Unveiling the Dynamics of Educational Equity: Exploring the Third Type of Digital Divide for Primary and Secondary Schools in China

Abstract

The COVID-19 pandemic has accelerated the integration of online learning into primary and secondary education. However, gaps persist in academic research, particularly in understanding its impact on educational equity within the third-type digital divide. This study conducted an equity-focused review to assess online learning’s impact on primary and secondary education within this context. It developed a theoretical framework integrating elements from schooling and home environments to explore equity implications in online learning. Building on this, the study proposed and validated a conceptual model using structural equation modeling (SEM), analyzing data from 1236 students in Shenzhen, China. The study found that both school investment and family involvement indirectly influence students’ online learning outcomes through complete mediating effects on students’ online learning engagement. Family investment slightly outweighs school education in its influence on outcomes. Consequently, online education within the environmental divide potentially hinders educational equity, necessitating caution with large-scale online education initiatives. This study fills research gaps on the digital divide in the third environment, leveraging China’s pandemic experience with online education. It also integrates school education and family input to examine the impact of large-scale online learning and its associated strategies on educational equity, providing insights into the promotion of educational equity.

1. Introduction

In response to the global shift towards digitalization and the accelerated evolution of Internet technology, online learning has emerged as a prominent feature in educational discourse. Embraced by institutions, online learning offers advantages such as learner convenience, information management, and cost reduction [1]. The United Nations Sustainable Development Goals emphasize the importance of robust technological infrastructure to enhance online learning’s role in addressing educational challenges [2].

The COVID-19 pandemic caused unprecedented disruptions across societal, economic, and educational domains globally. Government responses, including strict lockdown measures, led to the widespread closure of educational institutions. UNESCO data from April 2020 reveal that over 1.5 billion students, representing 91.3% of learners worldwide, transitioned away from traditional classrooms [3]. In response, educational institutions swiftly adopted online modalities to ensure educational continuity. This shift, prompted by the pandemic, has driven significant advancements and the widespread adoption of online learning in primary and secondary education [4].

Amid the growing focus on online learning, research has predominantly centered on adult college students, providing insights into various areas such as learning activity design [5], the factors influencing achievement [6], and the integration of interactive technologies [7]. Despite the widespread adoption of online learning due to the COVID-19 pandemic, there remains a notable lack of empirical evidence to guide this transition in primary and secondary education contexts. The use of online learning in primary and secondary education is rapidly increasing, yet practical research in this area is scarce. Understanding this demographic is crucial in the post-epidemic online learning landscape.

The existing literature highlights the benefits of online learning in promoting educational equity in primary and secondary education contexts. Online learning can target marginalized groups by offering flexible learning opportunities for individuals unable to access traditional classes, such as those with disabilities or facing challenges like hospitalization or bullying [8,9,10,11]. It also enhances educational access by providing students with broader resources [12] and fosters educational equity through personalized learning experiences [13].

However, scholars have identified challenges hindering educational equity through online learning [14], including the “homework gap”, where students lack connectivity for assignments and face-to-face interaction [15,16,17], and the digital divide, which refers to disparities in access to information technologies [18]. Online learning can exacerbate inequalities among students with varying internet access and support, thus impeding educational equity [19].

The present study aims to address two main inquiries: (1) Does online learning influence educational equity? (2) What specific modalities and mechanisms affect the equity of online education for primary and secondary school students? To tackle these questions, this study utilizes survey data from students across five primary and secondary schools in Shenzhen, China, collected in 2023. By examining the impact mechanisms of school and familial environments on educational equity, it clarifies the impact of online learning on educational equity through logical deduction. Subsequently, it empirically investigates the role of online learning in shaping educational equity using the gathered data.

2. Theoretical Foundations and the Framework Establishment

2.1. The Relationship between the Third Type of Digital Divide and Educational Equity

The term “digital divide” refers to disparities in the access to and utilization of electronic information and technology, driven by various social identities [20]. This divide exacerbates existing inequalities rooted in factors such as age, gender, race, and socio-economic status [21]. Online learning models are seen as promising for fostering equitable development in the digital age [22], leading to scholarly interest in the relationship between online learning and the digital divide.

The digital divide is commonly categorized into three levels: the physical divide, the skills divide, and the environmental divide [23]. The physical divide refers to the gap between those with and without access to computers and the Internet [24]. The skills divide involves the digital competencies needed to use information and communication technologies (ICTs) effectively [25]. Lastly, the environmental digital divide, also referred to as the outcome digital divide, encompasses variations in Internet access and benefits among users across different regions, economic classes, and levels of social development [26].

While existing research on online learning primarily focuses on the physical and skills divides [18,27], the environmental digital divide remains underexplored. This gap may stem from the challenge of measuring educational environments and online education outcomes, which are less tangible indicators [28].

During the COVID-19 pandemic in China, online learning became a mandatory educational method led by the government, with free access to abundant online resources provided by educational institutions [29]. This unique national context significantly reduced the impact of the first and second types of digital divides. Specifically, the Chinese government’s initiatives ensured widespread access to online learning materials, effectively eradicating the first type of digital divide [30]. Furthermore, while online learning platforms (such as Zoom and Tencent conferences) and teaching methodologies (including live teaching, recorded lectures, and hybrid approaches) exhibit diversity [31,32], ultimate authority lies with the educational institutions, resulting in minimal disparities in student utilization [29]. Additionally, schools played a critical role in supporting students’ online learning experiences during the pandemic, offering assistance in navigating digital resources and platforms to overcome barriers related to digital proficiency. Consequently, the second type of digital divide among Chinese students was substantially mitigated during this period.

The third type of digital divide presents a significant and persistent challenge compared to the physical and skills gaps. Addressing the environmental divide is crucial due to its broader impact on societal inequalities [26,33]. This divide focuses on understanding and improving digital inequalities within individuals’ environments to ensure the fair access to and use of digital technologies, fostering inclusive and sustainable societal development. Given the current research gaps in understanding the environmental divide, the opportunity to investigate this issue during China’s COVID-19 period, and its paramount importance, this study will focus on examining the relationship between online learning and educational equity among a substantial cohort of Chinese primary and secondary school students within the context of the environmental digital divide.

2.2. The Relationship between Learning Outcomes and Educational Equity

The concept of educational equity has a long history, representing an ancient pursuit within human society. However, “learning outcomes” and “educational equity” are interrelated yet distinct concepts in the field of education. Their relationship entails the question of whether educational evaluation and distribution are equitable in society. It is well established that intelligence, or more simply, talent and personal qualities, follow a normal distribution within the population [34]. Therefore, assuming that the ultimate goal of education is to foster student learning and enable students to realize their full potential, student learning outcomes should ideally exhibit a random distribution across race, ethnicity, and social class [35]. In summary, when academic learning outcomes are evenly and randomly distributed, educational equity is evident; conversely, when learning outcomes are differentiated, the development of educational equity is impeded.

2.3. The Relationship between School Education and Educational Equity

The school education environment, often referred to as the school climate in academic circles, has garnered attention since the 1950s with the emergence of organizational climate research, prompting researchers to delve into its dynamics [36]. ln the late 1970s, researchers established the link between the school climate and student achievement, revealing a strong correlation and proposing a process model elucidating how the school climate influences achievement [37]. Subsequent scholarly endeavors have further enriched this understanding, with mounting evidence indicating that a conducive school environment can enhance students’ learning outcomes [38]. In general, a safe, supportive school environment is conducive to students’ emotional, social, and academic well-being [39].

Educational equity has long been a normative goal of educational policy, with national governments striving to align school missions with the advancement of educational equity [40,41]. The progression of schools in promoting educational equity has evolved through three main stages: the quantitative stage, quality stage, and process stage. Initially, achieving educational equity involved expanding the number of schools and improving the education system [42]. Subsequently, the focus shifted to providing educational programs that met the minimum standards [43]. In the third-generation stage, the emphasis moved from education itself to the educational process, evaluating aspects like school time utilization, the teaching quality, curriculum content, and classroom practices [44]. Empirical research demonstrates that proactive school-based interventions can reduce the educational inequalities related to social factors [45]. Schools equipped with adequate teachers, facilities, curriculum resources, and supportive environments enhance student learning outcomes and narrow the educational disparities among students from diverse socioeconomic backgrounds [46]. Furthermore, as school education emphasizes both the quantity and quality of education, it often serves as a means to achieve horizontal equity in educational equity, as defined by Berne and Stiefel [47].

2.4. The Relationship between Family Environment and Educational Equity

The impact of the family environment on learning outcomes and educational equity has been extensively studied in academic research as a significant predictor of children’s development and academic success. Coleman’s 1968 study on equal educational opportunities, conducted one year after the passage of the Elementary and Secondary School Act (ESEA) in 1965, highlighted that family background has a more substantial influence on children’s educational achievement than school [48]. This finding sparked widespread debate and marked the formal exploration of family roles in educational circles. Early discussions regarding the family environment and education predominantly focused on genetics, analyzing the influence of parents on children’s development through innate factors, including inherited cognitive abilities [49]. Subsequent research increasingly emphasized the nurturing role of the factors acquired by parents. Numerous empirical studies have confirmed that families with higher social, cultural, economic, and political capital can provide superior educational opportunities for their children [50,51]. Additionally, family investments in education resources, parental education, expectations, cultural participation, support, and school engagement significantly enhance students’ academic performance across all educational stages [52].

The relationship between family education and educational equity can be traced back to the metaphor of “parents are the problem” [53], reflecting early education policies that believed families with non-dominant backgrounds could not effectively ensure their children’s well-being, leading to educational inequality. With the education reform movement in the late 1990s, the theme of “parents as partners” gained popularity [54], promoting the connection between parents and schools to foster students’ educational growth [55]. However, educational norms at this stage primarily reflected the expectations of white, middle-class families, not aligning with the daily practices of parents from non-dominant backgrounds, exacerbating the educational disadvantages experienced by children from marginalized families [56]. Scholars often draw on the concept of cultural capital to elucidate the unequal development of education caused by family factors, positing that advantaged families from higher socioeconomic backgrounds are more adept at transforming their family resources into cultural capital, creating a unique educational environment that promotes differentiated development in children [57,58]. In summary, due to its differential nature in terms of education quantity and quality, family education often impedes vertical equity in educational equity, as defined by Berne and Stiefel [47].

2.5. The Establishment of the Theoretical Framework of Educational Equity

Drawing from prior research on the environmental digital divide and educational equity (e.g., [26,33]), learning outcomes and educational equity (e.g., [35]), school education and educational equity (e.g., [40,41]), and family environment and educational equity (e.g., [56]), this study constructs a theoretical framework diagram, as depicted in Figure 1. It illustrates that investment in school education resources and the family environment can enhance students’ academic performance (+). Schools provide public and equitable educational resources and opportunities, which can reduce the academic performance gaps between students and foster educational equity (+). Conversely, differences in the family environment can lead to disparities in educational resources and opportunities, widening the academic performance gaps among students and posing a significant challenge to educational equity (−).

Based on the established theoretical framework, this study proposes the following inferences. It aims to explore the potential impact on educational equity by assessing the influence on learning outcomes.

Corollary 1: If learning outcomes are predominantly influenced by family factors, the advancement of educational equity will be impeded.

Corollary 2: If learning outcomes are primarily influenced by school-related factors, educational equity will be promoted.

3. Literature Review and Hypotheses Development

3.1. School Investment

In comparison to traditional learning, students in online environments exhibit significantly lower retention and completion rates than their face-to-face counterparts [59]. This disparity may arise from the isolating nature of online learning, technological challenges, academic expectations, and pressures from other aspects of student life [60]. Hence, fostering a supportive and engaging online teaching environment is essential for enhancing student learning outcomes and, consequently, retention rates [61]. This study incorporates school investment as a pivotal variable in evaluating the school environment. The inclusion of school education in the theoretical framework and the utilization of school investment in the conceptual model stem from the notion that school investment can be perceived as a specific manifestation or implementation of school education. The extent of school investment in online learning reflects the level of school education and directly influences the quality and effectiveness of school education [62]. Therefore, school investment was chosen as a variable in this study, aiming to scrutinize more precisely the impact of school education on learning outcomes and educational equity.

Scholarly findings suggest that school investment can directly impact students’ online learning outcomes. Borup et al. [63] conducted a semi-structured interview survey involving 11 teachers, revealing that teacher engagement in K-12 school investment—such as designing and organizing learning activities, providing one-on-one guidance, facilitating home–school communication, fostering a conducive learning environment, motivating student participation in learning activities, and assessing student learning behaviors—can significantly enhance students’ academic performance. Nityasanti et al. [64], through questionnaires, observations, documents, and interviews conducted across five primary schools in Kediri County in Indonesia, found, via a quantitative n-gain analysis of retrieved data, a substantial positive correlation between the schools’ supplementary teaching efforts during online learning and the enhancement of students’ academic, social, and personal competencies.

Further research indicates that school investment also impacts student engagement in learning. Pan and Shao [65], utilizing structural equation modeling and bootstrap analysis in a questionnaire survey of 312 college students enrolled in Chinese college English courses, demonstrated that teachers’ online feedback, an aspect of school investment, significantly correlates with students’ online learning engagement and motivation. Stone and Springer [61] synthesized research endeavors across 16 Australian universities, highlighting the significance of online “teacher presence”. Their findings underscored the pivotal role of teacher–student communication, interactive course design, and a supportive, encouraging teaching environment in fostering effective student participation in online learning. In summary, this study posits the following research hypotheses:

H1. 

School investment significantly impacts learning outcomes.

H2. 

School investment significantly impacts online learning engagement.

3.2. Family Involvement

Compared to traditional classroom learning, the transition to online learning at home has revolutionized the educational landscape, transforming it from a collective school environment to a personalized learning space that merges the home with the Internet. In this new educational paradigm, online learning has redefined the roles of both teachers and parents, underscoring the significance of family involvement in education and sparking considerable interest in the relationship between family educational participation and online learning [66]. Consequently, this study regards family involvement as a crucial variable in the family environment. The inclusion of the family environment in the theoretical framework and family involvement in the conceptual model is grounded in the understanding that family participation has a particularly profound impact on the family environment within the context of online education. By examining how family involvement influences the efficacy of online learning, we can gain deeper insights into the influence of the family environment on educational equity.

On one hand, scholars have observed that family involvement directly influences students’ online learning outcomes. Kang et al. [67] analyzed data from the 2020 China Family Panel Studies (CFPS) using ordered probit regression and structural equation modeling to investigate the factors affecting the academic performance of Chinese K-12 students during the COVID-19 pandemic. Their empirical findings revealed a significant positive relationship between parental involvement and academic self-efficacy, which in turn impacted academic performance. Ricker et al. [68], utilizing learning management system (LMS) data from three K-12 full-time virtual schools in the United States, assessed the impact of parental involvement on mathematics achievement. Despite controlling for various factors known to affect educational outcomes, they found that parental involvement positively influenced mathematics performance across the elementary, middle, and high school levels.

On the other hand, due to the unique challenges posed by online learning, which requires greater self-management and self-motivation compared to traditional classroom education, family involvement plays a critical role in shaping students’ learning engagement. Family involvement significantly predicts students’ learning engagement. Gao et al. [69] conducted a questionnaire survey among 1317 Chinese college students, analyzing the impact of family support on students’ online learning engagement. Their results demonstrated that families can effectively and positively influence students’ online learning engagement by fostering a conducive home learning environment. Similarly, Purnomo et al. [70] conducted a cross-sectional survey among students in five primary schools in Indonesia, finding a significant positive relationship between parental involvement and students’ online learning engagement in mathematics. In summary, this study posits the following research hypotheses:

H3. 

Family involvement significantly impacts learning outcomes.

H4. 

Family involvement significantly impacts online learning engagement.

3.3. Online Learning Engagement

In traditional learning, the roles of teachers and families are pivotal in facilitating learning. However, online learning has disrupted this conventional paradigm by granting learners greater autonomy and flexibility. Consequently, online learning imposes higher demands on learners’ engagement in their education [71]. Learning engagement, defined as the degree of effort exhibited by students in actively participating in their studies, encompasses behavioral engagement (active participation and completion of course requirements), cognitive engagement (the development of self-regulated learning strategies), and emotional engagement (making positive or negative contributions to learning) [72]. Thus, this study incorporates online learning engagement as a key variable in order to investigate the relationship between online learning and educational equity within the context of the third type of digital divide.

Scholars have observed that students’ engagement in online learning significantly influences their learning outcomes and cognitive development levels. Curtis and Werth [73] conducted a qualitative study on a full-day K-12 virtual school in the western United States, finding through interviews that self-motivated, fully engaged, and responsible children are more likely to transition from traditional to online learning environments. Additionally, students’ engagement in online education impacts their achievements in the online environment. Lowes et al. [74] utilized structural equation modeling to analyze high school students’ participation in 12 online courses provided by Pamoja Education, an International Baccalaureate course provider. They found a significant correlation between high school students’ academic performance in the courses and their active and positive online learning behaviors. You [75] conducted a questionnaire survey among undergraduate students at a university in Shandong Province, China, revealing through regression analysis that the three dimensions of students’ online learning engagement (behavioral, emotional, and cognitive) significantly influence students’ academic completion levels. Consequently, this study posits the following research hypothesis:

H5. 

Online learning engagement has a significant positive impact on learning outcomes.

Based on the established research hypotheses, this study further delineates the conceptual model required, as illustrated in Figure 2.

4. Method

4.1. Research Design

4.1.1. Sample Data Acquisition and Processing

This study employs structural equation modeling (SEM) to analyze the research questions and primarily utilizes questionnaires to collect sample data for the empirical analysis. The author developed an “Online Learning Questionnaire” (see Appendix A below) based on established scales, which underwent rigorous review by primary and secondary school principals in Shenzhen and underwent thorough proofreading before the survey commenced. Permission to conduct the study was obtained from the school principal, and the survey was conducted using the “Questionnaire Star” web platform tool.

Shenzhen, as one of China’s most developed cities, boasts advanced scientific and technological infrastructure and a high socioeconomic status [76,77]. With four years of practical online learning experience (from December 2019 to December 2022) during the COVID-19 pandemic in China, Shenzhen frequently implemented large-scale online learning initiatives for primary and secondary schools, enabling the physical and skills gap issues in the digital divide to be isolated and the environmental gap issue, which is the core concern of this study, to be focused on.

Against the backdrop of China’s COVID-19 epidemic prevention and control policies, Shenzhen’s primary and secondary schools commenced online learning in December 2022, which continued until the semester’s end. Offline teaching was set to fully resume in Shenzhen in the spring semester of 2023. Consequently, this study was conducted following the offline final exams at the onset of the spring semester, ensuring participants’ exposure to online learning. This approach makes the study more reflective of the online learning situation within the current technological and social environment. To mitigate the potential viewing angle limitations arising from technical constraints, the questionnaires were distributed and completed by students in the school’s computer room during information sessions.

4.1.2. Research Objects

To ensure the broad applicability and reliability of the research results, the conclusions of this study were designed to be persuasive and instructive, serving as a reference for various types of schools and educational institutions. The research design encompassed diverse school types, including one private primary school, one public primary school, one international primary school, one private junior high school, and one public junior high school in Shenzhen. International junior high schools were excluded due to their distinct curriculum systems, which differ from traditional schools and hinder the measurement of learning outcomes at a comparable level. However, international primary schools’ education and examination systems closely resemble those of traditional schools, thus warranting their inclusion.

All aforementioned schools belong to the same educational group, facilitating control over potential factors affecting the study, such as management systems and teaching philosophies. This enhances the internal validity of the research and ensures the reliability of the conclusions. Since differences in school types are not the primary focus of this study, the core emphasis lies in examining the differences in learning outcomes between school education and the family environment. Moreover, due to the study’s length constraints, differences in school types will be further explored in subsequent research.

The omission of primary school students in grades 1–3 from the questionnaire survey is justified for several reasons. Firstly, cognitive ability was considered; the study selected students in grades 4 to 9 as samples, as researchers believe that students within this grade range typically possess higher cognitive abilities and understanding levels, enabling them to independently comprehend and complete questionnaires. Secondly, the ‘double reduction policy’ prevalent in China’s basic education stage emphasizes a reduction in students’ learning burdens in grades 1–3, with no paper-and-pencil exams conducted, rendering the measurement of learning outcomes unfeasible [78]. Thirdly, international standards were referenced during sample selection, particularly the OECD’s SSES2023 China survey data, which indicated that the youngest age of students surveyed should be 10 years old, corresponding to fourth-grade students [79]. Thus, our sample selection aligns with international survey standards.

4.1.3. Data Cleaning

Online questionnaire surveys present various challenges, including a lack of supervision and guidance, technical issues, respondent distraction, and difficulties in understanding and interpretation, which pose greater research risks compared to offline questionnaire surveys [80]. To address these challenges, this study employed a rigorous data cleaning approach.

Initially, the author implemented a stringent criterion where each question required a minimum response time of 2 s. Responses falling below this threshold were deemed indicative of inadequate attention and were consequently regarded as low-quality data, warranting deletion. A total of 505 such inadequate responses were removed, resulting in 2118 remaining datasets. Subsequently, the author applied a criterion whereby responses to major sections of the questionnaire should not be identical. Highly similar responses were considered indicative of insufficient engagement and were thus excluded. This step led to the removal of 882 additional questionnaires, leaving 1236 valid samples.

While this strict screening process resulted in a lower questionnaire recovery rate, the researcher contends that it ensures the quality of the data used for analysis, thereby enhancing the accuracy and reliability of the research findings. Furthermore, the ample amount of original data enabled the study to maintain a substantial sample size even after the screening process, thereby supporting robust research outcomes.

4.1.4. Data Description

A descriptive analysis was conducted on 1236 valid samples, with the findings presented in

Table 1. Descriptive analysis of sample data.

Item	Category	Stage of Study
		Elementary School (Grades 4–6)		Junior High School (Grades 7–9)
		Frequency	% of Total	Frequency	% of Total
Gender	Male	376	30.42%	204	16.50%
	Female	415	33.58%	241	19.50%
	Total	791	64.00%	445	36.00%
School type	Public	335	27.10%	182	14.72%
	Private	236	19.10%	263	21.28%
	International	220	17.80%	0	0.00%
	Total	791	64.00%	445	36.00%

. Overall, the gender distribution among the 1236 respondents was relatively balanced, albeit with a higher number of primary school respondents compared to junior high school students. However, considering that the study encompassed five schools, with three at the primary level, the distribution of students across school stages remains relatively consistent. The diversity in gender, schooling stage, and school type among the respondents, coupled with the even distribution of numbers, suggests that the sample data obtained from the survey are relatively representative.

4.2. Variable Measurement

4.2.1. Learning Outcomes

Due to the need to uphold anonymity, it was not feasible to obtain students’ standardized scores directly from the school and correlate them with individual questionnaire responses. Consequently, students’ self-evaluation of their performance in these final exams following their participation in online learning was selected as an indicator of their learning outcomes. Additionally, we assessed the reliability of the indicator using Cronbach’s alpha.

To ensure the alignment of students’ assessments with the research objectives, the study opted to utilize larger-scale assessment metrics such as final exam scores. These scores typically carry a degree of authority and objectivity within the school’s academic framework, offering a more accurate reflection of students’ academic achievements.

The survey was conducted immediately after the release of the test results to mitigate the impact of time-related bias and subjective memory. This external research design enhances the precision and reliability of students’ self-perceived learning outcomes, ensuring their consistency with the study’s objectives. Moreover, some researchers [81] have demonstrated a positive correlation between students’ self-assessment of learning outcomes and their actual academic performance. Thus, this study contends that the self-evaluation of learning outcomes by students in this context effectively captures their outcomes in online learning.

A reverse-scored seven-point Likert scale questionnaire was used to gauge students’ perceptions of the impact of online learning on their academic achievement. To check the reliability of the instrument, factor analysis was conducted using SPSS 26.0, with principal axis factoring and optimal oblique rotation (kappa = 4). The scale demonstrated high internal consistency reliability (Cronbach’s α = 0.869), and the KMO measure yielded a value of 0.735, with Bartlett’s test of sphericity indicating significance.

4.2.2. Online Learning Engagement

To measure the latent constructs in our model, this study used the student online learning engagement scale adapted from Guo [82]. And we assessed the reliability of the instrument using Cronbach’s alpha. This scale includes three dimensions: behavioral engagement, cognitive engagement, and emotional engagement, consisting of 19 items scored on a seven-point Likert scale. Exploratory factor analysis was conducted using SPSS 26.0. The scale demonstrates good reliability (Cronbach’s α = 0.944).

4.2.3. School Investment

This study measures school investment as an indicator of the school environment’s influence on students’ online learning. Additionally, we assessed the reliability of the instrument using Cronbach’s alpha. The variable “school investment” is measured using items from the “School Online Teaching Investment” section of the “Online Learning Questionnaire for Primary and Secondary Schools” by Wang et al. [71]. The measurement content includes students’ perceptions of the school’s support for online learning platforms and tools, the quality of school management and services during online learning, teachers’ commitment to online teaching, and the school’s arrangements for online learning courses and tasks. The scale consists of 6 items rated on a seven-point Likert scale. An exploratory factor analysis was conducted using SPSS 26.0, demonstrating that the scale has strong reliability (Cronbach’s α = 0.943).

4.2.4. Family Involvement

To measure the latent constructs in our model, we adapted instruments from the literature. In addition, we checked the reliability of the instrument with Cronbach’s alpha values. The family involvement scale in this study was adapted from the “Parental Engagement Scale” by Wu and Yao [83]. It includes items related to participation in parenting education, home–school exchanges, volunteer participation in school activities, home tutoring, participation in decision-making, and the participation of children in community activities and life care. The scale consists of 25 items and is scored using a seven-point Likert scale. Exploratory factor analysis was conducted using SPSS 26.0. The scale demonstrates good reliability (Cronbach’s α = 0.922).

5. Results

5.1. Measurement Mode

A confirmatory factor analysis (CFA) was used to assess the fitness, validity, and reliability of the model’s constructs. The CFA was carried out in AMOS 24.0 software. We obtained a good model fit (χ² = 4924.733, df = 1312, χ²/df = 3.754, SRMR = 0.0466, TLI = 0.907, CFI = 0.911, RMSEA = 0.047), which indicated a good match between the variables and their respective constructs.

The validity and reliability measures attained for the model are presented in

Table 2. Test for validity and reliability.

	CR	AVE	MSV	MaxR(H)	Learning Outcomes	Online Learning Engagement	School Investment	Family Involvement
Learning outcomes	0.869	0.689	0.080	0.874	0.830
Online learning engagement	0.890	0.731	0.446	0.902	0.283 ***	0.855
School investment	0.944	0.736	0.381	0.950	0.153 ***	0.617 ***	0.858
Family involvement	0.821	0.540	0.446	0.856	0.154 ***	0.688 ***	0.536 ***	0.735

Note: *** p < 0.001.

. Convergent validity can be assessed through composite reliability (CR) and the average variance extracted (AVE) [84]. CR values exceeding 0.7 and AVE values surpassing 0.5 are indicative of satisfactory convergent validity [85]. In the measurement model, the CR values for all constructs exceed 0.7. Specifically, learning outcomes exhibit a CR of 0.869, online learning engagement exhibits a CR 0.890, school investment exhibits a CR 0.944, and family involvement exhibits a CR 0.821. Furthermore, the maximum reliability values (MaxR(H)) surpass 0.7, affirming reliability. The AVE values for all constructs exceed 0.5, as displayed in Table 3. The square root of the AVEs (diagonal fields in Table 2) exceeds the correlations between constructs, confirming discriminant validity [84]. Additionally, the maximum shared variance (MSV) remains below the AVE values, supporting discriminant validity.

5.2. Common Method Variance

Common method bias (CMB) occurs when responses are influenced by the data collection instrument rather than the actual measured construct. Given that the data for this study were collected via an online survey, Batista-Foguet et al. pointed out the possibility of CMB [86]. In SEM analysis, an accurate model is developed to explain the observed data. If common method bias exists but is not controlled for, the explanatory power of the model may be affected because the SEM model may misinterpret the common method variance as the true relationship between variables [87].

In the process of SEM analysis, the measurement model needs to be established first, followed by the structural model. The measurement model evaluates the relationship between each observed variable and the underlying constructs, while the structural model assesses the relationships between these underlying constructs. Because CMB involves controlling for bias between observed variables, it is typically addressed after the measurement model is established.

This study employed Harman’s single-factor test to assess the presence of common method bias [88]. It conducted an exploratory factor analysis on all measures related to the variables in the model, utilizing principal component analysis without any rotation extraction. The maximum factor variance explanation rate was 29.108%, which fell below the critical value of 40%, suggesting the absence of a significant common method bias issue in this study.

To further investigate common method bias, confirmatory factor analysis was conducted. This study utilized AMOS 24.0 to construct a structural equation model with a single-factor structure, ensuring that all measurement indicators loaded only on one common factor. The fitting results of the model were then computed. However, none of the single-factor CFA model fitting indices met the required standard (χ²/df = 12.412, RMSEA = 0.096, CFI = 0.495, SRMR = 0.0916, TLI = 0.480, NFI = 0.475), indicating that the issue of common method bias was not severe [89].

5.3. Question 1: The Impact of Online Learning on Educational Equity

Drawing upon the theoretical framework regarding the influence of school and family environments on educational equity, this article formulated a structural equation model. In this model, school investment and family involvement were designated as independent variables, online learning engagement was the mediating variable, and learning outcomes was the dependent variable.

The structural equation model (SEM) assessing online learning in primary and secondary schools underwent rigorous fitness testing. Table 3 presents the test outcomes, revealing key metrics such as χ²/df = 3.983, RMSEA = 0.049, CFI = 0.904, and PNFI = 0.834. An evaluation against established SEM goodness-of-fit criteria, as delineated by Wu [90], demonstrated a commendable alignment between the hypothesized path relationships and the observed data. Consequently, the constructed model and its associated assumptions emerged as robust and reflective of the empirical reality.

Table 3. The adaptation results of the SEM model.

Statistical Testing	Absolute Fitness Indices			Value-Added Adaptation Indices			Parsimonious Fitness Indices
	χ2/df	RMSEA	GFI	NFI	IFI	CFI	PNFI	PCFI	PGFI
Adaptation standard	2~5	<0.08	>0.8	>0.9	>0.9	>0.9	>0.5	>0.5	>0.5
Parameter	3.983	0.049	0.846	0.876	0.904	0.904	0.834	0.861	0.776
Other parameters: n = 1236; χ2 = 5231.334; df = 1315; p < 0.001;

Table 3. The adaptation results of the SEM model.

Statistical Testing	Absolute Fitness Indices			Value-Added Adaptation Indices			Parsimonious Fitness Indices
	χ2/df	RMSEA	GFI	NFI	IFI	CFI	PNFI	PCFI	PGFI
Adaptation standard	2~5	<0.08	>0.8	>0.9	>0.9	>0.9	>0.5	>0.5	>0.5
Parameter	3.983	0.049	0.846	0.876	0.904	0.904	0.834	0.861	0.776
Other parameters: n = 1236; χ2 = 5231.334; df = 1315; p < 0.001;

Path analysis was conducted on the online learning model, and the results are presented in

Table 4. The Model Path Test.

Hypothesis	Std Estimate	S.E.	C.R.	p	Results
H1. School Investment→Learning Outcomes	−0.030	0.042	−0.772	0.440	Rejected
H2. School Investment→Online Learning Engagement	0.459	0.020	14.923	***	Supported
H3. Family Involvement→Learning Outcomes	−0.055	0.068	−1.273	0.203	Rejected
H4. Family Involvement→Online Learning Engagement	0.493	0.035	13.007	***	Supported
H5. Online Learning Engagement→Learning Outcomes	0.309	0.085	6.126	***	Supported

Note: *** p < 0.001.

. According to the joint significance test, school investment significantly and positively predicts online learning engagement (β = 0.459, p < 0.001), while family involvement also significantly and positively predicts online learning engagement (β = 0.493, p < 0.001). Moreover, online learning engagement significantly and positively predicts learning outcomes (β = 0.309, p < 0.001). However, the impact paths of school investment and family involvement on learning outcomes were found to be not significant.

Hence, it can be inferred that online learning engagement serves as a full mediator in the relationship between school and family input and learning outcomes. In addition, the R-squared values for online learning engagement and learning outcomes are 0.452 and 0.065, respectively (see Figure 3 below).

This means that approximately 45.2% of the variance in online learning engagement and only 6.5% of the variance in learning outcomes can be explained by the independent variables used in the model. The explained variance in the online learning outcomes in primary and secondary schools is statistically low but still significant for educational equity. Educational outcomes are influenced by various complex factors, so a low explained variance is common in educational research. This indicates that our model partially explains differences in the learning outcomes and highlights the importance of these factors in educational equity. In addition, even small effects can be significant in educational equity. Differences in family involvement and school input can lead to some students receiving more resources and support, amplifying inequality. These small effects can accumulate over time and within large-scale educational environments, leading to serious disparities. Additionally, the context and baseline of the study must be considered when interpreting the explained variance. In online education, significant differences in family and school input may reflect the unique contributions of these factors to learning outcomes. The digital divide and inequality of educational resources in the research context may further amplify this effect.

In summary, the conceptual model constructed in this study, based on the theoretical framework, successfully passed the path test and model fitness test. This indicates that the educational equity theoretical framework established in this study is applicable to online learning in primary and secondary schools within the context of the third type of digital divide. Moreover, it demonstrates that both home and school environments during online learning can influence the equitable development of education.

5.4. Question 2: The Specific Impact Mechanism of Online Learning on Educational Equity

This study employs the Bootstrap method to test and analyze the SEM model’s mediation effect of the school and family environment on learning outcomes. Based on the test analysis results, the specific impact mechanism of online learning on educational equity is assessed. Although the path analysis demonstrated the impact of school input and family support on learning outcomes, as well as the mediating role of participation, using the Bootstrap method to test the mediating effect can provide additional statistical evidence. The Bootstrap method can calculate the standard error and confidence interval of the effect, enabling a more accurate assessment of whether the mediation effect is significant. Additionally, it enhances the credibility of the research results as it does not require normality and large samples as premise assumptions, and it is more efficient than traditional mediation effect analysis methods such as the stepwise regression method and the Sobel test, ensuring higher accuracy and scientific rigor. Furthermore, the Bootstrap method generates a large number of samples through repeated sampling, which can more reliably estimate the distribution of parameters, thereby enhancing the robustness of the findings.

This study repeatedly sampled 3000 individuals at a 95% confidence interval. The analysis results are presented in

Table 5. The bootstrap test of the mediation effect.

Path (Std)	Std Estimate	Standard Error	Bias-Corrected Confidence Interval (90%)		p Value	Effect Ratio
			Lower	Upper
The school effect	0.142	0.027	0.095	0.200	**	48.30%
The family effect	0.152	0.030	0.098	0.220	***	51.70%
The total effect	0.294	0.053	0.197	0.404	**	100%

Note: ** p < 0.01, *** p < 0.001.

. The point estimate of the model path “school investment → online learning engagement → learning outcomes” is 0.142, with a bias correction confidence interval of (0.095, 0.200). The model’s bias correction interval excludes 0, indicating a significant path mediation effect. Similarly, the point estimate of the model path “family involvement → online learning engagement → learning outcomes” is 0.152, with a bias correction confidence interval of (0.098, 0.220), also showing a significant path mediation effect. Furthermore, the point estimate of the total effect of the model is 0.294, with a bias correction confidence interval of (0.197, 0.404). Once again, the model’s bias correction interval excludes 0, suggesting a significant path mediation effect.

This study found that the impact of family involvement and school investment on learning outcomes is mediated through their effect on the learning input, thereby indirectly influencing learning outcomes. It was also found that school investment contributed to 48.30% of the total effect, attributed to the indirect path of “school investment → online learning engagement → learning outcomes”. However, family involvement accounted for 51.70% of the total effect, attributed to the indirect path of “family involvement → online learning engagement → learning outcomes”.

6. Discussion

6.1. The Opposite Effect of Online Learning: Family’s Ascendancy and School’s Recession

In the context of the environmental digital divide, this study explores the combined influence of family and school environments on shaping online learning outcomes in primary and secondary schools, offering novel insights into the development of educational equity in large-scale online learning settings. The findings reveal that both family and school environments exert similar pathways and effects on the learning outcomes of primary and secondary school students during online learning. This aligns with Tao et al.‘s research [91], which underscores the necessity of joint support from teachers and parents to guide students’ self-regulation processes in online learning.

Moreover, within the context of the environmental digital divide, research indicates that the home environment has a greater influence on student achievement than the school environment, signaling a shift in online learning dynamics from traditional school-centered influences to family-centric influences. This shift may stem from the decline in face-to-face interactions as teachers transition from offline to online teaching models, altering the traditional dynamics of school education [92,93,94,95]. Consequently, the family factors impacting the learning and holistic development of primary and secondary school students are magnified. This finding echoes the conclusions drawn by Liu et al. [96], who also underscore the pivotal role of parents in supporting their children’s online learning endeavors.

6.2. The Critical Role of Online Learning Engagement and Reinforcing Family Education

As proposed in this research model, within the context of the third type of digital divide, learning outcomes in online learning ecosystems are influenced by both direct and indirect factors. Among these, online learning engagement emerges as a direct and effective predictor of learning outcomes. Furthermore, environmental factors such as students’ families and schools do not directly impact students’ learning outcomes but rather indirectly shape them through the intermediary role of “online learning engagement”. This resonates with the personalized nature of online learning, which emphasizes and reinforces students’ individual learning behaviors [97,98].

The findings of this study further validate why, in the third type of digital divide scenario, online learning outcomes tend to emphasize the influence of family dynamics. This aligns with the personalized nature of online learning, which emphasizes and reinforces students’ individual behaviors in learning [99,100]. Consequently, engaging in online learning demands heightened self-management and self-control capabilities from students, distinct from the structured supervision and guidance prevalent in traditional offline classrooms. This highlights the need for students to possess increased autonomy and flexibility in their learning journeys, thereby amplifying the supportive and constructive influence of family education investment [71]. Consequently, the pivotal role of online learning investment in enhancing learning outcomes serves to accentuate the significance of family education, further intensifying its impact on the differentiation of educational quality and quantity. This cumulative effect profoundly influences the equitable advancement of education.

6.3. The Adverse Impact of Online Learning on Educational Equity in Third-Type Digital Divide

Amidst the online learning process, the traditional influence of formal education led by schools is disrupted, yielding ground to the influence of families. By analyzing the theoretical framework established in this study regarding online learning’s impact on educational equity, it becomes evident that during the formation of educational outcomes in widespread online learning settings for primary and secondary school students, familial factors predominantly shape the influencing dynamics. Consequently, the online learning model amplifies the differentiation in family education roles, potentially impeding the advancement of educational equity.

Previous research has often analyzed the impact of students’ online learning outcomes from the perspective of the “access gap” and “usage gap” differences between students’ online learning. It has been found that under the circumstances of physical and skills gaps, online learning exacerbates the inequality of students’ learning outcomes [101,102]. This study found that even when the first-level and second-level digital divides are relatively isolated, the existence of the third-level digital divide situation still impacts the development of educational equity. Network information technology promotes the fair and high-quality development of education, an inherent requirement of the education informatization era [103]. However, large-scale online learning exposes potential equity risks. Therefore, educational policymakers need to be cautious of the haphazard development of large-scale online education, especially those learning models that highly integrate formal school education and online learning.

7. Contributions

7.1. Evaluating Online Education’s Impact on Equity amid the Environmental Digital Divide

This study holds significant importance as an analysis of the impact of online education on educational equity within the framework of the third type of digital divide, particularly by integrating theoretical arguments at both the family and school levels. Existing research on the impediments of online education to educational equity primarily focuses on the first and second types of digital divides. However, in the era of rapid technological advancement, the third type of digital divide is emerging as a predominant concern. This divide exhibits the characteristics of extensive and influential group effects. By leveraging the advantages of China’s online education during the COVID-19 pandemic, this study systematically discusses online education in Chinese primary and secondary schools within the context of the third digital divide, effectively addressing deficiencies and gaps in the current research.

7.2. Bridging School and Family Dynamics for Assessing Equity in Online Learning

Furthermore, from a theoretical standpoint, numerous controversies persist in the discussion regarding the relationship between online learning and educational equity. However, existing research predominantly focuses on the characteristics of online learning from the perspectives of both application (e.g., family environment) and provision (e.g., school support) [104,105,106]. Few studies have conducted comprehensive assessments, simultaneously evaluating these dual aspects and their collective impact on educational equity. This study bridges this gap by integrating school education, the family environment, and previous research to construct a theoretical framework for assessing their impact on educational equity. Consequently, it logically infers the potential outcomes of large-scale online learning on the development of educational equity. This pioneering study combines school education and family contributions to analyze the implementation pathways and impact of large-scale online learning on educational equity, offering novel insights into mechanisms for promoting educational equity.

7.3. Examining Online Learning Dynamics in Primary and Secondary Schools

Moreover, this study holds significant value as it systematically examines the impact of online learning in primary and secondary schools, particularly focusing on educational equity. Current discussions surrounding the impact of online learning predominantly center on higher education, inadvertently overlooking the pivotal role of primary and secondary school students in the post-pandemic era. The study encompasses a diverse range of school settings among surveyed students, categorized based on their mode of operation (public, private, and international) and schooling period (primary and junior high school stages). With a large sample size and diverse sample sources, this study comprehensively evaluates online learning participation, facilitating the dissemination of research outcomes. By delving into the dynamics of influence within primary and secondary school contexts, this study addresses this oversight, providing insights that are relevant to the fundamental aspects of online learning development. Consequently, it serves as a crucial foundation for understanding the online learning ecosystem within primary and secondary schools.

8. Limitations and Future Directions

8.1. Limitations

8.1.1. Challenges of Concept Nesting in Research

While integrating school input and family involvement into broader concepts like school education and family environment may seem straightforward, this alignment does not guarantee identical behavior in research settings, potentially leading to confusion. Moreover, these factors are not exhaustive in their impact on respective environments, risking oversights. Lastly, integrating school input and family engagement within broader concepts might limit the nuanced understanding of these environments.

8.1.2. Limitations in Sample Generalization

This study’s applicability extends to regions susceptible to environmental digital divides, akin to Shenzhen’s context. Consequently, its scope is confined to areas sharing characteristics such as an advanced technological infrastructure, economic prosperity, ample educational resources, and active educational reforms. Thus, this study may find resonance in similarly developed Chinese cities and countries with comparable socioeconomic profiles and online education implementations.

8.2. Future Directions

8.2.1. Detailed Analysis of Educational Environment Dimensions

Due to space limitations, this study focuses on analyzing how online education affects student learning outcomes and explores the impact of school investment and family involvement on educational equity. Our research design addresses these core issues rather than performing a detailed analysis of specific dimensions. Additionally, while the family involvement scale used in this study is rich and multi-level, the school involvement scale is relatively simple. Analyzing family involvement in depth without a corresponding analysis of school investment would result in an unbalanced study. Therefore, we comprehensively scored and analyzed these factors as a whole. Discussing specific forms within the joint educational environment can enhance interpretation and provide specific recommendations. Future research may use measurement tools with consistent and comprehensive dimensions for detailed analysis.

8.2.2. Comparative Analysis of School Stages and Types

While this study’s sample encompasses primary and junior high school students from diverse school types, the article primarily focuses on differential analyses of school education and the impact of the family environment on learning outcomes and educational equity. Due to space constraints, specific comparative analyses across school periods and types were not conducted. However, variations in online learning adaptability, study habits, and outcomes may exist across different school stages and types, influenced by distinct learning environments and motivations. Future research could integrate a comparative analysis.

8.2.3. Methodological Considerations in Research Selection

This study relies on self-reported questionnaires that were analyzed through structural equation modeling in order to address the research questions. However, when exploring educational equity in online education, quantitative research alone may not suffice for in-depth analyses. Future research could integrate qualitative analysis methods. Additionally, quantitative investigations into online education and equity may incorporate variables related to social identity factors (such as minority groups or socioeconomic status), establish control groups, and analyze specific educational equity gaps.