Introduction

In recent decades, China’s social structure has undergone fundamental changes (He, 2016), with urban residents now outnumbering rural residents (Li et al. 2022). These changes in the urban‒rural population structure have resulted in significant shifts in residents’ socioeconomic status, urban‒rural relationships, occupational structure, production models, and lifestyles (Cui et al. 2019, Lu et al. 2021). As China’s economy continues to grow and living standards improve, new trends in socioeconomic stratification have emerged, including a widening income gap, increased wealth concentration, a stronger tendency towards class solidification, and heightened social tensions. These trends are likely to hinder social harmony and development (Li and Zhu, 2017). Therefore, it is necessary to uncover the current state and key factors affecting residents’ socioeconomic status and class mobility.

Existing research shows that residents’ economic status and class mobility are influenced by factors such as educational attainment, institutional framework, and regional characteristics (Dustmann, 2004; Restuccia and Urrutia, 2004; Rodríguez et al. 2008). As the most important asset and source of wealth for most households, housing not only serves as an indication of social and economic status but also directly impacts residents’ economic status and class mobility. On one hand, homeownership enables households to accumulate wealth (Turner and Luea, 2009; Wainer and Zabel, 2020), expand social networks (Glaeser and Sacerdote, 2000), and increase political capital (Holian, 2011; Yoder, 2020), which ultimately elevates income levels and social status. However, due to the “housing mortgage slave effect” that would be discussed in this paper in details, the immobility of housing in physical space and the substantial pressure from rapid increases in housing prices can hinder households from pursuing better job opportunities and higher-return investment activities. Consequently, homeownership may be detrimental to the income growth and economic status improvement. Additionally, home maintenance expenditures also reduce household income (Mok, 2018). In short, the impact of homeownership on residents’ socioeconomic status depends on the balance between the “wealth effect” from homeownership and the “housing mortgage slave effect”.

In view of this, our study investigates the relationship between homeownership and changes in household economic status, contributing several new insights to the literature. First, while existing literature focuses on the impacts of household characteristics and institutional factors on residents’ economic status (Chetty et al. 2020; Corak, 2013; Torche, 2015; Johnson, 2013; McLeod, 2013), little attention has been given to housing, the most significant asset for households. This study directly examines the impacts of homeownership on changes in household economic status, and key findings remain robust after addressing endogeneity issue. Second, we explore the underlying mechanism through which homeownership affects residents’ economic status. We argue that due to the “housing mortgage slave effect”, homeowners are locked into their current jobs and therefore have fewer opportunities for higher income and more promising positions compared to renters (Chen and Hu, 2019; Li and Wu, 2014). Households burdened by mortgages are hesitant to change jobs, even when the new opportunities are more promising. Additionally, the long-term mortgage repayment pressure may also induce risk-averse behaviours among households. Third, this study examines the heterogeneity in the relationship between homeownership and changes in household economic status. Particularly, the differential roles of intergenerational financial support in the nexus of homeownership and social mobility are explored in the analysis.

Literature review

Social stratification remains a critical and intensively studied topic within economics, sociology, and political science. Previous research has predominantly concentrated on process analyses of social stratification (Kerckhoff, 2001), exploring phenomena such as the acquisition of social class and changes in social class status. The former explains the reason for individuals’ attainment at certain classes, while the latter reveals the cause for individuals’ class change during a certain period of time. The social class of individuals or households is typically represented through various indicators, including social status, household wealth, household income, educational achievements, occupation, and occupational position (Chen et al. 2022). Traditional research on social mobility has primarily focused on intergenerational class change, comparing educational and occupational achievements between parents and their offspring to assess the distribution of opportunities and the permeability of class structures within societies (Sørensen and Grusky, 1996). Additionally, an expanding corpus of literature now addresses interclass mobility, particularly as it relates to shifts in income (Chetty et al. 2020; Corak, 2013; Torche, 2015). Most scholars have adopted the new structuralism theory to analyse the dynamics of social stratification. This theoretical approach reinterprets concepts from institutional economics-such as dual economy, market segmentation, and internal labour markets —within a sociological framework to better understand and delineate social classes (Johnson, 2013; McLeod, 2013). New structuralism emphasizes the significant role of institutional arrangements in shaping social stratification, asserting that employment types and work arrangements are influenced by various institutional factors, rather than being determined solely by market forces.

Housing serves as a significant indicator of a household’s social class in modern society (Hu et al. 2020). It is argued that housing, being a fundamental necessity, more accurately reflects the social status and class of a household compared to occupation and educational attainment (Hu et al. 2024). Extensive literature underscores the socio-economic advantages of homeownership for both households and individuals. DiPasquale and Glaeser (1999) found that homeowners tend to invest more in social capital and exhibit enhanced civic engagement compared to renters. A considerable body of research highlights the positive impacts of homeownership on wealth accumulation (Turner and Luea, 2009; Wainer and Zabel, 2020), investment portfolio choices (Chetty et al. 2017; Cocco, 2005), labour mobility and market participation (Chen et al. 2023; Fu et al. 2016), home maintenance and decoration (Harding et al. 2000), social relationships (Glaeser and Sacerdote, 2000), political and social activities (Holian, 2011; Yoder, 2020), and personal well-being, including self-esteem and happiness (Yu et al. 2022). Moreover, the influence of homeownership extends to children’s behaviour, as evidenced by studies from Boehm and Schlottmann (1999), Green and White (1997), and Haurin et al. (2002). Coulson (2002) and Dietz and Haurin (2003) have extensively discussed the benefits associated with homeownership and addressed related concerns. Notably, Coulson and Li (2013) quantified a specific benefit, estimating that transitioning from renting to owning a home can generate an equivalent value of approximately 1,300 USD. Further evidence from China demonstrates significant increases in political engagement, financial market participation, subjective happiness, entrepreneurship, and household consumption following homeownership (Zhang and Yan, 2022; Huang et al. 2022; Chen et al. 2020; Hu, 2013). Therefore, as an important indicator of households’ wealth and status, homeownership contributes to elevating household income levels through various socioeconomic channels. These include the “wealth effect,” which may be realized through mechanisms such as housing equity extraction, home equity loans, or the collateral effects on borrowing capacity (Bhutta and Keys, 2016; Haurin and Moulton, 2017; Pérez-Martín et al. 2018; Giambona et al. 2014).

Homeownership, despite its associated wealth benefits, may paradoxically suppress increases in household income due to the high-value nature of housing assets and the resultant “housing mortgage slave effect”. This effect refers to the significant financial burden that mortgage obligations impose on homeowners, often crowding out discretionary spending. Previous studies have demonstrated that the “housing mortgage slave effect” not only reduces consumption in areas unrelated to housing but also increases the financial strain as house prices rise, thus limiting households’ ability to manage other expenses. Moreover, the “housing mortgage slave effect” manifests in homeowners’ general reluctance to pursue high-risk investments or entrepreneurial ventures compared to renters (Chen and Hu, 2019; Li and Wu, 2014). This risk aversion suggests that the lock-in effect of homeownership—where homeowners are deterred from seeking high-income opportunities or job mobility—might be a principal factor undermining potential income improvements. Moreover, we believe that purchasing a housing in a particular area not only has the “mortgage slave” effect due to loan restrictions, but also a “work area restriction” effect. When we buy a house in a certain place, due to the immovable nature of the house in physical space, our range of work choices is limited by the region, just as the restrictions on people’s work and migration choices by the cultivated farmland in the agricultural era.

Thus, the impact of homeownership on economic status of households hinges on a critical trade-off between the “wealth effect” derived from homeownership and the adverse “housing mortgage slave effect”. When the “wealth effect” surpasses the “housing mortgage slave effect”, homeownership correlates positively with enhancements in household economic status. Conversely, when the “wealth effect” is overshadowed by the “housing mortgage slave effect”, the relationship between homeownership and economic status becomes negative.

However, research indicates that the liquidity of housing wealth through cash-out refinancing is notably more restricted in China compared to developed nations, consequently diminishing the likelihood of reallocating resources from the real estate sector to other forms of investment and consumption (Li and Wu, 2014; Chen et al. 2020). Li and Wu (2014) observed that both the property and financial markets in China are relatively underdeveloped, making it challenging for homeowners to liquidate housing wealth for alternative investments without selling their properties. Moreover, the complexity of selling a house and starting a business in China is exacerbated by continuously rising housing prices, which skew preferences toward purchasing rather than selling property. This preference is driven by two primary factors. First, the consistent rise in house prices over the past decade has entrenched the belief among the Chinese population that real estate is a highly secure investment, often yielding returns superior to other investment options. Second, Chinese people have a tradition of valuing homeownershipFootnote 1, and homes are one of most important “status” goods in China. In both urban and rural areas, owning a home is frequently considered a prerequisite for marriage—a view reinforced by intensifying competition in the marriage market due to demographic shifts (Wei and Zhang, 2011). A 2010 survey in Beijing highlighted marriage pressure as the foremost reason young people cite for acquiring a homeFootnote 2, suggesting that rising housing costs not only encourage savings but also deter activities like entrepreneurship that might postpone homeownership.

In summary, in China, a house is both a channel for investment and a fundamental requirement for marriage. The escalating costs associated with homeownership compel individuals to prioritize buying a home over initiating a business when family resources are limited. This scenario illustrates the crowding-out effect of rising housing prices on other economic activities (Li and Wu, 2014). Such institutional deficiencies ambiguously impact the direction of the wealth effect in China, complicating the dynamics between real estate investments and broader economic activities.

Research design

Sample selection and data sources

The data used in this study were derived from the China Family Panel Studies (CFPS) for the years 2010, 2012, 2014, 2016, and 2018. The CFPS, a key project funded by 985 Program of Peking University of China, is conducted by the Institute of Social Science Survey of Peking University. This initiative is supported by the National Population and Family Planning Commission of China, the National Bureau of Statistics of China, and academic institutions such as Shanghai University, Sun Yat-sen University, and Lanzhou University. Additionally, the University of Michigan offers critical guidance and technological support in survey design. Initially piloted in Beijing, Shanghai, and Guangdong in 2008 and 2009, the CFPS officially commenced in 2010 and has been conducted biennially since. The baseline survey in 2010 involved interviews with nearly 15,000 households across 25 provinces, municipalities, and autonomous regions in China, excluding Inner Mongolia, Xinjiang, Tibet, Hainan, Ningxia, and Qinghai. The CFPS collects data at the community, family, and individual levels, providing comprehensive insights into the socio-economic, demographic, educational, and health aspects of China, thereby offering high-quality data for academic research and policy analysis. Recent studies based on the CFPS data have been published in prestigious journals worldwide (Yang and Mukhopadhaya, 2016; Zhang et al. 2014; Luo et al. 2020).

This study uses CFPS data for several reasons: First, the CFPS includes extensive samples with detailed data on individual and household characteristic information such as educational attainment, age, sex, hukou registration status, ethnicity, political affiliation, and income. Second, the CFPS provides longitudinal data at the individual, household, and community levels, enabling the creation of balanced panel data over a significant period (five years) with large and diverse samples. While micro-level surveys on Chinese households are common, few facilitate the construction of such comprehensive panel data. Third, the CFPS covers nearly all major cities on the Chinese mainland. The representativeness of the samples is ensured through sophisticated sampling techniques including multistage probability, implicit stratification, and probability-proportional-to-size sampling.

After obtaining the raw CFPS data, we selected and cleaned the data via the following procedures. First, even though the CFPS covers both urban and rural areas of China, given the limitations in property ownership in rural areas, where housing can only be inherited or transacted within village collectives (Ho, 2017), this study focuses exclusively on urban samples. Second, to avoid interference from outliers, this study winsorizes variables such as income; that is, samples at 1% of each tail are trimmed. Third, individuals who missed one of the five rounds of the survey were excluded to construct five-year balanced panel data. Fourth, key variables with missing values were excluded. Following these procedures, we obtained 13,384 observation samples with 3,346 individuals from 26 Chinese provinces.

Variable measurement

The dependent variable in this study is the change in household income level, measured by the household income between two consecutive surveys. The measurement process involves several specific steps: First, households are ranked based on their income in ascending order and divided into ten levels. Level one represents the lowest income, while level ten represents the highest income. Second, changes in household income level are defined as the difference between household income level in the current survey and the previous survey. A positive difference indicates an increase in income level, while a negative difference indicates a decrease in income level. A greater difference signifies more substantial changes in income level.

This study employs the dummy variable of homeownership, i.e., the independent variable, which is constructed based on respondents’ answers for the homeownership type of their current dwelling. Specifically, the CFPS investigates the ownership type of the housing in which the respondents live, and the responses include full ownership (households hold property ownership certificates, state-owned land use certificates, and real estate tax payment certificates); shared ownership with an employer (residents partially own the home based on the proportion of the payment they make. The employer pays the outstanding costs of the home and holds partial ownership. If residents do not hold a state-owned land use certificate, their housing is counted as shared ownership with an employer. Most employers hold state-owned land-use certificates for housing that they previously owned after the housing reform. Such housing is categorized as “shared ownership with an employer”); rented; rent-free housing provided by the government; rent-free housing provided by an employer; housing provided by parents or children; and housing of relatives. Residents living in a dwelling with full ownership or shared ownership with an employer are considered to be homeowning households (equal to 1) and home-renting households (equal to 0), respectively. The control variables mainly include household characteristic factors such as marriage status, age, sex, hukou registration status, political affiliation i.e. whether member of China Communist Party (CCP), and educational attainment, which have been suggested to be associated with social mobility.

Descriptive statistics

The descriptive statistics of the key variables are shown in Table 1. The homeownership rate of the sample is 90.38%, which is in line with what has been mentioned in the literature for homeownership in urban areas of China. The full sample mean value of income level changes is 0.0928, implying relatively moderate income level changes in general. The number of households experiencing upward income level changes and those experiencing downward income level changes are basically the same. However, the standard deviation of income level changes is 2.7321, indicating large differences among households in terms of income level changes. Moreover, the mean value of homeowners’ income level changes (0.0661) is much lower than that of renters (0.3432). Therefore, descriptive statistics demonstrate a lower possibility of income level improvement for homeowning households. In addition, Table 1 shows significant differences between homeowners and renters in terms of individual and household characteristics. Therefore, it is necessary to control such factors in the estimation models.

Table 1 Descriptive statistics of the key variables.
Full size table

Model design

This study uses changes in household income level to measure changes in household economic status. We examine the impacts of homeownership on residents’ economic status through an ordinary least squares (OLS) model. Then, a myriad of methods, including panel model, propensity score matching, and variable substitution, are employed to ensure the robustness of the estimation results. The baseline model is set as follows:

$$\begin{array}{l}{{Mobility\; of\; income\; class}}_{{ijt}}={\beta }_{0}+{\beta }_{1}{{Homeowner}}_{{ijt}}\\+\delta X+{\sigma }_{j}+{\tau }_{t}+{\varepsilon }_{{ijt}}\end{array}$$
(1)

where \({{Mobility\; of\; income\; class}}_{{ijt}}\) denotes the income level changes of respondent i from community j in year t. Its value ranges from −9 (the income level of households at the current survey is 10 while it becomes 1 at the next survey) to 9 (the income level of households at the current survey is 1 while it becomes 10 at the next survey). A mean value of income changes of 0 implies no change in income levels. \({{Homeowner}}_{{ijt}}\) is the dummy variable for homeownership. If the respondent is living in a dwelling with “full ownership” or “shared ownership with an employer”, it equals 1, and 0 otherwise. X is the control variable for household characteristics, including marital status, age, sex, hukou registration status, political affiliation, and educational attainment. \({\sigma }_{j}\) and \({\tau }_{t}\) are community fixed effects and year fixed effects, respectively. \({\varepsilon }_{{ijt}}\) is the residual error. \({\beta }_{1}\) is the estimated coefficient of interest, which reflects the relation between homeownership and change in residents’ social class.

Although time-invariant unobservable factors at the community level are controlled for in Model (1) through community fixed effects, unobservable factors at the individual level may cause bias in the estimated results. By adding unobservable factors \({U}_{i}\) to Model (1), we construct Model (2) as follows:

$$\begin{array}{l}{{Mobility\; of\; income\; class}}_{{ijt}}={\beta }_{0}+{\beta }_{1}{{Homeowner}}_{{ijt}}\\+\delta X+{U}_{i}+{\sigma }_{j}+{\tau }_{t}+{\varepsilon }_{{ijt}}\end{array}$$
(2)

where \({U}_{i}\) is the unobservable characteristic factors of individuals and households, which remain constant over time. If such factors exert vital impacts on household income, the estimation results of Model (1) may be inefficient (when \({U}_{i}\) is unrelated to X) or biased (when \({U}_{i}\) is related to X). Therefore, we employ both random effect models and fixed effect models to resolve this issue in Model (2). When \({U}_{i}\) is unrelated to X, the random effect model is more efficient. When \({U}_{i}\) is related to X, the fixed effect model is more efficient.

The models shown above are based on the hypothesis that control variables impose linear impacts on outcome variables. If the relationship between these two variables is nonlinear, the above-mentioned model settings would be inefficient, which could lead to bias in the estimation results. Moreover, individuals’ homeownership decisions are affected by individual and household characteristics. For instance, residents who have a high expectation of housing appreciation or a rigid demand for housing are more likely to purchase housing. This means that there may be large differences among homeowning households and home-renting households in terms of initial endowments. The observed difference in income level changes may be derived from differences in initial endowments rather than homeownership. To resolve such potential model setting bias and selection bias, this study used propensity score matching (PSM) for remeasurement in the baseline model. In addition, we used an instrumental variable (IV) approach to address endogeneity issues that may arise due to reverse causality.

Empirical results

Benchmark results

The regression results of the baseline model are presented in Columns (1)–(5) of Table 2, including the estimated coefficients, standard errors, and significance levels. We tested the relationship between homeownership and changes in household income level by progressively incorporating control variables and year/community fixed effects into the OLS model. Column (1) displays the results from the most basic model configuration, which excludes control variables and fixed effects. The coefficient for homeownership is −0.2770, significant at the 1% level, indicating a negative association when other factors are not accounted for. Column (2) introduces controls for individual and household characteristics that influence income changes, such as marital status, age, sex, hukou registration status, political affiliation, and educational attainment. The findings suggest that homeowners are significantly less likely to experience income level changes compared to renters. Column (3) accounts for community and year-fixed effects. The results reaffirm the negative correlation between homeownership and improvements in household income levels, even when controlling for time-invariant regional factors and temporal trends.

Table 2 Homeownership and household income level changes: benchmark results.
Full size table

Furthermore, to address potential bias stemming from omitted variables at the individual and household levels, this study employs a panel model to re-estimate the relationship between homeownership and household income level changes. Random effect models are more effective when unobservable factors at the individual and household levels are irrelevant to control variables, and fixed effect models are more effective otherwise. Therefore, this study reports both the estimated results of the fixed effect model and random effect model in Columns (4) and (5). Both models reveal that homeownership is associated with a negative impact on income level improvement, significant at least at the 5% level. This finding underscores the consistently lower likelihood of income level improvement among homeowners compared to renters, even when accounting for unobservable individual and household factors.

Endogeneity and robustness checks

Despite this study introducing as many levels of control variables as possible to mitigate the bias caused by omitted variables, the benchmark results may still be affected by endogeneity issues. A pertinent issue to highlight is the housing as the most important asset for most households. Housing purchase decisions are closely linked with household characteristics. Households typically base these decisions on expected returns and their actual conditions, positioning homeowners and renters differently in terms of initial status. Consequently, the selection into “homeowners” or “renters” is not random. This non-randomness could introduce systematic selection bias into the estimated results. Additionally, there may also be reverse causality between homeownership and changes in household economic status, and households with improved household economic status may also be more inclined to purchase a home. Finally, since it is not possible to include everything that affects changes in the economic status of households in the model, the resulting omitted variable problem may also lead to endogeneity issues.

Therefore, to address the endogeneity issues as much as possible, this study employs the land supply area of each province in China as an instrumental variable for housing property. This choice is justified because land resources, crucial inputs in the housing market, directly influence housing prices and, subsequently, the acquisition of housing property rights by households. Moreover, under China’s public land ownership system, the use and transfer of land in each province are rigorously regulated by the Measures for the Administration of Annual Plans on the Utilization of Land (Order of the Ministry of Land and Resources of the People’s Republic of China No. 66). In addition, the implementation of China’s land bid invitation, auction, and listing system grants local governments a near-monopolistic control over land supply, making land transfer a relatively exogenous government policy variable.

In this study, the land supply area of each province in China for the years 2010, 2012, 2014, 2016, and 2018 serves as an instrumental variable in an extended regression model to estimate the impact of homeownership on changes in household economic status. The regression results, displayed in Table 3, reveal that even after the introduction of the instrumental variable, the relationship between homeownership and changes in household economic status remains significantly negative at the 1% level. This finding corroborates the benchmark regression results, suggesting that homeownership does not facilitate improvements in the economic status of households.

Table 3 Endogeneity tests: using instrument variable analysis.
Full size table

Employing propensity score matching

This study employed propensity score matching to estimate the impacts of homeownership on income level changes, thereby to further validate the robustness of the benchmark regression results. After calculating the propensity score of individuals, we used different matching approaches to match the treated group and the control group. However, a consensus has yet to be reached on which matching approach delivers the most efficient results. To ensure the robustness of the matching results, this study utilized various matching approaches, including nearest neighbour matching, radius matching, and kernel matching. The estimation results derived from these various matching methods are presented in Table 3. The coefficient for homeowners across different matching methods are consistently similar and significantly negative at the 1% level, implying that the estimated results are robust against potential self-selection and model specification biases. Additionally, the quality of the propensity score matching was rigorously tested. The results affirmed that the matching quality in this study is high (c.f. Table 4), ensuring that the matching process effectively mitigates potential biases in the analysisFootnote 3.

Table 4 Robustness tests: employing propensity score matching.
Full size table

Replacing the independent variable

In the initial phase of this research, household income was categorized into ten ascending levels. Recognizing the potential subjectivity of this classification, we sought to mitigate the impact of income level categorization on the estimation results. To this end, household income was reclassified into additional categories, specifically into 7, 8, and 9 levels, as well as 11, 12, and 13 levels. For each classification, we calculated the changes in income levels, defined as the difference in income level between consecutive surveys. The regression analyses, incorporating these variations in income level classifications, are presented in Table 5. Across all classifications, the variable representing homeownership was consistently and statistically significant, negatively impacting at the 5% level. This consistency suggests that the estimated results are robust and not significantly influenced by the method of income level classification.

Table 5 Robustness tests: replacing the independent variable.
Full size table

Controlling for “perceived” price increase potential

In this study, we hypothesize that homeowners might forgo certain opportunities for income improvement to maintain their homeowner status. Homeowners are clever. They know that such a small sacrifice would be compensated for by the rising prices in the real estate market. Even in a difficult year, owning a home is still important for a stable and decent life. Therefore, it is very important to control for a “perceived” price increase potential in the model. To control for a “perceived” price increase potential, we add the annual growth rate of regional housing prices in the past year, the average annual growth rate of regional housing prices in the past two years, or the average annual growth rate of regional housing prices in the past three years to the baseline model. The results, as shown in Table 6, show that the homeownership coefficient remains negative and statistically significant at the 1% level, even after these adjustments. This finding suggests a persistent negative correlation between homeownership and changes in household economic status, even when controlling for potential wealth accumulation effects due to rising property values.

Table 6 Robustness tests: controlling for “perceived” price increase potential.
Full size table

Using the latest CFPS 2020 data

To enhance the robustness of our findings, we also incorporated the most recent data from the China Family Panel Studies (CFPS) for the year 2020. We refined our dataset by excluding individuals who were not consistently interviewed across all six survey rounds from 2010 to 2020, thereby constructing a balanced panel dataset spanning six years. The results from the fixed effects model, as presented in Table 7, indicate that homeownership is significantly and negatively associated with changes in household economic status. This finding is consistent with our benchmark regression results.

Table 7 Robustness tests: using the latest CFPS 2020 data.
Full size table

Mechanism analysis

The results above suggest that homeownership is not conducive to improving the economic status of households. This may be because of the immobility of housing in physical space and the enormous pressure of housing loans, resulting in the “housing mortgage slave effect” and preventing households from easily moving to seek better job opportunities and engaging in higher-return investment activities. We therefore attempt to test the “housing mortgage slave effect” by examining the impact of homeownership on job mobility. If there is a “housing mortgage slave effect”, then the job mobility of the households would decrease. This study uses data from CFPS 2010 and 2020 to verify this mechanism. We use the CFPS 2010 and 2020 data because only the data from these two rounds of surveys cover respondents’ job change information. The questionnaire contains a survey question “Is your current job your first job?”. If the respondent’s current job is his or her first job, the value equals 1; otherwise, it equals 0. If the respondent’s current job is the respondent’s first job, we perceive that the respondent’s job mobility is considered relatively low. The regression results, as presented in Table 8, indicate that homeowners have a much higher likelihood to remain in their first job as compared to renters, signifying a significantly lower level of job mobility.

Table 8 Homeownership and job mobility.
Full size table

Further analysis

The heterogeneity of intergenerational financial support

The CFPS survey data provide information on intergenerational financial support, thus this study can analyse how the financial supports received by adult children from their parents (including their grandparents) influences their housing mortgage pressures. The results of the regression, as shown in Table 9, show that homeownership is significantly and negatively related to changes in the economic status of households without intergenerational financial supports but has a nonsignificant impact on the economic status changes of households with intergenerational financial supports. This suggests that financial supports can reduce the burden of home mortgages associated with the acquisition of housing property, thereby mitigating the negative correlation between homeownership and the economic status changes of households. This finding has significant policy implications that will be discussed in the concluding section.

Table 9 Heterogeneity of intergenerational financial support.
Full size table

Conclusions and policy implications

This study examines the impact and mechanism of homeownership on household economic status changes using household-level microsurvey data from the CFPS for the years of 2010, 2012, 2014, and 2016. Three main findings are achieved. First, homeownership is not conducive to improving the economic status of households. This conclusion remains robust after addressing endogeneity issue using the instrumental variable (IV) approach, as well as a series of robustness tests, including propensity score matching, controlling for “perceived” price increase potential and substituting independent variables. The likely cause of this finding is that the immobility of housing in physical space and the enormous pressure of housing loans, resulting in a “housing mortgage slave effect”, which prevent households from easily relocating for better job opportunities or engaging in higher-return but risky investment activities. Mechanism analyses further confirm that homeownership reduces household job mobility. Finally, homeownership is significantly and negatively related to changes in the economic status of households without intergenerational financial supports but has a nonsignificant effect on those with intergenerational financial support. These findings suggest that financial support can reduce the burden of home mortgages associated with home acquisition, therefore mitigating the negative correlation between homeownership and household economic status.

The findings of this study have important policy implications. First, this study adds a new perspective to the debate on whether the government should promote homeownership, considering its effects on changes in economic status. While homeownership is often lauded for its material and spiritual benefits, this paper highlights that it can also impose a significant mortgage burden on families, potentially deterring the pursuit of higher income opportunities. Second, this study offers policy guidance for reducing the negative socioeconomic effects of homeownership from the perspective of the “housing mortgage slave effect”. If the adverse effects of homeownership stem primarily from high mortgage burdens, they can be mitigated by policies aimed at lowering mortgage interest rates and extending the duration of mortgages. Third, the paper calls for the development of a healthy capital market to help households accumulate assets through means other than housing investment. In emerging economies, household asset tends to be highly concentrated on housing (Badarinza et al. 2019). For example, in China, housing asset account for more than 70% of household assets (Wan et al. 2021). This high concentration is largely due to the lack of investment channels other than housing, exacerbated by a long-term stagnation in the stock market, leading to excessive demand for housing investment (Badarinza et al. 2019; Wan et al. 2021). The finding in this paper that financial support can mitigate the negative correlation between homeownership and household economic status suggests that enhancing the attractiveness of financial investment could benefit both the housing market and the financial market. Finally, to migrate the homeownership trap of social mobility, it is crucial to extend housing options other than owner-occupied homes. In emerging economies, one of the reasons for the high housing prices is that public services and social resources are highly tied to homeownership. Consequently, the demand for homeownership is not only driven by the need for stable housing but also access to public services. Therefore, the government should strive hard to promote equal rights across renters and homeowners in terms of access to public services. In 2016, the Central Economic Work Conference of China pledged to implement a housing policy that emphasizes “housing is built for living in, not for speculation”. This study reveals that excessive demand towards homeownership could hinder households’ economic status improvement, endorsing policies to curb on housing speculation.

However, our study still has some limitations. First, the relationship between homeownership and social mobility examined in this paper occurs during a unique period characterized by both a sustained economic boom and significant economic transformations. Whether the correlation observed in this paper changes across different economic cycles requires further examination. Second, the analysis coincides with an unprecedented housing boom, during which the allure of home asset appreciation might overshadow or offset the loss in economic status growth. Third, the current analysis is conducted within the context of China, a country with a strong cultural preference for homeownership, where households may be willing to sacrifice economic status for non-economic benefits of owning a home. Future studies could consider examining different countries or different housing market cycles to explore the relationship between homeownership and social mobility in more general contexts.