How Does Parental Mediation Impact Children’s Academic Performance Within the Family System? Evidence from a Nationwide Survey in China

Abstract

With the accelerating pace of societal digitalization, parental mediation practices have emerged as a critical mechanism for facilitating children’s socialization within the family system. Based on the 2022 Chinese Minors’ Digital Life and Online Protection Survey dataset, this study empirically examined the current state of parental mediation and its influencing mechanism on school-aged children’s academic performance. The results indicated that the level of parental mediation demonstrated a significantly positive influence on students’ academic performance. This finding remained robust when subjected to rigorous validation methods, including instrumental variable analysis, propensity score matching, and double machine learning techniques. Meanwhile, the mediation models showed that children’s e-learning behavior and civic engagement played parallel mediating roles in the relationship between parental mediation and academic performance. Moreover, heterogeneity analysis further revealed that family socioeconomic status negatively moderated the indirect pathways through which parental mediation influenced students’ academic performance. Our research suggested that active parental mediation could not only be beneficial for children’s academic performance but also be helpful to narrowing developmental inequalities across different social classes.

1. Introduction

The pervasive adoption of digital technologies—such as mobile networks, artificial intelligence, and cloud computing—has fundamentally transformed the way we communicate, work, and perceive the world, ushering in an era of unprecedented connectivity and intellectual evolution. Undoubtedly, these innovations have not only restructured the macro-level socioeconomic system but have also profoundly influenced micro-level family dynamics, particularly in parent-child relationships [1,2]. The primary socialization process of minors is thus inevitably embedded within a digitalized environment. In China, where the present study was conducted, more than 95% of children and adolescents aged 6 to 19 had used the Internet for instant messaging and online entertainment [3]. To mitigate the potential negative impacts of the online environment on minors’ development, parents are increasingly expected to proactively supervise, guide, and regulate the digital content their children access. As a result, family parenting practices are undergoing a digital transformation, shifting from offline to online domains. Understanding how to promote the healthy growth of minors through effective parental mediation has become a central concern in family education within the digital age [4,5].

Theoretically, parental mediation is an abstract concept that arises from the deep integration of digital media with traditional parenting practices in the family system [6]. A large body of literature indicates that different parental mediation strategies can have varied effects on educational development of children and adolescents. Specifically, during daily parent-child interactions, when parents actively explain media content to their children, it may enhance students’ cognitive development and help reduce the risks of cyberbullying and privacy breaches [7]. Meanwhile, parental restrictions on children’s exposure to online media have been found to be beneficial in mitigating pathological Internet-based behaviors; however, excessive restrictions may provoke media rebellion among adolescents and exert adverse effects on their mental health [8]. Moreover, the co-use of online media between parents and children could facilitate access to educational resources and improve learning efficiency, thereby positively influencing children’s academic progress [9]. In a word, parental mediation plays a crucial role in shaping minors’ educational outcomes and ought not to be overlooked or undervalued. In addition, it is worth emphasizing that the influence of parental mediation may extend beyond a mere direct effect, as it could operate through specific behavioral mechanisms that amplify its overall effects. Admittedly, the family constitutes a complex microsystem within the ecological framework of child development [10]. Thus, from the perspective of family systems theory, the behaviors of individual family members and their respective consequences can only be fully understood in relation to one another [11,12,13]. If we view parental mediation as a specific manifestation of a digital media-based relationship, then in the process of implementing media interventions, parents may affect their children’s behavioral tendencies through emotional interactions. For instance, some scholars have argued that parental mediation contributes to shaping children’s behavioral habits, such as e-learning behavior and civic participation, which in turn indirectly affect children’s developmental outcomes [14,15]. In light of this, we assume that incorporating the relevant behavioral characteristics of school-aged children into the analytical framework is essential for further elucidating the micro-level mechanisms through which parental mediation influences students’ educational outcomes.

In this study, based on a nation-wide cross-sectional survey dataset, we aim to investigate the current status of parental mediation and its impact on academic performance of Chinese primary and secondary school students. We may make marginal contributions in the following ways. On one hand, we employ a rigorous empirical approach to mitigate the issues such as reverse causality, self-selection bias, and the nonlinear interference of confounding factors (i.e., family background and school characteristics). On the other hand, it may also enrich our understanding of the underlying mechanisms linking parental mediation to students’ academic performance through conditional process analysis. Furthermore, in terms of public policy realm, this study suggests that social policies aimed at promoting equitable child development should be taken into account under the context of digitalized family dynamics.

The remainder of this article is structured as follows. First, we review the literature on parental mediation and academic performance, from which we derive our research hypotheses. Second, we outline the research design, including the data source, variable measurement, and analytical strategies. Third, we present the baseline regression results along with a series of robustness checks. Finally, we discuss the empirical findings and propose policy recommendations to inform more effective family parenting practices in the digital era.

2. Literature Review and Research Hypotheses

2.1. Parental Mediation and Academic Performance

Academic performance serves as a critical indicator that reflects the knowledge, skills, and cognitive abilities acquired by individuals through the process of school-based socialization [16]. During the periods of childhood and adolescence, parental involvement has been found to exert a sustained impact on minors’ interpersonal relationships and educational outcomes [17]. However, the widespread application of digital technology is transforming the growth environment for children and adolescents. In fact, although digital media platforms provide expanded opportunities for interpersonal communication and leisure activities, unregulated or excessive usage may result in unintentional adverse outcomes, including addiction, diminished sleep quality, and strained interpersonal relationships [18,19]. A recent study pointed out that the use of digital devices for communication and entertainment purposes (e.g., watching television, browsing social media, and playing video games) has been shown to negatively affect students’ academic performance [20]. One possible explanation for this adverse effect is the time-displacement hypothesis, which posits that screen time encroaches upon and reduces the time allocated to academic-related activities. Therefore, as the gatekeepers of their kids’ media exposure, parental mediation is of critical importance to maximize the positive effects of digital media while minimizing potential negative influences [21].

Drawing on the typological framework originating from the early television era [22], scholars have proposed three primary forms of parental mediation: (1) Instructive mediation involves parents guiding children to critically understand and interpret media content through dialogue, discussion, and explanation; (2) Restrictive mediation refers to the establishment of rules concerning the duration and type of media content accessed by children; (3) Co-use describes situations in which parents and children engage with the same media content simultaneously, although discussion may not necessarily occur. Empirical research indicates that media intervention strategies can have differentiated impacts on students’ educational outcomes. To be specific, active mediation typically entails engaging in open dialogue and discussion with children about their media use, which may help foster a more objective and rational approach to media consumption and promote the development of critical thinking skills [21]. Hence, active mediation strategies (e.g., the establishment of family online guidelines) would be associated with multiple positive outcomes, including enhanced comprehension of media content [19]. Restrictive parental mediation, such as establishing rules to regulate their children’s screen time, can save more time for academic engagement and consequently enhance children’s school performance [23,24]. Nevertheless, restrictive interventions are not always effective. Actually, excessive restrictions may increase the appeal of prohibited media content, provoke children’s psychological resistance, and lead to problematic media use that ultimately hinder their academic performance [18,21]. Furthermore, the co-use of media content for educational purposes between parents and children mostly demonstrates a remarkable protective effect in mitigating problematic media use within family life [20].

As for the current study, given the rapid pace of digital transformation, Chinese parents’ awareness of media mediation remains relatively underdeveloped, and the current level of parental mediation is more likely to be inadequate rather than excessive. Additionally, the potential adverse effects of excessive media intervention may not manifest immediately but could emerge after a certain time lag. For example, regulations concerning the use of technology during childhood may influence an individual’s academic performance in university [25]. As such, although several studies have indicated that excessive media intervention may have negative effects on the development of minors, we still argue that in contemporary Chinese families, a higher level of parental mediation is more likely to result in the appropriate use of digital media, thereby exerting a positive influence on underage students’ educational achievement. Accordingly, we propose:

H1: 

Parental mediation can contribute positively to the academic performance of Chinese primary and secondary school students.

2.2. The Mediating Role of E-Learning Behavior

Digital technology per se is a double-edged sword, and its impact largely determined by the way it is applied [26]. Existing evidence suggests that educational online activities have a more positive effect on students’ human capital development compared to recreational media usage (e.g., online gaming) [27]. Therefore, an essential mechanism through which parental mediation influences academic outcomes may lie in its potential to guide minors toward using digital media for learning purposes rather than solely for entertainment. Under Chinese context, most parents initially acquire digital devices for their children primarily to support educational development, whilst active parental mediation has been found to be positively associated with the use of online learning resources [28]. That is to say, by means of constructive communication and guidance, parents could effectively direct their children’s focus toward the educational capabilities of the Internet [27]. Furthermore, increased parental support was also linked to greater student engagement in academic tasks and a higher likelihood of using the Internet for information-seeking purposes [29]. When parents have imposed restrictions on recreational online activities, children tended to allocate their limited online time to educational activities that are permitted or encouraged [27]. In addition, it has also been found that the collaborative use of smartphones by parents and children for learning purposes might be considered highly effective, as it focused on interactive collaboration and aligned with the developmental characteristics of adolescents and emphasizes [20]. In a nutshell, parental mediation seems to effectively shift children’s use of digital media from a primarily entertainment-focused activity toward one that is more oriented towards productive and educational outcomes. Based on this observation, we propose the following hypothesis:

H2a: 

Parental mediation promotes e-learning behavior among Chinese primary and secondary school students.

Learning-oriented digital practices, such as utilizing digital media to complete school assignments or acquire cutting-edge knowledge, are also positively correlated with students’ academic performance [30]. In fact, digital learning platforms would provide students with diverse learning approaches that can stimulate students’ self-directed learning interest and creativity, thereby significantly improving knowledge retention [31]. While recreational Internet use has been found to negatively affect academic performance, a substantial body of literature indicates that educational Internet use could enhance students’ digital literacy and overall learning competence [26,28,29]. As a result, it could be inferred that e-learning behavior may be beneficial for enhancing the academic performance of school-aged children. Based on the aforementioned analysis, we propose the following hypotheses:

H2b: 

E-learning behavior positively contributes to the academic performance of Chinese primary and secondary school students.

H2c: 

E-learning behavior mediates the relationship between parental mediation and academic performance of Chinese primary and secondary school students.

2.3. The Mediating Role of Civic Engagement

Civic engagement refers to the behaviors that individuals demonstrate during social interactions, such as helping, cooperating, sharing, and showing concern for public issues, which align with societal expectations [32]. In the digital age, prior research has shown that media addiction is associated with a decline in the quality of real-life interpersonal interactions [28]. Therefore, timely parental mediation plays a crucial role in fostering children’s capacity for social participation [33]. In fact, there was evidence that a higher level of parental media intervention could help reduce children’s exposure to antisocial and harmful media content and encourage them to participate in socially beneficial activities (e.g., volunteer club involvement) [25,34]. Additionally, when parents and children engage in communication regarding the use of digital devices, it would also enhance parent-child interaction, which should be conducive to creating a positive emotional atmosphere within the family and further promoting children’s empathy and prosocial personality development [30]. As such, we propose:

H3a: 

Parental mediation promotes civic engagement among Chinese primary and secondary school students.

In the meantime, prior empirical studies have indicated that civic engagement behaviors are related to a range of positive developmental outcomes, such as enhanced emotional and cognitive abilities and improved mental health. These outcomes are generally considered essential prerequisites for achieving academic success [35,36]. Moreover, students’ participation in public activities can foster a sense of social cohesion and contribute to the accumulation of social capital. Consequently, when confronted with academic pressure, they are more likely to receive encouragement and support from their peer groups, which may ultimately enhance their academic performance [37]. Therefore, in conjunction with the judgment of H3a, the impact of parental mediation on academic performance may be mediated through civic engagement behavior, leading to the following hypotheses:

H3b: 

Civic engagement positively contributes to the academic performance of Chinese primary and secondary school students.

H3c: 

Civic engagement mediates the relationship between parental mediation and academic performance of Chinese primary and secondary school students.

2.4. The Moderating Role of Family Socioeconomic Status

Family parenting is essentially a process through which parents transfer their own socioeconomic capitals across generations, and as a result, parenting practices reflect the characteristics of social stratification [12]. Parental mediation, which integrates digital technology with parenting practices, is inherently connected to the process of social class differentiation. Previous research has indicated that parents from lower socioeconomic backgrounds tended to pay less attention to the potential risks associated with digital media, leading to a wider digital gap between them and their children [7]. In contrast, parents with higher socioeconomic status generally possessed stronger digital literacy skills and were more capable of providing their children with a secure and high-quality digital learning environment [30,38]. Following this line of thought, those children from higher social class actually require less media intervention, which may in turn diminish the direct effect of parental mediation on academic performance. At the same time, parents from higher social class tend to hold higher educational expectations for their children. They often arrange for their children to attend offline tutoring sessions or extracurricular courses during their free time. This form of shadow education would inevitably reduce the time children have available for online self-directed learning and participation in social activities. Consequently, in higher-class families, such high educational expectations may also reduce the positive impact of parental mediation on two mediating variables, i.e., e-learning behavior and civic engagement. Based on the above analysis, this paper further proposes a set of hypotheses related to the heterogeneous effects of parental mediation:

H4a: 

Family socioeconomic status has a moderating effect on the direct path through which parental mediation influences the academic performance of Chinese primary and secondary school students.

H4b: 

Family socioeconomic status has a moderating effect on the indirect path of parental mediation influencing e-learning behavior among Chinese primary and secondary school students.

H4c: 

Family socioeconomic status has a moderating effect on the indirect path of Parental mediation influencing civic engagement among Chinese primary and secondary school students.

In summary, to systematically investigate the influencing mechanism of parental mediation on academic performance, we propose a moderated parallel mediation model (see Figure 1).

3. Materials and Methods

3.1. Data

The data utilized in this study were obtained from the 2022 Chinese Minors’ Digital Life and Online Protection Survey. Between May and July 2022, the research team from the Institute of Sociology at the Chinese Academy of Social Sciences conducted a nationwide questionnaire survey targeting students aged 6 to 17 years old and their parents in 29 provinces or municipalities directly under the Central Government. The objective of the survey was to investigate the current status and characteristics of Chinese minors with regard to family background, digital literacy, and educational development. To be specific, the survey utilized a multi-stage sampling approach to collect questionnaire information from students and their parents. Initially, 3 to 4 cities at the prefecture level were chosen from each province. Subsequently, four schools were selected from each of these cities, including two elementary schools, one junior high school, and one high/technical school. In these schools, a simple random sampling method was employed to select a single class for the questionnaire survey. In the present study, after excluding samples with missing data on key variables (e.g., parental mediation indicators), a total of 5209 parent-child dyads were retained for subsequent statistical analysis.

3.2. Variables

The explained variable involved children’s academic performance. Since academic performance was based on different schools and grades, to reduce assessment variability and enhance comparability among different schools and grades, ordinal indicators can be used to measure the overall situation of students’ academic performance [39]. Specifically, we intend to utilize a multi-indicator approach to assess academic performance. The first objective indicator was the students’ academic rank within the class, while the remaining two subjective indicators reflected the satisfaction levels of students and their parents regarding academic outcomes. Each of these three indicators was measured on a scale ranging from 1 to 5. For analytical convenience, a common factor was extracted through principal component analysis, which yielded a cumulative variance contribution rate of 69.0%, suggesting that this composite factor retained a substantial portion of the information contained in the original indicators. Subsequently, the Min-Max standardization technique was applied to transform the factor score into a continuous variable ranging from 0 to 100, with a higher score indicating better academic performance.

Parental mediation served as the core explanatory variable in this study. The questionnaire assessed parents’ degree of agreement with the following statements on a Likert scale: (1) I teach my child how to protect their online privacy; (2) I know how to locate appropriate online learning resources for my child; (3) I teach my child how to evaluate the authenticity of online information; (4) I know how to set video or short-video platforms to “youth mode”; (5) I know how to monitor my child’s gaming accounts through tools provided by game platforms; (6) I know how to prevent my child from making unauthorized online purchases using my account; (7) I know how to locate my child’s online accounts (e.g., gaming accounts); (8) I know how to track the online content my child has accessed; (9) I am capable of using the digital tools and software that my child uses. Consistent with the previously established conceptual framework [22], items (1) through (3) assess instructive mediation, items (4) through (6) correspond to restrictive mediation, and items (7) through (9) reflect the co-use of online media content. The response options for these nine items ranged from “strongly disagree = 1” to “strongly agree = 5”. Cronbach’s alpha coefficient for the scale was 0.931. We calculated the average of these nine items, with higher values indicating a greater level of parental mediation.

E-learning behavior and civic engagement were two mediating variables in this study. With respect to e-learning behavior, the survey investigated students’ frequency of using internet resources to engage in seven specific activities, such as searching for information and participating in online courses. These seven items were measured using a five-point Likert scale (ranging from never = 1 to always = 5). The Cronbach’s alpha coefficient for these items was 0.885, indicating a high level of internal consistency. The average score of the seven items was then calculated, with a higher value reflecting a greater intensity of e-learning behavior. Regarding the operationalization of civic engagement variable, the questionnaire asked students to indicate the frequency with which they participated in six specific activities during the past year, such as volunteering and reporting social issues to schools, the government or the media. The Cronbach’s alpha coefficient for these six items was 0.889. We computed the average score across the six items, with higher values indicating a greater degree of civic engagement.

To minimize the influence of confounding factors, we incorporated a set of control variables, i.e., gender, household registration, attributes of the school, stage of schooling (as a proxy variable for age), and family socioeconomic status. We utilized the father’s educational level, the mother’s educational level, and annual household income to measure family socioeconomic status. For analytical convenience, the principal component method was applied to extract a common factor from these three indicators, with a cumulative variance contribution rate of 71.6%. The factor score was then transformed into a continuous variable ranging from 0 to 100 by means of Min-Max standardization approach. Descriptive statistical results of the research variables are presented in

Table 1. Summary statistics for the research variables.

Variables	Coding	Mean (Percentage)	Standard Deviation
Student’s gender	Male = 1, female = 0	48.3%	-
Student’s household registration	Urban registration = 1, rural registration = 0	44.8%	-
Schooling stage
Primary school	Yes = 1, no = 0	29.7%	-
Junior high school	Yes = 1, no = 0	26.6%	-
High school/technical school	Yes = 1, no = 0	43.7%	-
School attribute	Key school = 1, non-key school = 0	37.7%	-
Family socioeconomic status	0 to 100	29.555	26.370
Academic performance	0 to 100	49.820	20.255
Parental mediation	1 to 5	3.103	0.851
E-learning behavior	1 to 5	2.760	0.831
Civic engagement	1 to 7	2.855	1.249

.

3.3. Analytical Procedure

The empirical analysis was conducted in four sequential steps. The first step involved performing a baseline analysis using Ordinary Least Squares (OLS) regression model, which served to preliminarily analyze the magnitude and direction of the impact of parental mediation on academic performance. Secondly, we employed a variety of statistical schemes to validate the robustness of the OLS regression results: (1) an instrumental variable (IV) approach was applied to address reverse causality; (2) propensity score matching (PSM) was utilized to correct for self-selection bias; and (3) double machine learning (DML) method was implemented to account for potential nonlinear confounding effects on dependent variables. Thirdly, the conditional process analysis was harnessed to examine the mediating roles of e-learning behavior and civic engagement. Fourthly, we introduced interaction terms to further investigate the moderating effect of family socioeconomic status on the influential pathways.

4. Research Results

4.1. Baseline Analysis Using OLS Regression Model

Table 2. OLS regression analysis on the determinants of academic performance.

	Model 1	Model 2	Model 3
Parental mediation	2.558 *** (0.328)	1.905 *** (0.331)	1.587 *** (0.330)
Gender (female = 0)		0.213 (0.555)	0.175 (0.552)
Household registration (rural = 0)		1.278 * (0.595)	0.995 (0.621)
School attribute (non-key school = 0)		−0.693 (0.582)	−0.779 (0.577)
Schooling stage (primary education = 0)
Secondary education		−5.191 *** (0.759)	−5.712 *** (0.765)
High school/technical school education		−6.323 *** (0.709)	−6.886 *** (0.725)
Family socioeconomic status			0.020 * (0.012)
E-learning behavior			1.405 *** (0.340)
Civic engagement			1.715 *** (0.226)
Constant	41.884 *** (1.055)	47.638 *** (1.286)	39.813 *** (1.570)
R2	0.012	0.033	0.051
F value	60.796 ***	29.855 ***	30.811 ***

Notes: The standard error of the non-standardized regression coefficient is presented in parentheses; * denotes p < 0.1 and *** denotes p < 0.01.

demonstrates the results regarding the effect of parental mediation on students’ academic performance, based on the traditional OLS regression model. We employed a nested modeling approach: Model 1 included only the core explanatory variable, whereas Model 2 and Model 3 progressively incorporated relevant control variables. As shown in Table 2, the marginal effect of parental mediation on academic performance was statistically significant at the 1% level across all models, providing initial support for Hypothesis 1. In addition, it could be seen from Model 3 that there was no significant difference in students’ academic performance across different gender or household (urban vs. rural) groups. To some extent, this finding reflects the trend of gender and regional equalization of public educational resources. Meanwhile, there appeared to be significant disparities in academic performance among students from different socioeconomic backgrounds (p < 0.1), suggesting that the issue of social stratification continues to exist in students’ educational development and cannot be overlooked in contemporary education systems. It is worth noting that the analytical results pointed out an unstable effect of parental mediation across three models, likely due to the presence of confounding variables. As such, in the subsequent section, we will use machine learning methods to mitigate the potential interference of these confounding factors on the estimation outcomes.

4.2. Robustness Assessment of Baseline Analysis Results

4.2.1. Test of Reverse Causality Using Instrumental Variable

OLS regression analysis based on cross-sectional data may face challenges in addressing potential reverse causality. Actually, most of Chinese parents place significant emphasis on their children’s education and are highly attentive to academic performance. In this regard, parents may intensify their involvement in regulating children’s digital media use due to their expectation of achieving better educational outcomes. Consequently, parental mediation could be an endogenous factor. To address this potential endogeneity issue, we employed the provincial-level mean of parental mediation as an instrumental variable. This variable satisfies the two key conditions for a valid instrument: relevance and exogeneity. On one hand, it reflects the aggregate level of parental mediation within a region and is thus correlated with parenting practices in that region. On the other hand, provincial-level characteristics are unlikely to directly influence individuals’ learning behaviors, which meets the exogeneous hypothesis of instrumental variable.

Table 3. IV regression analysis on the determinants of academic performance.

	First-Stage Regression DV: Parental Mediation	Second-Stage Regression DV: Academic Performance
Parental mediation	—	7.050 * (3.656)
Provincial parental mediation level	0.633 *** (0.096)	—
Other variables	Controlled	Controlled
Weak instrument test	F value = 43.117; p < 0.01
Dubin-Wu-Hausman test	Durbin χ2 =2.358; Wu-Hausman F value = 2.370; p > 0.1

Notes: The standard error of the regression coefficient is presented in parentheses; DV = Dependent Variable; * denotes p < 0.1 and *** denotes p < 0.01.

presents the results of a two-stage least square regression analysis. The first-stage regression results indicate that the provincial-level parental mediation variable exerted a statistically significant influence on individual-level parenting behavior at the 1% significance level, satisfying the relevance condition for instrumental variables. Meanwhile, the F-statistic from the weak instrument test exceeded 10, suggesting that the instrumental variable was sufficiently strong. In the second stage, the regression results demonstrate that, after accounting for potential endogeneity, parental mediation continued to have a significant positive effect on academic performance at the 10% level. Notably, the Durbin-Wu-Hausman test failed to reject the null hypothesis that parental mediation was exogenous (p > 0.1), implying no significant evidence of endogeneity or reverse causality. That is to say, the conclusion that higher levels of parental mediation contribute positively to children’s academic performance is reasonably well-supported.

4.2.2. Correction of Self-Selection Bias

In addition to the reverse causality issue, this study may also be subject to self-selection bias, as the level of parental mediation could result from parents’ self-selection based on characteristics of children, families, and schools. To address this issue, we will employ the propensity score matching method to construct a counterfactual framework, thereby mitigating the influence of self-selection bias.

More specifically, to enhance the distinction between levels of parental mediation, a mean score above 4—corresponding to the threshold for selecting “agree” on the 5-point scale—was used to indicate a high level of parental mediation and was coded as 1; otherwise, it was coded as 0. Then, based on observable sociodemographic and school-related characteristics, a binary logit regression model was applied to calculate the probability of high-level parental mediation (i.e., the propensity score). Next, by means of the propensity score estimates, multiple matching techniques were used to construct a comparable treatment group (high level of parental mediation) and control group (low level of parental mediation) to reduce self-selection bias. Lastly, using the matched samples, the average difference in academic performance between the treatment and control groups computed to estimate the average treatment effect on the treated (ATT). Figure 2 illustrates the balance of covariates between the treatment and control groups before and after one-to-one nearest neighbor matching. As shown in Figure 2, significant imbalances between the two groups were evident before matching, and these discrepancies were substantially reduced after the matching procedure.

Table 4. Estimation results of propensity score matching.

Matching Methods	Treatment Group	Control Group	ATT
Nearest neighbor matching (1:1)	54.082	48.963	2.550 ** (t value = 2.300)
Nearest neighbor matching (1:5)	54.082	50.527	3.555 *** (t value = 4.020)
Radius matching within caliper	54.082	51.062	3.020 *** (t value = 3.700)
Kernel matching	54.082	50.875	3.207 *** (t value = 3.960)
Local linear regression matching	54.082	51.133	2.949 *** (t value = 2.620)

Notes: ** denotes p < 0.05 and *** denotes p < 0.01.

describes the ATT estimates for the two sample groups after propensity score matching using five matching strategies. Although the ATT values varied slightly across the different methods, all estimates consistently indicated that high-level parental mediation had a positive effect on children’s academic development, reinforcing the findings of the baseline analysis.

4.2.3. Adjustment of Nonlinear Relationships Using DML Method

Chernozhukov pointed out that the traditional OLS regression model constrained the relationships between analytical variables to a linear specification, thereby failing to fully account for potential nonlinear influence of confounding variables [40]. Machine learning algorithms can offer a robust framework for addressing high-dimensional controls and identifying causal relationships in statistical modeling. Therefore, we intend to harness Double Machine Learning (DML) method in order to enhance the estimation accuracy. The DML approach uses machine learning algorithms to predict both the independent and dependent variables. By applying the Neyman orthogonal rule, the corresponding prediction residuals are obtained, leading to a more reliable estimation of the effect of parental mediation on academic performance.

Table 5. Partial regression analysis based on double machine learning method.

Machine Learning Algorithm	Regression Coefficient	Robust Standard Error	95% Confidence Interval
Lasso with Cross-Validation	1.586 ***	0.363	[0.874, 2.297]
Random Forest	1.829 ***	0.371	[1.101, 2.557]
Support Vector Machines	2.194 ***	0.366	[1.476, 2.912]

Notes: The regression coefficient represents the median value obtained from 101 repeated samplings; *** denotes p < 0.01.

showed the estimation results obtained from running DML programs using Stata 18.0 in conjunction with Python 3.10 software. Following the Chernozhukov’s recommendations [36], the samples were split in a ratio of 1:5 for cross-validation purposes, and the coefficient estimates were derived through repeated sampling conducted 101 times. More specifically, we employed three mainstream machine learning algorithms—Lasso with Cross-Validation (LASSOCV), Random Forest (RF), and Support Vector Machines (SVMs)—to assess and validate the robustness of the partial regression analysis results. As presented in Table 5, regardless of the algorithm employed, the regression coefficient of parental mediation remained positively significant at the 1% statistical level. This finding provided strong evidence that parental mediation could contribute to enhancing students’ educational development. In addition, the estimated value obtained using LASSOCV—which is particularly effective in capturing linear relationships between variables—was found to be comparable to the estimate in Model 3 of Table 2. In the meantime, the estimated coefficients derived from nonlinear modeling approaches (i.e., RF and SVM algorithms) yielded values of 1.829 and 2.194, respectively, which were relatively higher than the results from linear models. To sum up, although the results presented by different modeling specifications exhibit slight variations, their 95% confidence intervals largely overlap, consistently confirming the importance of parental involvement in shaping children’s media-related behaviors in the digital age. In the next section, we will conduct in-depth conditional process analysis to further identify the underlying mechanism linking parental mediation to students’ academic development.

4.3. Examination of the Mediating Roles of E-Learning Behavior and Civic Engagement

Prior empirical analysis indicates that the prediction results of different modeling approaches are broadly consistent. Consequently, we intend to proceed using Hayes’s conditional process analysis method under the assumption of linearity among variables, as it provides a more straightforward basis for explanation [41].

The findings of the mediation effect analysis were stated in

Table 6. Mediation analysis results based on bootstrap approach.

Influence Paths	Indirect Effect Value	Bootstrap 95% CI	Percentage (Indirect Effect/Total Effect)
Path (a): parental mediation → e-learning behavior → academic performance	0.133	[0.058, 0.215]	7.05%
Path (b): parental mediation → civic engagement → academic performance	0.165	[0.081, 0.271]	8.75%
Sum of indirect effects: Path (a) + Path (b)	0.298	[0.173, 0.431]	15.8%

Notes: The total effect of parental mediation on academic performance was 1.886, with a bootstrap 95% confidence interval ranging from 1.237 to 2.535.

. Firstly, after controlling for other variables, the influence of parental mediation on the two mediating factors—e-learning and civic engagement—was found to be statistically significant, with effect sizes of 0.095 (p < 0.01) and 0.096 (p < 0.01), respectively. These results provide support for Hypotheses 2a and 3a (for the sake of brevity, the regression results with the two mediating factors as dependent variables were omitted here). Secondly, as shown in Model 3 of Table 2, the regression coefficients of e-learning and civic engagement on academic performance were 1.405 (p < 0.01) and 1.715 (p < 0.01), respectively. Hence, Hypotheses H2b and H3b were both validated. Finally, after performing 2000 random bootstrap samplings, the indirect effects of the two mediating pathways and their respective contributions to the total effect were calculated. As presented in Table 6, mediating pathway (a) accounted for 7.05% of the total effect, whereas mediating pathway (b) accounted for 8.75%. Moreover, Model 3 of Table 2 also showed that the direct effect of parental mediation on academic performance was statistically significant (regression coefficient = 1.587, p < 0.01). Accordingly, e-learning behavior and civic engagement were verified to partially mediate the relationship between parental mediation and students’ academic performance. In other words, parental mediation not only directly promotes students’ academic development but also exerts significant indirect effects through these two mediating pathways, thereby supporting Hypotheses H2c and H3c. As a result, the proposed parallel mediation model has been substantiated by the survey data.

4.4. Examination of the Moderating Role of Family Socioeconomic Status

Previous literature highlights that parental mediation practices are inherently related to the process of social stratification [26,35]. Then, will the identified influence of parental mediation on academic performance vary according to differences in family socioeconomic status? To explore this further, we introduce an interaction term to examine the moderating effect of family socioeconomic status. The results of Model 1 in

Table 7. Test results of the moderating effect of family socioeconomic status.

	Model 1 DV: Academic Performance		Model 2 DV: E-Learning Behavior		Model 3 DV: Civic Engagement
	Regression Coefficient	Bootstrap 95% CI	Regression Coefficient	Bootstrap 95% CI	Regression Coefficient	Bootstrap 95% CI
Parental mediation	2.092 *** (0.510)	[1.092, 3.091]	0.132 *** (0.021)	[0.091, 0.174]	0.176 *** (0.032)	[0.113, 0.238]
Family SES	0.070 * (0.040)	[−0.008, 0.149]	0.004 ** (0.002)	[0.001, 0.007]	0.009 *** (0.003)	[0.004, 0.014]
Parental mediation × Family SES	−0.016 (0.013)	[−0.041, 0.008]	−0.0012 ** (0.0005)	[−0.0022, −0.0002]	−0.003 *** (0.001)	[−0.004, −0.001]
Other variables	Controlled		Controlled		Controlled

Notes: Other variables in this table were consistent with those reported in Table 2; * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01; DV = Dependent Variable.

indicate that the regression coefficient of interaction term was not statistically significant, suggesting no disparity in the direct impact of parental mediation across families with varying levels of socioeconomic status. H4a was not supported by our materials. In Model 2 and 3 of Table 7, however, the regression coefficients of the interaction terms were both statistically significant. Such findings show that family socioeconomic status exerted a moderating effect on mediating pathways of e-learning behavior and civic engagement. Thus, Hypotheses H4b and H4c were verified.

To visually represent the moderating effect of family socioeconomic status and its influencing boundary, we attempted to generate Johnson-Neyman conditional effect diagrams. As depicted in Figure 3, with the rise in family socioeconomic status score, there was a discernible decline in the conditional effect of parental mediation on students’ e-learning behavior. Additionally, the threshold interval of the moderating effect was quantitatively measured based on Johnson-Neyman method. The finding reveals that when family socioeconomic status score dropped below 69.348 (i.e., cumulative percentile from low to high at about 89.6%), the 95% confidence interval for the conditional effect remained positive but gradually decreased. However, once the score exceeded 69.348, the 95% confidence interval for the conditional effect began to include zero point, suggesting that within this statistical range, the moderating effect of family socioeconomic status was no longer statistically significant. Similarly, Figure 4 demonstrates that the positive impact of parental mediation on civic engagement diminished as the family socioeconomic status increased; once its score surpassed approximately 49.085 (i.e., cumulative percentile from low to high at about 74.8%), the conditional effect of parental mediation started to lose its statistical significance. Therefore, it could be inferred that the role of parental mediation in facilitating minors’ behavioral development seemed to be limited among higher socioeconomic families. These results suggest that, on one hand, family socioeconomic status does not constrain the direct impact of parental mediation on students’ academic performance. That is to say, a higher degree of parental mediation generally contributes to the academic development of children across various social classes. On the other hand, family socioeconomic status may exert a significant negative moderating effect on the two mediating pathways. Compared to families with higher socioeconomic status, those with lower socioeconomic status experience a relatively greater marginal utility from parental mediation in fostering children’s behavioral habits, leading to a stronger indirect positive influence on their academic development. In this sense, if parents from lower socioeconomic backgrounds place greater emphasis on digital parenting and guide their children to effectively harness digital media, it could help alleviate the developmental disparities experienced by children from disadvantaged family backgrounds.

5. Discussion

To date, the rapid proliferation of emerging digital technologies has enabled children to conveniently interact with the external world via mobile devices. A childhood characterized by mobile phone usage and mediated through the Internet is gradually supplanting one that was traditionally grounded in face-to-face social interactions. However, due to the open and anonymous nature of the Internet, parents often possess limited knowledge regarding their children’s online behaviors, which could inevitably lead to numerous safety risks to children’s development. Therefore, parental mediation becomes increasingly important in children’s socialization process. To some extent, compared to family structure factors such as intergenerational living arrangements, parental mediation strategies appear to be more malleable and amenable to improvement, offering greater potential for guiding targeted interventions aimed at reducing the cyber risks faced by children [42].

Based on the 2022 Chinese Minors’ Digital Life and Online Protection Survey dataset, the present study thoroughly examines the characteristics of parental mediation practices among Chinese parents and explores the mechanisms through which these practices influence children’s academic performance. Our research confirms the significant influence of parental mediation on children’s academic development. This finding underscores that parent-child dynamics exert a profound effect on individual growth outcomes [20]. In this regard, parents can play an active role in fostering healthy media consumption habits in children, ensuring that sufficient time and resources are dedicated to academic pursuits. In addition, parent-child communication also contributes to the establishment of a nurturing and supportive family environment, which in turn enhances children’s ability to maintain greater focus on achieving their educational objectives [43].

In the meantime, our results found that e-learning behavior served as a significant mediator in the relationship between parental mediation and students’ academic performance. Indeed, the influence of digital technology on academic performance is predominantly contingent upon its mode of use—specifically, whether it is employed for educational purposes or solely for entertainment [26]. When parents engage in discussions about digital content with their children and recommend high-quality websites and applications, they not only convey valuable information but also shape children’s perception of the internet as a primary tool for learning. This finding highlights the significance of parental mediation as a critical external factor in guiding children toward constructive online activities. In addition to e-learning behavior, civic engagement also played a mediating role that facilitated children’s academic development. By employing positive mediation strategies, parents can function as gatekeepers, guiding their children toward prosocial media content and encouraging participation in altruistic activities, such as community volunteer services. Civic involvement may enhance individuals’ social competence and empathy, which is beneficial for cultivating a supportive social-ecological environment conducive to academic achievement [35].

Empirical analysis further showed that family socioeconomic status exerted a negative moderating effect on the aforementioned mediating pathways. This novel insight points out the potential of parental mediation as a positive force for mitigating educational disparities across social classes. It has long been established that families with higher socioeconomic status, by virtue of their advantages in economic, cultural, and social capitals, are better positioned to provide their children with better access to digital resources [30]. However, this study provides an alternative explanatory perspective. In high socioeconomic status families, children’s academic development is typically supported by a range of favorable factors, including economic stability, abundant cultural capital, and access to high-quality educational resources. According to the principle of diminishing marginal utility, under conditions of resource saturation, parental mediation—although still beneficial—yields relatively limited additional gains. In contrast, for low socioeconomic status families facing resource constraints, effective parental intervention may serve as a critical lever for altering children’s developmental trajectories. This is particularly evident in the Chinese context. Irrespective of socioeconomic status, Chinese parents commonly hold high educational aspirations for their children and prioritize their children’s education as a central family objective [38]. Consequently, when parents with a lower socioeconomic status actively engage in their children’s digital activities and encourage constructive social participation, the compensatory effect of such involvement becomes more pronounced, thereby contributing to improved academic outcomes.

6. Conclusions, Implications, and Study Limitations

6.1. Conclusions

With the accelerating pace of societal digitalization, increasing scholarly attention has been devoted to examining family parenting practices within the digital media environment and their impact on the academic development of school-aged children. From the perspective of family systems theory, parent-child interaction constitutes a central element of family education. By constructing a moderated parallel mediation model, this article may contribute to a deeper comprehension of the intergenerational impacts of parental mediation within a specific family system and offer valuable theoretical insights for addressing inequalities in the educational development of school-aged children.

The main research findings were as follows. First, the level of parental mediation demonstrated a significantly positive influence on students’ academic performance. After controlling for confounding factors using different machine learning algorithms, this conclusion was found to be robust. Second, mechanism tests revealed that e-learning behavior and civic engagement played parallel mediating roles in the relationship between parental mediation and students’ academic performance: on one hand, parents could extend their strict mediation strategies to the online environment, guiding children to develop active e-learning habits, which in turn enhanced their academic outcomes; on the other hand, parental mediation might encourage children to use digital media responsibly to access prosocial content while minimizing exposure to antisocial information, thereby fostering civic engagement and improving children’s overall competence. Third, heterogeneity analysis further revealed that the effect of parental mediation on children’s e-learning behavior and civic engagement was more pronounced in families of lower socioeconomic status than in those of higher socioeconomic status. This finding indicated that family socioeconomic status negatively moderated the indirect pathways through which parental mediation influenced students’ academic performance.

6.2. Theoretical and Practical Implications

From an academic perspective, the aforementioned findings provide theoretical implications for understanding the factors that influence students’ academic development. Firstly, our study underscores the critical role of parental media intervention in shaping children’s developmental trajectories especially under a digitalized context. To be precise, we drew on empirical data to validate the inter-relationships among parental mediation, children’s e-learning behavior, civic engagement, and academic performance. This approach enables the establishment of a novel conceptual framework that enhances the understanding of intergenerational influences on children’s educational achievement. Secondly, existing studies predominantly adopt a typological analytical strategy, emphasizing the distinctions among instructive, restrictive, and co-use mediation practices. However, the nine-item parental mediation scale utilized in this study showed a high level of internal consistency, suggesting that these mediation strategies were not necessarily mutually exclusive but exhibited a degree of complementarity. This insight facilitated a more comprehensive understanding and interpretation of the concept of parental mediation. Thirdly, previous research has primarily focused on the fact that parents with lower socioeconomic status may encounter parenting dilemmas due to their unfamiliarity with digital technology, low confidence in managing their children’s online risks, and limited ability to provide academic support stemming from their own insufficient knowledge reserves. However, the existing literature has overlooked the “negative selection effect” of socioeconomic status. Actually, we found that parental mediation could more significantly enhance children’s e-learning and civic participation behaviors from lower socioeconomic backgrounds, thereby potentially contributing to the reduction of educational inequalities across different social classes.

Drawing upon the empirical findings discussed above, we propose to offer the following policy recommendations. To begin with, given that parental mediation demonstrates a positive effect on students’ academic performance, parents should place greater emphasis on the pivotal role of media intervention strategies in the domain of children’s development. They are encouraged to actively participate in their children’s developmental journey with a greater sense of responsibility and inclusiveness, which contributes to the creation of a supportive online environment conducive to children’s behavioral and educational growth. Then, this study reveals that e-learning and civic participation behaviors evidently impact students’ academic performance. As such, parents should guide their children in making good use of online learning platforms and accessing valuable information that supports the development of a prosocial personality. This approach may contribute positively to the enhancement of children’s overall competencies. Last but not least, considering the negative selection effect associated with parental mediation practices, relevant government agencies should provide digital training programs for parents from lower socioeconomic backgrounds. These programs would not only raise parents’ awareness of guiding their children’s digital activities but also equip them with the necessary skills to effectively access and utilize free or low-cost online educational resources. By doing so, children from disadvantaged backgrounds can achieve greater upward mobility through improved parenting practices, thereby contributing to the reduction of developmental inequalities in the digital age.

6.3. Study Limitations

As an exploratory study, this article has certain limitations. First, although we have applied statistical techniques to help reduce potential endogeneity concerns, the cross-sectional analysis still lacks the rigor necessary to establish causal relationships between variables. Second, due to the design constraints of the questionnaire items, the way we measured the research variables may not be as precise as desired. Third, this study focused on the parallel mediating effects of e-learning and civic engagement but did not explore the possible interaction between these two factors or how they might jointly influence students’ academic development. In future research, we will strive to enhance the validity of our findings by incorporating longitudinal data and in-depth qualitative materials to more accurately elucidate the underlying mechanisms connecting parental mediation and children’s educational development.