Internet Use, Market Transformation, and Individual Tolerance: Evidence from China

The study examines whether and how Internet use affects individuals' tolerance in China, where Internet penetration rose from 45.8% in 2013 to 75.6% in 2022. While prior work has explored links between media use and tolerance largely outside China, little is known about these relationships within China or about moderating and mediating mechanisms in this context. The research asks: Does Internet use positively or negatively affect public tolerance in China? If there is an effect, what moderating mechanisms explain it? Using the Chinese Social Survey (CSS) 2015–2021 and matched provincial statistics, the paper situates the question within China's market transformation, proposing that divisions in university education, labor market (within-system danwei vs. outside-system), and household registration (Hukou) may moderate Internet use’s impact on tolerance. It further posits that perceptions of fairness (opportunity and outcome fairness) may act as mediating or suppressing mechanisms. The study contributes by focusing on China’s context and by testing moderation and fairness mechanisms linking Internet use to changes in social attitudes.

The literature delineates types of tolerance—political, social, and moral (Vogt)—and their determinants. Political tolerance is influenced by demographics (age, education), cognitive ability, political knowledge, social networks, and context. Social and moral tolerance are affected by social capital, institutions, macroeconomic conditions, and prevailing social attitudes. Education, as cultural capital (Bourdieu), is central to tolerance formation through cognitive and multicultural exposure. Media effects research indicates that media framing shapes attitudes (Lippmann); exposure to negative content can induce distrust (Robinson & Appel; Putnam), whereas positive development information can bolster government trust (Li). Internet media differs from traditional media in content diversity and user-generated production, potentially broadening exposure and, in China’s regulated environment, emphasizing positive social values. Yet misinformation can erode trust (Kraut et al.) and rumor dynamics can manipulate public sentiment (Na et al.), underscoring the role of governance and gatekeeping. Prior cross-national studies have reported mixed effects of Internet use on tolerance, motivating examination within China’s unique market transformation context. The study advances hypotheses: H1—higher Internet use predicts greater tolerance; H2—effects are moderated by divisions in university education, labor market (system vs. non-system), and Hukou; H3—effects involve fairness perceptions (opportunity and outcome) as potential mechanisms.

Data: The study uses Chinese Social Survey (CSS) data (2015–2021) covering adults 18–69 across 31 provinces, merged with China Statistical Yearbook data on provincial Internet indicators (2015–2021; note: provincial Internet coverage not available for 2019 and 2021). After cleaning, N = 33,547. Measures: - Dependent variables (tolerance): From CSS items asking whether respondents can accommodate (1) premarital cohabitants, (2) LGBTQ+ groups, (3) people begging for food/money, (4) persons released from prison, (5) persons with different religious beliefs. Responses: very unacceptable, less acceptable, more acceptable, very acceptable. Factor analysis yields two dimensions: Moral Tolerance (MT: premarital cohabitation, LGBTQ+) with KMO 0.572, Bartlett p<0.000; Social Tolerance (ST: beggars, ex-prisoners, different faiths) with KMO 0.696, Bartlett p<0.000. Item scores are averaged per factor and normalized to 0–100 (higher = more tolerant). - Independent variable (Internet use): (1) Binary use (use_d: yes=1, no=0). (2) Frequency of using Internet to browse current political news and participate in online discussions (scale: never=1 to almost every day=6). For robustness, a 0–6 composite is formed with non-users coded 0 and users retaining 1–6 (use_c). - Moderator variables (market transformation divisions): (a) University education completion (edu_dum: college or above=1, else=0); (b) Employment within system (system: government/army/state-owned/collective=1; otherwise=0); (c) Hukou (urban/non-agricultural=1, rural/agricultural=0). - Mediator variables (fairness): Perceived social fairness measured by 8 CSS items (college entrance exam, political rights, law enforcement, employment opportunities, public health, wealth/income distribution, social security/pensions, urban–rural rights). Rated 1–5 (very unfair to very fair). Aggregated into Opportunity Fairness (OpF: exam, political rights, law enforcement, employment) and Outcome Fairness (OcF: healthcare, wealth/income distribution, social security, urban–rural rights) by averaging item scores. - Controls: Gender, age, CPC membership, family assets (log), Hukou locality, rural/urban residence, survey year, and survey region (east/middle/west). Analytic strategy: - Baseline estimation: OLS regressions for MT and ST; Seemingly Unrelated Regression (SUR) to account for correlated disturbances across MT and ST equations. - Moderation testing: Include interaction terms between Internet use (use_d) and each division variable (edu_dum, system, hukou) in separate models for MT and ST; visualize margins. - Mediation/suppression testing: Three-step regressions: (1) tolerance on Internet use; (2) fairness (OpF, OcF) on Internet use; (3) tolerance on Internet use and fairness simultaneously; compare Internet use coefficients to assess mediation vs suppression. - Endogeneity: Addressed by Instrumental Variables (IV) using 2015 and 2017 subsamples due to data availability. Instruments: provincial Internet coverage (Int_cov), mean Internet user ratio in respondent’s community (M_c_users), regional inclusive finance index (Fin_index). Tests include Hausman, first-stage relevance, Sargan underidentification, Hansen J overidentification, weak-IV F-tests. Also apply Propensity Score Matching (PSM) with 1:1 matching, radius, and kernel matching to compare treated (Internet users) vs controls (non-users) and assess balance (standardized deviations).

- Descriptives: Mean MT = 51.93, mean ST = 31.645 (0–100 scale). 49.1% Internet users; 18.2% college-educated; 10.5% CPC; 33.7% rural Hukou; 10.7% within-system employment. From 2015 to 2021, ST trends upward; MT declines then rises and remains higher than ST each year. - Baseline effects (OLS): Internet users vs non-users have higher MT (β=2.115, p<0.01) and ST (β=0.871, p<0.01). Higher Internet use frequency associates with higher MT (β=0.403, p<0.01) and ST (β=0.145, p<0.05). - SUR: Confirms positive effects; correlation between MT and ST disturbances ρ=0.265 (p=0.000). Equality of coefficients across equations rejected (χ²=10.95, p<0.01), indicating differential impacts on MT vs ST. - Controls (selected): College education strongly raises MT (β≈4.199, p<0.01) and ST (β≈11.562, p<0.01). Age reduces MT (β≈−0.157, p<0.01) and ST (β≈−0.358, p<0.01). CPC membership lowers ST (β≈−3.028, p<0.01). Urban (non-agri) Hukou increases MT (β≈0.896, p<0.01) and ST (β≈1.015, p<0.01). Within-system employment lowers ST (β≈−2.052, p<0.01). Higher family assets increase MT and ST. - Endogeneity (IV, 2015 & 2017): First stage shows M_c_users (≈0.220–0.221, p<0.01) and Fin_index (≈0.005, p<0.01) positively predict Internet use; Int_cov negative small effect (−0.003 to −0.004, p<0.10). Tests support instrument validity: Sargan underidentification χ²≈86.94–89.36 (p=0.000); weak-IV Wald F≈29.76–30.94 (p=0.000); Hansen J ≈22.68–28.95. Second stage: Internet use increases MT (β=2.202, p<0.10) and ST (β=3.166, p<0.05), supporting H1. - PSM: Before matching, treated–control gaps: MT 5.133 (t=23.51***), ST 9.694 (t=38.23***). After matching: MT differences remain significant (≈2.17–2.991 across methods), ST differences remain significant (≈2.849–3.691), confirming robustness of positive effects. - Moderation (interaction models): For ST, all three interactions with use_d are positive and significant at 1%: use_d×edu_dum β=7.331*** (SE=0.900); use_d×system β=2.328*** (SE=0.885); use_d×hukou β=4.297*** (SE=0.527). For MT, interactions with education and system are not significant; use_d×hukou shows a negative effect (β=−1.135**, SE=0.464). Thus, market transformation divisions moderate the Internet–ST relationship (H2 supported for ST; limited for MT). - Mediation/suppression (fairness): Internet use frequency reduces perceived fairness: OpF β=−0.021*** (SE=0.002); OcF β=−0.032*** (SE=0.003). Higher fairness predicts higher tolerance: MT on OpF β=1.577***; MT on OcF β=1.386***; ST on OpF β=0.849***; ST on OcF β=1.312***. When fairness is included with Internet use, the Internet use coefficients increase relative to baseline (suppression), indicating that OpF and OcF function as suppressors rather than mediators. H3 not supported as mediation; supported as suppression effect. - Overall: Internet use significantly and robustly increases both MT and ST among Chinese adults; effects are stronger/differential across social divisions for ST, and fairness perceptions suppress (mask) part of the positive association.

The findings address the core research question by demonstrating that Internet use in China is positively associated with both moral and social tolerance, even after accounting for endogeneity through IV and PSM approaches. The stronger and systematically moderated effects on social tolerance highlight the role of China’s market transformation—education, labor market segmentation, and Hukou—in shaping how digital exposure translates into acceptance of diverse groups and behaviors. Internet use appears to narrow some gaps (e.g., elevating social tolerance among within-system employees) while widening others (e.g., larger ST gap by education and between urban/rural Hukou). The suppression role of fairness suggests that exposure to inequality-related information may lower perceived fairness, which would otherwise dampen tolerance; controlling for fairness reveals a larger positive Internet effect. In contrast to some cross-national studies reporting negative associations, the Chinese context—characterized by rapid infrastructure expansion, broader access, and governance of online content—may facilitate diversified information access and foster more tolerant attitudes. These results underscore the relevance of digital technologies to social mentality and high-quality development, indicating that shifts in social attitudes accompany China’s digital transformation.

Using CSS 2015–2021, the study concludes: (1) Social tolerance rose steadily over time, while moral tolerance dipped then rose, remaining higher than social tolerance each year; (2) Internet use exerts a significant positive effect on both moral and social tolerance, robust to SUR, IV, and PSM; (3) Market transformation divisions moderate the Internet–tolerance link for social tolerance—Internet use reduces the impact of labor market segmentation by raising ST among within-system employees, but widens ST gaps by education level and between urban and rural Hukou; these moderation patterns do not fully extend to moral tolerance; (4) Opportunity and outcome fairness act as suppressors, not mediators, in the Internet–tolerance relationship. The work clarifies digital technology’s role in social attitude change and suggests that policymakers should attend to evolving social attitudes alongside promoting high-quality development.

- The study relies on multi-period cross-sectional survey data (CSS 2015–2021), which may leave residual endogeneity due to omitted variables or measurement error despite controls, IV, and PSM. - Provincial Internet coverage data were unavailable for 2019 and 2021, restricting IV analyses to the 2015 and 2017 subsamples. - The construction of MT and ST indices is based on specific item groupings; alternative operationalizations might yield different factor structures. - Internet use measures emphasize political news browsing and online discussions; other forms of Internet engagement were not explicitly modeled.