The power of Internet: from the perspective of women's bargaining power

Amidst China's rapid socio-economic advancement, women have gained substantial sway in household bargaining processes, evidenced by high involvement in decisions on loans, business, and house purchases and a rise in financial decision-making authority. Simultaneously, Internet adoption has surged (1.079 billion users; 76.4% penetration in 2023). While prior research has linked women's bargaining power to self-esteem, education, property rights, and income, the role of technological factors like the Internet remains underexplored. This study asks whether the Internet causally improves women's intra-household bargaining power and through what mechanisms. Using the 2010 National Survey on Chinese Women's Social Status (NSWSS) and the 2014 China Family Panel Studies (CFPS), the paper situates Internet use as a resource within Kabeer's resource-agency-achievement framework, with shifts in bargaining power reflecting agency. It applies instrumental-variable methods to identify causal effects and causal mediation analysis to quantify mechanisms via gender beliefs, employment, and income. The study finds that Internet use significantly enhances women's bargaining power overall and particularly in rural areas. Contributions include: integrating Internet use into women's empowerment frameworks; identifying causal effects and decomposing direct/indirect effects to inform precise policy; and offering implications for developing countries amid digital transformation. The paper proceeds with literature review and hypotheses, data and methods, empirical results, and conclusions with policy implications.

The literature on intra-household decision-making has shifted from Becker's unitary model, which assumes pooled resources and a single household utility, to collective models (Apps and Rees; Chiappori) that recognize heterogeneous preferences and bargaining based on individual resource endowments. Empirical work generally supports collective models and resource endowment theories, highlighting income, employment, and occupation as key determinants of bargaining power, though some studies note backlash or resistance that can mute benefits from women's resource gains. Cultural norm theory emphasizes external constraints—gender norms, religion, social norms—that shape bargaining dynamics; patriarchal norms can limit women's welfare gains despite improved fallback positions and may even induce backlash (e.g., domestic violence) during empowerment gains. Conceptually linking Internet use to bargaining power, the paper posits three pathways: (1) employment—Internet expands job opportunities, remote work, networking, and entrepreneurship; (2) income—Internet facilitates higher earnings via expanded labor markets, e-commerce, skill upgrading, and flexibility; and (3) gender beliefs—exposure to diverse perspectives, online communities, education, role models, and activism fosters more egalitarian attitudes. Hypotheses: H1, Internet use enhances women's bargaining power; H2, employment status, income level, and gender beliefs mediate this effect.

Data: Primary analysis uses NSWSS 2010, a nationally representative cross-sectional survey of Chinese adults (excluding Hong Kong, Macau, Taiwan) covering women's health, education, economic status, marriage/family, legal awareness, and gender beliefs. The sample is restricted to married, non-retired women; after handling missing values, N = 5408. For robustness, CFPS 2014 adult, family relations, and family economy data are used, linked to CFPS 2010 for variables like years married and parents' education; after cleaning, N = 5751. Outcomes: Two measures of women's bargaining power. (1) Actual_power (subjective): 0 = husband holds authority, 1 = joint negotiation, 2 = wife holds authority. (2) Bargain_power (objective): sum of six decision domains (daily expenditure; staple purchases; house purchase/construction; production/business; investment/loans; children's education/schooling), each coded 0 = husband dominates, 1 = nearly equal, 2 = wife dominates; total 0–12. Treatment: Internet_use (NSWSS): frequency of daily Internet use coded 0 (never) to 4 (3–8 hours), excluding the >8 hours category (0.4% of sample). In CFPS, Internet_use_cfps is binary (use vs. not use). Mediators: Employment (1 if engaged in paid work/labor), Income (log[annual income + 1], including labor, rental, property income, transfers), and Gender_belief (NSWSS): index from nine items on gender equality scored to yield 12–36 (higher = more egalitarian). Community-level gender beliefs (Gender_belief_community) proxy local norms via leave-one-out averages. Controls: Individual (urban hukou, property ownership, education years, age), family (spouse property, educational and age differences, years married, spousal support index, natal family economic advantage, parents' education), and community (gender-belief norms). CFPS adds interpersonal relationships, family harmony emphasis, and community family-ethics norms; definitions provided in the paper. Models: Baseline ordered logit and OLS regressions: Actual_power = β11*Internet_use + Xβ12 + ε; Bargain_power = β22*Internet_use + Xβ23 + ε. To address endogeneity (reverse causality and omitted variables), instrumental-variable ordered probit (IV-oprobit) is used. Instrument: interaction of (province-level fixed telephones per 10,000 and post offices per 10,000 in 1984) with the 2010 community Internet penetration rate, capturing historical telecommunications infrastructure and peer effects; argued to satisfy relevance and exogeneity. First-stage strength assessed via Cragg-Donald/Wald F. Mediation: Counterfactual causal mediation analysis (Imai-Keele-Tingley framework) estimates ACME and ATE for each mediator with sequential ignorability, using: M = α1 + θ1*Internet_use + γ1 X + μ1; Y = α2 + θ2*Internet_use + ρ*M + γ2 X + μ2; then computes δ(1) = Y(1, M(1)) – Y(1, M(0)). Heterogeneity: IV-based estimates stratified by urban vs. rural. Robustness: CFPS IV-oprobit using the same instrument; outcome Bargain_power_cfps (sum over five decision domains: expenditure allocation, savings/investments/insurance, housing, childrearing, high-value durables; 5–15). Provincial fixed effects included throughout; robust standard errors reported.

- Descriptive: In NSWSS (N=5408), women exhibit relatively lower bargaining power (Actual_power mean 0.808; Bargain_power mean 5.577). Internet_use mean 0.592 indicates low/infrequent use. - Baseline associations (NSWSS): Internet_use positively associated with both outcomes. OLS coefficients: +0.028 (SE 0.010) for Actual_power; +0.082 (SE 0.033) for Bargain_power, controlling for individual, family, community covariates and province effects. Ordered logit results are consistent. - Endogeneity correction (NSWSS IV-oprobit): Instrument strongly predicts Internet_use (first-stage coefficient 1.244, SE 0.155; Cragg-Donald Wald F ≈ 63.5). Second-stage shows positive, larger effects of Internet_use on Actual_power and Bargain_power than baseline, suggesting baseline underestimation and reflecting a local average treatment effect. - Sub-domains (NSWSS): IV estimates show significant positive effects of Internet_use on five of six decision domains: daily expenditure (+0.472, SE 0.134), staple purchases (+0.304, SE 0.153), housing decisions (+0.383, SE 0.151), production/business (+0.308, SE 0.154), investment/loans (+0.347, SE 0.156); children's education/schooling is positive but not significant (+0.246, SE 0.162). - Mediation (NSWSS): All three mediators significantly transmit effects. For Actual_power: ACME shares ≈ 3.21% (employment), 9.21% (income), 9.92% (gender beliefs). For Bargain_power: 2.96% (employment), 10.55% (income), 12.95% (gender beliefs). Average shares reported in the paper: ≈ 3.085% (employment), 9.88% (income), 11.435% (gender beliefs). - Heterogeneity: Effects are concentrated in rural areas; urban coefficients are small and not significant. Rural IV estimates: Internet_use significantly increases Actual_power (+1.001, SE 0.233) and Bargain_power (+0.762, SE 0.263). - Robustness (CFPS, N=5751): Instrument strong (first-stage Internet_use_cfps coefficient 0.362, SE 0.069; Cragg-Donald Wald F ≈ 26.9). Second-stage confirms significant positive effect of Internet use on women's comprehensive bargaining index. Sub-dimensions all significant and positive: expenditure allocation (+1.976, SE 0.482), savings/investments/insurance (+2.238, SE 0.407), housing (+1.159, SE 0.683, p<0.10), childrearing (+2.006, SE 0.480), high-value durables (+1.377, SE 0.632).

The findings provide causal evidence that Internet use enhances women's intra-household bargaining power. This addresses the research question by demonstrating that technology—beyond traditional determinants like education and income—operates as a resource that shifts women's agency within households. Mechanistically, the Internet improves employment prospects and income opportunities and fosters more egalitarian gender beliefs; each channel contributes significantly to the total effect, with gender beliefs playing the largest role. The stronger effects in rural areas reflect greater marginal exposure to new information and gender-equality norms compared to urban areas, where such norms are already more established. Instrumental-variable results mitigate concerns about reverse causality and omitted variables, and larger IV estimates suggest baseline attenuation and a local average treatment effect among compliers induced by historical telecom infrastructure and community penetration. Robustness across two independent national datasets and across multiple decision domains underscores the general validity of the results within the Chinese context. These findings highlight the Internet as a potent lever for shifting intra-household dynamics and improving women's welfare.

Using nationally representative data (NSWSS 2010; CFPS 2014), the study shows that Internet use significantly strengthens women's bargaining power in household decisions. Addressing endogeneity via an IV-oprobit framework confirms a causal effect, which operates through increased employment, higher income, and more egalitarian gender beliefs. Effects are especially pronounced in rural areas, consistent with larger informational gains and shifts in attitudes. Policy implications include: (1) improving Internet infrastructure and affordability, particularly in rural regions, including subsidized access or community Internet centers; (2) ensuring access to high-quality online educational and informational resources that support skill development and gender equality; and (3) promoting Internet-based industries and digital-skills training to expand compliant, income-generating opportunities for women. Recognizing the regulatory environment, the extent of empowerment is conditioned by Internet governance and content accessibility; policies that enable open access to empowering content can amplify gains. Future research should leverage newer waves of data to track evolving dynamics and assess external validity across diverse cultural contexts.

- Timeliness of data: Main analyses use NSWSS 2010 and CFPS 2014; results may not fully capture current conditions amid rapid digital development. The authors note gradual evolution in family dynamics, but call for updated surveys for contemporary validation. - Generalizability: Findings are derived from China; family power dynamics are culturally contingent. Applicability to other cultural contexts, especially beyond East Asia, may be limited. - Policy environment: China's Internet governance (e.g., censorship) could constrain access to information and platforms that facilitate empowerment, potentially moderating effects. This contextual factor complicates extrapolation and underscores the importance of regulatory environments. - Cross-sectional limitations in NSWSS restrict dynamic inference; although IV and mediation analyses address causality and mechanisms, unobserved factors or violations of sequential ignorability cannot be entirely ruled out.