2.1. The impact of the digital divide on the mental health of farmers

2.2. Research hypotheses

As previously discussed, existing research on the impact of internet use on residents’ mental health has yielded relatively divergent conclusions. Some scholars argue that internet application promotes residents’ mental health, while others hold the opposite view [42,43]. This study posits that due to China’s relatively late start in internet development and significant urban-rural disparities, the depth and breadth of internet technology application among rural residents in China remain to be improved. Therefore, we hypothesize that higher levels of internet use are conducive to improving the mental health of rural residents. Conversely, a deepening digital divide would negatively affect the mental health of rural residents in China. Specifically, the digital access divide restricts rural residents’ channels for information search, hindering their access to information beneficial for improving mental health. The digital use divide exacerbates differences in internet usage capabilities among individuals, preventing rural residents from leveraging the low-cost and high-efficiency advantages of internet technology, thus negatively impacting mental health. Additionally, excessive internet use may also adversely affect the mental health of rural residents. Based on this, we propose Hypothesis H1:

H1: All levels of the digital divide, including the digital access divide, digital use divide, and digital utility divide, will have adverse effects on the mental health of rural residents.

Self-Assessed Social Class Perception and Self-Assessed Economic Status Mechanisms.The most direct impact of widespread internet technology is that it significantly broadens residents’ social networks, enabling them to learn more about the economic and knowledge levels of individuals from different social classes, thereby prompting them to reassess their own social class. However, existing research has not reached a consensus on this effect. Some researchers adopt a positive stance, arguing that the internet provides an open and information-sharing platform for rural residents [44], where individuals often compare themselves with groups of lower social or economic status, which may enhance their self-assessed social class and economic status. Conversely, others argue that individuals tend to compare themselves with higher-status groups, potentially lowering their self-assessed social class or economic status [45].

Additionally, a series of studies have shown a strong association between mental health and social class. For example, Xu suggest that individuals of higher social class maintain better mental health because higher social status enhances their sense of social recognition, driving improvements in mental health [46]. Similarly, an improved self-assessed economic status also promotes better mental health. Therefore, we propose Hypotheses H2 and H3:

H2: All levels of the digital divide reduce the mental health of rural residents by lowering their self-assessed social class perception.

H3: Different levels of the digital divide reduce the mental health of rural residents by decreasing their self-assessed economic status perception.

Self-Assessed Sense of Fairness Mechanism.Scholarly consensus on how internet use influences residents’ perception of fairness remains elusive. Some scholars argue that internet use may negatively affect individuals’ sense of fairness. For instance, research has found that media in developed countries often propagate information about their own economic prosperity, which can confuse residents of developing countries and deepen their perception of social inequality and uneven development [47,48]. Conversely, others contend that internet penetration provides a platform for diverse viewpoints, conducive to enhancing individuals’ understanding of fairness and democracy Building on the arguments of Chen, Guo, and Benvenuti, who contend that governments can raise rural residents’ perceived sense of fairness by rigorously censoring media content, filtering negative online information, and actively guiding public opinion [49–51], the present study posits that the persistence of the digital divide is likely to exert a deleterious effect on rural residents’ sense of fairness.

Furthermore, research has found that social injustice increases residents’ depression levels or reduces their happiness, ultimately harming their mental health [52]. Thus, we propose Hypothesis H4:

H4: The digital divide reduces the mental health of rural residents by lowering their self-assessed sense of fairness.

Fig 1 outlines the research logic of this paper. As shown in Fig 1, this paper aims to investigate the impact of the digital divide at different levels on the mental health levels of rural residents. The digital divide primarily includes the digital access divide, the digital use divide, and the digital utility divide. On this basis, this paper seeks to explore whether self-assessed social class perception, self-assessed economic status, and self-assessed sense of fairness mediate the adverse effects of each level of the digital divide on rural residents.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The research logic of this article.

https://doi.org/10.1371/journal.pone.0341264.g001

3.1. Sources of data

The data for this study were selected from the 2020 and 2022 waves of the China Family Panel Studies (CFPS), a national, comprehensive, and large-scale research survey initiated by the Institute of Social Science Survey (ISSS), Peking University. By tracking data at the individual, household, and community levels, CFPS reflects changes in China’s society, economy, population, and health. The survey employs stratified sampling, with respondent samples covering 25 provinces/autonomous regions in China, ensuring relatively strong representativeness of the data. Since its launch in 2008, CFPS has completed 7 rounds of surveys as of 2024, with the most recent round conducted in 2022. Given the rapid development of the internet in China, to ensure the timeliness and reliability of research conclusions, this study uses data from the two most recent CFPS waves. Compared with earlier years, the 2020 and 2022 surveys include questions on the extent to which rural residents experience the digital utility divide, enabling a deeper exploration of the impact of the digital divide on the mental health of China’s rural residents, which is of special and profound significance for the investigation.The China Family Panel Studies (CFPS) has been approved by the Institutional Review Board of Peking University (approval No. IRB00001052-14010). All respondents provided written informed consent prior to data collection. The CFPS data are de-identified and publicly released by the Institute of Social Science Survey at Peking University. As this study uses only anonymized secondary data, no additional ethical review is required.

The initial target sample size for CFPS2022 was 16,000 households, using implicit stratified sampling and multistage equal probability sampling. Specifically, administrative regions were first selected as primary sampling units. Second, administrative villages were selected as sampling units based on the first step. Finally, households were selected as the terminal sampling units. During sampling, researchers constructed the sampling frame using the map address method and adopted cyclic sampling with random starting points to ensure the reliability of sampling results.

In our study, we carefully handled missing data and considered the use of sample weights to ensure the robustness and representativeness of our analyses. Missing data were addressed through stringent procedures to preserve data integrity. Specifically, we excluded observations that were not successfully matched across the 2020 and 2022 waves of the China Family Panel Studies (CFPS) and dropped cases with missing values on key independent or dependent variables, yielding a final analytic sample of 11,114 valid observations.

Regarding weights, the CFPS data set does not provide sampling weights; consequently, our analyses rely on unweighted data. We have evaluated the representativeness of our sample and consider it to adequately reflect the characteristics of rural Chinese residents, given that CFPS employs a stratified sampling design covering 25 provinces/autonomous regions. To mitigate potential biases arising from the absence of sampling weights and from item non-response, we control for a comprehensive set of individual- and household-level covariates and conduct extensive robustness checks. These measures jointly enhance the reliability and validity of our findings. Moreover, a number of published studies likewise omit weight adjustments and still obtain widely accepted results; in causal-inference settings such as ours, the influence of weighting is typically of secondary importance.

3.2. Psychometric properties of composite variables

Principal-component analysis (PCA) was used to construct composite indices for mental health, digital use, and digital utility (Table 1). The mental-health scale was derived from eight items; four principal components were retained, explaining 74.3% of the total variance (first component = 39.9%). Item loadings ranged from 0.24 to 0.43. Cronbach’s α = 0.80, indicating good internal consistency; mean inter-item correlation = 0.30 (range 0.17–0.52). The composite score was approximately normal (M = 0.00, SD = 1.00). The digital-use-divide index was derived from four items; three principal components were retained, explaining 86.2% of the variance (first component = 36.5%). Loadings ranged from 0.15 to 0.69. Cronbach’s α = 0.35, denoting modest reliability; mean inter-item correlation = 0.06 (range 0.04–0.11). The composite was roughly normal (M = 0.00, SD = 0.66). The digital-utility-divide index was derived from five items; the retained components explained 79.2% of the variance (first component = 36.5%). Loadings ranged from 0.26 to 0.52. Cronbach’s α = 0.73, indicating good internal consistency; mean inter-item correlation = 0.36 (range 0.16–0.50). The composite score was approximately normal (M = 0.00, SD = 0.98). Table 2 presents the detailed component loadings for each item. For mental health, “felt down” and “everything an effort” loaded positively on the first component, contributing to the overall score. For digital use, “online gaming” and “online learning” showed salient loadings on the first component. For digital utility, “importance for work” and “importance for entertainment” loaded on the first component. These patterns support the factorial validity of the indices, indicating that the items adequately capture the intended constructs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Psychometric properties of PCA-based composites.

https://doi.org/10.1371/journal.pone.0341264.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Loadings for PCA-based composites.

https://doi.org/10.1371/journal.pone.0341264.t002

3.3. Variables

3.3.4. Mediating variables.

Drawing on related literature and focusing on self-assessed social class perception, self-assessed economic status, and self-assessed sense of fairness, this study investigates the mechanisms through which the digital divide affects the mental health of rural residents in China [53,57].

Self-Assessed Social Class Perception:Respondents were asked, “On a scale of 1 to 5, how would you rate your social status in your local area?” A variable “class” was created to describe respondents’ self-assessed social class perception, with higher values indicating a higher perceived social class.

Self-Assessed Economic Status:Respondents were asked, “On a scale of 1 to 5, how would you rate your income level in your local area?” A variable “condition” was created to describe self-assessed economic status, with higher values indicating a higher perceived economic status.

Self-Assessed Sense of Fairness:Respondents were asked, “On a scale of 0 to 10, how would you rate the wealth gap in our country?” A variable “fair” was created to measure respondents’ self-assessed sense of fairness, with higher values indicating a higher perceived level of social fairness.

Descriptive analyses are presented in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Descriptive analysis.

https://doi.org/10.1371/journal.pone.0341264.t003

4.1. Benchmark results

Because several independent variables as well as the dependent variable are constructed with principal-component analysis and the data are the 2020–2022 CFPS panel, we estimate a two-way fixed-effects model that simultaneously controls for individual-specific and year-specific unobserved heterogeneity, thereby mitigating omitted-variable bias. The specification is:

(1)

In Equation (1), “Mental Health Levels” is the dependent variable, representing mental health status. It is derived from principal component analysis (PCA) of collected questions, resulting in a variable ranging from −7.7752 to 2.3401. This variable represents the mental health of individual i at time t, with higher values indicating better mental health. C is the constant term.

“digital divide_it” is the independent variable, representing the digital divide score for individual i at time t, which includes the digital access divide, digital use divide, and digital utility divide. αi is the individual-specific intercept that controls for the impact of individual characteristics on mental health. λ_t is the time-specific intercept that accounts for time-related influences on mental health. β represents the regression coefficient, and γ denotes the coefficients of the control variables. X_it refers to a set of control variables, and ε_it is the error term.

Table 4 shows the impact of the digital divide at different levels on the mental health of rural Chinese residents, with results presented in columns (1) to (3). As indicated in Table 2, the digital access divide has a negative effect on mental health at the 1% significance level, suggesting that rural residents without internet access have lower mental health levels. The digital use divide also shows a negative impact at the 1% level, indicating that less frequent internet use correlates with lower mental health. Similarly, the digital utility divide is significantly negative at the 1% level, implying that when rural residents perceive the internet as less important, their mental health is negatively affected. Thus, Hypothesis H1 is supported.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Benchmark regression.

https://doi.org/10.1371/journal.pone.0341264.t004

As shown in columns (1) to (3) of Table 4, among the control variables, variables like car ownership are consistently insignificant, while education and marriage have significant effects. Variables such as family size and housing are significant only in certain digital divide contexts.To sharpen the estimated impact of each digital-divide layer on rural residents’ mental health, Table 5 reports 95% confidence intervals and standardized coefficients. Columns (1)–(3) show that the 95% CIs for the digital access divide (−0.1959, −0.0237), the digital use divide (−0.1886, −0.0982) and the digital utility divide (−0.1468, −0.0703) are entirely below zero, underscoring the precision and reliability of the negative effects. Columns (4)–(6) present standardized coefficients that permit effect-size comparisons across metrics. The standardized beta for the digital access divide is −0.0537 (p < 0.05), indicating a medium effect. The corresponding values for the digital use divide and the digital utility divide are −0.0946 and −0.1070 (both p < 0.01), respectively. Thus, the digital utility divide exerts the largest relative influence, followed by the digital use divide, while the digital access divide shows the smallest standardized impact. These results imply that—although all three layers harm mental health—interventions that heighten rural residents’ perceived value of internet use (i.e., reduce the digital utility divide) are likely to yield the largest mental-health gains. Standardized coefficients therefore offer a clear empirical basis for prioritizing policies that enhance perceived internet utility alongside continued efforts to improve access and usage skills.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Digital divide’s impact on rural residents’ mental health (with 95% CIs & Std. Coeff.).

https://doi.org/10.1371/journal.pone.0341264.t005

4.2. Robustness tests

Robustness Test 1: Replacing the Dependent Variable. Instead of using principal component analysis (PCA) to measure mental health levels based on responses to questions like “How often have you felt down in the past week?”, the mental health level was measured using the average of responses to these questions, with results shown in Table 6.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Robustness tests.

https://doi.org/10.1371/journal.pone.0341264.t006

Robustness Test 2: Replacing the Independent Variable. The digital access divide was measured by the question “Do you use mobile devices to access the internet?”. The digital use and utility divides were measured using averages instead of PCA, with results presented in Table 6.

Robustness Test 3: Replacing Control Variables. Education was redefined on a scale from 1 (below secondary vocational school) to 6 (doctorate). The central and western regions were combined into a non-eastern region (coded as 0), with the eastern region coded as 1. Family size was redefined as 1 for families with ≤6 members and 2 for families with >6 members. Results are in Table 6.

Robustness Test 4: Replacing the Econometric Model. The two-way fixed-effects model was replaced with a random-effects model, and the results are shown in Table 6.

Robustness Test 5: Trimming Outliers. Samples in the top and bottom 5% of the age distribution were trimmed, with results presented in Table 6.

In summary, as shown in Table 3, the digital divide at all levels (access, use, and utility) remains significantly negatively associated with the mental health of rural Chinese residents, regardless of whether the dependent variable, independent variable, control variables, or econometric model are replaced, or whether outliers are trimmed. This confirms the robustness of the benchmark regression results in Table 6.

4.3. Endogeneity test

When exploring the impact of the digital divide on the mental health of rural Chinese residents, researchers must carefully address endogeneity issues to prevent systematic biases in research results.

First, the problem of omitted variables in model specification requires attention. Potential factors such as individual psychological qualities may simultaneously influence both the digital divide and mental health. However, due to researchers’ oversight, these factors are often ignored, ultimately leading to deviations between research results and reality. This study mitigates this issue by incorporating as many control variables as possible to reduce biases caused by omitted variables.

Second, measurement errors in data can also affect the robustness of final results. Respondents’ self-reports on internet usage and its consequences may contain biases stemming from cognitive differences, memory errors, or social desirability. Similarly, respondents’ evaluations of their own mental health levels may be inaccurate, prone to emotional fluctuations or social comparison effects. Effectively identifying and correcting measurement errors is a key problem that must be solved in empirical research.

Finally, the presence of bidirectional causality may exacerbate endogeneity. Bidirectional causality means that the digital divide not only affects the mental health of rural residents but also that their mental health status may in turn influence their evaluations of internet access, usage, and utility, impacting the accuracy of research results. To address this, this study employs the instrumental variable (IV) method. Drawing on research by Li and Peng, we use the mean value of the digital divide among the same age group in the respondent’s village as the instrumental variable [54,58]. For example, when testing the endogeneity of the digital access divide, we use the mean value of the digital access divide among the same age group in the respondent’s village as the IV. The theoretical justification for treating the village-level, same-age-cohort mean of each digital-divide dimension as an exogenous shock rests on the supply-side determinants of rural internet access. Experimental evidence shows that China’s “Universal Service” fibre-to-the-village programme allocated roll-out quotas to administrative villages by administrative decree, generating discontinuous jumps in network availability that are orthogonal to household-level unobservables [59]. Because the “last-mile” investment is negotiated by village committees and subsidised by county governments, the proportion of a specific age group within a village that remains unconnected, uses the internet infrequently, or attributes low utility to it is driven primarily by these exogenous infrastructure thresholds rather than by individual psycho-social traits. The cohort mean therefore captures a community-level supply shock that satisfies the relevance condition and is systematically unrelated to the idiosyncratic component of mental-health outcomes.

Second, social-norms and reference-group theory provide complementary support for the exclusion restriction. Zhang demonstrate that, once network coverage is in place, peer visibility within the village creates a threshold effect: adoption accelerates as soon as the peer-use rate crosses a critical point, independently of individual socioeconomic status [60]. Since beliefs about mental health in rural China are formed through local rather than cross-village comparisons, the village–age-group mean influences psychological well-being only by shaping an individual’s own digital access, intensity of use, and perceived utility. The three instruments thus isolate exogenous community-level variation in the digital access, use, and utility divides, enabling causal identification of each divide’s effect on mental health.

The IV results are reported in Table 7. After instrumenting, the digital access divide continues to exert a negative effect on rural residents’ mental health. In the first stage the F-statistic equals 723.72—far above the conventional threshold of 10—so the weak-instrument test is easily passed. The instrument is positively and significantly related to the access divide at the 1% level, confirming relevance. The Hausman statistic is 0.02 with a p-value of 1.00, indicating that the access divide can be treated as exogenous. The IV estimate remains negative and significant at the 10% level, corroborating robustness; the first-stage partial R² is 0.7013, showing that the instrument explains a large share of the variation in the endogenous regressor.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Endogeneity test.

https://doi.org/10.1371/journal.pone.0341264.t007

For the digital use divide, the first-stage coefficient is 0.9890 (significant at 1%), and the F-statistic is 718.68, again rejecting the weak-instrument hypothesis. The second-stage coefficient is negative and significant at 1%. The Hausman test yields 1.40 (p = 0.9992), so the use divide is also exogenous. The partial R² reaches 0.7425, reaffirming instrument strength.

Turning to the digital utility divide, the first-stage coefficient equals 0.8980 (significant at 1%), with an F-statistic of 615.12. The IV estimate is −0.0427 and significant at 10%. The Hausman statistic is 0.29 (p = 1.00), and the partial R² is 0.7534, demonstrating strong explanatory power.

Taken together, the instruments perform well: partial R² values of 0.7013, 0.7425 and 0.7534 imply a high correlation with the respective endogenous variables, while the large first-stage F-statistics and high Hausman p-values jointly establish instrument exogeneity and validity. Thus, the baseline findings are robust to endogeneity concerns: all three layers of the digital divide significantly undermine the mental health of rural Chinese residents.

4.4. Sub-dimension analysis

The disaggregated results in Table 8 show that the frequency with which rural residents play online games has no discernible effect on their mental-health score. By contrast, skills related to online learning, e-shopping, and the use of new-media platforms (short videos) are all positively and significantly associated with better mental health; respondents who report lower proficiency in these domains exhibit markedly poorer psychological well-being. Although online gaming has expanded rapidly and is widely recognised, rural residents—who on average have lower educational attainment and limited digital literacy—often lack the competencies needed to engage meaningfully with games; this plausibly explains the null finding for gaming frequency. The remaining dimensions of the digital-use divide are therefore more salient for rural residents, because they govern the ability to access information safely and conveniently and to improve daily living. Deficiencies in these practical skills translate directly into lower mental-health scores.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Sub-dimension analysis of the digital use divide.

https://doi.org/10.1371/journal.pone.0341264.t008

Table 9 turns to the sub-dimensions of the digital-utility divide. Perceived importance of the internet for “staying in touch with family and friends” is negatively associated with mental health at the 1% level: the higher the subjective value placed on online interpersonal contact, the worse the psychological outcome. Utilities attached to work, entertainment, learning and daily life, however, are statistically insignificant. A likely explanation is the socio-economic context of rural China: villages are relatively isolated and economically under-developed, prompting many working-age adults to migrate to cities. When the internet fails to deliver the expected relational benefits—i.e., when the utility divide in the social-connectivity domain is wide—feelings of loneliness intensify and mental health deteriorates. Lower average schooling and lagging local economic development may also attenuate the psychological relevance of internet utilities outside the interpersonal sphere, producing the non-significant results observed for the other domains.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Sub-dimension analysis of the digital utility divide.

https://doi.org/10.1371/journal.pone.0341264.t009

5.1. Heterogeneity analysis

Age Heterogeneity. Drawing on Zhao’s research, the sample was divided into two groups: young respondents (under 45 years old) and middle-aged and elderly respondents (45 years old and above) [61]. The results in Table 10 indicate that, among younger respondents, poorer mental health is driven primarily by the digital access and digital use divides. In sharp contrast, for middle-aged and elderly villagers the digital use divide alone exerts the most pronounced negative effect. This age-pattern likely reflects the life-course realities of rural China: older adults need a broader repertoire of online skills—job search, mobile banking, government-service portals—to secure or retain off-farm employment, so deficiencies in use directly curtail their labour-market opportunities and, in turn, depress psychological well-being. Younger villagers, while more digitally native, still face fundamental barriers to entry (device affordability, village-level fibre availability); once on-line, they demand advanced skills for study, social media and e-commerce. Hence inadequate access and low usage intensity reduce their information-gathering efficiency and generate the observed mental-health penalty.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 10. Heterogeneity analysis.

https://doi.org/10.1371/journal.pone.0341264.t010

Gender Heterogeneity. Following Li’s (2025) methodology, all respondents were divided into two groups by gender for in-depth analysis [55]. The results show that men’s mental health is impaired mainly by the digital use and digital utility divides, whereas women are adversely affected across all three layers. Along both the access and use dimensions, the negative mental-health effect is significantly larger for women; along the utility dimension, the effect is larger for men. These patterns plausibly reflect regional disparities in economic development and gendered educational expectations. Women, who on average face lower access to and fewer opportunities for internet use, tend to focus on its benefits and are less aware of potential downsides; hence each additional barrier to connectivity or usage translates into a sharper psychological penalty. Men, enjoying greater access and usage experience, are more attuned to the risks and limitations of online activity, so the perceived (low) utility of the internet exerts a stronger negative influence on their well-being.

Educational Heterogeneity. Referencing Lu’s research, the sample was divided into two groups based on educational attainment: “junior college or higher” (higher education) and “below junior college” (lower education) [9]. The mental-health penalty of the digital divide is concentrated among the low-educated, for whom access and use divides reduce information retrieval efficiency and job access. Among the highly educated, the utility divide predominates, constraining perceived advancement opportunities and thereby generating sharper psychological losses.

Regional Heterogeneity. Following Yang’s research, respondents were divided into two groups by province: eastern region (residing in eastern China) and non-eastern region (central and western China) [8], with results presented in Table 10. Empirical findings reveal marked regional heterogeneity in the relationship between the digital divide and rural residents’ mental health. Specifically, the digital access divide exerts a statistically significant negative effect on rural residents in eastern China (coefficient equals negative 0.145, p value less than 0.05). This result is consistent with the region’s advanced economic development and high internet penetration, conditions that render connectivity indispensable for schooling, employment, and daily life; exclusion therefore constitutes a barrier to opportunity and, consequently, a threat to psychological well-being. In non-eastern regions—where internet infrastructure is less developed and online services are not yet integral to every domain—the access divide does not significantly influence mental health. Nevertheless, among non-eastern residents who are connected, both the digital use divide and the digital utility divide continue to impede work efficiency and life convenience, producing inferior mental-health outcomes.

Further disaggregation shows that the use and utility divides operate differently across regions. In the east, the estimated coefficient of the digital use divide is negative 0.191 (p value less than 0.001) and that of the digital utility divide is negative 0.105 (p value less than 0.001), indicating that even when infrastructure is available, inability to exploit digital resources and low perceived importance of the internet both depress psychological states. In non-eastern areas the corresponding coefficients are negative 0.116 and negative 0.111 (both p values less than 0.001), demonstrating that these divides are equally harmful. Lower average schooling and slower economic growth in non-eastern regions likely magnify the damage, because residents possess fewer skills with which to convert digital access into tangible benefits.

5.2. Mechanism testing

In this subsection, drawing on the mechanism testing methods proposed in prior studies, we examine the mechanisms through which the digital divide affects the mental health of rural residents in China [53,57]. Specifically, we analyze the mediating roles of Self-Assessed Social Class Perception, Self-Assessed Economic Status, and Self-Assessed Sense of Fairness. The detailed results are presented in Table 11.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 11. Mechanism Test.

https://doi.org/10.1371/journal.pone.0341264.t011

Panel A of Table 11 examines whether self-assessed social-class perception mediates the effect of the digital divide on the mental health of rural Chinese residents. Column (1) shows that the digital access divide exerts a significant negative effect on mental health (coefficient = −0.110). Yet column (2) indicates that the access divide does not significantly predict self-assessed social class; hence, no mediating role is detected. Turning to the digital use divide, column (4) reports a significant negative coefficient of −0.143 on mental health. Column (5) reveals that the use divide also significantly lowers self-assessed social class. After controlling for the latter in column (6), the coefficient on mental health remains significant and only marginally attenuates to −0.142, implying that the use divide harms mental health partly by depressing perceived social status.

Finally, column (7) documents that the digital utility divide significantly reduces mental health (coefficient = −0.109, p < 0.01). Column (8) shows the same divide significantly lowers self-assessed social class. Introducing the mediator in column (9) leaves the negative effect on mental health virtually intact (coefficient = −0.107), confirming that the utility divide operates, at least partially, through diminished perceptions of social standing.In summary, the digital usage and utility divides influence mental health by altering Self-Assessed Social Class Perception, while the digital access divide does not. Thus, Hypothesis H2 is partially supported.

Panel B of Table 11 examines whether Self-Assessed Economic Status mediates the relationship between the digital divide and mental health. Columns (7) to (9) show that the digital utility divide has a significantly negative effect on mental health. Specifically, column (7) indicates a coefficient of −0.109, while columns (8) and (9) demonstrate that the digital utility divide significantly reduces Self-Assessed Economic Status. When controlling for Self-Assessed Economic Status, the coefficient of the digital utility divide on mental health remains robust, decreasing slightly to −0.107. This suggests that the digital utility divide negatively affects mental health by reducing Self-Assessed Economic Status. In contrast, the digital access and usage divides do not exhibit this mediating effect. Therefore, Hypothesis H3 is partially supported.

Panel C of Table 11 explores whether Self-Assessed Sense of Fairness mediates the relationship between the digital divide and mental health. Columns (1) to (3) show that the digital access divide negatively affects mental health, with a coefficient of −0.110 in column (1). Columns (2) and (3) further indicate that the digital access divide significantly reduces Self-Assessed Sense of Fairness. When controlling for Self-Assessed Sense of Fairness in column (3), the coefficient of the digital access divide on mental health remains negative and significant, decreasing slightly to −0.108. This demonstrates that the digital access divide negatively impacts mental health by reducing Self-Assessed Sense of Fairness. The digital usage and utility divides, however, do not exhibit this mediating effect.

In sum, the digital access divide undermines the mental health of rural Chinese residents primarily by eroding their perceived sense of fairness; the digital use divide does so by diminishing their self-assessed social-class position; and the digital utility divide exerts its adverse effect through both lowered self-assessed social class and reduced self-assessed economic status.

Several mechanisms underlie these patterns.First, insufficient connectivity curtails access to information and online social participation, fostering doubts about distributive justice and, consequently, psychological distress.

Second, as labour markets increasingly reward digital competencies, the inability to use the internet productively restricts employment opportunities, prompting negative evaluations of one’s own social standing. Third, perceiving the internet as unimportant intensifies feelings of social exclusion: the widening status gap implied by the utility divide breeds pessimism about one’s relative position, while diminished capacity to engage in e-commerce or digital entrepreneurship constrains income growth and lowers perceived economic status, ultimately damaging mental well-being [62,63].

7.1. Policy implications

The results of this study hold significant value for the advancement of the Digital China strategy and public health interventions. The Digital China strategy aims to achieve high penetration and high-quality development of the Internet. The heterogeneity analysis in this study demonstrates the impact of the digital divide on the mental health of different groups of rural residents, which will provide references for the precise advancement of the Digital China strategy. For example, the focus of high penetration development should be shifted to the central and western regions or middle-aged and elderly groups. The focus of high-quality development should be on rural areas in the eastern region or rural residents with higher education levels, thereby reducing the waste of efficiency. As an important part of public health interventions, the response to mental health issues is also of great significance for its optimization. The results show the differences in the mechanisms through which different levels of the digital divide affect the mental health of rural residents, which will provide fundamental intervention pathways for public health interventions to alleviate the mental health problems of digital vulnerable groups such as rural residents, thereby enhancing the effectiveness of interventions.

The results of this study also provide certain value enlightenment for grassroots policy-making. The results show the differences in the impact of the digital access divide, digital use divide and digital utility divide on the mental health of different groups of rural residents, which is conducive to the local governments to propose specific and targeted measures for different groups of rural residents.

Age-Specific Interventions. For rural youth,the strategy should involve expanding mobile internet coverage in remote areas and integrating digital skills training into pre-employment services. This includes incorporating skills like online marketing, remote work software, and short-video creation into training programs, and establishing a digital skills certification system to link skill acquisition with better job opportunities. This approach can enhance digital engagement and mitigate the psychological harm caused by access and use divides. For middle-aged and elderly rural residents, emphasizing the internet’s role in maintaining intergenerational connections and reducing caregiving burdens can alleviate their resistance to digital technology and the psychological impact of the use divide.

Gender-Specific Interventions.For women, efforts should focus on breaking down barriers to internet access and use, providing sustained digital tool training to reduce the psychological impact of these divides. For men, the emphasis should be on addressing the adverse effects of potential internet risks, enhancing their recognition of the internet’s value and safety, and thereby mitigating the negative impact of the digital utility divide.

Education-Specific Interventions.For rural residents with lower education levels, promoting basic digital skills training can help overcome cognitive barriers and reduce the psychological impact of the digital divide. For those with higher education, advanced digital skills training, such as digital marketing analysis, should be promoted to create more high-quality job opportunities and prevent skill-demand mismatches.

Region-Specific Interventions.In eastern rural areas, improving internet speed and stability while increasing the frequency of digital applications in daily life is essential. In central and western rural regions, priority should be given to ensuring basic internet access and optimizing base station construction and signal strength, along with long-term plans for digital skills enhancement. These measures can effectively reduce the adverse effects of the digital divide.

7.2. Academic implications

First, academic research should be closely integrated with societal realities, focusing not only on theoretical inquiry but also on practical relevance when addressing research questions. When studying factors affecting the mental health of rural residents, researchers should prioritize the representativeness and generalizability of sample selection. When examining the impacts of the digital divide, dividing it into hierarchical dimensions using reasonable criteria can yield more detailed and precise conclusions, helping to reduce sampling errors and ensure the robustness of findings [6].

Additionally, as internet technology continues to develop, the scope and content of the digital divide have evolved into a complex and diverse field [14]. Academic research must keep pace with these changes by incorporating relevant influencing factors and optimizing existing research frameworks, which is crucial for conducting comprehensive discussions.

Finally, there is a need for thorough research on how the digital divide affects the mental health of rural Chinese residents, including investigations into differential impacts across different age groups, regions, and occupations. Exploring potential mediating mechanisms will provide support for problem-solving and dynamically understanding the issue.