Women’s empowerment and child anthropometric failures across 28 sub-Saharan African countries: A cross-level interaction by Gender Inequality Index

Abstract

Background

Child undernutrition remains a major global health issue, particularly in sub-Saharan Africa (SSA). Given the important role mothers play in early childhood health and development, we examined how individual-level women’s empowerment and country-level Gender Inequality Index (GII) are jointly related with child undernutrition in SSA.

Methods

We pooled recent Demographic and Health Surveys from 28 SSA countries. For 137,699 children <5 years old, undernutrition was defined using anthropometric failures (stunting, underweight, wasting). Women’s empowerment was assessed using three domains of Survey-based Women’s EmPowERment (SWPER) index: attitude to violence, social independence, and decision-making; and country-level gender inequality was measured using GII from United Nations Development Programme. Three-level logistic regression was conducted to examine the joint associations of SWPER and GII as well as their interactions with child anthropometric failures, after adjusting for sociodemographic covariates.

Results

Overall, 32.85% of children were stunted, 17.63% were underweight, and 6.68% had wasting. Children of mothers with low-level of empowerment for all domains of SWPER had higher odds of stunting (attitude to violence: OR=1.15; 95% CI, 1.11-1.19; social independence: OR=1.21; 95% CI, 1.17-1.25; decision-making: OR=1.16; 95% CI, 1.12-1.20), and consistent results were found for underweight and wasting. Independent of women’s empowerment, country-level GII increased the probability of underweight (ranging ORs=1.46; 95% CI, 1.15-1.85 to 1.50; 95% CI, 1.18-1.90) and wasting (ranging ORs=1.56; 95% CI, 1.24-1.97 to 1.61; 95% CI, 1.27-2.03). Significant interaction was found between women’s empowerment and country-level GII for stunting and underweight (p<0.05).

Conclusions

In SSA countries with greater gender inequality, improving women’s social independence and decision-making power in particular can reduce their children’s risk of anthropometric failures. Policies and interventions targeted at strengthening women's empowerment should consider the degree of gender inequality in each country.

Highlights

• •Children of mothers with low-level of empowerment had higher odds of stunting, underweight and wasting
• •Country-level gender inequality increased the probability of underweight and wasting
• •Each domain of women’s empowerment had varying importance depending on country-level gender inequality
• •In more gender-equal countries, women’s gender-related attitudes mattered more for children’s health
• •In more gender-unequal countries, women’s social independence and decision-making power mattered more for children’s health

1. Introduction

Child undernutrition remains an important global health challenge, with about 148 million children under 5 years suffering from stunting in 2022 (United Nations International Children’s Emergency Fund [UNICEF], World Health Organization [WHO], & World Bank [WB], 2023). According to the UNICEF, the world is currently off-track to meet the global stunting target, specifically in Western and Middle Africa (UNICEF, WHO, & WB, 2023). Notably, the rates of hunger are severe in sub-Saharan Africa (SSA) (Akombi et al., 2017; Heemann et al., 2021), where nearly one-third of all undernourished children lives in (Akombi et al., 2017). Inadequate nutrition in early childhood is known for short- and long-term effects such as child mortality (Black et al., 2013), low cognitive skills (Walker et al., 2007), restricted economic potential (Hoddinott et al., 2008), and chronic diseases in adulthood (Victora et al., 2008).

Findings from empirical studies on social determinants (e.g., maternal education, residence, and household wealth) of child nutrition outcomes (deSouza et al., 2022; Li et al., 2021; Li et al., 2020; Smith et al., 2005) align with the UNICEF framework outlining the social, economic, and cultural factors as fundamental in affecting children’s health status (UNICEF, 2021). A factor recently gaining attention is women’s empowerment, commonly defined as a process by which women who have been denied the ability to make strategic life choices acquire such an ability (Kabeer, 1999). Promoting women’s empowerment is found to influence women’s nutritional status in terms of the variety of food types they consume in a day (Malapit et al., 2015) and their body mass index (Jones et al., 2020; Malapit et al., 2015). Their empowerment is also considered an important driver of child health since mothers are key agents in safeguarding the interests of children (Ekbrand & Halleröd, 2018; Santoso et al., 2019).

There are several pathways through which women’s empowerment can influence child growth. As Kabeer (1999) highlighted, access to education, waged employment, or other forms of social capital represent potential resources that empower women to exercise their agency. This empowerment may manifest in increased control over household assets and the freedom to engage with neighbors outside the home. These factors, in return, can enhance a woman’s ability to adeptly negotiate resource allocation, prioritizing her child’s nutrition. However, evidence on the association between women’s empowerment and child nutrition outcomes is quite mixed (Mekonnen et al., 2021; Onah, 2021; Quisumbing et al., 2021; Santoso et al., 2019; Yaya et al., 2020). A recent systematic review of 62 epidemiologic studies found that associations of women’s empowerment with stunting and wasting were not statistically significant in roughly 80% of cases (Santoso et al., 2019). Such inconsistent findings could be partly attributed to the diverse operationalization of women’s empowerment, ranging from using unvalidated measures (Yaya et al., 2020) to operationalizing women’s empowerment with a single indicator (e.g., decision-making power) (Ibrahim et al., 2015; Rahman et al., 2015). Given the complex, inter-relational, and multidimensional nature of women’s empowerment (Pratley, 2016; Santoso et al., 2019), developing and validating a standardized measure remains challenging (Ewerling et al., 2020).

The current study uses a recently proposed measure of women's empowerment that enables cross-country comparison: Survey-based Women’s EmPowERment (SWPER) (Ewerling et al., 2020). Initially developed and validated for use in African countries based on Demographic Health Surveys (DHS) (Ewerling et al., 2017), SWPER was updated for global use in its subsequent iteration (Ewerling et al., 2020). Alike the development theories suggested by previous literature on women’s empowerment (Kabeer, 1999; Miedema et al., 2018), SWPER conceptualizes women’s empowerment based on three domains: attitude to violence, social independence, and decision-making. Attitude to violence reflects women’s intrinsic agency, which is manifested through their gender beliefs and attitudes towards husbands’ use of violence; social independence represents enabling preconditions that allow a woman to stand up for herself and gain more power (e.g., education and access to media); and decision-making represents instrumental agency or ability to make choices on her own. SWPER was found to have associations with maternal and child health outcomes such as antenatal care utilization (Arroyave et al., 2021) and immunization (Johns et al., 2022). In respect to child growth, a significant role of SWPER was found in South Asia (Onah, 2021) and Ethiopia (Mekonnen et al., 2021), but whether this association applies to a larger set of SSA countries is yet to be tested.

So far, the majority of studies have been restricted to considering women’s empowerment at individual-level (Mekonnen et al., 2021; Onah, 2021; Quisumbing et al., 2021) when looking at its relationship with maternal and child health. At the same time, there is a need to understand the societal gender system that simultaneously structures and, in turn, is influenced by women’s empowerment at individual-level. Specifically, gender inequalities are reflected in policies, legislations, and budgetary allocations at the national level, which can structure individual’s daily lives (Marphatia et al., 2016). These structural inequalities can be strengthened by restrictive gender norms among individuals, while they can also reproduce such norms by compromising with the current gender system (Cislaghi & Heise, 2020). To monitor the extent of gender inequality at the national level, the United Nations Development Programme (UNDP) has been developing various indices since 1995 (Gaye et al., 2010). Gender Inequality Index (GII), in particular, is known to well represent the overall gender gaps in multiple sectors such as labor force participation and political engagement (Gaye et al., 2010).

Taken together, we sought to examine the associations between women’s empowerment and child anthropometric failures using a validated measure of women’s empowerment at individual-level and gender inequality at country-level, based on the pooled data from nationally representative surveys across 28 countries in SSA. Specifically, the current study aimed to explore the following research questions:

• 1)What are the associations between each domain of individual-level women’s empowerment (attitude to violence, social independence, and decision-making) and child anthropometric failures, after adjusting for country-level gender inequality and other covariates?
• 2)What are the associations between country-level gender inequality and child anthropometric failures, after adjusting for individual-level women’s empowerment and other covariates?
• 3)Does the relationship between individual-level women’s empowerment and child anthropometric failures vary by country-level gender inequality?

2. Methods

2.1. Data sources and study population

DHS are nationally representative surveys designed to collect data on a wide range of public health topics with a special interest in fertility, child mortality, and indicators of access to maternal and child health across 90 countries. In DHS, samples are selected through a two-stage stratified cluster sampling method. Details on the sampling and data collection process are described elsewhere (Corsi et al., 2012).

In the current study, we selected the most recent surveys in SSA conducted post-2013 to avoid surveys that are more than a decade apart in the pooled analysis. A total of 28 countries included in the analysis are listed in Table 1. The unit of analysis was children aged under 5 years, and their eligibility for inclusion was as follows: (1) currently alive, (2) mothers are currently married or in union (i.e., a precondition of women participants for most items included in SWPER), and (3) selected for anthropometric measurements. Of the 153,044 eligible mother-child pairs, approximately 10% (N=15,345) were excluded for missing data, resulting in analytic sample of 137,699 mother-child pairs (Supplementary Figure 1).

Table 1.

Descriptive summary of women’s empowerment and child anthropometric failures across 28 sub-Saharan Africa countries from Demographic and Health Surveys (2013∼)

Countrya	Survey year	N	SWPERb,c						GII	Child anthropometric failuresc
			Attitude to violence		Social independence		Decision-making			Stunting	Underweight	Wasting
			Mean	SD	Mean	SD	Mean	SD	Mean	Prevalence (%)
Pooled (Weight not applied)		137,699	0	1	0	1	0	1	0.59	32.85	17.63	6.68
Angola	2015-2016	4,394	0.41	0.80	-0.01	0.96	0.54	0.74	0.53	36.56	17.72	5.09
Benin	2017-2018	10,437	0.28	0.86	-0.07	0.92	-0.21	0.96	0.61	31.79	16.12	4.83
Burundi	2016-2017	5,471	-0.12	0.97	0.22	0.86	0.34	0.88	0.51	55.34	28.65	5.01
Cameroon	2018	3,573	0.36	0.83	0.05	1.07	-0.13	1.08	0.57	29.43	11.45	4.58
Chad	2014-2015	8,660	-0.69	1.02	-0.58	0.78	-0.55	0.87	0.70	40.46	30.11	13.85
DR Congo	2013-2014	6,816	-0.43	0.94	-0.04	0.88	-0.11	0.93	0.64	42.87	22.53	7.67
Ethiopia	2016	8,205	-0.47	1.09	-0.37	0.85	0.47	0.83	0.53	38.04	23.14	10.00
Gambia	2019-2020	3,279	-0.04	0.94	0.02	1.04	-0.36	0.92	0.61	16.68	11.43	5.28
Ghana	2014	2,359	0.33	0.84	0.35	1.08	0.43	0.77	0.55	17.49	10.63	5.20
Guinea	2018	3,162	-0.59	1.01	-0.39	0.92	-0.38	0.98	0.61	30.81	15.50	8.96
Kenya	2014	7,608	0.23	0.83	0.35	0.98	0.38	0.83	0.54	25.84	10.98	3.75
Lesotho	2014	842	0.43	0.71	0.67	0.86	0.67	0.61	0.56	28.41	9.43	3.28
Liberia	2019-2020	1,589	0.21	0.86	-0.05	0.90	0.48	0.85	0.65	27.76	10.02	3.29
Madagascar	2021	4,524	0.40	0.65	0.00	0.99	0.71	0.59	0.56	38.94	22.65	7.09
Malawi	2015-2016	4,255	0.66	0.53	0.00	0.79	0.26	0.87	0.58	36.21	10.94	2.68
Mali	2018	7,664	-0.66	0.92	-0.35	0.83	-0.92	0.78	0.68	26.79	18.24	9.01
Mauritania	2019-2021	6,891	0.31	0.86	-0.02	1.11	0.16	0.99	0.63	25.35	16.46	6.41
Namibia	2013	742	0.38	0.83	1.03	1.19	0.69	0.72	0.48	19.22	12.97	6.50
Nigeria	2018	10,538	0.29	0.97	0.19	1.30	-0.23	1.03	0.67	36.79	21.44	6.83
Rwanda	2019-2020	3,148	-0.05	1.05	0.91	0.98	0.54	0.73	0.39	31.97	7.33	0.94
Senegal	2019	4,877	-0.14	1.13	0.06	1.08	-1.07	0.74	0.53	17.80	14.04	8.17
Sierra Leone	2019	3,440	-0.05	0.93	-0.09	0.94	-0.35	1.05	0.64	29.30	12.92	5.50
South Africa	2016	395	0.81	0.31	1.37	1.02	0.96	0.47	0.42	24.61	3.15	2.49
Tanzania	2015-2016	7,594	-0.23	1.03	0.22	0.90	0.01	0.89	0.57	33.67	13.43	4.66
Togo	2013-2014	2,854	0.31	0.84	0.07	0.97	-0.24	0.91	0.60	26.11	15.32	6.56
Uganda	2016	3,682	0.18	0.85	0.10	0.92	0.29	0.86	0.54	28.15	9.56	3.40
Zambia	2018	6,460	-0.02	1.04	0.11	0.87	0.41	0.83	0.54	33.50	11.41	4.19
Zimbabwe	2015	4,240	0.40	0.74	0.48	0.89	0.73	0.62	0.54	25.33	7.47	3.43

Abbreviation: SD, standard deviation. GII, Gender Inequality Index (range 0∼1). DR Congo, Democratic Republic of the Congo.

^aThe countries are listed in an alphabetical order.

^bSWPER scores are standardized estimates having a mean of 0 and a standard deviation of 1 in the pooled dataset.

^cSampling weights were applied for country-specific estimates of SWPER and child anthropometric failures.

2.2. Measures

2.2.1. Child anthropometric failures

For the indicators of child undernutrition status, we used three anthropometric failure outcomes: stunting, underweight, and wasting. Generally, stunting reflects a chronic or recurrent nutritional deficiency in early childhood, while wasting reflects more acute and recent undernutrition. Underweight is a composite measure of both acute and chronic undernutrition. In DHS, trained field staffs used Shoff Boards to measure the length of children younger than 2 years lying down (length) and the height of children older than 2 years standing up. Weight was measured using a SECA digital scale (Assaf et al., 2015). Following the WHO Child Growth Standards (WHO, 2006), stunting was defined as height-for-age z-scores < -2 standard deviation (SD) from the median, underweight as weight-for-age z-scores < -2 SD, and wasting as weight-for-height z-scores < -2 SD.

2.2.2. Individual-level women’s empowerment

We used the SWPER global index developed by Ewerling et al. (2020), a single consistent survey-based measure of women’s empowerment that allows cross-country comparisons (Ewerling et al., 2020). We extracted 14 items from DHS and computed the scores of each domain (attitude to violence, social independence, and decision-making) using the factor loadings presented in Ewerling et al. (2020), where a higher score indicates higher empowerment in each domain. In the attitude to violence domain, five items showing women’s opinion on whether a husband beating his wife is justified in several situations (e.g., wife goes out without telling husband, wife argues with husband) were included. Higher empowerment in the attitude to violence domain indicates a higher degree of disapproval for wife-beating. In the social independence domain, items representing women’s preconditions for empowerment (e.g., women’s educational attainment, age at first birth, and frequency of reading newspapers or magazines) were included. In the decision-making domain, items showing women’s participation in decisions in specific settings (e.g., women’s health care and large household purchases) were included. Each item and its coding structure can be found in Supplementary Table 1. The scores of each domain were standardized using mean and SD derived from our pooled dataset and divided into terciles (high, medium, and low) following the approach proposed by previous SWPER-based studies to secure comparability and consistency (Ewerling et al., 2020; Wendt et al., 2022).

2.2.3. Country-level gender inequality

GII proposed by UNDP is an index that assesses the position of women in a society using three distinct dimensions: empowerment, labor market, and reproductive health (Gaye et al., 2010). The empowerment dimension is based on two indicators: shares of parliamentary seats held by each sex and gender differences in at least secondary education. The labor market dimension reflects labor participation rates by each sex. The reproductive health dimension is based on the maternal mortality ratio and adolescent birth rate. In its original version, GII ranges from 0 to 1 (as shown in Table 1), with higher values indicating worse gender inequality within the country. We obtained GII values corresponding to the survey year of DHS for each country. For our regression analysis, we rescaled it to range from 0 to 10 to enhance interpretability.

2.2.4. Individual-level covariates

Because most of the socio-demographic variables (e.g., women’s age and education years) were already included in the development of SWPER, the control variables in this study were limited to age of child (0–4), sex of child (female/male), number of children aged under 5 within the household (0-27), place of residence (rural/urban), household wealth index (poorest, poorer, middle, richer, and richest), and survey year (2013–2021). We selected these covariates based on previous studies (Abreha et al., 2020; Onah, 2021) regarding the association between women’s empowerment and child health. The household wealth index in DHS was developed using principal component analysis in which household ownership of assets (e.g., televisions, cars, and bicycles), dwelling characteristics, type of water access, and sanitation facilities were considered to develop a composite measure of households’ cumulative living standards. The quintiles of wealth index were created based on each survey data, which provide information on relative wealth within a country.

2.3. Data analysis

First, we conducted a weighted descriptive analysis for SWPER and child anthropometric failures. When presenting country-specific estimates, we used sampling weights that apply to each country. Second, we examined the association between each domain of SWPER and GII with respect to child anthropometric failures by using multilevel logistic regression given the multilevel structure of the data where individuals (level-1) are nested within clusters (level-2), and countries (level-3). Specifically, the probability of each child anthropometric failure (y_ijk=1) was modeled as:

$$\text{Logit}\left({\mathrm{\pi }}_{\text{ijk}}\right)={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\text{SWPER}}_{\left(\text{Low}\right)\text{ijk}}+{\mathrm{\beta }}_2{\text{SWPER}}_{\left(\text{Medium}\right)\text{ijk}}+{\mathrm{\beta }}_3{\text{GII}}_{\mathrm{k}}+{\mathrm{\beta }}_4{\mathrm{C}}_{\text{ijk}}+\left({\mathrm{\mu }}_{0\text{jk}}+{\mathrm{\upsilon }}_{0\mathrm{k}}\right)$$

where i, j, and k index are the level-1, level-2, and level-3 units, respectively; C is a vector of individual-level covariates included as fixed effects: age of child, sex of child, number of children aged under 5 within the household, place of residence, household wealth index, and survey year; μ_0jk and υ_0k are the random effects assumed to be normally distributed with a mean of 0 and a variance of σ²_u0 and σ²_υ0, respectively. In this part, we conducted a sensitivity analysis where we utilized other country-level indices (i.e., Gender Development Index (UNDP, 1995) and Global Gender Gap Index (World Economic Forum, 2006)) to determine if the association remains consistent.

Cross-level interaction between each domain of SWPER and GII was assessed as:

$$\text{Logit}\left({\mathrm{\pi }}_{\text{ijk}}\right)={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\text{SWPER}}_{\left(\text{Low}\right)\text{ijk}}+{\mathrm{\beta }}_2{\text{SWPER}}_{\left(\text{Medium}\right)\text{ijk}}+{\mathrm{\beta }}_3{\text{GII}}_{\mathrm{k}}+{\mathrm{\beta }}_4{\mathrm{C}}_{\text{ijk}}+{\mathrm{\beta }}_5{\text{SWPER}}_{\left(\text{Low}\right)\text{ijk}}*{\text{GII}}_{\mathrm{k}}+{\mathrm{\beta }}_6{\text{SWPER}}_{\left(\text{Medium}\right)\text{ijk}}*{\text{GII}}_{\mathrm{k}}+\left({\mathrm{\mu }}_{0\text{jk}}+{\mathrm{\upsilon }}_{0\mathrm{k}}\right)$$

Lastly, we examined the country-specific associations between SWPER and child anthropometric failures. In country-specific analyses, countries with less than 1000 observations (Lesotho, Namibia, and South Africa) were excluded from the analysis on all three outcomes, and we further dropped Liberia and Rwanda from the analysis on wasting since the number of children who had wasting was too small (<100).

Stata MP version 17 (StataCorp, College Station, TX) and MLwiN 3.05 software were used for the analysis. For multilevel logistic models, 2^nd-order penalized quasi-likelihood (PQL) estimation was used. In the event of failure in model convergence, 1^st-order marginal quasi-likelihood (MQL) estimation was used instead.

3. Results

3.1. Descriptive statistics

Overall, 32.85% (N=45,234) of children were stunted, 17.63% (N=24,283) were underweight, and 6.68% (N=9,205) had wasting. The prevalence of child anthropometric failure varied across countries, from 16.7% in Gambia to 55.3% in Burundi for stunting, 3.2% in South Africa to 30.1% in Chad for underweight, and 0.9% in Rwanda to 13.9% in Chad for wasting (Table 1). In 7 countries (Mali, Chad, Guinea, Demographic Republic of Congo (DR Congo), Senegal, Sierra Leone, and Gambia), more than 30% of women had low-level of empowerment for all three domains of SWPER (Figure 1). South Africa had the highest prevalence of high-level of women’s empowerment, ranging from 83.2% (social independence) to 85.1% (attitude to violence). In terms of GII, the values ranged from 0.39 (Rwanda) to 0.70 (Chad), with higher values indicating higher gender inequality within a country (Table 1).

Figure 1.

Level of women’s empowerment for each domain across 28 sub-Saharan Africa countries from Demographic and Health Surveys (2013∼)

3.2. Pooled analyses

3.2.1. Stunting

Compared to children of mothers with high-level of empowerment, children of mothers with low-level of empowerment for all domains of SWPER had higher odds of stunting (OR=1.15; 95% CI=1.11-1.19 for attitude to violence, OR=1.21; 95% CI=1.17-1.25 for social independence, and OR=1.16; 95% CI=1.12-1.20 for decision-making) after adjusting for individual-level sociodemographic covariates and country-level GII (Table 2 and Supplementary Table 2). In the same model, country-level GII was not statistically significant. These results were consistent when we used Gender Development Index and Global Gender Gap Index instead of GII (Supplementary Table 8 and 9). For all domains of SWPER, a statistically significant interaction was found with country-level GII (likelihood ratio test p-value<0.05) (Figure 2 and Supplementary Table 3). For the attitude to violence domain, the gap in probabilities of stunting between high- versus low-level of empowerment was widest in countries with the lowest gender inequality (25% vs 30%), whereas it narrowed down as the country-level gender inequality increased (32% vs 33%). In contrast, for both the social independence and decision-making domains, the gaps in probabilities of stunting widened as country-level gender inequality increased. For example, for the social independence domain, the probability of stunting was similar for all children regardless of the mother’s empowerment level in countries with the lowest gender inequality (28%), whereas the gap in probabilities of stunting became substantial in countries with the highest gender inequality (30% vs 35%).

Table 2.

Association between women’s empowerment, Gender Inequality Index, and child anthropometric failures (Pooled dataset N=137,699)

		Stunting	Underweight	Wasting
		OR [95% CI]b	OR [95% CI]b	OR [95% CI]b
Attitude to violence	High (ref)	1	1	1
	Medium	1.12*** [1.09-1.16]	1.13*** [1.08-1.17]	1.05 [0.98-1.12]
	Low	1.15*** [1.11-1.19]	1.16*** [1.11-1.21]	1.09** [1.02-1.17]
Gender Inequality Index (GII)a		1.10 [0.87-1.38]	1.50*** [1.18-1.90]	1.61*** [1.27-2.03]
Social independence	High (ref)	1	1	1
	Medium	1.12*** [1.09-1.16]	1.05* [1.01-1.10]	0.98 [0.92-1.05]
	Low	1.21*** [1.17-1.25]	1.15*** [1.11-1.20]	1.06+ [0.99-1.13]
Gender Inequality Index (GII)a		1.08 [0.86-1.35]	1.48** [1.17-1.88]	1.60*** [1.27-2.03]
Decision-making	High (ref)	1	1	1
	Medium	1.08*** [1.05-1.12]	1.11*** [1.06-1.15]	1.07* [1.01-1.15]
	Low	1.16*** [1.12-1.20]	1.24*** [1.19-1.29]	1.20*** [1.12-1.29]
Gender Inequality Index (GII)a		1.08 [0.86-1.36]	1.46** [1.15-1.85]	1.56*** [1.24-1.97]

+ p<0.10; *p<0.05; **p<0.01; *** p<0.001.

Abbreviation: OR, odds ratios. CI, confidence intervals.

^aGender Inequality Index is rescaled into the range of 0∼10.

^bAdjusted for age of child, sex of child, number of children aged under 5, place of residence, household wealth index, and survey year.

Figure 2.

Predicted probabilities of child anthropometric failures as a function of women’s empowerment and Gender Inequality Index (Pooled dataset N=137,699)

Abbreviation: GII, Gender Inequality Index. LR, Likelihood Ratio.

*Statistically significant interaction

3.2.2. Underweight

Children of mothers with low-level of empowerment for all domains of SWPER had higher odds of underweight (OR=1.16; 95% CI=1.11-1.21 for attitude to violence, OR=1.15; 95% CI=1.11-1.20 for social independence, and OR=1.24; 95% CI=1.19-1.29 for decision-making) in the fully adjusted model (Table 2 and Supplementary Table 4). In the same model, an incremental increase in country-level GII also significantly increased the odds of child underweight by approximately 50%. For all domains of SWPER, a statistically significant interaction was found with country-level GII (likelihood ratio test p-value<0.05) (Figure 2 and Supplementary Table 5), with a similar pattern observed for stunting.

3.2.3. Wasting

In the fully adjusted model, low-level of empowerment in the attitude to violence (OR=1.09; 95% CI=1.02-1.17) and decision-making domain (OR=1.20; 95% CI=1.12-1.29) were associated with higher odds of wasting (Table 2 and Supplementary Table 6). For the social independence domain, the association was marginally significant (OR=1.06; 95% CI=0.99-1.13). In the same model, an incremental increase in country-level GII also significantly increased the odds of child underweight by approximately 60%. For all domains of SWPER, there was no significant interaction with country-level GII (Figure 2 and Supplementary Table 7).

3.3. Country-specific analyses

Overall, the association between women’s empowerment and child anthropometric failures was consistent across countries (Figure 3). Low-level of empowerment in at least one domain of SWPER was statistically significantly associated with increased odds of child stunting in 17 out of 25 countries. In Nigeria, all three domains of SWPER were significantly associated with child stunting (OR=1.30; 95% CI=1.15-1.48 for attitude to violence, OR=2.03; 95% CI=1.79-2.31 for social independence, and OR=1.59; 95% CI=1.40-1.80 for decision-making). This pattern was similar for underweight. In 12 out of 25 countries, low-level of empowerment in at least one domain of SWPER was statistically significantly associated with increased odds of child underweight. Regarding child wasting, the country-specific associations with women’s empowerment were relatively weak.

Figure 3.

Country-specific association between women’s empowerment and child anthropometric failures (a) Stunting (b) Underweight (c) Wasting

Countries are ordered in an ascending order of odds ratios for low-level of empowerment versus high-level of empowerment.

Countries with sample size of less than N<1000 observations were excluded from this analysis (Lesotho, Namibia, and South Africa). For (c), countries with less than 100 cases of wasting were further excluded (Liberia and Rwanda).

4. Discussion

From our multilevel analysis of 28 SSA countries, we found that the prevalence of child anthropometric failures remains persistently high and is complexly linked with individual-level women’s empowerment as well as country-level gender inequality. Stunting was the most common type of anthropometric failure across all countries considered in this analysis (32.85%). The level of women’s empowerment was particularly low in Chad and Mali, where early marriages (>50%) (UNICEF, 2022) and fertility rates (>5.0) (Central Intelligence Agency, 2023) are also known to be high. In Chad, 60% of women receive no formal education at all (Deubel & Boyer, 2017), and in Mali, married mothers were more likely to be in favor of the continuation of female genital mutilation/cutting practices compared to other African countries (Coll et al., 2022). The low social status of women and restrictive gender norms were also reflected in the high gender inequality measured by GII in Chad (0.70) and Mali (0.68).

The role of individual-level women’s empowerment on child anthropometric failures

Our study demonstrated that three domains of women’s empowerment were significantly associated with children’s anthropometric failures in SSA using a globally validated and comprehensive measure that enables cross-country comparisons. In doing so, our findings extend prior studies that have relied on single indicators such as maternal education (Bbaale, 2014) to assess their association with child nutrition, and emphasize the importance of simultaneously considering various facets within each domain of empowerment (e.g., age at first birth and spousal age differences in the social independence domain). Our findings also align with SWPER-based research in South Asia (Onah, 2021) and Ethiopia (Mekonnen et al., 2021). In Ethiopia where child feeding practices are generally inadequate, women who were empowered in all three domains of SWPER had a lower risk of their children experiencing stunting and underweight (Mekonnen et al., 2021). The current study is the first to pool a large set of SSA nations and test the pooled and country-specific associations between SWPER and child anthropometric failures. Although data constraints caused some variation, our analyses for each country in SSA showed consistent trends of positive association between women’s empowerment and child stunting and underweight. These notable associations draw attention to the deeply entrenched patriarchal norms and traditional gender roles in SSA that restrict women’s access to school, labor market, and healthcare services, which in turn contributes to poor child health and growth.

Although interrelated, each domain of women’s empowerment may influence child anthropometric failures differently. Women who justify spousal abuse, for instance, may be more vulnerable to domestic violence which is known to increase the likelihood of child abuse and neglect (Lee et al., 2004), and ultimately deprive them of proper care needed for optimal physical growth. Social independence pertains to women’s enlightenment and knowledge of the need for pre- and post-partum care (Anik et al., 2021), which helps women to adopt nutritionally balanced diets and hygienic practices that impact children’s growth. Women with greater decision-making power may effectively interact with family members (Anik et al., 2021) and move freely outside the home, gaining more opportunities to learn appropriate child-feeding practices and health-seeking behaviors (Chi et al., 2024; Mekonnen et al., 2021)

The role of country-level gender inequality on child anthropometric failures

Furthermore, this study took advantage of the multilevel framework by including country-level gender inequality in the analysis. The findings show that gender parity at the country-level independently influences child anthropometric failures over and above women’s empowerment at the individual-level, and vice versa. Yet, it should be noted that GII was significantly associated with underweight and wasting but not with stunting, which is a sign of long-term undernutrition. The insignificant association between GII and child stunting may be due to the nature of GII. As GII is an annual indicator and focuses on the current state of gender inequality within a specific year, it may not offer a comprehensive view of the prolonged and deep-rooted challenges that women face, which could contribute to children’s long-term health.

A few previous studies have found significant associations between country-level gender equality and child nutrition after adjusting for the mother’s socioeconomic characteristics (Burroway, 2016; Ekbrand & Halleröd, 2018). Aligning with these studies, our results imply the presence of a contextual effect, such that in countries with more equal gender systems, children are overall healthier, regardless of their mother’s gender-related attitudes, education level, or general autonomy within the household. In a more gender-equitable society, women have the opportunities to engage in roles initiating positive changes in social institutions and encouraging the growth of public services that help themselves and other women, which can help further reduce the risk of child anthropometric failures (Burroway, 2016).

Interactions between women’s empowerment and gender inequality

More interestingly, we found dynamic relationships between individual-level empowerment and country-level gender inequality through the distinct interaction effect between GII and different domains of SWPER. In countries with highly unequal gender systems, women’s perception of domestic violence appeared to be less relevant in ensuring their children’s growth. Instead, in these countries, enhancing mothers’ social independence and decision-making power was found to be more salient to prevent children’s anthropometric failures.

Especially in SSA region where men’s supremacy marks gender systems, several structural barriers against women should be noted. Women may be limited in seeking health care for children due to a lack of control over financial resources (Dougherty et al., 2020) or due to socially constructed gender beliefs and attitudes (Azad et al., 2020). Also, they are often constrained in accessing markets where they can buy fresh food for their children due to the high burden of unpaid family work (Porter, 2011). In these social contexts, merely holding gender equitable attitudes alone may not empower women with the practical resources they need to better care for children’s physical growth and may not suffice to make tangible differences at least in the short run. Rather, in these contexts, women who have attained a high level of education or have a greater say over household activities are more likely to have the ability to provide adequate care for children. Indeed, these interpretations should be understood cautiously since the strategies to improve women’s empowerment over different domains are closely interrelated and are not mutually exclusive. Nevertheless, our results suggest a need for different interventions to empower women in contexts of high gender inequality.

5. Limitations and strengths

Our study has a few important limitations to note. First, the cross-sectional nature of the survey hindered us from establishing causality. Although reverse causation seems highly unlikely, there may be unmeasured and residual confounding bias by cultural traditions or religious practices. For example, deeply ingrained religious and cultural beliefs are found to be associated with traditional gender attitudes (Reitz et al., 2015) and child health outcomes (Karlsson, 2019). However, given the various religions within each country and the absence of this data in three countries (Mauritania, South Africa, and Tanzania), we were not able to adjust for religion in the main analysis. Future studies using longitudinal datasets are needed to test whether women’s empowerment impact child undernutrition and their underlying mechanisms. Second, while we used the latest and the most comprehensive measure of women’s empowerment, SWPER is also not free of limitation as it relies on questions available in the DHS and hence may not fully capture the multidimensionality of women’s empowerment (Ewerling et al., 2020). Future surveys may consider including questionnaires of other important feature of empowerment such as self-efficacy, time-use, and legal rights to thoroughly evaluate women’s empowerment in low- and middle- income countries. Yet, our study may have partially captured the political dimension of women's empowerment at the country-level, as the GII estimates take into account of the proportion of parliamentary seats held by each gender. Third, our analysis was restricted to women who are married or in a union since most information on indicators of SWPER was only collected among partnered women in the DHS. Thus, our results cannot be generalized to women who are single, widowed, or divorced, who may differently experience empowering processes and nurturing care. Fourth, our complete case analysis excluding deceased children at the time of the survey may have led to an underestimation of the association of interest since mothers of deceased children are likely to have a lower level of empowerment, as suggested by previous evidence (Doku et al., 2020). Lastly, heterogeneity in the country-specific analysis may be due to the different contexts across countries and inconsistent sample size ranging from N=1,589 (Liberia) to N=10,538 (Nigeria). Similarly, the relatively weak association for wasting may be due to the small number of children who had wasting, ranging from 125 cases in Ghana to 1,257 cases in Chad. Despite the above limitations, our study has unique strengths.

To our knowledge, this study is the first study to demonstrate the joint associations of individual-level women’s empowerment and country-level gender inequality with child anthropometric failures from a multilevel perspective. Given the mixed evidence on the association between women’s empowerment and child undernutrition, our findings provide a comprehensive and up-to-date evidence by utilizing the indicator of women’s empowerment that has been globally validated based on a large set of countries in SSA. Furthermore, the present study advances our understanding of the role of women’s empowerment on child undernutrition by discovering which domains of empowerment were more critical depending on country-level gender inequality. To gain a deeper understanding of the mechanisms underlying women's empowerment, gender systems, and child undernutrition, future studies should look into interactions with other socioeconomic (e.g., household wealth) and demographic variables (e.g., child sex).

6. Conclusions

All women and girls have the right to fully participate at all levels of educational, economic, political, and public spheres, and lead a fulfilling life. In addition to this intrinsic value of women’s empowerment, our findings suggest that women’s empowerment also have instrumental value in promoting children’s health in SSA. Policies and interventions aimed at promoting women’s empowerment should consider each country’s level of gender inequality in order to prioritize specific domains of empowerment and promote its mitigating impact on child undernutrition in SSA.

Ethical Statement

The DHS data are not collected specifically for this study and no one on the study team has access to identifiers linked to the data. These activities do not meet the regulatory definition of human subject research. As such, it was granted an exemption from ethical approval (KUIRB-2023-0245-01).

CRediT authorship contribution statement

Yun-Jung Eom: Writing – original draft, Methodology, Investigation, Formal analysis, Data curation. Hyejun Chi: Writing – review & editing, Investigation. Sohee Jung: Writing – review & editing, Investigation. Jinseo Kim: Writing – review & editing, Investigation. Joshua Jeong: Writing – review & editing, Investigation. S.V. Subramanian: Writing – review & editing, Investigation. Rockli Kim: Writing – review & editing, Supervision, Investigation, Conceptualization.

Declaration of Competing Interest

None.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Multimedia component 1

Data availability

The authors do not have permission to share data.