The Impact of Agricultural Hukou on Migrants’ Home Purchasing in Destination Cities of China

Abstract

The dual Hukou system, originating in China’s planned economy period, structured Chinese society into separate urban and rural segments, thereby generating distinct sets of rights and benefits for agricultural and non-agricultural residents regarding land, social security, education, and healthcare. Urban home purchase is a pivotal indicator of social integration for rural–urban migrants in destination cities. While the literature has extensively examined migrants’ residential conditions in China, the institutional impact of the agricultural hukou system—a core constraint—on their urban homeownership, along with its underlying mechanisms and heterogeneity, remains underexplored. To address this gap, this study adopts a twofold approach: theoretically, it employs the separating equilibrium model in housing markets with incomplete information to verify that agricultural hukou acts as an institutional barrier to migrants’ local home purchases; empirically, it uses data from the China Migrants Dynamic Survey (CMDS) and applies the Fairlie decomposition method to quantify the constraint effect. The empirical results suggest that agricultural hukou exerts a 29.72% suppressive effect on migrants’ urban home purchase behavior. This effect operates indirectly by weakening migrants’ long-term settlement intention, which serves as a mediating variable. Moreover, the hindrance of agricultural hukou varies heterogeneously across groups, differing in education level, generational cohort, and regional distribution. To advance the fair and sustainable development of the real estate market, we advocate accelerating hukou reform by decoupling public services from residence status, fostering inclusive urbanization, and ensuring equitable development of housing markets.

1. Introduction

In January 1958, the Household Registration Regulations of the People’s Republic of China formally institutionalized the division of urban and rural residents into “agricultural hukou” and “non-agricultural hukou” categories. The urban–rural dual structure system, centered on the separate administration of urban and rural hukou, is a product of China’s planned economy era. “Agricultural hukou” is traditionally tied to rural areas, land contract rights, and rural-specific welfare, while “urban hukou” is associated with urban residency and access to urban public services such as education, medical care, and housing subsidies. In developing countries, advancing the citizenization of rural-to-urban migrants and narrowing the urban–rural gap hold positive strategic significance for implementing the United Nations Sustainable Development Goals (SDGs). As early as December 2009, the Central Economic Work Conference emphasized that “resolving the employment and household registration issues of qualified rural migrant workers in cities and towns should be a key task in advancing urbanization, with relaxed hukou restrictions in medium and small cities and towns.” In 2016, the State Council further eased household registration restrictions except in megacities and established a points-based hukou registration system. Nevertheless, urban–rural hukou barriers remain prevalent to date. According to official data from the National Bureau of Statistics (NBS) and the Ministry of Public Security (MPS) of China, the permanent population urbanization rate rose from 17.92% to 66.16% between 1978 and 2023; in contrast, the hukou-based urbanization rate stood at only 48.3%, leaving approximately 252 million rural-to-urban migrants without local hukou. The disparity between these two rates is primarily driven by rural-to-urban migrants. Specifically, due to the distinction between agricultural and non-agricultural hukou statuses, rural-to-urban migrants face discriminatory treatment in multiple aspects in cities, such as employment opportunities, wage levels, and public services.

Nevertheless, three critical research gaps remain unaddressed in the existing literature. First, it is unclear whether agricultural hukou status creates disparities in home purchasing difficulty between rural-to-urban and urban-to-urban migrants in their destination cities. Second, the extent to which agricultural hukou identity constrains these migrants’ home purchasing behavior in their destination cities. Third, potential heterogeneity in these hukou-induced disparities between the two migrant groups has yet to be explored. Understanding these issues is essential for forecasting potential housing demand under China’s ongoing hukou and urbanization reforms.

The potential marginal contributions of this study include the following: (1) Marginal contribution in research perspective: Unlike previous studies that focus on housing differences between migrant and local populations, this study, based on the perspective of urban–rural hukou stratification within urban migrant populations, can more accurately identify the degree of hindrance caused by rural hukou to migrant populations’ participation in the local housing market. (2) Marginal contribution in expanding the signaling theory model: This study constructs a hukou signaling model in the housing market with imperfect information, which theoretically confirms that rural hukou is an institutional factor hindering migrant populations from purchasing housing in cities. (3) Marginal contributions in empirical method and conclusions: The Fairlie decomposition method for discrete choice models adopted in this study has rarely been used in previous housing-related studies. Compared with the traditional linear regression method, the Fairlie decomposition method can quantify the disparity in the average expected probability of home purchasing between agricultural and non-agricultural hukou migrants by disaggregating it into two components: the “explained” or “endowments effect,” attributable to differences in observable characteristics such as income and education, and the “unexplained” or “coefficients effect,” which captures structural differences and discrimination, including that linked to hukou status itself. This approach effectively isolates the specific contribution of agricultural hukou discrimination to the homeownership gap between rural-to-urban and urban-to-urban migrants.

The remaining part of this study is structured as follows: Section 2 presents a literature review and commentary. Section 3 proposes research hypotheses based on the signal model of the housing market with imperfect information. Section 4 introduces the basis for selecting the empirical model, defines relevant variables, and conducts descriptive statistical analysis. Section 5 includes basic regression, mediating effect tests, and robustness tests; discusses the unequal effects of agricultural hukou on home purchasing among rural–urban and urban–urban migrants in cities based on decomposition results; and further examines the heterogeneous characteristics of the coefficient effect. Section 6 presents the conclusions and implications.

2. Literature Review and Commentary

Housing is the most fundamental issue that migrant populations must first address in their destination cities. Migrant populations in cities may accept temporary voluntary or involuntary unemployment, but they cannot go a single day without a place to live.

Over the past few decades, housing inequality among international migrants has emerged as a key research topic in migration studies. Scholars have focused on examining cross-country differences in immigrants’ homeownership rates [1,2,3]. Some scholars have argued that regional housing price increases can create arbitrage expectations, thereby attracting immigrant inflows [4,5]. However, housing tenure choices can influence migrants’ relocation costs and regional mobility [6,7]. The residential stratification model has further confirmed that discrimination and biases faced by migrants in the housing market restrict their access to homeownership [8].

International migration typically flows from developing countries to developed countries. A subset of studies has adopted the ecological footprint as the metric of consumption. These studies have demonstrated that such migrants exert a significant impact on both the consumption level and spatial structure of the destination cities’ housing market [9]. The pressure of purchasing housing amid high prices also motivates individuals to migrate from high-housing-price countries to those with lower prices [10]. Additionally, a subset of studies has focused on metropolitan internal migration, linking migration flows to analyses of the spatial structure of the housing market. These studies have found that urban variations in housing utility and housing prices are important factors shaping migrants’ housing consumption patterns [11,12,13].

However, this international migration pattern differs significantly from the “rural-to-urban” migration of China’s rural population in the context of the unique hukou system. In the process of China’s urbanization, most studies on hukou and migrant populations’ housing have focused on comparing residential segregation and differences in housing conditions between rural-to-urban migrant populations (RUMPs) and local urban residents [14,15]. Scholars have examined a nationally representative 2013 household survey and used the mother’s hukou status as an instrumental variable. Their empirical analysis found that household heads with rural hukou were approximately 20 percentage points less likely to be homeowners in cities than comparable household heads with local urban hukou [16]. The hukou system and the social welfare system linked to it are regarded as important factors impeding the citizenization of RUMPs [17,18,19].

Regarding empirical research on economic inequality across different groups, the “Taste Discrimination Model” established by Becker marks the beginning of modern discrimination economics [20,21]. Since then, scholars have used various econometric methods to measure the degree of discrimination in the labor market. Among these, the Mincer earnings regression model [22,23,24] and the linear Oaxaca–Blinder decomposition method [25,26] are regarded as classic approaches for estimating this degree and have been widely adopted by scholars. However, this decomposition method—widely applied in labor economics models—produces biases when directly used in discrete models [27]; thus, scholars have extended this method to nonlinear models [28,29,30].

In the rental housing market, extant research has documented the existence of racial discrimination, which serves to reproduce racial inequality and residential segregation [31,32]. Scholars have identified that such discrimination is influenced by factors such as age, gender, cognitive skills, and local market knowledge, and have found that the discriminatory social norm is widely recognized among real estate agents [33]. However, in China’s real estate market, few studies have quantified the magnitude of discriminatory barriers imposed on migrant populations’ access to homeownership due to their rural hukou status.

3. Theoretical Model and Hypothesis

This study assumed that in a housing market with imperfect information, the true housing purchase capability of migrants as potential buyers is T_i. Sellers cannot directly observe the true housing purchase capability of migrants; instead, they can only obtain the signal value $T_i^j*$ of this capability based on documents provided by migrants or intermediaries, such as those related to the potential buyers’ hukou, income, and educational background:

$$T_i^j*=T_i^j+u_i^j$$

(1)

where j serves as the hukou identifier: j = 0 or 1 when individual i is an agricultural or urban–urban migrant, respectively. $u_i^j$ is a random disturbance term that is independent of $T_i^j$ and follows a standard normal distribution (0, ε_j²). It is assumed that ε₀² = ε₁² = ε².

In a market with perfect information, the market equilibrium condition is where the seller’s offer matches the buyer’s capability. However, in a market with imperfect information, the market equilibrium condition shifts to a separating equilibrium [34,35] based on the signal T* transmitted between sellers and migrants with different hukou statuses.

$$P_i^j=E(T_i^j/T_i^j*)=E(T^j)+{\displaystyle \frac{var(T^j)}{var(T^j)+var{u}^j}}[T_i^j*-E(T^j*)]$$

(2)

Let ${\beta }^j={\displaystyle \frac{var\left(T^j\right)}{var\left(T^j\right)+{\epsilon }^{j^2}}}$. Then, Equation (2) can be further rearranged as

$$P_i^j=E(T_i^j/T_i^j*)=\underset{①}{\underbrace{(1-{\beta }^j)E(T^j)}}+\underset{②}{\underbrace{{\beta }^j{T}_i^j*}}$$

(3)

where it can be seen that under the market equilibrium, housing prices are linearly correlated with the signal value $T_i^j*$ obtained by sellers. It represents a weighted average of the mean true housing purchase capability E(T^j) and the signal value of housing purchase capability $T_i^j*$. The lower the degree of information asymmetry between buyers and sellers, the higher the weight of $T_i^j*$. In Equation (3), ① denotes the group effect of different hukou groups, and ② denotes the individual capability effect. To compare the differences in housing purchase difficulty faced by urban–urban and rural–urban migrants, it is necessary to further discuss the strength of the relationships of intercepts and slopes between different hukou groups in Equation (3):

(1)

E(T⁰) < E(T¹): The expected housing purchase capability of urban–urban migrants is higher than that of rural–urban migrants. The true housing purchase capability of migrants in the destination city mainly depends on expected income, financing constraints, access to public services, settlement intention, etc. There are significant differences between rural–urban and urban–urban migrants in these aspects: ① Differences in hukou nature lead to gaps in expected lifetime income. Most urban–urban migrants first receive higher education or accumulate strong management and professional skills in cities before engaging in cross-regional employment. ② Differences in hukou nature result in disparities in information transaction costs during housing purchase financing. From the perspective of credit demand, issues such as excessively high transaction costs and information asymmetry may also prevent rural households from obtaining the expected amount of funds. Transaction costs and financing constraints directly affect the intertemporal allocation of household housing assets, exacerbating the difficulty for agricultural migrants to obtain financing for homeownership in cities. ③ Differences in hukou nature lead to disparities in access to local social security and public services. Welfare elements such as education and medical security that are strongly attached to the hukou system strengthen the precautionary savings motive of agricultural migrant households to cope with greater uncertainties in their amphibious urban–rural lives. This, to a certain extent, inhibits housing purchase intention, crowds out housing consumption, and lowers the rate of urban housing purchases among agricultural migrants.

(2)

Var (T_i⁰) > Var (T_i¹): The disparity in housing purchase capability is greater among agricultural migrants. Due to limited educational levels, migrants’ choices of destination cities and occupations are highly temporary and periodic. The incomes of agricultural migrants lack long-term stability and are prone to fluctuations due to external economic environment shocks. In addition, uncertainties arising from the long-distance amphibious life between urban and rural areas act as strong interference factors affecting the housing purchase capability of agricultural migrants in cities.

Based on the above discussion, it can be concluded that the intercept terms and slope terms of different hukou groups in the linear equation of Equation (3) have the following relationships, respectively: [(1 − β₀) E(T⁰)] < [(1 − β₁) E(T¹)] and β₀ > β₁. Even if sellers observe the same housing capability signal from rural–urban and urban–urban migrants, rural–urban migrants will face higher housing quotes and greater difficulty in purchasing housing due to the explicit label of agricultural hukou. Based on the above model analysis, we propose the first hypothesis as follows:

Hypothesis 1.

Agricultural hukou reduces the actual residential purchasing rate of migrants in the destination cities.

Agricultural migrants strongly depend on rural social networks tied by kinship and geography. Additionally, differences in culture, region, and living habits significantly weaken their sense of identity and belonging to cities. Compared with urban–urban migrants, agricultural hukou undermines migrants’ long-term settlement intention in cities, which serves as a leading variable for final housing purchase decisions. Thus, the second hypothesis is proposed:

Hypothesis 2.

Agricultural hukou affects the residential purchasing rate by weakening migrants’ long-term settlement intention, which acts as an intermediary variable.

Due to imbalanced allocation of urban–rural educational resources, most agricultural migrants have low educational levels, leading to high randomness and temporariness in their choices of destination cities and occupations. By contrast, urban–urban migrants often receive higher education or accumulate strong management and professional skills in cities before seeking cross-regional employment. Migrant groups with high human capital tend to have higher incomes and stable expectations, and they are more likely to purchase housing in cities, forming strong “rigid demand” for urban housing. In the urban housing market, educational attainment, as an explicit positive signal to sellers, can reduce information asymmetry between buyers and sellers and weaken the obstruction of agricultural hukou to urban residential purchasings. Therefore, the third hypothesis to be tested is proposed:

Hypothesis 3.

The obstructive effect of agricultural hukou on migrants’ residential purchases in destination cities decreases with the improvement in educational levels.

Against the backdrop of accelerated intergenerational transition among rural-to-urban migrant populations—with the post-1980s’ new generation gradually becoming the main body of this group—scholars have gradually conducted research on intergenerational differences among RUMPs. Demographic theory holds that intergenerational differences include age and cohort effects. The cohort effect is a group characteristic, and people of the same generation can be regarded as a single cohort [36]. New-generation migrants tend to live in cities and have a stronger desire to settle there [37,38]. Based on this, the fourth hypothesis to be verified in this study is proposed as follows:

Hypothesis 4.

The coefficient effect of agricultural hukou identity on residential purchases in destination cities is weaker for the “new generation” of agricultural migrants compared with the “older generation”.

The housing market is significantly regional, with average transaction prices showing a pattern of “higher in the east and lower in the west” due to differences in regional economic development and vitality. Migrant worker families have higher housing affordability in central and western regions, but this is lower in eastern regions, resulting in varying degrees of difficulty in purchasing housing across regions. Moreover, economically developed eastern regions have a higher degree of openness, stronger market awareness among buyers and sellers, and more effective coordination between financial and housing markets. Thus, the following hypothesis is proposed:

Hypothesis 5.

In economically developed and highly open regions, the obstructive effect of agricultural hukou on migrants’ residential purchases in destination cities decreases.

4. Empirical Models and Data

A discrete model is constructed to identify the impact of agricultural hukou on local housing purchase behavior, followed by the application of a nonlinear model decomposition method to further quantify the contribution of agricultural hukou to differences in housing purchase choices between different migrant groups.

4.1. Selection of Discrete Model

It is assumed that the potential factors determining whether migrant i can purchase housing locally depend linearly and additively on observable and unobservable factors $X_i^j$ and ${\epsilon }_i^j$, respectively. Here, j serves as the hukou identifier: j = 0 or 1 when individual i is an agricultural or non-agricultural migrant, respectively. The observed probability of the migrant purchasing housing locally is

$$Pr(y_i^j=1)=Pr\left[(X_i^j{\beta }^j+{\epsilon }_i^j)>0\right]=F(X_i^j{\beta }^j)$$

(4)

where $y_i^j$ is a binary choice variable indicating whether migrant individual i has made a local housing purchase, taking a value of 1 if they have and 0 if they have not. β^j represents the coefficient of observable factors. When there are multiple categorical variables among the observable variables $X_i^j$, each categorical variable needs to be converted into dummy variables before being included in the model. Assuming ${\epsilon }_i^j$ is a logistic distribution, Equation (4) is a Logit model. If the probability of $y_i^j=1$ is ${\pi }_i^j$, the odds of migrant individual i making a local housing purchase versus not making one is

$${\mathit{\Omega }}_i={\displaystyle \frac{{\pi }_i^j}{1-{\pi }_i^j}}=e^{X_i^j{\beta }^j}$$

(5)

4.2. Mediating Effect Model

In the Logit nonlinear binary probability model, the total effect cannot be decomposed into direct and indirect effects as in linear models. This study employs the KHB method for mediating effect analysis in nonlinear regression models to decompose the mediating effect by specifying the relationships among the dependent variable y, the initial independent variable x, and the mediating variable z [39]. The specific decomposition steps are as follows:

$$Logit[Pr(y *>0)]=b_{yxz}x+b_{yzx}z={\displaystyle \frac{{\beta }_{yzx}}{{\sigma }_e}}x+{\displaystyle \frac{{\beta }_{yxz}}{{\sigma }_e}}z$$

(6)

$$Logit[Pr(y *=1)]=b_{yx}x={\displaystyle \frac{{\beta }_{yx}}{{\tilde{\sigma }}_e}}x$$

(7)

$$z={\theta }_{zx}x+w$$

(8)

$$\mathrm{D}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}\text{:} DE=b_{yxz}={\displaystyle \frac{{\beta }_{yxz}}{{\sigma }_e}}$$

(9)

$$\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}\text{:} IE={\theta }_{zx}{b}_{yzx}={\displaystyle \frac{{\theta }_{zx}{\beta }_{yzx}}{{\sigma }_e}}$$

(10)

$$\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}\text{:} TE=DE+IE={\displaystyle \frac{{\beta }_{yx}}{{\sigma }_e}}={\displaystyle \frac{{\beta }_{yxz}+{\theta }_{zx}{\beta }_{yzx}}{{\sigma }_e}}$$

(11)

Since the scaling factor ${\sigma }_e$ has been unified at this time, the proportion of the mediating effect IE to the total effect TE can be measured:

$${\displaystyle \frac{IE}{TE}}={\displaystyle \frac{{\theta }_{zx}{\beta }_{yzx}}{{\beta }_{yxz}+{\theta }_{zx}{\beta }_{yzx}}}$$

(12)

4.3. Fairlie Decomposition Method for Nonlinear Models

4.3.1. Decomposition Method Based on Different Estimated Coefficients

Drawing on Fairlie’s nonlinear decomposition method [28], this study constructs counterfactual distributions to decompose differences in housing purchase intentions and actual housing purchase odds ratios among migrant populations with different hukou statuses. Advantages of this approach include

$${\overline{Y}}^A-{\overline{Y}}^U=\left[{\displaystyle \sum _{i=1}^{N^A}\frac{F(X_i^A{\stackrel{\wedge }{\beta }}^A)}{N^A}-{\displaystyle \sum _{i=1}^{N^U}\frac{F(X_i^U{\stackrel{\wedge }{\beta }}^A)}{N^U}}}\right]+\left[{\displaystyle \sum _{i=1}^{N^U}\frac{F(X_i^U{\stackrel{\wedge }{\beta }}^A)}{N^U}-{\displaystyle \sum _{i=1}^{N^U}\frac{F(X_i^U{\stackrel{\wedge }{\beta }}^U)}{N^U}}}\right]$$

(13)

where $X^j$ is the row vector of explanatory variables, ${\stackrel{\wedge }{\beta }}^j$ is the row vector of estimated coefficients, ${\overline{Y}}^j$ represents the average probability of the binary choice between renting and purchasing, and F is the cumulative distribution function of the logistic distribution. Equation (13) decomposes the total difference into two terms on the right-hand side: The first bracket represents the endowments effect, and the second bracket denotes the coefficient effect. In this study, it specifically refers to the differences in housing purchase willingness and actual housing purchase rate caused by the binary hukou system. The larger the proportion of the coefficient effect, the more serious the hindrance of agricultural hukou to the housing choices of the floating population. Similarly, total difference decomposition can be carried out by taking agricultural migrants as the reference group.

The nonlinear Fairlie difference decomposition method also has the index problem. The choice of different reference groups will lead to different decomposition results. To avoid the impact of this reference group problem on the empirical results and enhance the robustness of the output results, this study uses the estimated coefficients of the pooled sample ${\stackrel{\wedge }{\beta }}^P$ for difference decomposition [26,39]:

$${\stackrel{—}{Y}}^A-{\stackrel{—}{Y}}^U=\left[{\displaystyle \sum _{i=1}^{N^A}\frac{F(X_i^A{\stackrel{\wedge }{\beta }}^P)}{N^A}-{\displaystyle \sum _{i=1}^{N^U}\frac{F(X_i^U{\stackrel{\wedge }{\beta }}^P)}{N^U}}}\right]+\left[{\displaystyle \sum _{i=1}^{N^A}\frac{F(X_i^A{\stackrel{\wedge }{\beta }}^A)}{N^A}-{\displaystyle \sum _{i=1}^{N^A}\frac{F(X_i^A{\stackrel{\wedge }{\beta }}^P)}{N^A}}}\right]+\left[{\displaystyle \sum _{i=1}^{N^U}\frac{F(X_i^U{\stackrel{\wedge }{\beta }}^P)}{N^U}-{\displaystyle \sum _{i=1}^{N^U}\frac{F(X_i^U{\stackrel{\wedge }{\beta }}^U)}{N^U}}}\right]$$

(14)

where the total difference is decomposed into three parts: the first term on the right-hand side is the endowments effect; the second and third terms are both coefficient effects caused by hukou differences. Among them, the second term can be regarded as the unequal treatment that agricultural migrants receive in the housing market due to their hukou status, while the third term can be seen as the additional preferential treatment that non-agricultural migrants enjoy in the housing market.

4.3.2. Identification of the Contribution of Observable Variables to the Total Difference

When the values of other explanatory variables are fixed, the independent contribution of the distribution characteristics of a specific explanatory variable X_j (1 ≤ j ≤ n) in Equation (14) to the endowments effect can be further identified. The procedure involves the following steps:

First, since there are more agricultural migrants in this study than inter-urban migrants, the program randomly selects a subsample from the agricultural migrant group such that the sample size equals that of the inter-urban migrant group (N^A = N^U).

Second, following Fairlie’s nonlinear decomposition method [29], the estimated probabilities are sorted in ascending order for the two groups with equal sample sizes. This enables a one-to-one counterfactual pairing of individuals at the same rank position i (1 ≤ i ≤ N^U) between the agricultural migrant group and the inter-urban migrant group.

Finally, by summing the contribution degrees of observable variable X_jm across all rank positions, the contribution of the specific observable variable X_j to the total difference in renting–purchasing choices between groups can be identified.

Similarly, if the estimated coefficients of the inter-urban migrant group ${\stackrel{\wedge }{\beta }}^U$ in Equation (13) are replaced with those of the agricultural migrant group ${\stackrel{\wedge }{\beta }}^A$ and the pooled sample ${\stackrel{\wedge }{\beta }}^{\mathrm{P}}$, the contribution of hukou differences in the specific variable X_j to the disparity in renting–purchasing choices in Equations (13) and (14) can be identified. Additionally, the variance of each observable variable’s contribution to the total difference can be identified, which is not elaborated here.

4.3.3. Variable Description and Statistical Inference

The data used in this study is derived from the China Migrants Dynamic Survey (2016) conducted by the National Health and Family Planning Commission. The survey selected sampling points in areas with a relatively concentrated floating population across 31 provinces (autonomous regions and municipalities directly under the Central Government), adopting a stratified, multi-stage, probability proportional to size (PPS) sampling method. The China Migrants Dynamic Survey was discontinued after 2018, and the latest survey data from 2017 and 2018 fail to incorporate all the variables required for this study. Consequently, the 2016 dataset remains the most up-to-date one encompassing all the essential variables for this study. The total sample size of the survey was 169,000 households. After excluding samples with missing or ambiguous values in the selected independent variables, the final valid sample size was determined to be 150,258 households, including 127,385 in the agricultural transfer group and 22,873 in the inter-city mobility group. Stata 17 was used for the empirical analysis in this study. Stata 17 can run complex models and has sufficient capacity to handle sample data of approximately 170,000 observations.

The core explained variable of this paper, “whether housing purchase has been made locally”, serves as an objective indicator to measure the real participation of the floating population in the local housing transaction market. Differences in the actual housing purchase rates between the agricultural and non-agricultural transfer populations indicate the inequality of opportunity and inequality of outcome in the local housing market.

Considering that the factors influencing the housing purchase decisions of the floating population are more complex than those affecting employment choices—often involving dilemmas such as family migration and geographical/kinship ties, the individual trait effects in the explanatory variables are expanded into three parts: personal characteristics, family characteristics, and mobility characteristics (see

Table 1. Variable classification and definition.

Variable Category	Variable Name	Variable Meaning and Assignment
Explained Variables	Local Housing Purchase Intention	Yes = 1, No = 0
	Purchased Housing Locally	Yes = 1, No = 0
Core Explanatory Variables	Hukou Nature	Categorical variable: Agricultural = 0, Non-agricultural = 1 (the non-agricultural hukou category does not include small samples of two groups: those who changed from agricultural to non-agricultural status and those who changed from non-agricultural to urban status)
Intermediary Variables	Intention to Reside Locally Long-Term (>5 years)	Categorical variable: No or Uncertain = 0, Yes = 1
Explanatory Variables (Personal Characteristics)	Gender	Categorical variable: Female = 0, Male = 1
	Age	Continuous variable
	Education Categorical variable,	Categorical variable: Primary school or below = 1, Junior high school = 2, High school = 3, Junior college or above = 4
	Marital Status	Categorical variable: Unmarried = 0, Married = 1
	Housing Provident Fund	Categorical variable: No = 0, Yes = 1
	Endowment Insurance	Categorical variable: No = 0, Yes = 1
Explanatory Variables (Family Characteristics)	Monthly Family Income	Continuous variable
	Number of Biological Children	Continuous variable
	Family Size in Inflow Area	Continuous variable
Explanatory Variables (Mobility Characteristics)	Total Number of Migrations	Continuous variable
	Cumulative Duration of Migration	Categorical variable: Less than 1 year = 1, 1–4 years = 2, 5–9 years = 3, 10–19 years = 4, 20 years or more = 5
	Scope of Current Migration	Categorical variable: Inter-provincial migration = 1, Intra-provincial inter-city migration = 2, Intra-city inter-county migration = 3
	Inflow Area	Eastern Region = 1, Northeastern Region = 2, Central Region = 3, Western Region = 4 (the Eastern Region includes Beijing, Tianjin, Hebei, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian, Guangdong, and Hainan; the Central Region includes Shanxi, Anhui, Henan, Hubei, Hunan, and Jiangxi; the Western Region includes Inner Mongolia, Guangxi, Shaanxi, Gansu, Ningxia, Qinghai, Xinjiang, Sichuan, Chongqing, Yunnan, Guizhou, and Tibet; the Northeastern Region includes Liaoning, Jilin, and Heilongjiang)
	Migrated Alone	Categorical variable: Yes = 0, No = 1

).

This study focuses on the differential impact of urban and rural household registration on the housing choices of the floating population. Therefore, in the statistical description section, not only is the overall sample description provided, but the floating population is also divided into the rural–urban and urban–urban migrant groups based on their household registration status for description. The mean differences in each variable between the household registration groups are also presented in

Table 2. Statistical descriptions of variables and mean differences between groups.

Statistical Indicator	Full Sample		Rural–Urban Migrant Group		Urban–Urban Migrant Group		Mean Differences Among Household Registration Groups
	Mean	Variance	Mean	Variance	Mean	Variance
Local Housing Purchase Intention	0.235	0.424	0.223	0.417	0.297	0.457	−0.074
Whether Purchased Housing Locally	0.280	0.449	0.246	0.431	0.467	0.499	−0.221
Hukou Nature	0.152	0.359	0.000	0.000	1.000	0.000	1.000
Long-Term Local Residence Intention	0.546	0.248	0.263	0.249	0.663	0.224	−0.400
Male	0.523	0.500	0.524	0.499	0.517	0.500	0.007
Age	34.777	10.381	34.474	10.121	36.465	11.580	−1.991
Education Level	2.400	0.916	2.262	0.847	3.170	0.905	−0.908
Migration Work Experience	2.771	1.029	2.774	1.036	2.752	0.991	0.022
Endowment Insurance	0.541	0.498	0.519	0.500	0.662	0.473	−0.143
Housing Provident Fund	0.102	0.303	0.068	0.251	0.293	0.455	−0.226
Family Income (CNY)	6880	17,120	6514	6764	8919	40,814	−2405
Married	0.830	0.375	0.830	0.376	0.836	0.371	−0.006
Number of Biological Children	1.176	0.865	1.220	0.876	0.929	0.750	0.291
Migration Family Size	2.603	1.178	2.651	1.186	2.334	1.095	0.317
Number of Migrations	1.342	1.024	1.364	1.062	1.220	0.769	0.143
Not Migrated Alone	0.669	0.471	0.688	0.463	0.564	0.496	0.124
Parental Migration Experience	1.352	0.715	1.366	0.727	1.277	0.643	0.089
Migration Scope	1.682	0.751	1.682	0.753	1.680	0.739	0.002
Migration Region	2.419	1.317	2.433	1.321	2.339	1.287	0.094
Individual Sample Size	150,258		127,385		22,873		/

.

Based on the mean differences between the “rural–urban” and “urban–urban” migrant groups in Table 2, the following statistical information can be compared: The actual housing purchase rate of the “rural–urban” migrant group is 22.07% lower than that of the “urban–urban” migrant group. In the comparison of household head characteristics, the rural–urban migrant group shows the characteristics of younger migrants, lower educational level, lower coverage rate of security resources, and richer work experience. In the comparison of family structure characteristics, the agricultural transfer population families have lower monthly income, more children, and heavier family burdens. In the comparison of mobility characteristics, the agricultural transfer population group is more inclined to migrate “with the whole family” across provinces and cities to the most economically developed eastern regions, which also confirms that the mobility of the agricultural transfer population group has a strong employment orientation rather than a settlement orientation. From the perspective of parents’ migrant work experience, the proportion of the agricultural transfer population with parents who have migrant work experience is higher, and parents’ migrant work experience has a strong “demonstration effect” on the next generation.

5. Empirical Analysis

5.1. Regression Results

After controlling for three major categories of variables—individual characteristics, family characteristics, and mobility characteristics, this study conducts logistic regressions separately for the full sample and the rural–urban and urban–urban mobility groups. The logistic function form of Equation (4) above can be evolved into Models (1)–(3). The results are presented in

Table 3. Logistic model regression on housing purchase behavior of migrant population.

	Model (1) (Total)	Model (2) (Rural–Urban Migrants)	Model (3) (Urban–Urban Migrants)
Core explanatory variable: Hukou	1.798 *** (0.032)	/	/
Individual characteristic variables	Controlled	Controlled	Controlled
Household characteristic variables	Controlled	Controlled	Controlled
Migration characteristic variables	Controlled	Controlled	Controlled
Number of individual samples	150,258	127,385	22,873
Prob > chi2	0.000	0.000	0.000
Log likelihood	−76,405	−62,243	−13,895
Pseudo R2	0.142	0.124	0.121

Notes: Standard errors are in parentheses. *** Significance at 1% level.

.

The joint significance Wald endogeneity test was conducted for the three models in the above table. All explanatory variables passed the test at the 1% significance level, and there was no significant endogeneity. The VIF values of the three models were all less than 2.5, indicating no significant collinearity, and the equation tests were passed.

By comparing Models (2) and (3) in the above table, it can be seen that the actual housing purchase rate of inter-urban migrants in the local area is 1.52 times that of the agricultural transfer population. It is found that among the agricultural transfer population, those who are older, male, low-income, married, have more children, frequently move across provinces/municipalities, work in the eastern region, and without housing provident fund protection are the “most disadvantaged groups” in terms of purchasing housing in the cities where they work.

5.2. Decomposition of Differences in Housing Purchase Odds Ratios

To avoid the problem of path dependence in the ordering of observable variables [40], and to ensure that the distribution characteristics of randomly generated subsamples are as close as possible to those of the full sample, when extracting subsamples of “rural–urban” migrants with the same sample size as “urban–urban” migrants for sequential matching in the program, the subsample extraction process was completed using the random repeated Bootstrap method 1000 times.

Table 4. Fairlie decomposition results of differences in actual housing purchase incidence among migrants with different hukou statuses.

Samples Used for β Estimation Coefficients	(1) Rural–Urban Migrant Group (127,385 Households)		(2) Urban–Urban Migrant Group (22,873 Households)		(3) Mixed Sample Group (150,258 Households)
	Characteristic Differences	Contribution Rate	Characteristic Differences	Contribution Rate	Characteristic Differences	Contribution Rate
Gender	0.000 ***	0.13%	−0.001 ***	0.44%	−0.001 ***	0.23%
Age	−0.009 ***	3.85%	−0.007 ***	3.35%	−0.011 ***	5.00%
Educational level	−0.080 ***	36.38%	−0.066 ***	30.07%	−0.089 ***	40.39%
Work experience	0.002 ***	−0.85%	0.002 ***	−1.07%	0.002 ***	−1.04%
Medical insurance	0.002 ***	−0.77%	−0.003 **	1.25%	0.001 **	−0.48%
Housing provident fund	−0.027 ***	12.34%	−0.018 ***	8.03%	−0.027 ***	12.18%
Household income	−0.023 ***	10.24%	−0.023 ***	10.50%	−0.023 ***	10.40%
Marital status	−0.004 ***	1.90%	−0.003 ***	1.20%	−0.004 ***	1.64%
Number of children	−0.013 ***	5.82%	−0.001	0.56%	−0.012 ***	5.44%
Frequency of migration	−0.007 ***	3.19%	−0.009 ***	4.04%	−0.007 ***	3.29%
Non-solo migration	−0.000	0.15%	0.001 *	−0.64%	0.000	0.16%
Size of migrant household	0.015 ***	−6.98%	0.007 ***	−3.00%	0.014 ***	−6.27%
Parental work experience	0.001 ***	−0.31%	0.001 ***	−0.62%	0.001 ***	−0.38%
Migration scope	0.005 ***	−2.29%	0.0001	−0.04%	0.003 ***	−1.50%
Migration region	−0.002 ***	0.69%	−0.003	1.55%	−0.003 ***	1.24%
Endowments effect	−0.140	63.49%	−0.123	55.64%	−0.155	70.28%
Coefficient effect	−0.081	36.51%	−0.098	44.36%	−0.066	29.72%
Total	−0.221	100%	−0.221	100%	−0.221	100%

Notes: *, **, ***: Significance at 10%, 5%, and 1% levels, respectively.

shows that the estimated coefficients of the “rural–urban” migrant group, “urban–urban” migrant group, and mixed sample group were used as their weight values. The decomposition results based on Equations (13) and (14) correspond to columns (1)–(3) in Table 4.

Considering the issue of index benchmarking, this study focuses on analyzing the decomposition results in column (3) where the full-sample coefficients are used as weights. The actual housing purchase odds ratio of the agricultural transfer population group is 22.07% lower than that of the inter-urban migrant group. The coefficient effect, which cannot be explained by sample characteristics, accounts for 29.72%. The agricultural hukou has a significant expanding effect on the gap in housing purchase incidence among the transfer population, and the agricultural transfer population faces certain hukou barriers to some extent when participating in urban housing transactions. Thus, the first hypothesis is verified.

In the endowments effect of the mixed sample group in Table 4, the top three observable variables with positive degrees of contribution to the differences in housing purchase incidence among migrant populations with different hukou statuses, sorted from high to low, include the inter-group differences in educational level, household income, and housing provident fund system. Together, these three variables explain 62.79% of the total difference, where educational level, with a high value of 40.39%, becomes the characteristic variable with the highest contribution to the inequality in housing purchase willingness. This indicates that many agricultural transfer populations who have not received higher education will inevitably encounter educational thresholds in the strict household registration conditions of large and megacities if they cannot break through class solidification through education. Educational level and the institutional factors closely tied to it are the primary reasons for widening the gap in housing purchase incidence between “rural–urban” and “urban–urban” migrants. In addition, three characteristic variables, namely the size of the migrant household, parental work experience, and migration scope, have negative degrees of contribution to the differences in housing purchase willingness between the groups. This means that larger migrant household size, short migration distance, and parental work experience help narrow the gap in housing purchase odds ratios between the agricultural transfer population and inter-urban migrants.

5.3. Robustness Test

This study adopts the variable replacement method to verify the robustness of the regression model and the Fairlie basic decomposition model. With reference to the benchmark regression method, the core explained variable “whether to purchase housing” is replaced with “housing purchase willingness”. The empirical results are shown in

Table 5. Nonlinear logistic model regression on migrants’ willingness to purchase housing.

Variables	Willingness to Purchase Housing Locally
Model	Model (4) (Total)	Model (5) (“Rural–Urban” Migrant Group)	Model (6) (“Urban–Urban” Migrant Group)
Core explanatory variable: hukou	1.217 *** (0.022)	/	/
Individual characteristic variables	Controlled	Controlled	Controlled
Household characteristic variables	Controlled	Controlled	Controlled
Migration characteristic variables	Controlled	Controlled	Controlled
Number of individual samples	150,258	127,385	22,873
Prob > chi2	0.000	0.000	0.000
Log likelihood	−78,698	−64,885	−13,545
Pseudo R2	0.039	0.041	0.027

*** Significance at 1% level.

below.

The joint significance Wald test for multiple explanatory variables was conducted on Models (4)–(6). All explanatory variables passed the test at the 1% significance level. The VIF values of the three models were less than 2.5, indicating no significant multicollinearity, and the equation tests were passed. By comparing Models (5) and (6) in the above table, it can be seen that the willingness of the “urban–urban” migrant group to purchase housing locally is 1.07 times that of the “rural–urban” migrant group. Agricultural hukou is an explicit factor hindering migrants from purchasing a home in the inflow cities, which is consistent with the conclusion obtained from the benchmark regression model, demonstrating the robustness of the regression results.

Due to the issue of index benchmarking, this study only takes the results of columns (1) and (2) in

Table 6. Fairlie decomposition results of differences in willingness to purchase a home.

Selection of Samples for Coefficient β Estimation	(1) “Rural–Urban” Migrants Group		(2) “Urban–Urban” Migrants Group		(3) Total
	Characteristic Differences	Contribution Rate	Characteristic Differences	Contribution Rate	Characteristic Differences	Contribution Rate
Gender	0.000 *	0.03%	0.000	−0.01%	0.000	0.02%
Age	0.002 ***	−2.04%	0.004 ***	−5.68%	0.002 ***	−2.41%
Educational level	−0.053 ***	70.98%	−0.025 ***	33.44%	−0.050 ***	67.74%
Work experience	0.001 ***	−1.00%	0.000	−0.13%	0.001 ***	−1.02%
Medical insurance	−0.001 **	1.66%	−0.000	0.33%	−0.001 **	1.39%
Housing provident fund	−0.013 ***	18.02%	−0.006 **	7.99%	−0.011 ***	14.90%
Household income	−0.015 ***	20.13%	−0.008 ***	10.79%	−0.014 ***	18.56%
Marital status	−0.000	0.07%	0.000	−0.11%	0.000 **	−0.25%
Number of children	−0.009 ***	12.57%	−0.008 ***	10.41%	−0.009 ***	12.60%
Frequency of migration	−0.003 ***	3.49%	−0.001 **	1.69%	−0.002 ***	3.26%
Non-solo migration	0.001	−0.68%	0.001	−0.96%	0.000	−0.37%
Size of migrant household	0.014 ***	−19.51%	0.011 ***	−15.09%	0.015 ***	−19.71%
Parental work experience	0.001 ***	−0.77%	−0.001	0.61%	0.000 ***	−0.43%
Migration scope	0.006 ***	−7.70%	0.001 **	−0.64%	0.005	−6.66%
Migration region	−0.000	0.10%	−0.002 ***	2.19%	−0.001	1.09%
Endowments effect	−0.071	95.37%	−0.033	44.96%	−0.066	88.73%
Coefficient effect	−0.003	4.63%	−0.041	55.04%	−0.008	11.27%
Total	−0.074	100%	−0.074	100%	−0.074	100%

*, **, ***: Significance at 10%, 5%, and 1% levels, respectively.

as a reference for the robustness test of the housing purchase willingness model, and focuses on examining the difference decomposition results of column (3). The willingness of the agricultural transfer population to purchase a home is 7.39% lower than that of inter-urban migrants. It can be seen from the coefficient effect that the agricultural hukou factor reduces this willingness of the agricultural transfer population by 0.83%, contributing 11.27% to the total difference in willingness between the two groups of migrants. This conclusion is consistent with that obtained from the benchmark model, indicating the robustness of the difference decomposition results.

5.4. Empirical Analysis of Mediating Effect Decomposition

Taking the decision of whether to plan to settle locally for more than 5 years as the mediating variable, this study adopts the KHB mediating effect measurement method to decompose the total effect, direct effect, and mediating effect [41]. As shown in the following table, the coefficients of the three effects all pass the significance test at the 1% level, indicating that the willingness to settle long-term plays a mediating role in the impact of hukou on housing purchase willingness and incidence.

The empirical results of KHB decomposition for mediating effects in

Table 7. KHB decomposition results of mediating effect.

Explained Variables	Willingness to Purchase Local Housing	Having Purchased Local Housing
Total effect coefficient	0.227 *** (0.018)	0.639 *** (0.018)
Direct effect coefficient	0.165 *** (0.018)	0.574 *** (0.018)
Mediating effect coefficient	0.061 *** (0.004)	0.064 *** (0.005)
Proportion of mediating effect	27.01%	10.08%
Pseudo-R2	0.080	0.180

*** Significance at 1% level.

show that when taking the willingness to settle long-term as the mediating variable, the total effect coefficient of hukou on housing purchase willingness is 0.227, with the mediating effect accounting for 27.01%. Specifically, 27.01% of the total effect is attributed to the fact that “urban–urban” migrants have a stronger belief in long-term local settlement, which in turn strengthens their willingness to purchase local housing. The results in the last column of Table 7 indicate that the total effect coefficient of hukou on housing purchase behavior is 0.639, with the mediating effect accounting for 10.08%. This means that 10.08% of the total effect stems from the stronger belief in long-term local settlement among “urban–urban” migrants, which further drives them to purchase property locally. The mediating effect of long-term settlement intention on local home purchase intention is 27.01%, which is significantly higher than its effect on actual purchasing behavior. This is mainly because the occurrence of real home purchasing behavior, in addition to subjective willingness to buy a house, needs to consider more practical factors such as wealth, income level, income sustainability, and whether to return to hometown to take care of the elderly. More realistic factors have weakened the mediating effect of long-term settlement intention on the incidence of actual home purchases. This result is consistent with the second hypothesis.

5.5. Differences in the Contribution of Agricultural Hukou to the Gap in Housing Purchase Incidence Among Transferred Populations

Based on inferences 3 to 5 of the research hypotheses mentioned earlier, this section examines the changes in the contribution of agricultural hukou to the gap in housing purchase incidence among transferred populations from three perspectives: human capital, intergenerational transmission, and regional conversion.

5.5.1. Decomposition of the Gap in Housing Purchase Odds Ratios by Education Gradients

In domestic and foreign studies on the transfer of agricultural labor to non-agricultural industries, agricultural transferred populations with higher education levels possess greater human capital and will take the lead in completing cross-regional migration.

The difference decomposition results in

Table 8. Heterogeneity in the contribution of agricultural hukou to the gap in housing purchase rates among migrants by educational level.

	(1) Primary School and Below (21,206 Households)		(2) Junior High School (71,607 Households)		(3) Senior High School (33,595 Households)		(4) Junior College and Above (23,850 Households)
	Difference Value	Contribution Rate	Difference Value	Contribution Rate	Difference Value	Contribution Rate	Difference Value	Contribution Rate
Endowments effect	−0.084	40.70%	−0.082	45.86%	−0.084	52.14%	−0.088	64.96%
Coefficient effect	−0.123	59.30%	−0.097	54.14%	−0.077	47.86%	−0.047	35.04%
Total	−0.207	100%	−0.179	100%	−0.161	100%	−0.135	100%

show that with the improvement in educational level, the contribution of agricultural hukou to the gap in housing purchase incidence decreases, from 59.30% in column (1) to 35.04% in column (4). In the urban housing transaction market, educational level serves as an explicit positive signal transmitted to sellers, which can reduce information asymmetry between buyers and sellers. Agricultural transfer populations with higher education experience a significantly lower degree of hukou-related barriers in the housing transaction market. Thus, the third hypothesis is verified.

5.5.2. Decomposition of the Gap in Housing Purchase Odds Ratios by Generation

Demographic theory suggests that people of the same generation experience similar macro-institutional changes, thus exhibiting specific group characteristics, and there are systematic differences in consumption behaviors and preference choices among populations of different birth cohorts The cohort effect can influence the housing consumption preferences and purchase decisions of agricultural transfer populations through exogenous environmental channels.

The difference decomposition results in

Table 9. Intergenerational heterogeneity in the contribution of agricultural hukou to the gap in housing purchase rates among migrants.

	15–25 Years Old (Post-1990s) (28,258 Households)		26–35 Years Old (Post-1980s) (58,535 Households)		36–45 Years Old (39,845 Households)		Over 46 Years Old (23,620 Households)
	Difference Value	Contribution Rate	Difference Value	Contribution Rate	Difference Value	Contribution Rate	Difference Value	Contribution Rate
Endowments effect	−0.064	44.48%	−0.166	76.08%	−0.159	74.17%	−0.123	48.67%
Coefficient effect	−0.080	55.52%	−0.052	23.92%	−0.055	25.83%	−0.129	51.33%
Total	−0.144	100%	−0.218	100%	−0.214	100%	−0.252	100%

show that with the successive growth of generations, the contribution of agricultural hukou to the gap in housing purchase incidence (coefficient effect) presents a “U-shaped” characteristic. The post-1990s’ new-generation agricultural transfer population (15–25 years old) and the older-generation agricultural transfer population over 46 years old face a high degree of agricultural hukou barriers in the housing transaction market, with the contribution rates reaching 55.52% and 51.33%, respectively, indicating that hukou has become a substantial dominant factor widening the gap in housing purchase incidence. For the post-1980s’ new-generation agricultural transfer population (26–35 years old) and the older-generation agricultural transfer population under 45 years old (as defined in previous studies, the “post-1990s” and “post-1980s” are classified as the new-generation agricultural transfer population, while those born before 1980 are classified as the older-generation agricultural transfer population), the degree of hukou barriers in the urban housing transaction market decreases by approximately half, to 23.92% and 25.83%, respectively. Thus, the fourth hypothesis holds for the post-1980s’ new-generation agricultural transfer population.

5.5.3. Decomposition of Inter-Group Differences in Housing Purchase Odds Ratios by Region

Housing prices and the difficulty in purchasing housing vary across different regions. In accordance with the classification criteria mentioned earlier, the full sample is divided into four major regions (eastern, central, western, and northeastern China) based on the inflow destinations to decompose the gap in housing purchase rates between migrant populations with different hukou statuses.

The results of the regional difference decomposition in

Table 10. Regional heterogeneity in the contribution of agricultural hukou to the gap in housing purchase rates among migrants.

	Eastern Region (62,589 Households)		Central Region (26,767 Households)		Western Region (49,363 Households)		Northeastern Region (11,539 Households)
	Difference Value	Contribution Rate	Difference Value	Contribution Rate	Difference Value	Contribution Rate	Difference Value	Contribution Rate
Endowments effect	−0.208	77.02%	−0.093	62.28%	−0.126	57.77%	−0.044	50.29%
Coefficient effect	−0.062	22.98%	−0.056	37.72%	−0.092	42.23%	−0.043	49.71%
Total	−0.270	100%	−0.149	100%	−0.219	100%	−0.087	100%

indicate that the expanding effect of agricultural hukou on the internal gap in housing purchase rates among transferred populations shows a clear upward trend from east to west. In the economically active and highly open eastern region, the contribution of hukou differences (coefficient effect) to the gap in housing purchase rates between the two groups is lowest among all regions at 22.98%, while 77.02% of the gap can be explained by Endowments effect, such as inter-group differences in education, household income, and migration scope across hukou statuses. In the less economically developed western and northeastern regions, the contribution of agricultural hukou to the gap in actual housing purchase rates between agricultural transferred populations and inter-urban migrants rises to 42.23% and 49.71%, respectively. In the less economically developed and less open western and northeastern regions, the degree of information asymmetry between buyers and sellers increases, and there are more market interference factors in identifying individual characteristics, leading to a rise in the coefficient effect. Consequently, agricultural hukou further exacerbates the gap in housing purchase rates between agricultural transferred populations and inter-urban migrants in the inflow cities. Thus, the fifth hypothesis is verified.

6. Discussion

The lack of recent nationwide survey data on the housing conditions of migrant populations has limited research in this area. Future nationwide surveys would enable a more elaborate examination of the housing disparities between rural-to-urban and urban-to-urban migrants. This line of inquiry is particularly timely, given that recent investments in rural education and the growth of township economies may have mitigated the traditional disadvantage associated with an agricultural hukou.

Within urban rental markets, it remains an open question whether significant disparities in rental space, quality, and prices exist between these two groups, and to what extent agricultural hukou explains these differences. An empirical examination of rental decision-making among migrants would deepen our understanding of how institutional factors like hukou shape housing outcomes within urban rental markets.

Consequently, the future availability of nationwide rental data would be invaluable for advancing this research agenda.

7. Conclusions

Under the situation where the scale of migrant populations remains persistently high, this paper proposes research hypotheses based on the separating equilibrium model of the housing market with imperfect information. Using national-level microdata from the China Migrants Dynamic Survey (CMDS), Logistic regression analysis reveals that the actual homeownership odds ratio of rural-to-urban migrants is 22.07% lower than that of urban-to-urban migrants. Empirical results from Fairlie decomposition demonstrate that, compared with urban-to-urban migrants, rural hukou exerts an coefficient effect on the actual homeownership rate of rural-to-urban migrants as high as 29.72%. KHB decomposition for mediating effects illustrates that with long-term settlement intention as the mediating variable, the total effect coefficient of hukou on homeownership intention is 0.227, with the mediating effect accounting for 27.01% of the total effect. Furthermore, rural hukou’s amplifying effect on the homeownership gap among migrants in destination cities exhibits heterogeneity across education levels, generations, and regional mobility patterns.

The findings hold significant implications beyond China. With deepening globalization, migrant integration and settlement have become global challenges. Immigrants with different identity backgrounds may face varying degrees of difficulty in purchasing housing in their destination cities. To promote urban–rural integration, this study suggests that while accelerating the reform of the urban–rural hukou system, efforts should be made to reduce the linkage between hukou status and the allocation of social public service resources. Equalizing public service resources will enable migrant populations to enjoy fair housing rights and interests in local areas, thereby promoting the fair, stable, and sustainable development of the real estate market.