1. Introduction
Urban–rural relationships are among the most fundamental and significant aspects of human society, profoundly influencing the modernization process of a country or region [
1,
2]. Observing global population development patterns, the urbanization rates of developed countries tend to stabilize after exceeding 80% [
3]. Over the past four decades, amid ongoing economic and social progress, China has also undergone a rapid urbanization process. China’s rural-urban population mobility can generally be divided into two stages: The first stage involves rural laborers individually moving to cities, creating a “migrant worker flow”, which refers to a massive group of rural migrant workers. The second stage involves a shift towards family-based migration, gradually leading to a “student flow”, which consists of an increasing number of migrant children [
4]. Since Lewis (1954) proposed the dual-sector model, rural-urban labor mobility has attracted extensive attention from scholars, resulting in a well-developed theoretical framework and analytical model [
5,
6]. However, studies focusing on rural-urban student mobility during the compulsory education phase (hereafter referred to as “RUSM”) have been somewhat insufficient.
As the urbanization process accelerates, policies for the schooling of migrant workers’ children have improved, and rural families have pursued better educational opportunities. China’s RUSM rate has been increasing year by year, rising from 16.05% in 2012 to 30.22% in 2020 [
7]. RUSM in China shares many similarities with that in other countries, but it also exhibits some peculiarities due to China’s unique household registration (hukou) system [
8]. For instance, rural-urban population mobility in European countries was significantly tied to industrialization during the mid-20th century, with RUSM closely linked to rural-urban labor mobility [
9]. However, RUSM in China shows a trend that is not synchronized with rural-urban labor mobility. This asynchronous phenomenon is a unique feature of the Chinese context. In such cases, as China’s rural households recognize the value of investing in their children’s education, RUSM has become a crucial factor influencing their resource allocation decisions. For instance, with RUSM, some rural households have changed their labor model from “part-farming, part-working” to “part-working, part-accompanying”, leading many rural parents to relocate to cities for their children’s education [
10]. In addition to labor, RUSM also affects land and capital allocation, as well as vital material resources in rural households. Thus, it is evident that RUSM comprehensively impacts the allocation of the three core household resources of labor, land, and capital in rural households (
Figure 1).
Classic theories of population migration identify income disparities as a primary driver of population movements [
5,
6]. However, beyond seeking higher incomes, migrants can also be motivated by the pursuit of better public services. Tiebout’s “voting with their feet” theory was among the first to incorporate local public services into the utility model of residential choice [
11]. In recent years, research has increasingly focused on the importance of children’s needs in family migration decisions [
12]. For instance, a study by Bushin (2009) on rural-urban migration among families in the United Kingdom highlighted that children’s developmental prospects are a key driving factor in parental and familial migration decisions, including considerations for providing better opportunities for their children [
13]. Building on Becker’s framework of family economics, the fundamental principle of family educational investment is that parents consider their children’s utility while maximizing their own, aiming to optimize their overall family utility [
14]. Studies in these areas often discuss how migration for educational and familial reasons can lead to shifts in labor market participation and occupational changes [
15,
16].
Therefore, what impact have resource allocation adjustments, prompted by RUSM, had on the efficiency of resource allocation in rural households? A review of the literature reveals that the impact of RUSM on the efficiency of resource allocation in rural households is a double-edged sword [
17]. On the one hand, several studies show that population mobility driven by disparities in public services or comfort levels between regions can lead to negative consequences. For instance, it may redirect labor from high-productivity, low-comfort areas to low-productivity, high-comfort ones, with regional productivity deviating from its optimal level and rural welfare diminishing [
18]. RUSM could result in hidden unemployment and higher developmental risks [
19]. On the other hand, for developing countries, given the productivity gap between the urban and rural sectors, encouraging rural-urban population mobility can still enhance overall levels of productivity [
20,
21,
22]. In China, where the land and hukou systems constrain rural-urban labor mobility, optimal cross-sector labor allocation has not yet been achieved [
23,
24]. RUSM can be viewed as a mechanism to alleviate the barriers to rural-urban labor mobility. The labor mobility induced by student mobility, referred to as “mobility for education” [
25,
26], can potentially optimize labor resource allocation in rural households, improving agricultural production efficiency and overall productivity. Notably, empirical evidence is lacking on whether this mobility has optimized or distorted the efficiency of resource allocation in rural households.
Our research background and literature review show the theoretical and practical significance of studying the impact of RUSM on the efficiency of resource allocation in China’s rural households. Such research is also crucial for refining policies related to rural migrant settlement and the balanced allocation of urban and rural educational resources [
27]. If the current policy on RUSM results in a loss of efficiency in family resource allocation, this highlights the potential for policy refinement. This necessity points to a critical reassessment of existing frameworks to promote sustainable resource management, thereby improving overall wellbeing and advancing sustainable development objectives. Therefore, this paper specifically addresses the following three questions: Does RUSM optimize or distort the efficiency of resource allocation in China’s rural households? Does this impact vary among rural household groups with differing resource endowments? How do China’s rural households dynamically adjust their resource allocation after moving? To explore these questions, this study first developed an analytical framework based on the New Economics of Labor Migration and family strategy concepts, aiming to delineate the theoretical mechanisms through which RUSM affects resource allocation efficiency. This study then utilized nationally representative data from China Family Panel Studies (CFPS) and applied ordinary least squares (OLS), propensity score matching difference-in-differences (PSM-DID), and endogenous switching regression (ESR) models to examine the impact of RUSM on the efficiency of both labor resource allocation and agricultural production in China’s rural households. A heterogeneity analysis was subsequently conducted to identify which rural households, based on specific factor endowment characteristics, experience greater losses in resource allocation efficiency due to RUSM. Moreover, in-depth field interviews were conducted to analyze the dynamic adjustment mechanism of resource allocation among these households. Finally, the findings of this study are juxtaposed with those from existing research, its contributions are highlighted, and policy recommendations that are directly informed by the key research findings are formulated.
3. Data, Models, and Variables
3.1. Data Source
This study’s baseline estimation analysis utilized data collected as part of CFPS in 2012. CFPS is a comprehensive longitudinal survey that tracks and collects micro-level data across the individual, family, and community levels to reflect changes in society, the economy, the population, and education in China. The survey encompasses samples from 25 provinces, representing approximately 95% of the national population, ensuring its national representativeness. The empirical analysis did not include post-2014 data in the baseline estimation due to the absence of questions on agricultural land transfer areas in later questionnaires. This omission renders it impossible to accurately identify the actual cultivated land area of rural households and thus to estimate the agricultural production efficiency of rural households [
51]. This paper focused on rural household samples with students in the compulsory education stage (5 to 17 years old). After processing the original data, a total of 2990 valid samples were obtained.
The in-depth interview materials used in this study were sourced from D County, X Province, China, where this paper conducted a field survey for 20 days from 13 July to 1 August 2022. In line with academic ethical norms and specific contractual agreements, the identity of the sample county remains undisclosed. This confidentiality, mandated by an agreement signed with the county-level government before the commencement of the fieldwork, is critical for accessing detailed county data and securing the trust of the participants involved in the study. Ensuring their anonymity helps maintain the integrity of the research process, as all interactions with the study subjects were conducted through established local government networks. County D, nestled in the mid-western region of X Province in China, is the epitome of a typical agricultural county. As of the end of 2021, 28.79% of the county’s registered population was classified under urban household registration, contrasting with the 71.21% classified under rural registration. The demographic composition included 45.54% urban and 54.46% rural permanent residents. Over the decade spanning from 2010 to 2021, the county’s urbanization rate increased from 16.99% to 46.62%. Situated in a mountainous area with challenging natural conditions, County D’s rural populace faces substantial hurdles in achieving sustainable income growth through traditional agriculture. The county’s strategic proximity to City N, the provincial capital and its adjacency to the economically advanced Province D has catalyzed substantial migration. As of the end of June 2022, the total number of rural migrant workers from the county reached 277,100, with 61,700 working within the county, 97,200 working elsewhere within the province, and 118,200 working outside the province. The majority of the young workforce migrated across other provinces to Province D to work in the electronics manufacturing industry, while middle-aged and older workers mostly migrated to City N to work in construction. Some of the workforce was also engaged in entrepreneurial activities. By the end of 2021, there were 326 primary schools in the county, including 4 urban and 322 rural schools, and there were 25 middle schools, including 4 urban and 21 rural schools.
D County was selected as the sample county due to its typical and representative nature for researching RUSM issues in China. Geographically, D County is neither adjacent to the urban districts of its affiliated prefecture-level city nor is it the most remote county within the city. Regarding urban–rural income disparities, in 2021, the per capita disposable income ratio between urban and rural residents in D County was 2.6:1, which is close to the national average of 2.5:1 in China. In terms of educational development, D County met the national assessment criteria for balanced compulsory education development in 2020 and ranked mid-tier in provincial educational development. In terms of RUSM, the rates in D County for the years 2012, 2014, 2018, and 2020 were 14.79%, 19.98%, 19.55%, and 21.55%, respectively. These figures are comparable to the national rates measured through the CFPS database, which recorded mobility rates of 11.57%, 16.99%, 19.77%, and 22.61% for the same years. Overall, the situation of RUSM in D County and its impact on the efficiency of rural household resource allocation are indicative of broader national trends.
Participants for case interviews were randomly selected from rural household samples, with 40 parents of rural students who experienced urban–rural student mobility interviewed. Each interviewee received a one-to-one in-depth interview lasting no less than 30 min (the interview outline is provided in
Appendix A). The occurrence of RUSM among the interviewed families spanned from 2010 to 2022, aligning well with the CFPS data used in the qualitative analysis section. This consistency ensured that the findings and trends observed were reflective of broader data insights and representative of the period studied. The coding method for the interview records was XXYY-P (or N), in which XX represents the interviewee’s place of household registration (located township), YY represents the interviewee’s coding in the township, and P and N represent the identity of the interviewee as a parent of the student or another family member (mostly grandparents), respectively.
3.2. Model Setting
3.2.1. Basic Model: OLS Model
This study first used the ordinary least squares (OLS) estimation method to construct the following basic regression model:
where the subscript
i denotes the sample number among the rural households;
denotes the resource allocation efficiency of rural household
i;
denotes whether rural household
i has experienced RUSM;
is a set of control variables;
,
, and
are coefficients to be estimated; and
is the random disturbance term.
3.2.2. Robust Models: ESR Model and PSM-DID Model
The difficulty in the empirical analysis of this study lies in addressing the issues of self-selection and endogeneity. If RUSM decisions were random, making the groups comparable, the OLS model could yield valid conclusions. However, the choice of urban versus rural schooling, influenced by household resources, leads to self-selection issues. Alternatively, there is an endogenous problem of mutual causality between RUSM decisions and resource allocation efficiency. Therefore, this paper employed an endogenous switching regression (ESR) model and propensity score matching difference-in-differences (PSM-DID) estimation to deal with self-selection and endogeneity issues.
ESR Model
The ESR model simultaneously estimates the following three equations:
Equation (2) is the behavior equation, where indicates whether RUSM occurs for rural household i, denotes a series of control variables, is the coefficient to be estimated, and denotes the random disturbance term. Equations (3) and (4) are the outcome equations for the treatment group and the control group, respectively, where and denote the resource allocation efficiency for rural households that move or do not move, respectively, denotes a series of control variables, and are the coefficients to be estimated, and and are the random disturbance terms. In addition to at least one instrumental variable to ensure model identification, the variables in are generally consistent with those included in .
After estimating the coefficients of the ESR model, this paper estimated the factual and counterfactual levels of resource allocation efficiency for rural households that move or do not move, thereby calculating the average treatment effect of RUSM on resource allocation efficiency.
PSM-DID Model
The PSM-DID model is set as follows:
- (1)
PSM stage: The probit model is specified as follows:
where
denotes the probability of engaging in RUSM, and
is the normal cumulative distribution.
denotes the matching variables that influence the decision-making process of RUSM, and
denotes the matching rule for the characteristic variables, which, in this paper, is five-to-one nearest-neighbor matching.
- (2)
DID Stage: Subsequently, the DID model is established as follows:
where
denotes the resource allocation efficiency of rural household
i in year
t,
indicates whether RUSM is experienced,
and
are vectors of control variables that are constant over time and those that vary over time, respectively,
is the individual fixed effect,
is the time-fixed effect,
and
are the parameters to be estimated for the control variables,
is the random disturbance term, and
is the constant term.
3.3. Variable Selection
3.3.1. Dependent Variable: Resource Allocation Efficiency of Rural Households
Labor Resource Allocation Efficiency
Following the identification method of Bryan and Morten (2019), this paper used the average labor compensation of the household to reflect the labor resource allocation efficiency of rural households. The compensation per unit of labor in rural households was determined by dividing the sum of the household’s annual wage income and operating income by the number of labor force members in the household.
Agricultural Production Efficiency
Referencing the settings of [
52], this paper selected a stochastic frontier production function to measure the loss in agricultural production efficiency and then calculated the agricultural production efficiency. The measurement index can be seen in
Table 1.
3.3.2. Focus Variable: RUSM
RUSM was ascertained through the question “where is the child’s school located?” A response indicating the school’s location as “rural” was coded as zero, and selections of “county town”, “general city”, or “provincial capital city” were coded as one.
3.3.3. Instrumental Variable: School Consolidation Intensity
This paper selected “School Consolidation Intensity” as the instrumental variable for RUSM. The “School Consolidation Policy” is a significant initiative within China’s education system, specifically applied to schools at the compulsory education stage in rural areas. This policy focuses on merging small-scale schools that lack sufficient resources into larger educational hubs. The policy aims to enhance the efficiency of resource allocation and improve the quality of education, ensuring that educational facilities are more sustainable and better equipped to meet the needs of rural communities [
53]. The reason for this is that “school consolidation” is an exogenous policy that can directly change rural households’ choice of school location (their decision making in the context of RUSM), and such RUSM is “passive”, as opposed to being influenced by the household’s resource allocation. For instance, if the school in Family A’s area is closed, the children from Family A must move to school elsewhere, and this decision is not affected by the initial resource allocation of the household. Referencing the settings of Liang and Wang (2020), this paper measured the school consolidation intensity in different cities based on the change in the number of schools per student from 2000 to 2012. The city names in CFPS are restricted data, and the use of this variable has been officially authorized by the CFPS project office. The specific formula is as follows:
3.3.4. Control Variables
Drawing from existing research, the control variables selected for this study included the following: (1) Offspring characteristics, including their gender, age, and number of siblings. (2) Parental characteristics, including their age, educational level, and health status. (3) Family characteristics, including the family’s party membership status, religious beliefs, land area, productive assets, non-productive assets, financial assets, and non-mortgage financial liabilities. (4) External environmental characteristics, including village topographical features and locational attributes (
Table 2).
6. Discussion
This study examines the effects of RUSM decisions on subsequent resource allocation within China’s rural households, adopting a theoretical perspective rooted in “family strategy”. Previous research typically regards RUSM decisions as discrete subjects for analysis, examining factors that influence these decisions [
4], the living conditions of mobile and left-behind children [
8], the challenges faced, and opportunities for institutional improvements [
7] from a static viewpoint. This approach often overlooks the dynamic interactions between RUSM and the broader family decision-making process. This study argues that RUSM should be understood as a dynamic element of family strategy, reflecting rational choices made by China’s rural households based on their available resource conditions. These decisions profoundly influence subsequent family planning and resource allocation. The study finds that RUSM is not merely an isolated decision, but is an integral part of family dynamics, significantly impacting consumption patterns, production, and lifestyle choices within a family. This shift in resource allocation often results in inefficiencies, which affect economically vulnerable families in particular, undermining their capacity for economic growth and weakening their survival and development prospects. By focusing on family strategy, this research offers a nuanced view of the complexities of RUSM and its implications for China’s “urban–rural amphibious” families, thereby enriching the understanding of sustainable development challenges and resilience in rural China.
Second, this study highlights losses in resource allocation efficiency stemming from RUSM, a subject not extensively covered in previous research. While prior studies have primarily celebrated the benefits of RUSM—boosting urbanization rates [
12], decreasing the number of left-behind children [
19], and improving rural human capital [
33]—scant attention has been paid to its adverse impacts. Typically, discussions focus on direct costs, such as increased living and educational expenses [
27]. However, the literature from contexts outside of China indicates that population movement driven by regional disparities in public services may result in resource mismatches and reduce the efficiency of resource allocation in rural households [
18]; however, detailed empirical evidence for China remains sparse. This paper significantly enriches this discourse by shedding light on the hidden costs—namely the losses in resource allocation efficiency—associated with RUSM, supported by observational data from China. Our findings confirm that RUSM markedly diminishes the resource allocation efficiency of China’s rural households. While these inefficiencies appear at the micro level as rational economic decisions aimed at optimizing family utility, at the macro level, they manifest as regional production inefficiencies. Thus, addressing these inefficiencies through targeted policy intervention and optimization is imperative for enhancing sustainable development in rural areas.
Third, this study advances our understanding of losses in resource allocation efficiency among China’s rural households by employing a hypothesis of heterogeneity. Unlike the majority of existing research, which is anchored upon an assumption of homogeneity and neglects the varied resource constraints faced by rural households across different contexts [
10], this paper recognizes that these households might adopt diverse resource allocation strategies in response to RUSM. Such strategies can lead to varied impacts on resource allocation efficiency. By conducting a nuanced analysis of heterogeneity, this study precisely pinpoints the traits of rural households experiencing the most significant losses in resource allocation efficiency. This insight is crucial, as it equips China’s rural households with the knowledge to more accurately assess the phenomenon of RUSM and to make informed, rational decisions about student mobility and resource management. This approach not only enhances the sustainability of rural development but also supports the strategic planning and resilience of rural communities facing socio-economic transitions.
7. Conclusions and Policy Recommendations
Our conclusions can be summarized as follows: First, using large-scale, nationally representative data, this paper found that RUSM, a strategy employed by rural households to access quality urban educational resources, exerts a substantial negative impact on both the LRAE and APE of China’s rural households. This finding suggests that RUSM, chosen by China’s rural households to access superior urban educational resources, incurs a cost that has been previously underemphasized in the literature—the loss of resource allocation efficiency in rural households. Therefore, this paper constitutes a significant addition to the academic research on the relationship between population mobility and resource mismatch. Second, our heterogeneity analysis showed that China’s rural households characterized by lower-to-medium annual incomes, substantial productive fixed assets, limited labor force, lower-to-medium educational levels in their labor force, and smaller land holdings face heightened losses in resource allocation efficiency after moving. Therefore, this study precisely identifies groups experiencing greater losses in resource allocation efficiency, helping rural households to make informed decisions about RUSM and resource allocation. Third, by detailing how RUSM affects resource allocation in China’s rural households, this study shows that RUSM prompts a shift in the rural labor force away from sectors with optimal labor productivity, thereby diminishing LRAE. RUSM also impairs APE through agricultural labor force loss, the crowding out of productive agricultural investments, and disruptions in decisions about land cultivation and transfer.
These findings provide insights for China’s policymakers to improve strategies such as rural migrant settlement and the balanced allocation of urban and rural educational resources: Firstly, governments should focus on the development of rural education to mitigate the expenses incurred by rural households in accessing quality education. Secondly, considering that China’s resource-scarce rural households experience greater losses in resource allocation efficiency after moving on account of students in their family, governments should advise such households to make “careful” decisions regarding their RUSM. At the same time, governments could introduce differentiated educational support policies, such as providing additional educational aid, subsidies, or financial support to low-income migrant workers’ children, thereby helping these households to better cope with the challenges brought by RUSM. Thirdly, to address the adverse effects of RUSM on the occupational choices of the rural household labor force and agricultural production, governments could improve the skill levels and employment competitiveness of rural accompanying parents through employment training programs. By promoting rural industrial upgrading, providing social services, and improving the rural land market, governments can help rural households to better utilize and allocate their household resources after the loss of their agricultural labor force, thereby promoting the sustainable development of rural households.
Given the historical dual structure of urban and rural areas in China and the characteristics of progressive reforms, as well as the ongoing process of urbanization, it is crucial to acknowledge that the impact of RUSM on resource allocation in China’s rural households may vary across different institutional contexts. Therefore, it is important to note that this paper focuses on the phenomenon of RUSM from 2012 to 2022, and, thus, the conclusions and perspectives presented are primarily applicable to China within this timeframe. A limitation of this study is that it primarily explains the dynamic adjustment mechanisms of resource allocation based on observations from a single case study, without rigorous statistical validation. Consequently, the representativeness of these findings at a statistical level remains uncertain. Future research could build on the theoretical framework constructed in this paper, gather data on a larger scale, and conduct more rigorous testing of the dynamic adjustment mechanisms.