017. Also included here are the variables to be used in the regression analysis. The magnitude of the changes indicated are striking. For one, the overall headship rate declined by more than five points. Other components, such as educational attainment and personal income, also saw large increases. In particular, the average personal income—after adjusting for inflation—has increased by more than four points. Further, young adults are shown more likely to be employed and work in the private sector than before, in addition to the vast majority being in married couple households. Additionally, housing prices have increased over time, as was expected, and young adults are more likely to have access to the housing provident fund. Not reported in the table are the fact that the homeownership rate of the young adults, measured at the household level, increased from 56.3% in 2011 to 80.2% in 2017. Not surprisingly, the renter share of the total households decreased significantly over the same period. (Table 2) Regression results The headship patterns of 2013 and 2015 are consistent with the trend. Due to smaller intervals, the changes are also smaller. 4 19 Moving into the models for our analysis, Table 3 reports the binomial logistic regression results, which are shown in six columns. Section I provides the results from the model inclusive of age and trend only, Sections II and III report the results for human capital and economic factors, while Sections IV and V present the results for demographic and institutional attributes, respectively. The final model, Section VI, includes all concerned variables. (Table 3) Section I investigates key factors in household formation. The results shown here are largely consistent with those reported in Figure 2, wherein there are significant age differences in household formation. With respect to headship propensities, the odds of AG1 are only 35.2% that of AG2. As expected, those of AG2 have lost ground in household formation over time, however, relative to the reference group (AG2), those of AG1 have experienced more substantial losses. Shifting to Sections II, III, IV and V, we look at the influence of human capital, economic, demographic, and institutional explanations, respectively. Overall, having lower personal income and being unemployed lowers headship probabilities, while educational attainment has a positive effect. Shown in Section IV, controlling for demographic factors partially helped explain the age differences in headship rates and the declines over time. The results of institutional attributes, reported in Section V, are also significant. As expected, having access to the housing provident fund, similar to low interest loans, leads to higher headship propensities. Contrary to our original expectation, however, migrants have higher propensities than local residents. 20 The final column looking at Section VI acts as the full model. After controlling for all covariates, the following are still strong influencers of headship: age, year, the age and year interaction, personal income, marriage, gender, and institutional attributes. Conversely, housing prices are no longer statistically significant. Other results worth mentioning include: formerly married women have nearly 6 times the odds of being householders as those who are in married couple households; migrants are more than twice as likely to be householders than local residents; and having access to housing provident fund has a positive effect. Similar to the findings in Section I, those in AG2 had a more substantial decline in headship over the study period, while those in AG1 have fared even worse. This result supports the depreciation hypothesis, contrary to our expectations in hypothesis 2. Further, with respect to the household formation rates of young adults, the relative importance of socioeconomic and institutional factors appears to have declined over time. Surprisingly, however, young adults appear largely insensitive to housing prices, employment, housing market conditions, and education. These findings, which are inconsistent with the literature (Miron 1989; Yu and Myers 2010), and our first hypothesis, may be due to the complexity of China’s emerging housing market. Another noteworthy finding is that per capita homeownership rates increased, while per capita rentership rates—the percent of individuals who are the householders of renteroccupied housing—decreased over time. If the household-based measure were used, young adults’ homeownership rates would have increased from 56.3% in 2011 to more than 80% in 2017. However, much of the increase is due to the dramatic decline in the share of renters’ present, rather than the popular imagination of increasing housing prosperities among young 21 people in urban China (e.g., Seales 2017 ; Sito and Liu 2018). That is, the share of commercial and self-build units has increased over time, while a much smaller share is living in rental units. This rapid change in tenure structure is largely due to housing reform, wherein individuals sell public rental units to private individuals who, in most cases, are the existing tenants. To better understand the formation of specific types of households, we now present the results from our multinomial logistic analyses. Table 4 reports the regression results in two sections. The first includes only the age and period effect, while the second includes all the covariates. In each section, there are four columns. The first two columns report the probability of being a householder of either a rental unit or a self-build unit, respectively, while the other two report the probability of heading a special-access unit or a commercial unit. The baseline category is the probability of being a non-householder, which is omitted from the table. As such, each reported coefficient reflects the effect of a specific characteristic on one of the four types of headship status, relative to non-householders. (Table 4) Turning first to the age and period effect only model, the results shown in Section I are largely consistent with those of past studies. That is, the older individuals are, the higher the headship probabilities. Further, the 35-44 age group (AG2) became much less likely to be a renter householder, with only 22.3% of the odds in 2011, while the propensities for other types of headship status remain largely unchanged. In Section II, the results are quite different from those of the binomial logit model reported above. After controlling for the covariates, the age gaps shrink a bit. Here, those in (AG2) are much less likely to be a renter householder from 2011 to 2017. As well, relative to 22 those in (AG2), those in (AG1) have fared worse over time in terms of access to special access and commercial housing, in addition to being less likely to receive the welfare distribution of housing than before. This group also become less likely to head a commercial unit. Other key findings include college educated individuals’ strong affinity for commercial units — the most desirable form of housing in urban China – as compared to their clear aversion for self-build units, which aligns with the taste explanation. Further, personal income is positively associated with residential independence, while job sector plays a role in the type of headship acquired. For instance, those who are employed in farming are the most likely to head a self-build unit, while those who work in the public sector have higher propensities to head a special-access unit. Thus, although the system was officially phased out in the early 2010s, it seems that public sector workers have kept benefiting from the welfare distribution of housing. Moreover, the results for housing prices are expected. Those in the tier 2 regions are less likely to head a special-access or commercial unit, while higher housing prices are positively associated with renting. Demographic factors are also relevant, with the formerly married having a much higher headship propensity than the never married. Along the same lines, institutional factors play an important role, wherein migrants are 5 times more likely to head a renter household than local residents. This explains the high rate of household formation among migrants, which is again heavily skewed toward rental. Furthermore, those who have access to the housing provident fund are much more likely to head a special-access or commercial unit, which is akin to housing vouchers in the U.S. that increase housing demand and elevate headship rates (Borsch-Supan 1986). 23 Simulations Building on the regression results above, we rely on Blinder–Oaxaca non-linear decomposition techniques (Blinder 1973; Oaxaca 1973), using the multivariate regression coefficients in 2017, to simulate four counterfactual scenarios on composition, income, institutional factors, and housing prices. The goal is to quantify their relative importance in affecting household formation. Through this process, we acknowledge four key questions: (1) What would the headship rate be if the young adults in 2017 had the same characteristics of those in 2011 (i.e., lower levels of income and education, lower likelihoods of working in the private sector, lower housing prices, and lower likelihoods of having access to the housing provident fund)? The answer is that the overall headship would have declined even further, by 2.5 points. In other words, young adults punched above their weight in 2017 and had higher propensities than in 2011. (2) What would the rates be if the income level were twice as high as the actual level in 2017, but everything else remains the same? The headship rate would only be 0.2 points higher than the actual number, thereby illustrating the relatively small role that income has played in household formation. (3) What would the rates be if there were no institutional barriers (i.e., everyone had access to the housing provident fund and were all locals)? The headship rate would have been 5.9 points higher, as a larger share of young adults would have owned a commercial housing unit. In other words, with respect to household formation, market abilities are still secondary to institutional attributes, including hukou status, territorial affiliation, and access to the housing provident fund. 24 (4) What would the rates be if the average housing prices in 2017 were reduced to the 2011 level? In this case, the headship would have been 2.5 points higher.5 Overall, these simulation results demonstrate that institutional barriers have greater effects on young adults’ headship rates than doubling income or lowering housing prices. Thus, this group has faced more headwind in household formation over time. Robustness check One of the goals of our analysis is to ensure the stability of our regression results and to examine how our conclusions change when we adjust our samples and variable specifications. Using the 2013 and 2015 microdata, we examine the extent to which the coefficients differ from those reported in Tables 3 and 4. Due to the smaller gap between our observation periods, 2013 and 2015, the age differences are smaller, but the period effects are still evident. Compared to our previous findings, the headship probability for those aged 35-44 (AG3) in 2015 is only 76.8% of that in 2013. This is a significant decline. Further, the results of the covariates mirror those reported in Table 3, as shown in the Appendix, although income is no longer significant, indicating its relatively weak role in household formation. Overall, we find our regression results to be robust and our conclusions stand when we use two different micro samples in the robustness test. Conclusion Confucius said some 2,500 years ago that, among other major life events, he established himself at the age of 30 (SanShiErLi). This famous statement has accentuated the role of independence in young adulthood. This idea of independence has gained renewed attention 5 The simulation results are available upon request. 25 in recent years, as many worry about whether the younger generation is overly spoiled, the extent to which its shrinking cohort size will affect the housing market, and the potential for exclusion of vulnerable groups in the housing market (Lian 2014; Wang and Otsuki 2015; Cameron et al. 2013). This study has examined the growing deficit that young adults suffer in their housing pathways, illustrated the complexity of China’s emerging housing market, and outlined important variations in headship status. Our findings show that, while both market abilities and institutional attributes are strong determinants of household formation, their relative importance has declined over time with market abilities still playing second fiddle. This supports the depreciation hypothesis. Further, although economic factors, such as income and housing prices, matter greatly in household formation, they do not explain why the headship rates of young adults have declined in times of economic prosperity and rapid income growth. While it is true that young adults in urban China have quickly improved their economic fortune by rapidly increasing their income and education, they have not translated their newfound market advantages into an earlier start in their housing pathways. Thus, contrary to our first hypothesis, young adults would have had even lower headship rates if they had not improved their income and education. Contrary to our second hypothesis, the age differences in headship rates grew from 2011 to 2017, with the youngest age group showing the largest decline. This is due to their lower likelihood to have benefited from the welfare distribution of housing when compared to older groups, in addition to rising income not being a sufficient compensation. Furthermore, local young adults are detached from the rental sector due to the housing reform, through which 26 most rental housing units are sold to existing tenants. These decreasing headship rates are worrisome, especially given that those aged 25 to 34 have experienced decreases from 40% in 1982 to less than 20% today, which is in stark contrast to that seen in industrialized countries in the post-WWII period. In this way, the disappearance of renters has artificially inflated the homeownership rate, which is contrary to that reported by Haurin and Rosenthal (2007). Finally, contrary to our third hypothesis, migrants have higher headship rates than the local residents. Part of this can be explained by rural-urban migration, which has helped the overall headship rate of young adults from falling even further in urban China. Further analysis also shows that migrants dominate the rental sector, because institutional and financial barriers reduce migrants’ access to homeownership, and they have to have place to stay in urban China. As such, migrants may have crowded out local youth in the rental sector, squeezing them into homeownership. However, those who cannot afford homeownership are then unable to form independent households. Overall, our findings echo the methodological limitation of the binary headship measure reported previously (Willekens 2010). This limitation is amplified in urban China, where the hierarchical nature of headship status is not always clear. For instance, while homeownership is often regarded as a preferred choice over renting for migrants, owning a self-build unit is not only inconvenient for access to jobs, but also impractical due to institutional barriers. Nevertheless, this limitation can be mitigated by allowing multiple options in the headship status. Further, our findings highlight a divergence at the start of the housing pathway, affected by institutional attributes and augmented by socioeconomic and demographic factors. Within our sample, younger adults tend to start their housing pathway at a 27 progressively lower level, while renting is a more likely destination for most migrants who lack permanent status in urban China. Furthermore, employment plays an important role in the housing pathway. That is, those who hold public-sector jobs are twice as likely as those in the private sector to head a special-access unit. In contrast, those in farming are largely in selfbuild units, which is likely due to institutional barriers. In order to continuously grow the economy while maintaining stability, policymakers should look to facilitate the residential independence of young adults in urban China by reducing institutional barriers and making rental units more available and affordable. To facilitate residential mobility, more effort should be made to create housing that can be sold and resold. Our paper is, to our knowledge, the first to look at independent household formation in China. That said, there are limitations to this research. For one, even though the CHFS data is claimed to be nationally representative, the sample size is substantially smaller than that of national census data, and the sampling strategy may have been adjusted slightly over time. Hence, we cannot capture the subtle age gradient in housing transition. Furthermore, this study uses a trend study approach and, as a result, causal relationships may not always be identifiable. As such, there are several important avenues for further study. First, future research should track individual birth cohorts and longitudinal progress, which would allow for a more precise understanding of housing pathway trajectories. Our model fit statistics reveal that a good deal of residential behaviors remain unexplained. Second, this study could be extended across a longer period to better understand the evolution of household formation in both rural and urban China. Third, China has planned to introduce property tax in several regions, and it 28 would be interesting to know how (or if) this will affect residential behaviors. Fourth, further research should further examine cultural and socioeconomic factors, such as parental financial support and intergenerational cohabitation, in household formation in the context of mainland China. Fourth, since China has not fully experienced the so called “second demographic transition” (Lesthaeghe, 2014), it may be worthwhile to test the “depreciation” hypothesis in other countries and see if it helps explain the struggle of many young adults in the housing sector. Finally, it is important to know not only whether the improvement of housing conditions is associated with the departure of young adults from family homes, but also the extent to which migrants are competing with locals in the rental sector. This would help demonstrate whether declining household formation is by choice or by force. 29 References Aguila, Emma, Jung Ho Park, and Alma Vega. 2020. Living Arrangements and Supplemental Income Programs for Older Adults in Mexico. Demography:1-24. Albrecht, Stan L, and Tim B Heaton. 1984. Secularization, higher education, and religiosity. Review of Religious Research:43-58. Banister, Judith. 1987. China's changing population. Stanford, Calif.: Stanford University Press. Becker, Gary S. 1993. A Treatise on the Family. 9780674906990 vols: Harvard university press. Berk, Richard A, and Sarah Fenstermaker Berk. 1983. Supply-side sociology of the family: The challenge of the new home economics. Annual review of sociology 9 (1):375-395. Blinder, Alan S. 1973. Wage Discrimination: Reduced Form and Structural Estimates. Journal of Human Resources 8 (Fall): 436-455. Borsch-Supan, Axel. 1986. Household formation, housing prices, and public policy impacts. Journal of Public Economics 30 (2):145-164. Börsch-Supan, Axel. 1989. Household dissolution and the choice of alternative living arrangements among elderly Americans. In The Economics of Aging: University of Chicago Press. Burch, Thomas K. 1995. Theories of Household Formation: Progress and Challenges. In Household Demography and Household Modeling, edited by E. van Imhoff, A. Kuijsten, P. Hooimeijer and L. van Wissen. Boston, MA: Springer US. Burch, Thomas K., and Beverly J. Matthews. 1987. Household Formation in Developed Societies. Population and Development Review 13 (3):495-511. Buzar, Stefan, Philip E. Ogden, and Ray Hall. 2005. Households matter: the quiet demography 30 of urban transformation. Urban Studies 29 (4):413–436. Cameron, Lisa, Nisvan Erkal, Lata Gangadharan, and Xin Meng. 2013. Little emperors: behavioral impacts of China's One-Child Policy. Science 339 (6122):953-957. Carliner, Geoffrey. 1975. Determinants of household headship. Journal of Marriage and the Family 37 (1):28-38. Carliner, Michael. 2003. Headship rates and housing demand. Housing Economics 51 (6):8-11. Chan, Kam Wing. 2018. Urbanization with Chinese Characteristics: The Hukou System and Migration: Routledge. Chen, Junhua, Fei Guo, and Ying Wu. 2011. One decade of urban housing reform in China: Urban housing price dynamics and the role of migration and urbanization, 1995-2005. Habitat International 35 (1):1-8. Chen, Kaiji, and Yi Wen. 2017. The great housing boom of China. American Economic Journal: Macroeconomics 9 (2):73-114. Chen, Yvonne Jie, Zhiwu Chen, and Shijun He. 2019. Social Norms and Household Savings Rates in China. Review of Finance 23 (5):961-991. Cheng, Tiejun, and Mark Selden. 1994. The origins and social consequences of China's hukou system. China Quarterly 139:644-668. Clapham, David. 2005. The Meaning of Housing: A Pathways Approach. University of Bristol: Policy Press. Clark, William AV. 2012. Residential Mobility and the Housing Market. In The Sage handbook of housing studies, edited by D. F. Clapham, W. A. Clark and K. Gibb: Sage. Clark, William AV, Marinus C Deurloo, and Frans M Dieleman. 2003. Housing careers in the 31 United States, 1968-93: Modelling the sequencing of housing states. Urban Studies 40 (1):143-160. Clark, William AV, Youqin Huang, and Diachun Yi. 2019. Can millennials access homeownership in urban China? Journal of Housing and the Built Environment:1-19. Cobb-Clark, Deborah A. 2008. Leaving home: What economics has to say about the living arrangements of young Australians. Australian Economic Review 41 (2):160-176. Cooper, Daniel, and María José Luengo-Prado. 2018. Household formation over time: Evidence from two cohorts of young adults. Journal of Housing Economics 41:106-123. Day, Lincoln H., and Xia Ma. 1994. Migration and urbanization in China, Studies in Chinese environment and development. Armonk, N.Y.: M.E. Sharpe. Deng, Wen Jing, Joris SCM Hoekstra, and Marja G Elsinga. 2016. The changing determinants of homeownership amongst young people in urban China. International Journal of Housing Policy 16 (2):201-222. Deng, Yongheng, Joseph Gyourko, and Jing Wu. 2012. Land and house price measurement in China. National Bureau of Economic Research. Douglass, Mike. 2014. Afterword: Global householding and social reproduction in Asia. Geoforum 51:313-316. Elder, Glen H. 1975. Age Differentiation and the Life Course. Annual Review of Sociology 1:165190. Ermisch, John F. 1981. An Economic Theory of Household Formation: Theory and Evidence from the General Household Survey. Scottish Journal of Political Economy 28 (1):1-19. Fan, C Cindy. 2007. China on the Move: Migration, the State, and the Household: Routledge. 32 Fang, Yiping, and Zhanxin Zhang. 2016. Migrant household homeownership outcomes in large Chinese cities-the sustained impact of hukou. Eurasian Geography and Economics 57 (2):203-227. Fisher, Lynn M, and Jamie Woodwell. 2015. Housing Demand: Demographics and the Numbers Behind the Coming Multi-Million Increase in Households. In Mortgage Bankers Association Research Report. Forrest, Ray, and Ngai-ming Yip. 2012. Young people and housing: Transitions, trajectories and generational fractures: Routledge. Friedmann, John. 2005. China's urban transition. Minneapolis: University of Minnesota Press. Furlong, Fred. 2016. Household formation among young adults. FRBSF Economic Letter 2016:17. Gan, Li. 2014. Urbanization and the Housing Market in China. In Results from China Household Finance Survey. College Station, TX: Texas A&M University Southwestern University of Finance and Economics. Gan, Li, Zhichao Yin, Nan Jia, Shu Xu, Shuang Ma, and Lu Zheng. 2013. Data you need to know about China: Research Report of China Household Finance Survey 2012: Springer Science & Business Media. Glaeser, Edward, Wei Huang, Yueran Ma, and Andrei Shleifer. 2017. A real estate boom with Chinese Characteristics. The Journal of Economic Perspectives 31 (1):93-116. Goldscheider, Frances K., and Julie Da Vanzo. 1989. Pathways to independent living in early adulthood: Marriage, semiautonomy, and premarital residential independence. Demography 26 (4):597-614. 33 Goldscheider, Frances Kobrin, and Julie DaVanzo. 1986. Semiautonomy and Leaving Home in Early Adulthood Social Forces 65 (1):187-201. Goodman, John L. 1986. Economic determinants of household formations and living arrangements. In Working Paper 66 Economic Activity Section from Board of Governors of the Federal Reserve System (U.S.) Goodman, Laurie, Rolf Pendall, and Jun Zhu. 2015. Headship and Homeownership: What Does the Future Hold? Ham, Maarten van. 2012. Chpt3. Housing Behaviour. In The Sage handbook of housing studies, edited by D. F. Clapham, W. A. Clark and K. Gibb: Sage. Haurin, Donald R., Patric H. Hendershott, and Dongwook Kim. 1993. The Impact of Real Rents and Wages on Household Formation. Review of Economics and Statistics 75 (2):284293. Haurin, Donald R., and Stuart S. Rosenthal. 2007. The Influence of Household Formation On Homeownership Rates Across Time and Race. Real Estate Economics 35 (4):411–450. Hirayama, Yosuke. 2010. Housing Pathway Divergence in Japan's Insecure Economy. Housing Studies 25 (6): 777-797. Hochstenbach, Cody, and Willem R. Boterman. 2015. Navigating the Field of Housing: Housing Pathways of Young People in Amsterdam. Journal of Housing and the Built Environment 30: 257-274. Hu, Zhan, and Xizhe Peng. 2015. Household changes in contemporary China: an analysis based on the four recent censuses. The Journal of Chinese Sociology 2 (1):1. Huang, Xu, Martin Dijst, Jan van Weesep, Yixue Jiao, and Ying Sun. 2017. Residential choice 34 among rural–urban migrants after hukou reform: Evidence from Suzhou, China. Population, Space and Place 23 (4):e2035. Huang, Y., He, S. & Gan, L. (2021) Introduction to special issue: unpacking the Chinese dream of homeownership. Journal of Housing and the Built Environment 36 (1): 1–7. https://doi.org/10.1007/s10901-021- 09827-y Or, T.M. (2017) Pathways to homeownership among young professionals in urban China: the role of family resources, Urban Studies, 55(11): 2391-2407. https://doi.org/10.1177/0042098017714212 Jiang, Leiwen. 2006. Living conditions of the floating population in urban China. Housing Studies 21 (5):719-744. Jiang, Quanbao, Marcus W Feldman, and Shuzhuo Li. 2014. Marriage squeeze, never-married proportion, and mean age at first marriage in China. Population Research and Policy Review 33 (2):189-204. Kendig, Hal L. 1990. A Life Course Perspective on Housing Attainment. In Housing Demography: Linking Demographic Structure and Housing Markets, edited by D. Myers. Madison, WI: University of Wisconsin Press. Kent, Richard J. 1992. Household Formation by the Young in the United States. Applied Economics 24 (10):1129-1137. Kiefer, Len, Ajita Atreya, and Venkataramana Yanamandra. 41. Why is Adulting Getting Harder? Young Adults and Household Formation 201841]. Kobrin, Frances E. 1973. Household Headship and Its Changes in the United States, 1940-1960, 1970. Journal of the American Statistical Association 68 (344):793-800. 35 Repeated Author. 1976. The Fall in Household Size and the Rise of the Primary Individual in the United States. Demography 13 (1):127-138. Kuijsten, Anton. 1995. Recent trends in household and family structures in Europe: an overview. In Household demography and household modeling, edited by E. Imhoff, A. Kuijsten, P. Hooimeiger and L. v. Wissen. New York: Plenum Press. Lee, Kwan Ok, and Gary Painter. 2013. What happens to household formation in a recession? Journal of Urban Economics 76:93-109. Lee, Sharon M, and Barry Edmonston. 2014. Residential Independence of Elderly Immigrants in Canada. Canadian Journal on Aging/La Revue canadienne du vieillissement 33 (4):359-377. Leppel, Karen. 1986. A Trinomial Logit Analysis of Household Composition. AREUEA Journal 14 (4):537-556. Lesthaeghe, Ron, and Guy Moors. 2000. Recent trends in fertility and household formation in the industrialized world. Review of Population and Social Policy 9 (1):121-170. Lesthaeghe, Ron. 2014 The second demographic transition: A concise overview of its development. Proceedings of the National Academy of Sciences, 111(51), 18112-18115. Li, Sheng, Jie Li, and Alice Y Ouyang. 2017. Housing and household wealth inequality: Evidence from the People's Republic of China. ADBI Working Paper Series. Li, Si-Ming. 2012. Housing inequalities under market deepening: the case of Guangzhou, China. Environment and Planning A 44 (12):2852-2866. Repeated Author. 2017. Homeownership and housing consumption change in Urban China: Guangzhou under market transition. Urban Geography 38 (5):752-770. 36 Lian, Hongping. 2014. The post-1980s generation in China: Exploring its theoretical underpinning. Journal of Youth Studies 17 (7):965-981. Liang, Zai, Zhen Li, and Zhongdong Ma. 2014. Changing patterns of the floating population in China, 2000–2010. Population and development review 40 (4):695-716. Logan, John R., and Yanjie Bian. 1993. Inequalities in Access to Community Resources in a Chinese City (Tianjin). Social Forces 72 (2):555-577. Logan, John R., Yiping Fang, and Zhanxin Zhang. 2009. Access to housing in Urban China. International Journal of Urban and Regional Research 33 (4):914-935. Mackie, Peter K. 2016. Young people and housing: Identifying the key issues. International Journal of Housing Policy 16 (2):137-143 Mak, Stephen WK, Lennon HT Choy, and Winky KO Ho. 2007. Privatization, housing conditions and affordability in the People's Republic of China. Habitat International 31 (2):177192. McFadden, Daniel. 1978. Modeling the Choice of Residential Location. In Spatial Interaction Theory and Planning Model, edited by A. Karlqvist, L. Lundqvist, F. Snickars and J. W. Weilbull. Amesterdam: North-Holland. McKee, Kim. 2012. Young people, homeownership and future welfare. Housing Studies 27 (6):853-862. Michael, Robert T., Victor R. Fuchs, and Sharon R. Scott. 1980. Changes in the Propensity to Live Alone : 1950-1976. Demography 17 (1):39-56. Miron, John R. 1989. Household formation, affordability, and housing policy Population Research and Policy Review 8 (1):55-77. 37 Mulder, Clara H. 2013. Family dynamics and housing: Conceptual issues and empirical findings. Demographic Research 29:355-378. Mulder, Clara H, and Pieter Hooimeijer. 1999. Residential relocations in the life course. In Population issues: Springer. Mulder, Clara H, and Annika Smits. 2013. Inter-generational ties, financial transfers and homeownership support. Journal of Housing and the Built Environment 28 (1):95-112. Myers, Dowell. 1992. Analysis with Local Census Data: Portraits of Change. Boston: Academic Press. Repeated Author. 1999. Cohort Longitudinal Estimation of Housing Careers. Housing Studies 14 (4):473-490. National Bureau of Statistics of China. 2020. China Statistical Yearbooks. Nee, Victor, and Rebecca Matthews. 1996. Market Transition and Societal Transformation in Reforming State Socialism Annual Review of Sociology 22:401-435. Newman, Sandra, Scott Holupka, and Stephen L Ross. 2018. There's no place like home: Racial disparities in household formation in the 2000s. Journal of Housing Economics. Oaxaca, Ronald. 1973. Male-female wage differentials in urban labor markets. International economic review 14 (3):693-709. Paciorek, Andrew. 2016. The long and the short of household formation. Real Estate Economics 44 (1):7-40. Pampel, Fred C. 1983. Changes in the Propensity to Live Alone: Evidence from Consecutive Cross-sectional Surveys, 1960-1976. Demography 20:443-448. Peng, Xizhe. 2011. China’s Demographic History and Future Challenges. Science 333 38 (6042):581-587. Phua, Voon Chin, Gayle Kaufman, and Keong Suk Park. 2001. Strategic adjustments of elderly Asian Americans: Living arrangements and headship. Journal of Comparative Family Studies 32 (2):263-281. Pitkin, John. 1990. Housing Consumption of the Elderly: A Cohort Economic Model. In Housing Demography: Linking Demographic Structure and Housing Markets, edited by D. Myers. Madison, WI: University of Wisconsin Press. Pitkin, John R. , George S. Masnick, and John Brennan. 1990. Cohort Housing Trends in a Local Housing Market: The Case of Southern California. In Housing Demography: Linking Demographic Structure and Housing Markets, edited by D. Myers. Madison, WI: University of Wisconsin Press. Polanyi, Karl. The great transformation (1st Beacon paperback). Beacon Press 1957. Available from http://name.umdl.umich.edu/HEB03171 Ren, Qiang, and Rongqin Hu. 2016. Housing inequality in urban China. Chinese Journal of Sociology 2 (1):144-167. Riley, Matilda White. 1987. On the significance of age in sociology. American Sociological Review 52:1-14. Santi, Lawrence L. 1990. Household Headship among Unmarried Persons in the United-States, 1970-1985. Demography 27 (2):219-232. Scutt, David. 2017. NOMURA: China's housing bubble is spreading to its smallest cities. Business Insider Australia. Seales, Rebecca. 2017. The country where 70% of millennials are homeowners. BBC News. 39 Sito, Peggy, and Pearl Liu. 2018. China property: how the world’s biggest housing market emerged. South China Morning Post. Skaburskis, Andrejs. 1994. Determinants of Canadian Headship Rates. Urban Studies 31 (8):1377-1390. Skeldon, Ronald. 2011. China: An Emerging Destination for Economic Migration. Migration Information Source: Country Profiles. Smith, Lawrence B., Kenneth T. Rosen, Anil Markandya, and Pierre-Antoine Ullmo. 1984. The Demand for Housing, Household Headship Rates, and Household Formation: An International Analysis. Urban Studies 21 (4):407-414. Sobotka, T. 2008. The diverse faces of the Second Demographic Transition in Europe. In Childbearing Trends and Policies in Europe--Demographic-Research Special Collection, edited by J. H. T. Frejka, T. Sobotka, L. Toulemon. Song, Jing. 2010. Moving Purchase and Sitting Purchase: Housing Reform and Transition to Homeownership in Beijing. Housing Studies 25 (6):903-919. Song, Xi, and Yu Xie. 2014. Market Transition Theory Revisited: Changing Regimes of Housing Inequality in China, 1988-2002. Sociological Science 1:277. Sweet, James A. 1984. Components of change in the number of households: 1970–1980. Demography 21 (2):129-140. Sweet, James A. 1990. Changes in the Life-Cycle Composition of the United States Population and the Demand of Housing. In Housing Demography: Linking Demographic Structure and Housing Markets, edited by D. Myers. Madison, WI: University of Wisconsin Press. Szelenyi, Ivan, and Eric Kostello. 1996. The market transition debate: Toward a synthesis? 40 American Journal of Sociology 101 (4):1082-1096. Tang, Mingzhe, and N Edward Coulson. 2017. The impact of China's housing provident fund on homeownership, housing consumption and housing investment. Regional Science and Urban Economics 63:25-37. Tsai, I-Chun, and Shu-Hen Chiang. 2019. Exuberance and spillovers in housing markets: Evidence from first-and second-tier cities in China. Regional Science and Urban Economics 77:75-86. Waite, Linda J, Frances Kobrin Goldscheider, and Christina Witsberger. 1986. Nonfamily living and the erosion of traditional family orientations among young adults. American Sociological Review:541-554. Walder, Andrew G, and Xiaobin He. 2014. Public housing into private assets: Wealth creation in urban China. Social science research 46:85-99. Wang, Yourong, Yuyao Li, Youqin Huang, Chengdong Yi, and Jianyu Ren. 2020. Housing wealth inequality in China: An urban-rural comparison. Cities 96:102428. Wang, Yuna, and Toshio Otsuki. 2015. Do institutional factors influence housing decision of young generation in urban China: Based on a study on determinants of residential choice in Beijing. Habitat International 49:508-515. Wiemers, Emily E. 2014. The effect of unemployment on household composition and doubling up. Demography 51 (6):2155-2178. Willekens, Frans. 2010. Family and household demography. Encyclopedia of Life Support Systems, Demography 2:86-112. Wu, Jiawei, Zhou Yu, Yehua Dennis Wei, and Liuqing Yang. 2019. Changing distribution of 41 migrant population and its influencing factors in urban China: Economic transition, public policy, and amenities. Habitat International 94:102063. Wu, Lili, Yang Bian, and Wei Zhang. 2018. Housing ownership and housing wealth: new evidence in transitional China. Housing Studies:1-21. Wu, Lili, and Wei Zhang. 2018. Rural migrants' homeownership in Chinese urban destinations: Do institutional arrangements still matter after Hukou reform? Cities 79:151-158. Wu, Weiping. 2004. Sources of migrant housing disadvantage in urban China. Environment and Planning A 36 (7):1285-1304. Xie, Yu, and Yongai Jin. 2015. Household wealth in China. Chinese sociological review 47 (3):203-229. Xu, Yilan. 2017. Mandatory savings, credit access and home ownership: The case of the housing provident fund. Urban Studies 54 (15):3446-3463. Yu, Zhou. 2017. Macro effects on the household formation of China's young adults – demographics, institutional factors, and regional differences. International Journal of Housing Policy:1-29. Repeated Author. 2020. Homeownership attainment of adult children in urban China: parental attributes and financial support. Housing Studies:1-33. Yu, Zhou, and Dowell Myers. 2010. Misleading Comparisons of Homeownership Rates When the Variable Effect of Household Formation is Ignored: Explaining Rising Homeownership and the Homeownership Gap between Blacks and Asians. Urban Studies 47 (12):2615-2640. Zhou, Da. 2010. Housing Demand of Chinese Urban Migrants (Wuo Guo Cheng Shi Yi Min De 42 Zhu Fang Xu Qiu Fen Xi). Urban Problems (in Chinese) 176 (3):58-61. 43 Million Figure 1. The number of migrants and percentage urban residents, 1982-2017 70 250 60 200 50 150 40 30 100 20 50 0 Migrant population Percentage urban residents 10 1982 1987 1990 1995 2000 2005 2011 2017 0 Source: 1982, 1990, 2000, and 2010 Chinese National Population Census; 1987, 1995, 2005, and 2015 1% mini population census; China Statistical Yearbook, 2017. Based on the floating population statistics, floating population when a person resides in a location where he/she is not registered. Figure 2. Headship rates by age in urban China, 2011 and 2017 0.6 0.5 0.4 2011 0.3 2017 0.2 0.1 0.0 18-34 35-44 45-54 55-64 65+ Table 1. Headship rates by age in urban China, 2011 and 2017 18-34 35-44 45-54 55-64 65+ Total Obs. 2011 0.220 0.455 0.463 0.450 0.451 0.376 14,079 2017 0.145 0.402 0.492 0.487 0.493 0.399 48,975 Diff. -0.074 -0.054 0.029 0.037 0.043 0.022 Table 2. Summary statistics of young adults in urban China, 2011 and 2017 2011 Year Obs.: Headship Mea n 7,557 Std. Dev. 2017 17,304 Mea Std. n Dev. 0.311 0.463 0.258 0.438 18-34 0.613 0.489 0.566 0.496 35-44 0.387 0.489 0.434 0.496 Some college or higher 0.285 0.451 0.346 0.476 Middle or high school 0.521 0.500 0.496 0.500 Elementary school or lower 0.094 0.291 0.074 0.262 Still in school 0.100 0.301 0.084 0.277 2.831 5.223 4.649 6.217 Not employed 0.265 0.441 0.219 0.413 Employed in the public sector 0.148 0.355 0.124 0.330 Employed in the private sector 0.549 0.498 0.635 0.481 Employed in farming 0.038 0.192 0.021 0.145 Tier 1 region 0.282 0.450 0.131 0.338 Tier 2 region 0.325 0.468 0.377 0.485 Outside tier 1 and 2 region 0.394 0.489 0.492 0.500 Married couple 0.708 0.455 0.712 0.453 Never married man 0.157 0.364 0.163 0.370 Formerly married man 0.009 0.093 0.011 0.104 Never married woman 0.115 0.319 0.103 0.304 Formerly married woman 0.012 0.107 0.010 0.101 Local residents (whose hukou is registered in the local area) 0.778 0.416 0.694 0.461 Migrants 0.224 0.417 0.306 0.461 0.144 0.351 0.240 0.427 0.856 0.351 0.760 0.427 Age group Educational attainment Personal income (logged) Labor force status Housing market Marital status and gender Migrant status Access to housing provident fund Yes No Note: Personal income is adjusted for inflation. The sample is limited to the individuals who are 18-44 years old and who live in urban China at the time of survey. Table 3. Odds ratios of headship for young adults in urban China, from 2011 to 2017 Obs: 24,861 Pseudo R2: I 0.050 II 0.084 III 0.124 IV 0.074 V 0.102 VI 0.160 Variables Age group (omitted: 35-44) 18-34 0.352 *** 0.382 *** 0.408 *** 0.467 *** 0.329 *** 0.456 *** Year (2011 = 0; 2017 = 1) 0.796 *** 0.778 *** 0.698 *** 0.801 *** 0.669 *** 0.638 *** Age group relative changes (Year * Age group 18-34) 0.771 ** 0.739 *** 0.626 *** 0.781 ** 0.704 *** 0.641 *** Educational attainment (omitted: Middle or high school) Some college or better Elementary school or lower Still in school 1.548 *** 0.879 0.141 *** 1.076 0.848 0.284 *** Personal income (logged) 1.107 *** 1.093 *** Labor force status (omitted: Employed in the private sector) Not employed Employed in the public sector Employed in farming 0.659 *** 1.122 0.743 0.897 1.038 0.917 Housing market (omitted: the tier 2 region) The tier 1 region Outside the tier 1 and 2 regions 1.117 0.987 1.177 1.192 * Marital status and gender (ommited: Married couple) Never married man Formerly married man Never married woman Formerly married woman 0.428 *** 1.836 0.394 *** 4.200 *** 0.667 *** 2.424 0.681 ** 6.086 *** Migrant status (omitted: Local residents) Migrants (whose hukou is registered outside the local area) 2.439 *** 2.234 *** Access to housing provident fund (omitted: no access) Have access 3.199 *** 1.483 *** 0.565 *** 0.482 *** Intercept 0.836 *** 0.763 *** 0.565 *** 0.821 *** * p<0.1 **p<0.05 ***p<0.001 Two-tailed tests Note: Logit regression is used to estimate the headship model. The sample is limited to the individuals who are 18-44 years old and who live in urban China at the time of survey. Table 4. Multinomial logit relative risk ratios for young adults, from 2011 to 2017 I Obs: 24,861 Pseudo R2: 0.046 Rental Self-build Special access Age group (omitted: 35-44) 18-34 0.586 *** 0.140 *** 0.226 *** Commercial 0.197 Rental II Self-build Special access Commercial 0.246 *** 0.576 *** 0.287 *** 0.464 *** 0.413 *** Year (2011 = 0; 2017 = 1) 0.368 *** 0.982 0.902 1.087 0.223 *** 1.080 0.935 0.957 Age group change (Year * Age group 18-34) 0.845 1.576 * 0.802 0.937 0.775 1.419 0.593 ** 0.776 0.322 *** 0.830 0.276 0.869 0.975 0.212 ** 1.814 *** 0.654 0.644 1.108 *** 1.067 *** 1.110 *** 1.078 *** 0.953 e 1.110 e 0.182 *** 0.498 *** 0.764 2.329 *** 1.431 * 1.986 *** 0.474 ** 0.976 0.804 0.472 ** 1.073 0.885 0.931 0.702 ** 1.105 2.228 *** 1.715 *** 1.966 *** 1.061 2.026 1.200 12.478 *** 0.375 *** 3.491 *** 0.238 ** 0.607 0.702 3.898 *** 0.492 * 6.064 *** 0.291 *** 1.687 0.308 *** 5.371 *** Migrant status (omitted: Local residents) Migrants (whose hukou is registered outside the local area) 6.883 *** 0.196 *** 0.335 *** 0.821 Access to housing provident fund (omitted: no access) Have access 1.083 0.613 ** 2.099 *** 1.979 *** _ 0.098 *** 0.205 *** 0.030 *** 0.105 *** Educational attainment (omitted: Middle or high school) Some college or better Elementary school or lower Still in school e 1.027 e 0.866 i 0.216 ** Personal income (logged) Labor force status (omitted: Employed in the private sector) Not employed Employed in the public sector Employed in farming Housing market (omitted: the tier 2 region) The tier 1 region Outside the tier 1 and 2 regions Marital status and gender (omitted: Married couple) Never married man Formerly married man Never married woman Formerly married woman Intercept 0.276 *** 0.142 *** 0.118 *** 0.296 *** * p<0.1 **p<0.05 ***p<0.001 Two-tailed tests Note: Results are weighted. Multinomial ogit regression is used to estimate the headship model. Sample is limited to those who are 18-44 years old and who live in urban China at the time of the survey. The reference category: non householders. Appendix 1. Odds ratios of headship for young adults in urban China, 2011 and 2017 2011 0.169 7,557 2017 0.139 17,304 Pseudo R2: Obs: Variables Age group (omitted: 35-44) 18-34 0.444 *** 0.388 *** Educational attainment (omitted: Middle or high school) Some college or better Elementary school or lower Still in school 1.062 0.843 0.278 *** 1.225 *** 0.710 *** 0.209 *** Personal income (logged) 1.106 *** 1.015 *** Labor force status (omitted: Employed in the private sector) Not employed Employed in the public sector Employed in farming 0.968 1.011 0.925 0.395 *** 0.840 ** 1.093 Housing market (omitted: Tier 2 region) Tier 1 region Outside tier 1 and 2 region 1.211 1.254 * 0.903 1.037 Marital status and gender (ommited: Married couple) Never married man Formerly married man Never married woman Formerly married woman 0.711 ** 2.613 0.757 6.721 *** 0.488 *** 1.980 *** 0.236 *** 3.683 *** Migrant status (omitted: Local residents) Migrants (whose hukou is registered outside the local area) 2.464 *** 1.235 *** Access to housing provident fund (omitted: no access) Have access 1.517 ** 1.363 *** Intercept 0.361 *** 0.590 *** * p<0.1 **p<0.05 ***p<0.001 Two-tailed tests Note: Logit regression is used to estimate the headship model. The sample is limited to the indivudals who are 18-44 years old and who live in urban China at the time of survey. Appendix 2. Odds ratios of headship for young adults in urban China, from 2013 to 2015 Obs: Pseudo R2: 46,621 0.102 Variables Age group (omitted: 35-44) 18-34 0.463 *** Year (2013 = 0; 2015 = 1) 0.770 *** Age group relative changes (Year * Age group 18-34) 0.975 Educational attainment (omitted: Middle or high school) Some college or better Elementary school or lower Still in school 0.997 0.933 0.634 Personal income (logged) 0.993 Labor force status (omitted: Employed in the private sector) Not employed Employed in the public sector Employed in farming 0.512 *** 0.787 *** 0.797 ** Housing market (omitted: Tier 2 region) Tier 1 region Outside tier 1 and 2 regions 0.924 1.077 * Marital status and gender (omitted: Married couple) Never married man Formerly married man Never married woman Formerly married woman 0.877 1.996 0.797 3.947 Migrant status (omitted: Local residents) Migrants (whose hukou is registered outside the local area) 2.667 *** Access to housing provident fund (omitted: no access) Have access 3.057 *** Intercept 0.540 *** * *** ** *** * p<0.1 **p<0.05 ***p<0.001 Two-tailed tests Note: Logit regression is used to estimate the headship model. The sample is limited to the individuals who are 18-44 years old and who live in urban China at the time of survey. Income is adjusted for inflation.