Child marriage alone does not explain why Somali women accept wife-beating

A national analysis of survey data from Somalia has found that women who married before the age of 18 are not more likely than other women to justify wife-beating once the regions they live in and their socioeconomic circumstances are taken into account. The study, based on responses from 7,341 ever-married women aged 20 to 49 who took part in the 2020 Somali Health and Demographic Survey, was published in Frontiers in Global Women’s Health. Its central result runs counter to what a simple comparison of the two groups suggests at first glance, and it points instead to geography as the dominant factor shaping attitudes toward intimate partner violence in the country.

In the raw, unadjusted data, there was indeed a modest difference. Among women who had married before their 18th birthday, 59.0 percent said wife-beating was justified for at least one specified reason, compared with 55.0 percent among women who had married at 18 or later. That difference yielded a crude prevalence ratio of 1.07, with a 95 percent confidence interval of 1.00 to 1.15, a range that just barely excludes the value of 1 that would indicate no association. The confidence intervals of the two prevalence estimates themselves overlapped, hinting that the apparent gap might not survive closer scrutiny.

It did not. When the researchers fitted a survey-weighted Poisson regression with robust standard errors, adjusting for region of residence, type of residence, household wealth quintile, educational attainment, and current age, the association between child marriage and justification of wife-beating essentially vanished. The adjusted prevalence ratio was 1.01, with a 95 percent confidence interval of 0.95 to 1.08 and a p-value greater than 0.05. A parallel survey-weighted logistic regression produced an adjusted odds ratio of 1.02, with a confidence interval of 0.87 to 1.21, likewise crossing the null. In other words, once the social and geographic context of women’s lives was held constant, whether a woman had married as a child carried no measurable independent signal in her attitudes toward wife-beating.

What did carry a signal was region. Hiraan, Bay, Bakool, Woqooyi Galbeed, Awdal, Lower Juba, Bari, Middle Shabelle, Togdheer, Sool, Galgaduud, and Sanaag showed significantly elevated adjusted prevalence relative to Banadir, with ratios ranging from 1.25 to 1.91. In total, 12 of the 15 comparison regions differed significantly from the reference. The authors note that adjusted prevalence was up to 1.9 times higher in several regions, a spread they describe as pronounced.

The survey instrument asked women whether a husband is justified in beating his wife under seven specific circumstances: if she neglects the children, refuses to have sex with him, neglects household duties, wastes household resources, argues with him, goes out without telling him, or does not respect him. Endorsement was remarkably uniform across these items, ranging from 36.8 percent for the least-endorsed reason, that the wife does not respect her husband, to 40.8 percent for the most-endorsed, that she neglects the children. No single circumstance dominated. The authors interpret this pattern as suggesting a generalized rather than context-specific acceptance of wife-beating among those who endorse it at all. Overall, 56.7 percent of women in the analytical sample justified wife-beating for at least one of the seven reasons, with a 95 percent confidence interval of 53.5 to 59.9.

The analytical sample itself was drawn from a larger survey population. The researchers restricted the analysis to ever-married women aged 20 to 49 because exposure to child marriage, defined as first marriage before age 18, can only be fully ascertained for women who have passed that birthday; including 15-to-19-year-olds would have left some women’s exposure status incomplete. After excluding women with missing information on age at first marriage or on any of the seven justification items, 7,341 women remained. In the weighted sample, 43.9 percent had married before age 18 and 56.1 percent at 18 or older. Most women, 84.0 percent, had no formal education. Residence was fairly evenly split: 37.1 percent urban, 32.1 percent nomadic, and 30.9 percent rural. Mean current age was 30.7 years, and women endorsed a mean of 2.76 of the seven possible justifications.

Methodologically, the study leaned on modified Poisson regression rather than logistic regression for its primary estimates because the outcome was common, affecting more than half the sample, and odds ratios would have overstated the corresponding prevalence ratios. All analyses accounted for the complex, stratified, multistage cluster sampling design of the survey, incorporating individual sampling weights, primary sampling units, and sampling strata. Covariates were specified a priori on the basis of existing literature rather than statistical significance, and all were retained in the model regardless of their p-values. Model diagnostics offered some reassurance: the largest variance inflation factor was 2.9, with a mean of 1.7, indicating no problematic collinearity; the area under the receiver operating characteristic curve for the logistic model was 0.65, indicating modest discrimination; and modeling age in five-year categories instead of continuously left the child-marriage estimate essentially unchanged at an adjusted prevalence ratio of 1.01.

Not every covariate behaved identically across the two model types. The richest wealth quintile had a lower prevalence of justifying wife-beating than the poorest, with an adjusted prevalence ratio of 0.85 and a confidence interval of 0.73 to 1.01 that included the null, but the corresponding adjusted odds ratio of 0.67, with a confidence interval of 0.46 to 0.98, was statistically significant. Education and residence type, which showed clear gradients in the unadjusted data, justification ranged from 57.8 percent among women with no education down to 41.4 percent among those with higher education, and from 58.4 percent among nomadic women down to 54.5 percent among urban residents, lost statistical significance entirely in the adjusted models. The authors suggest this may reflect regional composition: women with little or no education, and those in rural or nomadic areas, may be concentrated in regions where conservative gender norms prevail, so that the crude gradients largely mirror regional norm environments rather than independent associations. They note they did not test this explanation formally.

The McFadden pseudo-R-squared of the logistic model was 0.046, a low figure indicating that even the full set of covariates explained only a modest share of the variation in justification of wife-beating. The authors attribute this to unmeasured community-level drivers. They also flag a selection concern: 3,460 eligible women, about 32 percent of the eligible pool, lacked data on the wife-beating items and were excluded from the analysis. If that missingness is related to child marriage or to attitudes, selection bias is possible. The survey also lacked direct variables capturing structural mechanisms such as local clan governance intensity, conflict displacement metrics, or community-level institutional access, limiting the authors’ ability to parameterize the precise drivers behind the regional variation they observed. Because the design is cross-sectional, temporal or causal relationships cannot be established.

The authors are careful to draw a bright line between attitudes and experiences. Their findings concern justification of wife-beating only and do not address whether child marriage is associated with the risk of actually experiencing intimate partner violence. Prior systematic reviews and multi-country analyses have consistently documented elevated rates of physical and sexual partner violence among women who married as children, and the authors do not dispute that literature. What their analysis adds is nuance about the attitudinal side: in adjusted cross-sectional models, a woman’s own marital history was not independently linked to her acceptance of wife-beating. The practical implication, they argue, is that prevention strategies targeting former child brides as an isolated demographic are unlikely to shift population-level attitudes. Because justification varied so markedly by region, they suggest interventions may need to focus on community-wide social norm change, engaging traditional clan elders, local religious leaders, and municipal governance bodies, particularly in high-prevalence regions such as Hiraan, Bay, and Bakool.

Subject of Research: The association between child marriage and justification of wife-beating among ever-married Somali women using 2020 Somali Health and Demographic Survey data

Article Title: Child marriage and justification of intimate partner violence among Somali women: a national analysis of the 2020 Somali health and demographic survey

Article References: Ali, Y. O. (2026). Child marriage and justification of intimate partner violence among Somali women: a national analysis of the 2020 Somali health and demographic survey. Frontiers in Global Women's Health, 7, Article 1991927. https://doi.org/10.3389/fgwh.2026.1991927

Image Credits: AI Generated

DOI: 10.3389/fgwh.2026.1991927

Keywords: child marriage, Somalia, wife-beating justification, intimate partner violence, Somali Health and Demographic Survey, regional variation, cross-sectional analysis, prevalence ratio, gender attitudes, women's attitudes, survey weighting, public health