The research team, led by Sanjiv Singh of the University of Petroleum and Energy Studies together with Jayantika Chakraborty and Sampurna Kundu, drew on the fifth round of the National Family Health Survey, conducted between 2019 and 2021. After applying the survey’s complex sampling weights and restricting the analysis to ever-married women aged 15 to 49 who completed the domestic violence module, the final weighted sample comprised 53,965 women. The outcome measure was a composite binary indicator capturing whether a respondent had experienced at least one act of physical, sexual, or emotional violence from her current or most recent husband at any point during the marriage, following the World Health Organization’s definition of intimate partner violence.

At the national level, the prevalence of intimate partner violence stood at 31.40 percent, consistent with the survey’s headline figure that 31.3 percent of married women aged 18 to 49 have experienced spousal violence. But the geographic contrast was dramatic. Kerala, India’s most educated state, reported a prevalence of roughly 12 percent, while Bihar, among the least educated, reported 42.6 percent. That nearly fourfold gap between two states within a single democracy frames the study’s central question: how much of the difference can be explained by the characteristics of women themselves, and how much by the normative and institutional environments they inhabit?

The statistical architecture of the study was correspondingly layered. Binary logistic regression estimated the associations between education, employment, and empowerment and the odds of reported violence, with interaction terms testing whether empowerment moderated those associations. Empowerment was measured using the SWPER global index, a validated tool for low- and middle-income countries built on three domains: attitudes toward wife-beating, social independence, and decision-making autonomy, each classified as low, medium, or high. The model showed adequate explanatory strength, with a likelihood ratio chi-square of 8677.36 and a pseudo R-squared of 0.1292. To probe the Kerala-Bihar gap, the team applied Fairlie decomposition, a nonlinear extension of the Blinder-Oaxaca technique that partitions a prevalence difference into the contributions of individual covariates. Generalized structural equation modelling added a test of whether empowerment mediates the pathway from education to violence.

The regression results confirmed the protective gradient for education. Compared with women who had no formal education, those with primary, secondary, and higher education had progressively lower odds of reporting intimate partner violence, with adjusted odds ratios of 0.83, 0.74, and 0.58 respectively. Husbands’ education mattered too: women whose husbands held higher education had 26 percent lower odds of reporting violence nationally. But in Bihar, women’s own education lost statistical significance, a pattern the authors read as evidence that the value of a structural resource depends on the normative conditions surrounding it. In Kerala, where female literacy is near-universal, education could not even be retained in the state model.

Employment told a different and darker story. Working women had higher odds of reported violence in every setting: non-working women had 13 percent lower odds nationally, 24 percent lower in Bihar, and 38 percent lower in Kerala. The pattern echoes what researchers call the male backlash hypothesis, in which a woman’s economic activity is perceived as a challenge to patriarchal authority and provokes retaliatory control. Earlier work in urban India had identified both a backlash effect and a so-called female guilt channel, in which working women are more likely to justify violence against themselves. The new analysis also found that when husbands were unemployed, their wives reported 22 percent lower odds of violence nationally, underscoring that the employment-violence link is relational rather than simply individual.

Two risk factors towered over everything else. Husbands’ alcohol consumption more than doubled the odds of intimate partner violence in all three settings, with adjusted odds ratios of 2.45 nationally, 2.58 in Bihar, and 2.44 in Kerala; among employed husbands who drank, the odds approached four times the reference nationally and reached 6.39 in Bihar. Childhood exposure to inter-parental violence was nearly as potent: women who had witnessed violence between their parents were more than three times as likely to experience partner violence themselves, with odds ratios of 3.09 nationally, 3.36 in Bihar, and 3.98 in Kerala. These findings highlight the intergenerational transmission of violence and suggest that substance-use interventions and early childhood programs could serve as critical entry points for prevention.

The empowerment results were the study’s most puzzling and, the authors argue, its most revealing. Relative to the most empowered women, those in the lowest categories of decision-making autonomy and of rejection of wife-beating had significantly lower odds of reporting violence, with odds ratios as low as 0.54 and 0.66 nationally. The most plausible interpretation is not that empowerment causes violence, but that empowerment changes disclosure. Women who internalize wife-beating as justified may set a higher threshold for recognizing and naming an act as violence in a survey, while women with greater marital authority may both recognize coercion more readily and, consistent with the backlash account, encounter more of it. Social independence behaved differently, declining in prevalence as independence rose, and it was the second-largest contributor to the Kerala-Bihar gap after women’s education.

The decomposition analysis quantified the state contrast. Women’s education explained 8.69 percent of the Kerala-Bihar difference in prevalence, the social independence domain 7.89 percent, exposure to parental violence 4.24 percent, husbands’ education 3.10 percent, and women’s employment 2.31 percent. Two characteristics, attitudes toward violence and husbands’ working status, contributed in the opposite direction, though neither reached statistical significance. Crucially, all measured factors together explained only about 18 percent of the gap, leaving the large residual to unmeasured contextual forces: governance quality, community gender norms, access to legal and health services, alcohol regulation, and state-level differences in the recognition and reporting of violence.

The authors are careful about what their cross-sectional design cannot establish. Causal ordering is unknown, employment may be endogenous to violence, and self-reported prevalence is almost certainly conservative, particularly where violence is normatively accepted. Still, the policy implications are pointed. Education and employment alone will not reduce violence without parallel shifts in household power and community norms; alcohol harm-reduction and primary care offer practical screening gateways; and breaking the intergenerational cycle demands interventions that reach children exposed to domestic violence. State-specific strategies, the study concludes, must pair individual empowerment with gender-transformative structural change, because the same resource can carry entirely different consequences under different normative regimes.

Subject of Research: Associations of women's education, employment, and empowerment with intimate partner violence in India

Article Title: Interrupting the violence: intersecting associations of education, employment, and empowerment with intimate partner violence in India

Article References: Singh, S., Chakraborty, J., & Kundu, S. (2026). Interrupting the violence: intersecting associations of education, employment, and empowerment with intimate partner violence in India. SN Social Sciences, 6(10), Article 494. https://doi.org/10.1007/s43545-026-01799-6

Image Credits: AI Generated

DOI: 10.1007/s43545-026-01799-6

Keywords: intimate partner violence, India, NFHS-5, women's empowerment, education, employment, SWPER index, Kerala, Bihar, male backlash, decomposition analysis, gender norms