The Impact of Housing Prices on Chinese Migrants’ Return Intention: A Moderation Analysis of Public Services

Abstract

Housing prices are a topic of significant social concern, and public services are a crucial factor influencing migrants’ return intentions. Based on the China Labour Force Dynamics Survey and China Real Estate Index database from 2012 to 2018, this study adopts probit model to explore the influence mechanism of housing prices on migrants’ return intentions and the moderating effect of public services. The results indicate that housing prices have a significant positive impact on migrants’ return intentions, and the level of public services negatively moderates the relationship between housing prices and migrants’ return intentions. Moreover, employing an instrumental variable approach to address the endogeneity of housing prices, the modeling results provide robust evidence of the significant and heterogenous impact of housing prices on return intentions among migrants. In particular, the positive impact of housing prices is mainly concentrated among single urban migrants without housing. Additionally, public services negatively moderate the positive impact of housing prices on return intentions among single rural migrants without housing. By elucidating the correlation between housing prices, public services, and return intentions among migrants, this study offers recommendations for policymakers regarding migration issues in urban development.

1. Introduction

Return migration has emerged as a significant socio-economic phenomenon influencing human settlement patterns both internationally and domestically [1]. According to the National Bureau of Statistics, the number of Chinese migrants reached 376 million by 2020, accounting for a quarter of the country’s total population [2]. Since the reform and opening up of the 1980s, the Pearl River and Yangtze River deltas, as focal points of economic liberalization, have attracted numerous migrants seeking better employment opportunities and higher wages [3]. However, since the beginning of the twenty-first century, China has seen a pronounced trend of return migration from urban to rural areas, particularly in central regions such as Hubei Province [4], and the household registration system and governmental policy on migration trends have had a significant influence [5]. China has released guidelines on orderly manufacturing sector transfers to assist regional development, and the rural revitalization strategy implemented in 2018 paid great attention to return migration, especially the return of entrepreneurship to hometowns [6].

Along with the global economic adjustment and the promotion of the industrial transformation strategy in China’s coastal areas, a new round of household registration system reform was launched to promote the integrated development of urban and rural areas [7]. The return of population has become a normal phenomenon in China and is changing national spatial patterns of population flow [8]. In this new context, questions such as whether population return can become an important engine for promoting the development of urbanization in central and western regions, whether the choices of returning migrants can reshape the spatial distribution pattern of the population, and whether it can have an important impact on the implementation of the strategy of coordinated development in the region have aroused widespread concern in Chinese society [9].

Return intention is a significant predictor of return migration and has been well examined [5,6,7,8,9]. Generally, scholars categorize the factors influencing migrants’ return intentions into two directions, “push” and “pull”, which exert crowding out and attraction effects on migrants in destination cities and ultimately determine their willingness to return [10]. Specifically, unaffordable housing prices can discourage people from migrating elsewhere as well as discouraging migrants from returning home [11]. A growing body of literature has documented that the correlation between housing prices and migration is contemporaneous and geographically interdependent, and rapidly increasing housing prices are likely to affect migrant movements [12]. In addition, public services increasingly contribute to the promotion of urbanization in China and can affect urban housing prices through resource aggregation and labor inflow [13]. Currently, most studies regarding the influence of public services and migration align with the “voting with their feet” concept, suggesting that improving the level of public services promotes inflow migration [14].

However, existing research has significant gaps in exploring migrants’ return intentions [4,5,6,7,8,9]. In China, scholars have focused on how housing prices affect settlement or migration intentions, but there is a lack of systematic analysis of the specific decision to “return from cities to hukou registration location (hometowns or rural areas)” [11,12,13,15,16,17]. The moderating role of public services in the relationship between housing prices and return intentions of migrants has been overlooked, as has the heterogeneous impact on groups differentiated by marital status, homeownership, and hukou status [12,13]. Additionally, the scarcity of long-term panel data at the macro level has limited reliable causal inferences [7,8,9,15]. Meanwhile, international studies based on developed-country contexts lack explanations for China’s unique return migration under the hukou system and urban–rural dual structure [18,19,20]. They also fail to deeply explore how public services mitigate the “push effect” of high housing prices [21,22,23]. This study fills these gaps by using nationally representative panel data to validate the positive impact of housing prices on return intentions of migrants and the negative moderating effect of public services for the first time.

This study introduces three key innovations. Theoretically, it constructs a moderated model of “housing prices–public services–return intentions”, revealing that public services weaken the push effect of high housing prices on return intentions of migrants by improving living quality and development opportunities (Figure 1). Methodologically, it addresses endogeneity through an instrumental variable (IV) approach and constructs a comprehensive public services index covering four dimensions (healthcare, education, transportation, etc.) using the entropy method, enhancing analytical rigor. From a demographic perspective, it highlights heterogeneous effects: the positive impact of housing prices is concentrated among single, non-homeowning migrants, while the moderating effect of public services is more pronounced among rural, single, and non-homeowning groups, providing precise evidence for understanding differentiated migration decisions. Additionally, by aligning with China’s “Rural Revitalization” policy, it proposes a coordinated intervention path combining housing policies and public service equalization, enhancing practical relevance.

Moreover, the research enriches migration theory in three ways. First, within the push–pull framework, it supplements the mechanism of “institutional pull” (public services) buffering “economic push” (high housing prices), offering a theoretical explanation for population mobility under China’s urbanization context [10,24]. Second, it expands the traditional application of Tiebout’s “voting with their feet” theory by demonstrating that public services not only act as an independent attraction factor but also moderate market-driven forces, bridging public economics and migration studies [14,25]. Third, through group analyses by hukou, marriage, and homeownership, it reveals interactions between micro-level individual characteristics and macro-level institutional environments, deepening understandings of why return intentions differ among groups facing similar economic pressures [26]. The proposed “institutional–market” dual-dimensional framework provides a generalizable theoretical reference for studying population mobility in developing countries’ urbanization processes.

2. Literature Review

2.1. Return Migration

“Return migration” refers to the relocation of migrants to their original location for permanent resettlement [27], while “return intention” indicates whether a migrant intends to return to their hometown and reflects the overall attitude to migration to predict future population flow laws and population patterns [5,6]. Multiple disciplines have enhanced our understanding of the return migration phenomenon within the framework of international migration theories, fundamentally considering return as an alternative type of migration [23,27]. The theoretical framework of return migration studies integrates multidisciplinary perspectives, unraveling the motivations and mechanisms of population migration across different dimensions [26,27,28].

At the macro-structural level, the push–pull theory highlights the combined role of “push factors” in origin regions—such as economic deprivation and political instability—and “pull factors” in destinations—like employment opportunities and public services—while also accounting for migration costs and institutional barriers [10,24]. At the meso-network level, social network theory suggests that kinship- and geography-based networks create a self-reinforcing migration mechanism by reducing information costs and providing support [29]. At the micro-decision level, neoclassical economic theory treats migration as a rational choice where individuals weigh costs (economic, psychological) against benefits (wage differentials). Todaro’s (1969) “expected income” concept explains migration amid urban unemployment, while human capital theory (Becker, 1964) views migration as a strategy to enhance human capital—high-skilled migrants gravitate toward high-return countries, while low-skilled migrants accumulate capital through transnational employment [30,31]. The new economics of migration theory extends to family decision making, positing that families send members to migrate to diversify economic risks and cope with market imperfections [32]. These theories offer macro-, meso-, and micro-level explanations, but real-world migration typically emerges from the interplay of multiple factors, requiring context-specific synthesis.

Additionally, Tiebout’s “voting with their feet” theory, rooted in public economics’ local government competition framework, focuses on domestic residential choices—distinct from international migration theories in scope, core questions, and assumptions [14]. While both involve population movement, the former examines the efficiency of local public goods provision, whereas the latter focuses on transnational migration drivers and structural impacts. Additionally, Borjas’ selective migration theory highlights human capital disparities: high-skilled migrants favor regions with robust public services, while low-skilled groups, restricted by institutional barriers like hukou, are more prone to return under high housing prices [26]. Borjas (1987) [26] links migration selectivity to human capital, explaining varied public service demands—insights that inform analyses of China’s stratified migrants’ return, where individual traits shape return intentions. This study employs the push–pull theory alongside Tiebout’s and Borjas’ theories to explore how public service access amplifies or mitigates the return-inducing effect of housing prices on China’s migrants [10,14,26].

2.2. Factors Influencing Return Intentions

Research identifies multifaceted drivers of return intentions: individual traits, family needs, social environment, economic conditions, policies, and housing attributes [8,9,10]. International studies focus on transnational return across diverse cultural, economic, and policy contexts, while domestic research centers on rural–urban or intercity flows [1,15,27]. Individual characteristics such as gender, age, marital status, income, and education significantly impact return intentions across generational divides [5,6,7]. In Western immigrant societies, second-generation migrants face complex motivations—shaped by ancestral identity and cultural conflicts—such as Turkish-German youth returning due to discrimination [33,34]. Meanwhile, new-generation migrants (born the after the 1980s) in China, with stronger urban adaptability and non-agricultural experience, exhibit weaker return intentions than their parents [35,36].

Low human capital often drives return due to urban labor market disadvantages [37,38]. International studies, set in market economies, consider exchange rates, taxes, and social security factors ill-suited to explain China’s “semi-urbanization”, where rural migrants retain land rights but lack urban welfare [39,40]. Research by Chinese scholars emphasizes regional economic gaps and living costs: high destination costs push migrants to return, while rural economic growth (e.g., non-farm jobs) and land endowments pull them back [6,8]. Furthermore, family responsibilities—caring for elders and children—unite global return motivations: African migrants in the West return due to caregiving pressures [41], while Chinese migrants are driven by “roots-return” traditions and family reunion needs, especially with left-behind children and empty-nest elders [17,42]. Notably, international research highlights cultural conflicts, racial discrimination, and values as barriers to belonging, while Chinese studies focus on hukou, family ties, and regional culture [31,34]. Germany’s Turkish migrants’ return intentions reflect a blend of economic rationality, transnational ties, identity reconstruction, and generational experience [34]. Gherghina et al. [43] proposed that the impression of discrimination, evaluation of public institutions’ efficacy, and sentiments of belonging are significant determinants of Romanian migrants’ return intentions [31]. In the Chinese context, hukou and associated exclusions—unequal resource access, service denial, and discrimination—boost return intentions, amplified by anti-immigrant policies [36,39]. Conversely, urban social security reduces return likelihood: migrants with urban health insurance are less likely to return than those relying on rural schemes [4,9].

2.3. Housing Prices, Public Services, and Return Intentions

Housing not only provides migrants in cities with essential security for life and development but also serves as a symbol of their status and identity [13]. International and Chinese research indicates that housing prices, as a core component of urban living costs, directly affect migrants’ residential decisions [16,17,18,19,20]. High housing prices increase survival pressures, reduce the relative utility of target cities, and thus inhibit labor flow to high-price areas [36,38,44,45]. Plantinga et al. (2013), analyzing US metropolitan data, found that high housing prices decrease the probability of individual inflow [46]. Rabe and Taylor (2012), using UK data, showed that housing price differences are a key factor in family migration decisions, with high-price areas significantly discouraging homeowner families from moving [47]. Dohmen [48] observed that rising housing prices increase living expenses for renters, reducing their migration intention, especially for low-income groups. In China, high housing prices similarly impact migration decisions through economic pressure [49,50,51]. For example, Chamon and Prasad [49] noted that high housing prices force migrants to save more and consume less, weakening their ability to settle long-term in cities. Furthermore, high housing prices exacerbate wealth gaps and social stratification: nearly half of immigrant respondents spend over 30% of their income on housing, face discrimination in the housing market, and have lower homeownership rates than local residents [20,52,53]. Thus, housing prices in destination cities act as a typical “push factor”, squeezing disposable income, reducing urban livability, and lowering the return intentions of migrants.

Hypothesis 1:

Housing prices have a positive effect on migrants’ return intentions.

Recently, many scholars have increasingly focused on access to public services and the role of public services in mitigating migration flows [54,55]. Tiebout’s (1956) “voting with their feet” theory highlights the impact of public services on residential choices, where local public goods influence living selection and social integration [14]. Hence, access to adequate housing and public services is crucial for migrants’ community integration [20,45]. Bayoh et al. (2006) found that the quality of local public services (e.g., schools, parks) is a key factor attracting homeowners to city centers [22]. The level of public services has a positive effect on individuals’ perceptions of social integration, enhancing their willingness to migrate to inflow cities [55]. High-quality public services enhance residential satisfaction and promote community integration, while high housing prices may serve as a barrier to accessing these services [22,56,57]. Public service levels also attract high-skilled labor, with public services and housing prices jointly influencing urban attractiveness [58]. Public services such as medical, educational, transportation, and environmental resources act as “pull factors”, mitigating the negative effects of high housing prices and collectively determining migrants’ return intentions [37].

Hypothesis 2:

Public services moderate the relationship between housing prices and migrants’ return intentions.

Borjas’ (1987) selective migration theory emphasizes that different human capital groups vary in their sensitivity to institutional environments: high-skilled migrants prefer regions with sound public services, while low-skilled groups are more likely to return under high housing prices due to limited access to institutional benefits (e.g., hukou restrictions) [26]. For example, highly educated individuals tend to move to cities with abundant job opportunities and good public services, even if housing prices are high [57,58]. This suggests a positive correlation between housing prices and public service levels, where high-quality public services can partially offset the crowding-out effect of high housing prices on labor [59]. Foote (2016) noted that rising housing prices generate wealth effects for homeowners but restrict mobility for non-homeowners, further exacerbating social stratification [56]. High housing prices widen wealth gaps between migrants and local residents, potentially causing emotional alienation and prejudice that hinder social integration [55]. Borjas’ (1987) theory of migration selection, emphasizing human capital-driven selectivity, explains diverse demands for public services among groups [26]. By applying this theory, individual characteristics such as gender, marital status, hukou status, and homeownership influence return intentions, providing a theoretical basis for analyzing stratified return migration of migrants in China.

Hypothesis 3:

The impact of housing prices on migrants’ return intentions is heterogenous.

3. Data and Methods

3.1. Data Source

Data were obtained from the China Labour Force Dynamics Survey (CLDS), a biennial survey, with selected waves from 2012, 2014, 2016, and 2018. This timeframe captures four waves of data collection conducted at two-year intervals. The survey is a nationally representative large-scale labor dynamic tracking survey designed and implemented by the Social Science Research Center of Sun Yat-sen University, focusing on the current situation of and changes in China’s labor force, and employing a multistage, multilevel sampling methodology for data gathering. In the data preprocessing stage, this study focuses on floating migrants who have lived in destination sites for more than half a year without local hukou [15]. To improve the reliability and stability of the data, we excluded samples with missing or abnormal data and winsorized continuous variables at the 99th and 1st percentiles. The final dataset in this study contained 6016 valid observations from 29 provinces, 259 cities, and 391 villages in China.

3.2. Variables

3.2.1. Dependent Variable

The dependent variable was migrants’ return intentions, which were elicited through the following survey questions: “Are you likely to settle locally in the future?” and “Do you still prepare to migrate for work?” Responses indicating “intending to settle in their hometown (including village and town near their hometown) in the future” or “having work experience of migration and unwillingness to migrate for job” were rendered as “having intention to return home.” On the other hand, responses indicating “having work experience of migration and preparing to migrate for job” were regarded as “being without return intention.” In this analysis, the return intention is constructed as a binomial variable.

3.2.2. Independent Variables

This study selected housing prices as an independent variable and collected the average housing prices at the city level from the China Real Estate Index database (http://www.crei.cn, accessed on 11 August 2024), which is hosted by the National Information Center and provides professional information and data services on macroeconomics and real estate.

3.2.3. Moderator Variables

To explore how the moderation effect of the public services level with high housing prices influences the return intentions of migrants, this study constructed a public services index evaluation system to measure the level of urban public services in multiple dimensions. Currently, scholars have not reached a consensus on how to measure public services. Thus, referring to previous research results, the level of public services was measured using 16 variables from five aspects (medical resources, educational resources, transportation resources, urban environment, and cultural resources), which cover the most representative types of urban public services [58,59]. The per capita supply index of public services, calculated using the permanent population, was used as a proxy variable for the level of public services in various fields in this study and was derived from the China City Statistical Yearbook. In addition, this study utilized the entropy method to calculate the basic public services level index for 70 cities across China in 2012, 2014, 2016, and 2018.

Table 1. Weights of indicators of public service.

Influencing Elements	Indicators	Weight
Healthcare resources	Hospitals per 10,000 inhabitants (unit)	0.061
	Hospital beds per 10,000 population (bed)	0.055
	Doctors per 10,000 population (persons)	0.065
Educational resources	Elementary school per 10,000 population (unit)	0.064
	Elementary school teachers per 10,000 population (person)	0.044
	Secondary schools per 10,000 population (unit)	0.038
	Secondary school teachers per 10,000 population (person)	0.042
Transportation resources	Road density of prefecture-level cities (km/km2)	0.025
	Buses per 10,000 population (coach)	0.130
	Highway freight transport volume (10,000 tons)	0.088
Urban environment	Greening coverage rate of built-up areas (%)	0.010
	Comprehensive utilization rate of industrial solid waste (%)	0.026
	Centralized industrial wastewater treatment rate (%)	0.017
	Innocuous treatment rate of municipal solid waste (%)	0.015
	Green space area in municipal districts (hectares)	0.153
Cultural resources	Total book collection of public libraries (thousand volumes)	0.160

presents the definitions and weights for each public service indicator. Then, the interaction term between housing prices and public services (housing prices × public services) is used as an explanatory variable to test the moderating effect of the impact of housing prices on migrants’ return intentions.

3.2.4. Instrumental Variables (IVs)

Owing to population mobility, the number of potential housing consumers may affect housing prices at the destination. Therefore, it is necessary to investigate the possible reverse causal relationship between housing prices and return intentions. In addition, unobservable variables may simultaneously affect the relationship between housing prices and return intentions, indicating potential endogeneity issues in housing prices.

Table 2. Descriptive statistics of variables.

Variables	N	Mean	S.D.	Min	Max
Dependent variables
Return intentions	6016	0.536	0.499	0	1
Independent variables
Housing prices	6016	8.594	0.408	7.759	10.12
Instrumental variables
Land supply elasticity	5988	0.0510	0.121	0	0.572
Land supply elasticity × Public services	5988	0.003	0.010	0.000	0.109
Fiscal decentralization	6016	0.015	0.012	0.001	0.088
Moderator variables
Public services	6016	0.0530	0.0380	0.00200	0.219
Individual-level variables
Age	6016	39.70	12.75	16	66
Gender	6016	0.554	0.497	0	1
Marriage	6016	0.826	0.379	0	1
Education	6016	1.479	0.730	1	3
Annual household income	6016	2.128	1.085	1	5
Hukou status	6016	0.820	0.384	0	1
Housing provident fund	6016	0.099	0.299	0	1
Self-rated health	6016	3.766	0.952	1	5
Subjective social status	6016	4.303	1.784	1	10
Homeownership	6016	0.730	0.444	0	1
City-level variables
Provincial capital city	6016	0.130	0.337	0	1
City tiers classification	6016	2.741	1.381	1	6
Ratio of the output value in the second industry	6016	48.87	9.707	24.27	73.19
Ratio of the output value in the third industry	6016	41.24	8.912	16.79	72.13
Per capita GDP	6016	10.89	0.503	9.770	11.89
Registered population	6016	5.818	0.763	3.045	7.099
Wages of active employees	6016	14.63	0.974	12.64	16.95
Year	6016	2015	1.790	2012	2018

Note: min = Minimum, max = Maximum, N = Observations, S.D. = Standard deviation.

presents a statistical description of the variables.

Importantly, this study uses land supply elasticity, the level of fiscal decentralization, and the interaction term between land supply elasticity and public services as IVs for housing prices to address endogeneity issues. Extant studies have employed land supply elasticity as an instrumental variable [60]. A larger share of steep land reduces land supply elasticity and housing prices rise faster in high-demand locations [36]. Following Saiz (2010), we construct land supply elasticity variables using an interaction term between the fraction of land with a slope exceeding 15° and the proportion of developed land [60]. Since Model 2 includes the interaction term between housing prices and public services (housing prices × public services), and housing prices is an endogenous variable, the interaction term is also endogenous. We use the interaction term between land supply elasticity and public services as an instrumental variable. Furthermore, against the background of China’s tax-sharing system, in regions where the level of fiscal decentralization expands, land finance demand surges. At the same time, housing and land prices also tend to advance rapidly [61,62]. Consequently, this study suggests that land supply elasticity, the interaction term between land supply elasticity and public services, and the level of fiscal decentralization are reasonable instrumental variables when using an IVprobit model to perform a two-stage regression. Specifically, the level of decentralization of local finance is measured as fiscal revenue within the budget of prefecture-level cities/fiscal revenue within the central budget, with data sourced from the China Urban Statistical Yearbook by the National Bureau of Statistics. The average price of the land for sale in 2012, 2014, 2016, and 2018 (in billions of yuan) is sourced from the China Land and Resources Statistical Yearbook.

3.2.5. Control Variables

According to existing research, the willingness to migrate to inflow cities was subject to the effects of individual-level variables but was also affected by city-level variables. Individual-level variables included age, gender, marriage, education, annual household income, hukou status, housing provident fund, self-rated health, subjective social status, and homeownership. City-level variables included the provincial capital city, city tiers classification, the proportion of added value of the secondary industry in GDP, the proportion of added value of the tertiary industry in GDP, the logarithm of per capita regional GDP, the logarithm of the registered population, and the logarithm of the total wages bill of active employees. Other control variables included a categorical variable for year.

Table A1. Definitions and descriptive statistics of variables.

Variables	Definition	N	Mean	S.D.	Min	Max
Dependent variables
Return intentions	Dummy variable: Return = 1 (base); Non-return = 0	6016	0.536	0.499	0	1
Independent variables
Housing prices	Continuous variables: The logarithm of the average price of commercial housing in each city (yuan/m2)	6016	8.594	0.408	7.759	10.12
Instrumental variables
Land supply elasticity	Continuous variables: Elastic land supply (%)	5988	0.0510	0.121	0	0.572
Land supply elasticity × Public services	The interaction term between elastic land supply and public services	5988	0.003	0.010	0.000	0.109
Fiscal decentralization	Continuous variables: Fiscal revenue within the budget of prefecture-level cities/fiscal revenue within the central budget (%)	6016	0.015	0.012	0.001	0.088
Moderator variables
Public services	Calculated by the entropy weight method based on the indicator system	6016	0.0530	0.0380	0.00200	0.219
Individual-level variables
Age	Continuous variables: Subtract the year of birth from the statistical year	6016	39.70	12.75	16	66
Gender	Dummy variable: Male = 1 (base); Female = 0	6016	0.554	0.497	0	1
Marriage	Dummy variable: Married = 1 (base); Single = 0	6016	0.826	0.379	0	1
Education	Categorical variables: No schooling/primary school/private tutoring/junior high school = 1; General high school/vocational high school/technical school/secondary technical school = 2; Junior college/undergraduate/master’s degree/doctorate = 3	6016	1.479	0.730	1	3
Annual household income	Categorical variables: 0–25,000 = 1; 25,000–50,000 = 2; 50,000–100,000 = 3; 100,000–200,000 = 4; 200,000 and above = 5 (yuan)	6016	2.128	1.085	1	5
Hukou status	Dummy variable: Agricultural household registration = 1 (base); Non-agricultural household registration = 0	6016	0.820	0.384	0	1
Housing provident fund	Dummy variable: Having housing provident fund = 1 (base); Without housing provident fund = 0	6016	0.099	0.299	0	1
Self-rated health	Continuous variables: Excellent health = 5; Good health = 4; Fair health = 3; Relatively poor health = 2; Very poor health = 1	6016	3.766	0.952	1	5
Subjective social status	Continuous variables: The highest score of “10” represents the top-most level; the lowest score of “1” represents the bottom-most level	6016	4.303	1.784	1	10
Homeownership	Dummy variable: Having homeownership = 1(base); Without homeownership = 0	6016	0.730	0.444	0	1
City-level variables
Provincial capital city	Dummy variable: Being a provincial capital city = 1 (base); Not being a provincial capital city = 0	6016	0.130	0.337	0	1
City tiers classification	Categorical variables: Fifth-tier cities = 1; Fourth-tier cities = 2; Third-tier cities = 3; Second-tier cities = 4; First-tier cities = 5; New first-tier cities = 6	6016	2.741	1.381	1	6
Ratio of the output value in the second industry	Continuous variables: The proportion of the added value of the secondary industry in GDP (%)	6016	48.87	9.707	24.27	73.19
Ratio of the output value in the third industry	Continuous variables: The proportion of the added value of the tertiary industry in GDP (%)	6016	41.24	8.912	16.79	72.13
Per capita GDP	Continuous variables: The logarithm of per capita regional GDP (yuan)	6016	10.89	0.503	9.770	11.89
Registered population	Continuous variables: The logarithm of the registered population (in ten thousand people)	6016	5.818	0.763	3.045	7.099
Wages of active employees	Continuous variables: The logarithm of the total wages bill of active employees (in ten thousand yuan)	6016	14.63	0.974	12.64	16.95
Year	Statistical year	6016	2015	1.790	2012	2018

in Appendix A demonstrates the descriptive statistics and definitions of all variables.

3.3. Model

3.3.1. Entropy Method

It is essential to allocate a weight to each indicator before evaluating public services. The entropy method is utilized to ascertain the weights of each index for standardized data [63]. The formula is as follows:

$${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }={\displaystyle \frac{{\mathrm{X}}_{\mathrm{i}\mathrm{j}}-\mathrm{m}\mathrm{i}\mathrm{n}\left({\mathrm{X}}_{\mathrm{j}}\right)}{\mathrm{m}\mathrm{a}\mathrm{x}\left({\mathrm{X}}_{\mathrm{i}}\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left({\mathrm{X}}_{\mathrm{j}}\right)}} (\mathrm{Positive} \mathrm{Indicator})$$

(1)

$${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }={\displaystyle \frac{\mathrm{m}\mathrm{a}\mathrm{x}\left({\mathrm{X}}_{\mathrm{j}}\right)-{\mathrm{X}}_{\mathrm{i}\mathrm{j}}}{\mathrm{m}\mathrm{a}\mathrm{x}\left({\mathrm{X}}_{\mathrm{i}}\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left({\mathrm{X}}_{\mathrm{j}}\right)}} (\mathrm{Negative} \mathrm{Indicator})$$

(2)

The first step is to standardize the value of indicators. Here, ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }$ represents the standardized value of the i-th assessing object on the j-th indicator, and ${\mathrm{X}}_{\mathrm{i}\mathrm{j}}$ denotes the original value.

$${\mathsf{\rho }}_{\mathrm{i}\mathrm{j}}={\displaystyle \frac{{\mathsf{\alpha }}_{\mathrm{i}\mathrm{j}}}{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathsf{\alpha }}_{\mathrm{i}\mathrm{j}}}}, {\mathrm{e}}_{\mathrm{j}}=-{\displaystyle \frac{1}{\ln \mathrm{n}}}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathsf{\rho }}_{\mathrm{i}\mathrm{j}}\ln {\mathsf{\rho }}_{\mathrm{i}\mathrm{j}}$$

(3)

$${\mathrm{w}}_{\mathrm{j}}={\displaystyle \frac{(1-{\mathrm{e}}_{\mathrm{j}})}{{\sum }_{\mathrm{j}=1}^{\mathrm{m}}(1-{\mathrm{e}}_{\mathrm{j}})}}$$

(4)

The second step is to calculate the weight. In Formulas (3) and (4), ${\mathsf{\rho }}_{\mathrm{i}\mathrm{j}}$ is the proportion of index j in the i-th year, ${\mathrm{e}}_{\mathrm{j}}$ is the entropy value of j indexes, and ${\mathrm{w}}_{\mathrm{j}}$ is the index weight:

$${\mathrm{S}}_{\mathrm{i}}={\sum }_{\mathrm{j}=1}^{\mathrm{m}}{\mathrm{w}}_{\mathrm{j}}{\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\prime }$$

(5)

The third step is to calculate the comprehensive assessment score (${\mathrm{S}}_{\mathrm{i}}$):

3.3.2. Probit Model

Since the explained variable return intentions is a binary variable, the probit model is used for estimation. The expression for this model is:

$$y_i^{\mathrm{*}}={\alpha }_0+{\alpha }_1{x}_i+{\alpha }_2{s}_i+{\epsilon }_i$$

(6)

$$y_i=\left\{\begin{array}{c}1, y_i^{\mathrm{*}}>0\\ 0, y_i^{\mathrm{*}}\le 0\end{array}\right.$$

(7)

As shown in Formula (6), $y_i^{\mathrm{*}}$ represents the latent variable of the return intentions. When $y_i^{\mathrm{*}}$ > 0, $y_i$ = 1 otherwise $y_i$ = 0. $x_i$ represents housing prices, $s_i$ are a series of control variables, mainly including age, gender, marriage, education, annual household income, hukou status, homeownership, city tiers classification, and registered population. ${\alpha }_0$, ${\alpha }_1$, ${\alpha }_2$ are parameters to be estimated, ${\epsilon }_i$ is a random disturbance term.

3.3.3. Moderating Model

In order to examine the moderating effect of public services level on the relationship between housing prices and return intentions [39,64], this study adds the moderating variable (public services level) and the interaction term between housing prices and public services (housing prices × public services) into Formula (6). And the basic expressions are shown in Formula (8):

$$y_i^{\mathrm{*}}={\beta }_0+{\beta }_1{x}_i+{\beta }_2{j}_i+{\beta }_3{k}_i+{\beta }_4{s}_i+{\epsilon }_i$$

(8)

where $j_i$ represents public services level and $k_i$ presents the interaction term between housing prices and public services.

3.3.4. IV Strategy

We constructed land supply elasticity, the interaction term between land supply elasticity and public services, and the level of fiscal decentralization as instrumental variables, using the IV probit model to carry out regression analysis [39,60,61,62]. The specific model settings are as follows:

$$j_i=c_0+c_1{z}_i+c_2{m}_i+c_3{t}_i+c_4{s}_i+v_i$$

(9)

$$k_i=c_0+c_1{z}_i+c_2{m}_i+c_3{t}_i+c_4{s}_i+v_i$$

(10)

$$y_i^{\mathrm{*}}={\beta }_0+{\beta }_1{x}_i+{\beta }_2{j}_i+{\beta }_3{k}_i+{\beta }_4{s}_i+u_i$$

(11)

$$y_i=\left\{\begin{array}{c}1, y_i^{\mathrm{*}}>0\\ 0, y_i^{\mathrm{*}}\le 0\end{array}\right.$$

(12)

where $y_i^{\mathrm{*}}$ represents the latent variable of the return intentions. $j_i$ and $k_i$ are endogenous variables. $z_i$, $m_i$, $t_i$ are instrumental variables. And $v_i$, $u_i$ are random disturbances items.

4. Result

4.1. The Spatial Distribution of Housing Prices

Figure 2 shows the spatial distribution of housing prices at the city level, based on data from the China Real Estate Index System for the years 2012, 2014, 2016, and 2018. All of them show a consistent spatial distribution characterized by high value decreasing from eastern to western regions. High-value areas are predominantly concentrated in the southeastern coastal cities and a few regional economic centers within the central and western regions. The spatial heterogeneity reveals disparities in housing prices in China.

4.2. Descriptive LISA Clusters of Housing Prices

Global and local spatial autocorrelation analyses were performed to detect spatial correlations between housing prices. Moran’s I statistic was used for the global spatial analysis to examine the presence or absence of spatial clusters in terms of housing prices in China. This study employed a global autocorrelation measure to describe and visualize the spatial distribution of housing prices based on Moran’s I statistic. The results demonstrated that the value of Moran’s I is positive and the p-value is below 0.001, which indicated that housing prices had a significant and positive spatial autocorrelation. Additionally, this research further used the local indicators of spatial association (LISA) map to detect the local autocorrelation. From a geographical perspective, the housing prices presented a spatially heterogeneous pattern at the city level (Figure 3). The high–high hot spots of housing prices were in Langfang–Tianjin, the Yangtze River Delta, the central part of the Fujian area, and the Pearl River Delta. Meanwhile, the low–low hot spots were found mainly in Henan, Shanxi, Shaanxi, Hubei, and Hunan provinces. The LISA and Moran’s I data indicated that there were significant spatial variabilities of housing prices in China.

4.3. Descriptive Statistics

Table 2 presents the descriptive statistics for all variables. The total number of samples counted was 6016. About half of participants were those with return intentions (53.6%). The average age of the total sample was 40 years old, the male-to-female gender ratio was 55.4:44.6, married persons accounted for 82.6% of the total sample, agricultural hukou accounted for 82% of the total sample, and the participants who were homeowners accounted for 73% of the total sample, whereas only 9.9% participants had a housing provident fund. In term of self-rated health, the mean value of self-rated health was above the median. However the mean value of subjective social status was below the median.

4.4. The Effect of Housing Prices and Public Services on Migrants’ Return Intentions

Table 3. Testing the mediation effect of public services on migrants’ return intentions.

	Model 1		Model 2
	Estimate	S.E.	Estimate	S.E.
Housing prices	0.196 **	(0.091)	0.400 ***	(0.115)
Public services			39.64 ***	(13.37)
Housing prices × Public services			−4.342 ***	(1.541)
Individual-levelvariables
Age	0.012 ***	(0.001)	0.0122 ***	(0.0017)
Gender (reference group: Female)
Male	0.127 ***	(0.036)	0.126 ***	(0.036)
Education level (reference group: Junior high school and below)
Senior high school and technical secondary school	−0.091 **	(0.046)	−0.099 **	(0.046)
Junior college and above	−0.421 ***	(0.067)	−0.421 ***	(0.067)
Marriage (reference group: Unmarried/divorced/widowed/cohabitation)
Married	0.417 ***	(0.056)	0.411 ***	(0.056)
Hukou status (reference group: Non-agricultural hukou)
Agricultural hukou	1.051 ***	(0.059)	1.052 ***	(0.059)
Annual household income (reference group: Annual household income (RMB 25,000 and below))
Annual household income (RMB 25,000–50,000)	−0.064	(0.049)	−0.067	(0.049)
Annual household income (RMB 50,000–100,000)	−0.042	(0.051)	−0.045	(0.051)
Annual household income (RMB 100,000–200,000)	−0.109	(0.079)	−0.107	(0.079)
Annual household income (RMB 200,000 and above)	−0.455 ***	(0.113)	−0.458 ***	(0.114)
Self-rated health	−0.062 ***	(0.020)	−0.062 ***	(0.020)
Subjective social status	−0.019 *	(0.010)	−0.020 *	(0.010)
Housing provident fund (reference group: None)
Yes	−0.156 **	(0.070)	−0.152 **	(0.071)
Homeownership (reference group: None)
Yes	0.008	(0.044)	0.008	(0.044)
City-level variables
Provincial capital city (reference group: None)
Yes	−0.175 **	(0.068)	−0.149 **	(0.071)
City tier classification (reference group: Tier-5 cities)
Tier-4 cities	0.128 **	(0.063)	0.110 *	(0.065)
Tier-3 cities	0.087	(0.073)	0.0541	(0.075)
Tier-2 cities	0.306 **	(0.123)	0.255 **	(0.124)
Tier-1 cities	0.984 ***	(0.257)	1.384 ***	(0.363)
New tier-1 cities	0.286 *	(0.158)	0.267	(0.169)
Ratio of the output value in the second industry	−0.003	(0.004)	−0.004	(0.004)
Ratio of the output value in the third industry	−0.012 **	(0.005)	−0.014 ***	(0.005)
Per capita GDP	−0.010	(0.078)	−0.061	(0.079)
Registered population	−0.053	(0.055)	−0.127 **	(0.064)
Wages of active employees	0.025	(0.064)	0.031	(0.065)
Year (reference group: 2012)
2014	0.005	(0.066)	0.004	(0.007)
2016	0.010	(0.078)	0.107	(0.078)
2018	0.048	(0.089)	0.0699	(0.090)
_cons	−1.391	(0.911)	−3.159 ***	(1.158)
N	6016		6016

Notes: Significance * p < 0.1, ** p < 0.05, *** p < 0.01. S.E. = standard error. Standard errors in parentheses.

presents the intercorrelations of housing prices and public services with migrants’ return intentions. First of all, consistent with existing studies and confirming Hypothesis 1, there is a significant negative relationship between housing prices and return intentions (β = 0.196, p < 0.05, Model 1) [38,65]. This is possibly due to the migrants’ difficulty in affording high housing prices and becoming homeowners. Housing pressures decrease migrants’ happiness and social integration [48]. Thus, they have to return to their original regions where housing prices are affordable.

In addition, public services have a significant moderating effect on the relationship between housing prices and return intentions. The coefficient for the interaction term between housing prices and public services consistently remains negative (β = −4.342, p < 0.01), indicating that the level of public services negatively affects the association between housing prices and migrants’ return intentions. This confirms the importance of public services in encouraging migrants to settle down and remain in Chinese cities. Better public services help migrants access more and better resources for working and living [66]. These results support Hypothesis 2 and echo existing conclusions regarding the role of public services that migrants tend to tolerate higher costs for a city with a higher level of public services [22].

Factors at the individual level and city level also affect migrants’ return intentions: at the individual level, marriage, age, male gender, and agricultural hukou are positive traits for migrants’ return intentions (model 1). However, education levels of senior high school and technical secondary school or junior college and above, agricultural hukou, self-rated health, subjective social status, annual household income (RMB 200,000 and above), and having a housing provident fund are notably negatively related with the return intentions of migrants. As for destination cities level, tier-1 cities exert a significantly positive effect on the return intentions of migrants. Nonetheless, provincial capital and ratio of the output value in the third industry are notably negatively correlated with the return intention of migrants.

4.5. Instrumental Variable Results

We used the IVprobit model for endogeneity analysis, and the results are shown in

Table 4. Effect of housing prices on migrants’ return intentions (instrumental variable approach (IV)).

	(1) First Stage	(2) First Stage	(3) Second Stage
	DV. = Housing Prices	DV. = Housing Prices × Public Services	DV. = Return Intention
Land supply elasticity	0.979 *** (0.036)	−0.023 *** (0.0027)
Land supply elasticity × Public services	−3.163 *** (0.620)	0.760 *** (0.047)
Fiscal decentralization	1.515 ** (0.793)	0.846 *** (0.060)
Housing prices	0.980 *** (0.036)		0.452 ** (0.229)
Housing prices × Public services		−0.023 *** (0.0027)	−15.24 *** (5.114)
Control	Yes	Yes	Yes
Constant	5.242 *** (0.114)	−0.153 *** (0.009)	−5.524 *** (1.727)
N	5982	5982	5982
R-squared	0.804	0.998
F-statistic	334.62	88.42
Wald x2			6.16 ** (p = 0.046)

Notes: Significance ** p < 0.05, *** p < 0.01. S.E. = standard error. Standard errors in parentheses.

. In the first stage, two ordinary least squares regressions were performed. The coefficients of the instrumental variables were significant at the 5% level, indicating that the instrumental variables satisfied the relevance condition. Meanwhile, the F-statistic computed for the weak identification test was higher than the critical value of 10, suggesting that the instrumental variables selected in this study were not weak instrumental variables. In the second stage of the IVprobit model, the Wald test of exogeneity was significant at the 5% level, indicating that housing prices and the interaction term between housing prices and public services (housing prices × public services) endogenously affect return intentions. It was essential to use instrumental variables to control endogeneity problems. After addressing endogeneity, housing prices have a greater impact on migrants’ return intentions in the IVprobit model (Table 4, Column 3). The second-stage result suggests that housing prices are associated with a 0.452-point increase in migrants’ return intentions (β = 0.452, p < 0.05), which is quite similar to Model 2 (Table 3). Meanwhile, the absolute value of the regression coefficient of the interaction term between housing prices and public services is 15.24 in the IVprobit model (β = −15.24, p < 0.01), which is greater than that in Model 2 (Table 3). This indicates that public services have a more significant moderating effect on the relationship between house prices and return intentions among migrants after addressing endogeneity.

4.6. Heterogeneity Differences

To further determine the impact of housing prices on migrants’ return intentions and the moderating effect of public services, this study examined whether the impact of housing prices differs by hukou status, marriage, homeownership, and gender (

Table 5. Heterogeneous Effects.

Variables	Married	Single	Agricultural Hukou	Non-Agricultural Hukou	Homeownership	Non-Homeownership
Model3
Housing prices	0.132	0.464 **	0.061	0.546 **	0.035	0.579 ***
	(0.099)	(0.227)	(0.096)	(0.273)	(0.110)	(0.176)
Control	Yes	Yes	Yes	Yes	Yes	Yes
FP test	0.333 *		0.485 **		0.544 ***
p-value	(p = 0.090)		(p = 0.028)		(p = 0.006)
Constant	−1.076	−2.227	−1.224	1.652	−1.562	−3.096 *
	(0.999)	(2.366)	(0.990)	(2.837)	(1.069)	(1.846)
Model4
Housing prices	0.314 **	0.872 ***	0.293 **	0.613 *	0.176	0.876 ***
	(0.126)	(0.279)	(0.122)	(0.337)	(0.144)	(0.216)
Housing prices × Public services	−3.665 **	−11.071 ***	−4.963 ***	−1.657	−2.632	−8.080 ***
	(1.707)	(4.304)	(1.642)	(5.156)	(1.890)	(0.300)
Control	Yes	Yes	Yes	Yes	Yes	Yes
FP test	0.558 *		0.320		0.701 ***
p-value	(p = 0.062)		(p = 0.169)		(p = 0.008)
Constant	−2.560 **	−6.578 **	−3.146 ***	1.143	−2.707 *	−5.321 **
	(1.281)	(2.878)	(1.221)	(3.531)	(1.401)	(2.273)

Notes: Standard errors are reported in parentheses below the corresponding regression coefficients. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. The p-values of FP tests are significant at 1% and 5% levels, respectively. The row above the p-values of FP tests presents the coefficient difference (b1–b0) after Fisher’s permutation test (bootstrap 1000 times). * p < 0.1, ** p < 0.05, *** p < 0.01.

). The results of grouped regression further confirmed the statistical significance of group differences; the p-values after Fisher’s permutation test (bootstrap 1000 times) were all statistically significant at the level of p < 0.1 in Model 3. However, in Model 4, the FP test p-value for hukou status was below the 5% significance level.

For marital status analyses, the positive impact of housing prices on return intentions is significant for “single” (β = 0.464, p < 0.05) but not significant for “married.” After taking the level of public services into account, the absolute value of the interaction coefficient between housing prices and public services for single migrants (β = −11.071, p < 0.05) was higher by 7.406 points than that for married migrants (β = −3.665, p < 0.05) in Model 4. This indicates a stronger negative moderating effect of public services on migrants’ return intentions for single than for married individuals. More specifically, a 1% increase in the public services level in the destination city leads to a 0.876 decrease in the score of the impact of housing prices on single migrants’ return intentions when all other variables are controlled for. Due to increasing housing prices in large cities, migrants who are more willing to marry intend to purchase houses and are prone to move back [17].

From the perspective of hukou status, the positive impact of housing prices on migrants’ return intentions is significant for migrants with a non-agricultural hukou (β = 0.546, p < 0.05) but not significant for migrants with an agricultural hukou. Regarding the effect of public services, a 1-point increase in the level of public services causes a decrease of 0.613 in the score of the impact of housing prices on return intentions for migrants with an agricultural hukou. However, such an interaction effect is not significant for migrants with a non-agricultural hukou, indicating that rural migrants exhibit a stronger propensity not to return home when the destination cities have greater comparative advantages in terms of public services compared with their original cities [36]. In the context of the large gap between urban and rural basic public services in China’s urban–rural integration development, rural migrants are sensitive to differences in public services between urban and rural areas, such as education and medical care, as well as differences in infrastructure [67].

In terms of homeownership, the positive effects of housing prices are only significant for migrants without homeownership (β = 0.579, p < 0.01). A house is regarded as a crucial foundation for spiritual support and marital stability [38]. Public services appear to be negatively associated with the impact of housing prices on the return intentions of those migrants without homeownership. A 1-point increase in the level of public services decreases the impact of housing prices on migrants’ return intentions by 0.876. For migrants without housing, renting or purchasing a house in a city with a high level of public services can be a good choice for them [57]. However, changes in housing prices affect their monthly housing expenditure, whether they choose to buy or rent.

Hence, this study found that the positive impact of housing on return intentions is significant for single but not for married migrants. With regard to hukou status, the results demonstrate a positive and statistically significant impact on the return intentions of migrants with non-agricultural hukou and an insignificant impact on the return intentions of migrants having non-agricultural hukou. Moreover, in relation to homeownership, the effects of housing prices are only significantly associated with the return intentions of those migrants without homeownership. Therefore, these results indicate that the positive impact of housing prices is mainly concentrated among single urban migrants without housing. Specifically, the level of public services primarily moderates the impact of housing prices on the return intentions of single rural migrants without housing.

4.7. Robustness Test

To ensure the reliability of the above findings, two robustness tests are performed (

Table 6. Subsample robustness tests.

Variables	Entire Sample	Subsample from 2012 to 2014
Housing prices	0.196 **	0.246 **
	(0.0905)	(0.122)
Controls	Yes	Yes
Constant	−1.391	−1.791
	(0.911)	(1.131)
N	6016	3799

Notes: Significance ** p < 0.05, S.E. = standard error. Standard errors in parentheses.

and

Table 7. Alternative methodology robustness tests.

Variables	Probit Model		Logit Model
	Coef.	Marginal Effect	Coef.	Marginal Effect
Housing prices	0.196 **	0.065 **	0.311 **	0.062 **
	(0.0905)	(0.0299)	(0.152)	(0.0302)

Notes: Significance ** p < 0.05, S.E. = standard error. Standard errors in parentheses.

for the test result). Firstly, we performed additional estimations using subsamples from 2012 to 2014. Table 6 reports the estimation results of the probit model for all samples and subsamples from 2012 to 2014. The results indicate that, after excluding the samples from 2016 to 2018, the regression coefficients for housing prices remain consistent in direction and significance with the conclusions drawn earlier. This finding confirms the results reported in Table 3. Secondly, logit and probit models rely on different distributional assumptions and, to test whether the results are sensitive to these assumptions, we replace the probit model with a logit specification. Table 7 shows that the regression coefficient for house prices remains consistent with the direction and significance of the probit model. The direction, significance, and magnitude of the marginal effects are consistent with those obtained from the probit model. This confirms that our findings are robust to different assumptions about the cumulative distribution function.

5. Discussion and Conclusions

5.1. Discussion

The role of housing prices in shaping migrants’ return intentions is rooted in both economic constraints and socio-cultural norms. As highlighted in the literature, the family social network in destination cities—operating as a form of social capital—strongly influences migrants’ settlement decisions, reflecting traditional Chinese family consciousness that values spatial proximity and mutual support [9,11]. For male migrants, the cultural imperative to purchase a house before marriage intensifies the pressure of unaffordable housing, as homeownership is often a prerequisite for establishing a family in urban areas [15,17]. This aligns with the finding that housing prices exhibit a significant positive association with return intentions (β = 0.196, p < 0.05), particularly among non-homeowners and single migrants, who lack the social and economic buffers of homeownership or familial co-residence. Conversely, female migrants often follow spousal migration patterns, illustrating how gendered family strategies moderate housing-driven return pressures, consistent with prior research on marital migration trajectories [11,12,13].

The crowding-out effect of rising housing prices, documented in studies showing that high costs erode migrants’ capacity to settle [36,49,50], directly links to the policy concern of “urban congestion” in megacities. This effect is exacerbated by long-standing urban–rural resource imbalances, as cities have historically drawn talent and land from rural areas. This process has enabled the development of robust infrastructure and public services that attract migrants, yet simultaneously pushed living costs to unaffordable levels for migrants [35,36,37]. Crucially, public services emerge as a countervailing force: cities with robust medical, educational, and transportation provisions (β = −4.342, p < 0.01) mitigate housing-induced return intentions among migrants, echoing literature that highlights education access for migrant children and healthcare subsidies as pivotal for long-term settlement [36,51]. This “push–pull” dynamic underscores the need for policies like affordable housing and public service equalization, as advocated in China’s 14th Five-Year Plan, to address the dual challenges of housing unaffordability and service inequality [4,44].

The heterogeneous effects among migrants of non-agricultural hukou include being more sensitive to housing prices, while migrants of agricultural hukou prioritize public services—reflect the enduring urban–rural dual structure [10,11]. Rural migrants, despite lacking formal urban entitlements, are drawn to cities by service advantages but constrained by housing costs, a tension consistent with studies on hukou-related welfare gaps [7,9]. This aligns with the policy call to relax hukou restrictions in medium and small cities, allowing rural migrants to access public services without being tied to unaffordable first-tier housing markets, as prior research has emphasized the role of hukou in shaping migration trajectories [39,50].

The paradox of megacities as both economic magnets (provincial capitals negatively associated with return intentions, β = −0.149, p < 0.05) and sources of “push” (first-tier cities positively associated with return intentions of migrants, β = 1.384, p < 0.01) is explained by literature on urban development disparities [40]. First-tier cities’ high housing prices and competitive pressures drive return intentions, while provincial capitals—with strong economies, large registered populations, and high tertiary industry ratios—retain migrants through agglomeration effects [9,10,11]. This mirrors findings that economic strength and urbanization level influence migration decisions, yet extreme urbanization in first-tier cities creates intolerable living costs, necessitating strategies like the “new urbanization” initiative to develop medium-sized cities as alternatives [51,59].

Individual attributes such as education, health, and income further modulate return intentions, consistent with literature on migrant competitiveness [10,33]. Higher education enhances labor market appeal, reducing return likelihood, while self-rated health and subjective social status—proxies for physical and mental adaptability—strengthen settlement intentions [17,56]. Households with annual incomes over RMB 200,000, as well as those with housing provident funds or homeownership, exhibit lower return intentions, confirming that economic resources and housing-related benefits are critical for urban integration [49,50]. These individual-level effects underscore the need for targeted policies, such as expanding provident fund coverage and improving labor market training, to support vulnerable groups.

In the Chinese context, these findings necessitate that policymakers design integrated policies combining housing affordability measures such as subsidized housing and property tax pilots with public service investments to create livable urban environments. Policymakers must also provide targeted support, including tailored interventions like affordable housing for single migrants and rural–urban service integration for agricultural hukou holders, to reduce return intentions among the productive labor force. Additionally, local governments should adopt decentralized development strategies to enhance public services in non-provincial capital cities and leverage fiscal incentives to redirect migrants from overcrowded metropolises to emerging urban centers, thereby fostering balanced regional growth. By bridging micro-level migration behavior with macropolicy, the study underscores that sustainable urbanization in China hinges on balancing social inclusion with economic opportunities: prioritizing public service equalization and housing security can transform cities from temporary workplaces into long-term homes, a goal that aligns with the nation’s vision of inclusive, resilient, and equitable urban development.

5.2. Conclusions

Using nationally representative data from China, this study examined the return intentions of migrants in destination cities, with a particular focus on the effects of housing prices and public services. This study highlights the importance of examining housing prices and levels of public services as concurrent push factors. It also analyses heterogeneity in explaining the return intentions of Chinese migrants. Using an instrumental variable approach to address endogeneity in housing prices, the study draws the following conclusions.

(1)

Housing prices positively affect migrants’ intentions to return to their hometowns. If housing prices in destination cities increase by 1%, the return intentions of migrants increase by 0.196%. Increasing housing prices have a crowding-out effect on the migrants in the destination city and push migrants to return to original cities. The results show that marriage, age, male gender, agricultural hukou, and tier-1 cities are significantly positively correlated with the return intentions of migrants.

(2)

Public services levels negatively moderate the housing prices–return intention relationship. A 1-point increase in the level of public services causes a decrease of 4.342 in the score of the impact of housing prices on return intentions for migrants. The high level of public services has a crowding-in and “pull” effect on the migrants in destination cities.

(3)

Heterogeneity exists in the impact of housing prices on migrants’ return intentions. From the perspective of individual factors, the positive impact of housing prices is mainly concentrated among single urban migrants without housing. It is hard for single rural migrants without housing with low human capital to settle down in a destination and they have to return to their hometowns. The findings reveal that levels of public services negatively moderate the positive impact of housing prices on the return intentions of single rural migrants without housing. Destination cities that offer better public services alleviate their intention to return to their hometowns.

Based on these findings, three policy recommendations are proposed: first, establish a coordinated “housing price–public service” regulation mechanism in large cities by integrating stable housing policies (e.g., expanding affordable housing supply, implementing equal rights for renters and homeowners) with inclusive public service systems (e.g., integrating migrant children’s education into local enrollment schemes). This dual approach prevents the vicious cycle of “service reduction” or “population loss–tax revenue decline” that may arise from solely focusing on housing price control without addressing migrants’ essential needs for education, healthcare, and social security. Second, design targeted interventions for vulnerable groups: provide “housing subsidies + vocational training” packages for single non-homeowners to enhance their housing affordability and labor market competitiveness while prioritizing rural hukou holders with expanded access to basic public services (e.g., universal healthcare and compulsory education) to bridge urban–rural welfare gaps. Third, strengthen the “inclusive competitiveness” of medium-sized, small, and provincial capital cities by leveraging tertiary industry growth—shown to negatively correlate with return intentions—and upgrading public service quality to create “cost-effective” migration destinations. These measures can divert population pressure from megacities and foster a “hierarchical and functionally complementary” urban system, where cities at different tiers attract migrants through a balance of living costs, employment opportunities, and service accessibility.

This study has several limitations that should be acknowledged. First, the database selected consists of cross-sectional data from 2012, 2014, 2016, and 2018. There may be missing variables or unobservable differences between individuals in statistical data collection. Second, the return intentions of migrants may change with economic growth and environmental improvement in home and migration locations. This study uses the database currently available through 2018. Although it has research significance, there is still a certain lag with the current society, which may affect the external validity of the research conclusions. At the same time, constrained by data availability, the measurement indicators used in this study cannot fully cover urban public services. The scope of this study only focuses on prefecture-level cities and does not cover the migration of counties. Consequently, future studies should use nationally representative tracking data to further explore the influencing mechanism of changes in housing prices and return intention. The migrants’ return intention should be analyzed more from the perspective of urban public services. Moreover, migrants from different countries, regions, or cultural backgrounds should be focused on. Additionally, future research should adopt a more diverse theoretical standpoint by considering additional housing-related factors (such as the differences between housing property rights, contracted land and other assets in the hometown and migration locations) and expand the analysis of the difference between the hometown and migration locations on the impact of migrants’ return intention.