The Ripple Effect: Unpacking the Impact of Early Educational Disruptions on Rural Migrant Children’s Learning in China

Abstract

This study investigates the effects of early educational disruptions on the academic achievements of rural migrant children in China. Using a life course theory framework, it examines whether early educational disruptions negatively affect children’s future cognitive abilities, academic performance and ability to cope with academic challenges, drawing on data from the 2013 to 2014 China Education Panel Survey. The analysis shows that early educational disruptions lead to a decline in the academic achievements of rural migrant children during their middle school years, affirming the assertions of life course theory. Notably, gender does not significantly modulate the influence of educational disruptions. Parental educational expectations and the management practices of schools emerge as critical moderators, with high parental expectations potentially exacerbating the adverse effects of frequent disruptions. Conversely, the experience of teachers and strict school discipline play protective roles, highlighting the importance of supportive learning environments and effective educational strategies. These findings are pivotal for social workers intervening with migrant children. They advocate for the development of comprehensive social work strategies—encompassing both curative and preventive measures—that proactively address the unique challenges faced by this population, fostering resilience and enhancing life trajectories in a context characterised by significant societal and educational transitions.

Introduction

Since the economic reforms of the 1970s, China has undergone significant industrialisation and urbanisation, leading to one of the largest rural-to-urban migrations in its history. This movement has seen a growing number of rural workers relocate to cities in search of better employment and living conditions, consequently starting their families in these new urban settings (Goodburn, 2009). The children of rural migrant labourers, living with their parents outside their officially registered rural permanent residence areas (hukou), are identified as ‘rural migrant children’ (Goodburn, 2009; Li et al., 2010). Due to the household registration system and policy barriers, the disparity between living location and registered domicile limits educational opportunities for children of urban migrant workers compared to their urban peers. As a result, these children often face difficulties accessing public schools and are more likely to attend either licensed private schools or ‘black’ unlicensed migrant-run schools. Both types of institutions not only typically offer poorer educational conditions but are also more susceptible to closures (Goodburn, 2009). Additionally, economic constraints, residency status, frequent relocations and parental job shifts contribute to educational disruptions like repeating grades, transferring schools or dropping out (Goodburn, 2009; Lu et al., 2015). Consequently, these children experience ongoing upheaval and instability throughout their formative years (Li et al., 2010).

Life course theory examines the profound influence of macro-level historical events and governmental actions on the individual trajectories of life, underscoring that the historical experiences of an individual’s early years largely dictate the outcomes of their later life (Settersten, 2018). Rural migrant children, as unintended consequences of governmental policies, experience life trajectories that are the outcome of their personal characteristics, actions, as well as the surrounding cultural frameworks and the prevailing institutional and structural conditions (Mayer, 2009). The tumultuous personal and educational experiences of these children are crucial determinants of their future educational and career achievements (Lu et al., 2015). Consequently, a deep understanding of these early life experiences is essential for predicting and enhancing future life outcomes (Mayer, 2009).

The focus of our article on rural migrant children in China not only sheds light on their unique challenges but also permits the examination of the universal validity of life course theory. By testing its applicability across diverse socio-economic and historical contexts, this study aims to enhance theoretical accuracy and promote a deeper cross-cultural understanding. Such insights are crucial for informing effective social work policies and practices globally. Based on extensive survey data from across China, this research investigates the effects of early educational disruptions on academic achievements during junior high. It also explores the moderating effects of variables at the individual, familial and school levels, and how these factors interact with educational disruptions to influence the educational outcomes of migrant children across micro, meso and macro levels. This nuanced analysis supports robust, evidence-based social work interventions that enhance practice effectiveness and promote the well-being of this vulnerable population, thereby benefiting research, practice and policy on a global scale.

Literature review

Experience of educational disruption

Educational disruption are defined as the phenomenon when a plan is created by an individual or school and interrupted by the unplanned with overlapping and intersecting effects along a continuum of physical, social and emotional well-being that potentially result in individualised trauma (Panther et al., 2021). Researchers have evaluated the impacts of various types of educational disruptions on children. For example, the study by Smith (1975) examined the economic impact of educational disruption, focusing on the costs and benefits related to excluding disruptive students—defined as those who have been suspended, expelled, retained or acknowledged by educational authorities as having voluntarily withdrawn (dropped out) from school. Grigg (2012) investigates the consequences of educational disruption and continuity by examining the impact of school transitions on student achievement. However, existing research on the impacts of educational disruptions presents divergent views.

On the one hand, Research suggests that educational disruptions negatively affect children’s developmental progress. First, students who frequently transfer schools often have lower academic achievements and are more likely to experience absenteeism and disengagement (Parke and Kanyongo, 2012). Secondly, such disruptions can reduce social engagement and damage social relationships (Gruman et al., 2008). Lastly, disruptions such as suspensions and retentions break the continuity of education, adversely affecting self-esteem and attitudes towards school, thereby increasing the risk of dropout (Alexander et al., 2003; Glick and Sahn, 2010).

Conversely, some studies suggest that educational disruptions can have positive effects on children. First, children migrating from less advantaged areas to better environments experience improved educational quality (Lee and Park, 2010). Secondly, experiences like repeating a grade or transferring schools are opportunities to strengthen children’s adaptability to new environments, positively impacting their cognitive and academic development (Scanlon and Devine, 2001). Thirdly, in short term, those who take academic breaks or repeat grades often receive higher acceptance from peers and better assessments of their academic capabilities from both teachers and peers (Gleason et al., 2007).

Relevant research on academic achievement

Academic achievement is defined as the degree to which a student meets established educational objectives in settings such as schools and universities (Steinmayr et al., 2014). The most frequently investigated and used variables to measure academic achievement in the fields of educational psychology and education are grade point average (GPA) and standardised achievement test scores (Steinmayr et al., 2014). However, the academic community recognises that the definition of academic achievement encompasses a broader spectrum. This includes cognitive abilities such as language, reading, writing, arithmetic and logical skills (Farkas, 2003), as well as non-cognitive abilities like self-efficacy and academic anxiety (Stankov and Lee, 2014). Exploring academic achievement from a multifaceted perspective yields a richer understanding. Therefore, this study adopts a broader conception of academic achievement, considering it a multidimensional concept that includes standardised test scores, cognitive abilities and non-cognitive abilities.

Research on factors influencing academic achievement predominantly focuses on individual characteristics, family background and school-related factors. Amongst individual characteristics, gender is a key determinant, with empirical evidence showing that female students often outperform their male counterparts academically (Farkas et al., 1990; Chambers and Schreiber, 2004). IQ and personal effort are also important to academic success and are closely related to intrinsic motivation (Gottfried, 1990; Mayes et al., 2009). Studies indicate that female students not only have higher levels of intrinsic motivation but also put more effort into their studies than male students (Veenstra and Kuyper, 2004).

In examining the impact of family factors on academic achievement, explanations are generally categorised into two dimensions: family capital and family processes. The family capital dimension attributes variations in academic achievement to economic, cultural and social capital. There is a consensus that parental education and family economic and social status positively influence children’s academic performance (Caldas and Bankston, 1997; Kirkup, 2008). Research on the family process dimension focuses on the effects of parents’ educational expectations and involvement. These expectations are significantly correlated with academic achievement (Pinquart and Ebeling, 2020), motivating parents to participate in their children’s education. Such involvement typically enhances academic success (Lee and Park, 2010).

School factors such as the physical environment, library size and administrative practices significantly impact academic performance (Heyneman and Loxley, 1983; Baumann and Krskova, 2016). At the classroom and teacher level, student–teacher ratios and class sizes are crucial. Higher ratios typically correlate with better student outcomes (Günçer and Köse, 1993), though some studies show conflicting results (Fuchs and Wößmann, 2008). Smaller class sizes can lead to more personalised attention and improved performance (Shin and Chung, 2009), but their impact on achievement may not be significant (Hoxby, 2000). Despite ongoing debates about the effects of student–teacher ratios and class sizes, broader research indicates that teacher quality and educational qualifications significantly influence student performance more than class size variations (Engin-Demir, 2009).

The present study

The life course perspective posits a linkage between early developmental experiences and long-term life outcomes, proposing that educational experiences during childhood and adolescence are critical in shaping both academic performance and lifelong developmental trajectories (Audas and Willms, 2001). Nevertheless, existing studies present contradictory conclusions about the impact of early experiences such as suspension, repetition, and transfer on subsequent academic performance and long-term development. Whilst the majority of research indicates that these experiences are detrimental factors affecting future development (Alexander et al., 2003; Parke and Kanyongo, 2012), some studies suggest that they represent opportunities for environmental improvement and can be beneficial for children’s future development (Rathmann et al., 2020). Moreover, research on educational disruptions, particularly concerning rural migrant children—a group highly susceptible to such disruptions (Goodburn, 2009)—remains remarkably scarce. Consequently, this research focuses on rural migrant children in China, aiming to investigate the influence of early educational disruptions on their later academic achievements. This study seeks to address the research void concerning the impacts of early educational disruptions on rural migrant children and to offer new social work strategies and insights for supporting this vulnerable group.

In addition, the consensus emerging from life course research highlights the pivotal role of life course analysis in facilitating the paradigm shift from curative to preventive interventions (Mayer, 2009). Consequently, this study delves into how various levels of academic achievement factors such as gender (Veenstra and Kuyper, 2004), parental educational expectations (Pinquart and Ebeling, 2020), teachers’ teaching experience (Engin-Demir, 2009), and school discipline management (Baumann and Krskova, 2016) can moderate the impacts of early educational disruptions. This inquiry assesses their potential efficacy in preventive interventions. The findings thus inform the formulation of protective policies tailored for rural migrant children and advocate for a transition amongst social workers who intervene with these children from curative to more effective preventive strategies.

Data and methods

Data and variable selection

This study analyses data from the 2013–2014 China Education Panel Survey (CEPS) administered by Renmin University of China. Ethical approval was not required for this study with human participants due to local legislation and institutional guidelines. Additionally, written informed consent was not necessary for participation in accordance with these regulations.

The year 2013–2014 serves as a critical juncture to examine the effects of recent policy reforms aimed at improving educational access for migrant children in public schools. These reforms, particularly the 2006 New Compulsory Education Law and the 2012 State Council’s Opinions on Pushing Forward the Balanced Development of Compulsory Education, have demonstrably reduced the risk of educational disruption by better integrating migrant education into the public system (National People’s Congress, 2006; Liu, 2012; The State Council, 2012). By analysing data from the 2013–2014 academic year, we capture the experiences of rural migrant children prior to the full implementation of these reforms. This timeframe minimises the potential confounding effects of the policy changes, offering a strong representation for the study’s thematic focus.

The survey targeted seventh and ninth-grade junior high school students, employing a multi-stage sampling methodology, with stratification based on variables such as the proportion of the mobile population and per capita educational attainment. In the first stage, twenty-eight county-level units were randomly selected across the nation. Subsequently, within these counties, 112 schools were identified using proportional stratified sampling techniques. From these schools, 438 classes were randomly selected for participation. Ultimately, approximately 20,000 respondents were surveyed from these classes. The survey gathered comprehensive data on students’ academic achievements, capabilities and school performance, as well as information about their family backgrounds, lifestyle habits, parent–child interactions and educational expenditures. This approach facilitated a thorough understanding of both the students’ and their families’ circumstances.

To select samples of rural migrant children, this study used two identifiers from the student database: ‘What is your hukou registration location? (in-county (or district) = 1, out-of-county (or district) = 0)’ and ‘What is your current hukou type? (rural hukou = 1, urban hukou = 2, resident hukou = 3, no hukou = 4)’ to identify the migrant child population (Ma, 2020). The term ‘in-county’ (or ‘district’) refers to respondents who are currently located at their hukou (household registration) address, indicating a non-migratory status. Conversely, ‘out-of-county’ (or ‘district’) describes respondents who are not at their registered hukou location, thus classified as being in a migratory state. The study excluded children registered in-county (or district) as non-migrants, as well as those with urban, resident, or no hukou status. Only children registered as out-of-county (or district) with a rural hukou were retained and categorised as rural migrant children. Ultimately, a data-set comprising 1,711 rural migrant children samples was obtained. After removing samples with missing values for the variables used in the study, 1,571 valid samples remained.

Measurements

In this study, academic achievement was operationalised through three distinct dependent variables: academic performance level (academic_level), cognitive ability (cog3pl) and the ability to cope with academic challenges (academic_ease). The academic performance level is measured through an ordinal variable, based on responses from the parents most acquainted with the students (Smith and Adams, 2006), rating the students’ performance in class (poor = 1, below average = 2, average = 3, above average = 4, excellent = 5). Cognitive ability is assessed using an interval variable derived from a specialised cognitive ability test developed by the CEPS. This comprehensive test measures various cognitive domains, including language, spatial reasoning, computation, and logical thinking. Standardised test scores for each student’s cognitive ability are directly available in the database. Measuring the ability to cope with academic challenges utilises an ordinal variable, aggregating students’ self-assessed ease levels across language, mathematics and foreign language courses (extremely easy = 1, fairly easy = 2, somewhat difficult = 3, very difficult = 4), subsequently adjusted to yield learning ease scores between 1 and 9.

This study examines the early educational disruptions experienced by migrant children, focusing on incidents that interfere with the continuity of their schooling as the primary independent variable. Specifically, it measures disruptive events during their elementary education from first to sixth grade. These include the number of times the children changed schools, repeated a grade or took academic breaks during this period. The responses to these three queries were aggregated to quantify the extent of educational disruption.

The control and moderating variables in this study include: student characteristic variables, family-level variables and school-level variables. Student characteristic variables encompass: whether English classes were attended during primary school (attended = 0, not attended = 1), whether boarding at school from Monday to Thursday (yes = 0, no = 1) and gender (female = 0, male = 1). Family-level variables include: family economic status (difficult = 1, average = 2, affluent = 3), total years of education attained by both parents (no education = 0, primary school = 6, junior high school = 9, technical/vocational school = 10, vocational senior high/regular high school = 12, junior college = 15, undergraduate degree = 16, graduate degree or above = 19) and parents’ educational expectations for their children (indifferent = 0, currently not attending = 6, junior high school graduate = 9, technical/vocational school = 11, vocational senior high = 11, regular high school = 12, junior college = 15, undergraduate = 16, graduate = 19, PhD = 22). School-level variables include the highest qualification of the class teacher (junior college and adult higher education = 0, formal undergraduate and graduate = 1), teaching experience of the class teacher (years), the status of youth misconduct within the school’s community (absent or minor = 0, significant or extensive = 1), the school’s current ranking within the county (district) as reported by school leadership (middle and below = 1, middle and upper = 2, best = 3) and the school’s disciplinary management as reported by the students’ class teacher (very lenient = 1, somewhat lenient = 2, moderate = 3, somewhat strict = 4, very strict = 5).

Table 1 summarises the descriptive statistics for the key study variables, including sample size, mean, skewness and kurtosis. The observed statistics for cognitive ability (mean = −0.04, skewness = 0.02), academic performance (mean = 3.10, skewness = −0.28) and coping abilities (mean = 7.27, skewness = 0.07) suggest distributions that are approximately normal. In contrast, educational disruption (mean = 1.24, skewness = 1.97, kurtosis = 9.21) and primary English attendance (mean = 1.04, skewness = 4.49, kurtosis = 21.15) show notable skewness, reflecting diverse student experiences.

Empirical strategy

Initially, models were constructed separately for each of the three dependent variables, utilising Ordinary Least Squares (OLS) to estimate cognitive abilities—a continuous variable—and Ordered Probit (Oprobit) models for the ordinal variables of academic performance and coping with academic challenges. This methodology provides a preliminary exploration of the correlations between educational disruptions experienced by migrant children and their academic achievements. Building on this, the study will evaluate the models’ basic assumptions using outcomes from parallel regression hypothesis tests with the Oprobit models. If these tests are satisfied, the dependent variables—academic performance levels and ability to cope with academic challenges—will be treated as continuous in subsequent analyses.

Secondly, Seemingly Unrelated Regression (SUR) is applicable in systems where dependent variables differ and seem independent, yet the error terms of the equations are correlated, thus warranting classification as an SUR system.

Given that unobservable factors like upbringing, intelligence and family educational approaches simultaneously influence students’ academic performance, cognitive abilities and coping abilities, the error terms associated with these variables are inherently correlated. Therefore, modelling these relationships separately could compromise econometric efficiency and lead to overestimated standard errors for the model coefficients. Consequently, this study assumes that the impact of early educational disruptions on various academic achievement variables amongst migrant children conforms to the SUR model’s assumptions. By jointly estimating these three equations, the explanatory power of the model is significantly enhanced.

Thirdly, after initial observations on how early educational disruptions are related to various academic achievements amongst migrant children, factor analysis was used to combine three academic achievement indicators into a single ‘academic achievement’ variable. Interaction terms between early educational disruptions and different student and school-level characteristics were developed to investigate potential moderating effects. These characteristics included the gender of migrant children, parental expectations for education, the teaching experience of the class teacher and school administration. The interaction terms were designed to reveal how these factors might relate to the connection between early educational disruptions and academic achievement in migrant children. With the introduction of a single dependent variable, the SUR model was unsuitable, leading us to use OLS regression. The F-test was applied to determine the statistical significance of the interaction terms, confirming the validity and necessity of their inclusion in the model.

Fourthly, our data exhibit a hierarchical structure, with students nested within classes, and classes nested within schools. This structure necessitates addressing potential violations of the independence assumption inherent in OLS regression. When observations are clustered within groups, error terms can be correlated, leading to underestimation of standard errors and potentially misleading inferences. To address this concern and enhance the robustness of our analysis, we use Hierarchical Linear Mode;ling (HLM). HLM is specifically designed for analysing data with nested structures. We utilise an unconditional means model or a one-way random effects analysis of variance model (Singer, 1998) within the HLM framework. These models account for the hierarchical nature of the data by estimating the variance components at each level of the hierarchy (individual, class and school). This approach yields more accurate standard errors and strengthens the reliability of our findings. $Yijk$ denotes the dependent variable of academic achievement or the three indices used to measure academic achievement, where i refers to individual migrant children, j and k refer to their class and school respectively, and $Childrenijk$⁠, $Class0jk$⁠, $School00k$ represent variables at the individual, class, and school levels, respectively.

Finally, to mitigate potential endogeneity concerns arising from the influence of early educational disruptions on the subsequent academic achievement of migrant children, this study uses the instrumental variable (IV) method. This approach is particularly suitable because experiences of early educational disruptions precede the attainment of academic achievement, thereby alleviating concerns about reverse causality. However, the model remains susceptible to the issue of omitted variables. Factors like early mental health, individual abilities and social support might correlate with both academic achievements and early educational disruptions, resulting in biased and inconsistent estimations within the model assessing the effects of educational disruptions on academic outcomes. Therefore, this study resorts to the instrumental variable method to address the inherent endogeneity issues within the model.

For the early educational disruption experiences of migrant children, we selected two instrumental variables, the first being the age at which the migrant child arrived at their current location. Typically, the delayed arrival of migrant children at their current locale, signifying multiple relocations, increases the likelihood of experiencing various educational disruptions. In contrast, native-born children are less likely to encounter such disturbances, thus meeting the relevance criterion. Furthermore, the arrival age at the current residence does not directly influence migrant children’s academic achievement, adhering to the exogeneity requirement.

The second instrumental variable selected is whether the migrant children were born before September. In China, children who turn six by August 31 are eligible to start school that year, whilst those born on September 1 or later must wait another year to enrol. Research shows that children born after the cut-off date are generally more mature, which leads to better school readiness and supports their learning in later grades (Zhang and Xie, 2018). This maturity is vital as children might not be fully ready—cognitively, physically, and psychologically—for formal education until a certain age (Zhang and Xie, 2018). Conversely, children born before the cut-off date face relative disadvantages and challenges in adaptation, including issues with self-identity and academic skills (Liu and Li, 2016), and a higher risk of early educational disruptions (Mühlenweg and Puhani, 2010). Consequently, using whether a child was born before September as an instrumental variable for educational disruption meets the relevance criterion. Since a birth date before September does not directly influence the academic achievements of migrant children in junior high, it also satisfies the condition of exogeneity.

Results

SUR model analysis results

Model construction using separate equations demonstrates (see Supplementary Table S1) that the p-values for the parallel regression tests of both Oprobit models exceed 0.05, leading to the acceptance of the null hypothesis and validating the tests. Subsequent analyses will treat the two dependent variables, academic performance level and ability to cope with academic challenges, as continuous for analysis. We continue to use the SUR model for the joint estimation of the correlations between educational disruptions and various academic achievement indicators in migrant children. Additionally, these three academic achievement variables are dimensionlessly processed and normalised to a scale of 0–10 to facilitate comparison between statistical outcomes. The Breusch–Pagan test for independence was applied to investigate correlations amongst residuals, with findings indicating that the p-values for the Breusch–Pagan tests amongst the disturbance terms of the three equations are less than 0.001. This result permits the rejection of the null hypothesis that assumes no correlation amongst the disturbance terms, confirming their contemporaneous correlation. Employing the SUR model thus enhances estimation efficiency (see Supplementary Table S2).

The SUR model analysis yields statistically significant findings (see Table 2). Educational disruption is associated with a decrease in all three student outcomes: cognitive ability (coefficient: −0.12, 95% CI: [−0.19, −0.06], p-value: <0.01), academic performance level (coefficient: −0.15, 95% CI: [−0.24, −0.07], p-value: <0.01), and coping ability (coefficient: −0.14, 95% CI: [−0.21, −0.08], p-value: <0.01). These negative coefficients and statistically significant p-values indicate that educational disruptions are substantially correlated with declines in student outcomes. The results obtained from HLM are similarly robust (see Supplementary Table S3).

Moderating factors analysis

Based on the analysis above, we found that educational disruptions are significantly and negatively correlated with all three academic achievement indicators for migrant children. To synthesise these findings further, we used Principal Component Analysis (PCA) to reduce the dimensions of academic performance, cognitive abilities, and ability to cope with academic challenges into a single composite academic achievement variable. The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy for these variables was 0.617, and the Bartlett’s Test of Sphericity returned a p-value of less than 0.001, indicating strong correlations amongst them. PCA extracted a single factor with an eigenvalue greater than 1, and a factor score was calculated using regression analysis, resulting in the formula: 0.394 × cog3pl + 0.472 × academic_level + 0.438 × academic_ease, representing a comprehensive measure of students’ academic achievement. For ease of model interpretation, this academic achievement score was normalised to a range of 0–10. After establishing this measure, we introduced variables across three different dimensions—individual, family, and school levels. These included gender, parental educational expectations, the class teacher’s teaching experience, and the school’s disciplinary management, to explore their moderating effects on the correlations between educational disruptions and academic achievement amongst migrant children. Statistical results for each model are presented in Supplementary Table S4.

Across all four models examining moderating factors, educational disruption exhibits a consistently negative correlation with academic achievement. The coefficients range from −0.13 to −0.15, and all have 95% confidence intervals that exclude zero, substantiating statistically significant negative correlations (p-value < 0.01).

We visualised the moderating effects within the regression model using graphical methods (Figures 1–4). Each display of moderating effects is divided into two parts. The upper part of the figure shows the conditional marginal effects of educational disruptions on academic achievement based on the values of different moderating variables, with the x-axis representing the centred values of the moderating variables, the y-axis displaying the marginal effects of educational disruptions, and the 95% confidence intervals are demarcated with dashed lines. The lower part of the figure is a frequency distribution graph or kernel density plot of the moderating variables, intended to display the distribution characteristics of the moderating variables.

Whilst Model 1 reveals a statistically significant negative correlation between gender and academic achievement (coefficient: −0.51, 95% CI: [−0.70, −0.33], p-value: <0.01), the interaction term between gender and educational disruption is not statistically significant (coefficient: 0.01, 95% CI: [−0.11, 0.12], p-value: 0.91). This suggests that the negative correlation of educational disruptions with academic achievement is invariant across genders.

Parental educational expectations are positively correlated with academic achievement in Model 2 (coefficient: 0.16, 95% CI: [0.13, 0.19], p-value: <0.01). However, the interaction between educational disruption and parental expectations is statistically significant (coefficient: −0.02, 95% CI: [−0.03, 0.00], p-value: 0.03). This indicates that higher parental expectations may exacerbate the negative correlation of educational disruptions with academic outcomes.

A teacher’s teaching experience exhibits a modest positive correlation with academic achievement in Model 3 (coefficient: 0.02, 95% CI: [0.01, 0.03], p-value: <0.01). The interaction term suggests that increased teaching experience may alleviate the negative relationship between educational disruptions and academic achievement (coefficient: 0.01, 95% CI: [0.00, 0.01], p-value: 0.01). These findings highlight the potential mitigating role of experienced teachers in contexts marked by educational disruptions.

School disciplinary management has a positive correlation with academic achievement in Model 4 (coefficient: 0.26, 95% CI: [0.13, 0.40], p-value: <0.01). The interaction between disciplinary management and educational disruption is also statistically significant (coefficient: 0.08, 95% CI: [0.01, 0.17], p-value: 0.04). This suggests that stricter disciplinary management can alleviate the negative correlation between educational disruptions and academic achievement to some extent.

Results from HLM affirm the robustness of the findings across these four models (see Supplementary Table S5).

Endogeneity problem

The results of the instrumental variable regression are reported in Supplementary Table S6. The Hausman test yields a p-value of 0.051, suggesting probable inconsistency between the Two-Stage Least Squares (2SLS) and OLS estimates and indicating potential endogeneity within the model, which justifies the use of instrumental variables. Furthermore, since the number of instrumental variables surpasses the number of endogenous variables, overidentification tests are essential to evaluate whether the instrumental variables are exogenous. The Sargan and Basmann tests, with p-values of 0.484 and 0.485, respectively, do not reject the null hypothesis (p > 0.05), confirming the exogeneity of the instrumental variables used in this study. Additionally, the study conducts a weak instrument test. According to the empirical rule proposed by Staiger and Stock (1997), in scenarios with a single endogenous variable, an F-value above 10 in the first-stage regression signifies a robust and significant correlation between the instrumental and endogenous variables. In this study, the F-test value in the first stage is 22.15, with a p-value less than 0.001, and Shea’s partial R² is 0.086, indicating that the instrumental variables meet the relevance criterion.

In the second stage of IV analysis, using 2SLS, Generalised Method of Moments appropriate for heteroscedastic conditions, and Limited Information Maximum Likelihood (LIML), which is less susceptible to weak instrument variables, educational disruption has been consistently observed to exert a significant negative influence, with a coefficient of −0.28 across all methodologies. This influence, supported by 95% confidence intervals ranging from −0.48 to −0.08 and a p-value of 0.01, underscores its detrimental effects. The close coefficient estimates between LIML and 2SLS suggest no weak instruments amongst the variables used.

Discussion

Our research demonstrates that early educational disruptions significantly compromise later academic achievements of migrant children, including cognitive ability, academic performance level and their ability to cope with academic challenges. This finding suggests a different perspective from the view that educational disruptions, often accompanied by improvements in environmental conditions, can afford migrant children access to higher quality education and thereby boost their academic achievements (Lee and Park, 2010). Furthermore, it is argued that such disruptions provide enhanced developmental opportunities in new settings (Scanlon and Devine, 2001; Gleason et al., 2007) and yield benefits that extend from the medium to the long term (Rathmann et al., 2020). However, our findings indicate that rural migrant children in China, despite moving from regions with substandard educational and living conditions to urban areas with more favourable environments, do not fully benefit due to the restrictions of the hukou system, authorities’ efforts to reduce the numbers of ’low quality’ migrants settling in cities, and discrimination in both schools and broader society (Goodburn, 2009). Educational disruptions in this context are not accompanied by improvements in environmental conditions and adversely affect their long-term development. Our conclusion is consistent with another vital implication of life course theory, namely that patterns of life course behaviours and outcomes vary systematically between societies (Mayer, 2004). In Chinese society, the long-term adverse effects of educational disruptions can be attributed to unique regional institutional characteristics.

Moreover, our gender analysis offers new insights into the conventional view that girls generally surpass boys across various academic dimensions, such as academic help-seeking behaviours, intrinsic motivation, and performance (Chambers and Schreiber, 2004; Cheng et al., 2023). However, does this imply that girls are less impacted by educational disruptions? Our findings suggest that gender differences do not significantly influence the effects of educational disruptions; that is, the academic achievements of girls are just as susceptible to these disruptions as those of boys. Additionally, the prevalent preference for sons in China often results in families providing more educational support to boys (Wang, 2005). Given that girls typically outperform boys academically, the educational support needs of rural migrant girls affected by educational interruptions are likely to be neglected. This observation highlights the imperative for practitioners working with migrant children in China to prioritise equity, particularly addressing the disparities faced by girls in the context of educational disruptions. It also advocates for the development of comprehensive support measures that cater to the needs of all migrant children, irrespective of gender.

Furthermore, whilst existing academic consensus generally holds that parental educational expectations are positively correlated with children’s academic achievements (Pinquart and Ebeling, 2020), our study adds depth to this understanding. Our exploration of the moderating roles at the family levels indicates that higher parental educational expectations positively impact migrant children’s academic achievement when they face fewer educational disruptions. Conversely, when educational disruptions are more frequent, increased parental educational expectations can have a detrimental effect. Typically, migrant children face higher educational expectations (Feliciano and Lanuza, 2016). However, our research demonstrates that in the context of educational disruptions, these expectations do not always prove effective. These insights offer specific guidance for designing strategic interventions. On a practical level, they suggest that policymakers and interveners for migrant children need to adjust their support strategies for migrant children based on the frequency of educational disruptions they experience. This includes, but is not limited to, adjusting parental educational expectations and optimising school management strategies.

Additionally, this study highlights the crucial role of teachers’ teaching experience and strict school management in mitigating the negative effects of educational disruptions on the academic achievement of migrant children. This underscores the value of high-quality teachers and effective school management in creating stable learning environments and providing support for migrant children. These findings align with those of Baumann and Krskova (2016) and Engin-Demir (2009), who observed similar outcomes in broader child population studies. This suggests that regardless of the context, teachers’ teaching experience and strict school management are critical determinants of children’s academic success.

Limitations

This study has several noteworthy limitations. First, this study’s data only include seventh and ninth graders in China, focusing solely on the educational disruptions and achievements of migrant children who remain in school. For those who have dropped out due to ongoing educational disruptions, our data-set provides no insights into the academic success they might have achieved if they had continued their education. Secondly, certain variables used in this study, such as academic performance levels, school rankings and disciplinary management, are inherently influenced by personal judgements. Consequently, these measures may not accurately reflect the objective standards typically used in external evaluations, introducing the potential for reporting biases. Finally, this study exclusively examines rural-to-urban migrant children in China. Future research could expand to include urban-to-urban migrant children, conducting comparative analyses to deepen our understanding of the impacts of early educational disruptions.

Conclusions and implications

Our article tests the validity of life course theory, emphasising that prior life history has strong impacts on later life outcomes (Mayer, 2009), and explores the shift from curative to preventive interventions. The findings suggest that disruptions in early education negatively affect children’s future academic achievements, corroborating the assertions of life course theory. Individual factors, family environments and school settings variably impact these effects and also serve as potential avenues for preventive interventions. Moreover, the study uses an econometric technique to address issues of endogeneity, thereby ensuring the generation of unbiased results.

This research offers a novel perspective for Chinese social workers on interventions targeting rural migrant children. It emphasises the necessity for social workers to recognise the life disturbances that stem from the early upheaval these children often experience. Identifying such risks enables social workers to proactively address the developmental challenges faced by rural migrant children through environmental enhancements and advocacy for structural reforms. Moreover, this research serves as a cautionary note for social workers globally. It is particularly relevant for those working with migrants in societies experiencing rapid urbanisation and a surge in international migration. Given that life course behaviours and outcomes vary according to social environments, social workers must remain vigilant to variations in institutional and social contexts and understand that some systems and environments may harm migrants. Furthermore, by integrating resources and addressing structural barriers, social workers can enhance the resilience of their clients against risks.

Supplementary material

Supplementary material is available at British Journal of Social Work Journal online.

Conflict of interest statement. None declared.

Data availability

The data used in this study are derived from the China Education Panel Survey conducted by the National Survey Research Center in 2013. The dataset is publicly available through the Chinese National Survey Data Archive. For access and more information, please visit https://ceps.ruc.edu.cn/.

References