Multidimensional Poverty, Gender, and Forced Displacement A Multi-Country, Intrahousehold Analysis in Sub-Saharan Africa

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. This paper examines multidimensional among

analysis of the Multidimensional Poverty Index (MPI) by examining individual-level deprivations of women and men in forcibly displaced households and host communities, as well as synthesizing intrahousehold dynamics of multidimensional poverty in forcibly displaced communities. The results provide insights into the educational constraints of boys and girls living in forcibly displaced households, the labor market inequalities experienced by men and women in these communities, and their differential access to legal documentation and employment as part and parcel of the forced displacement experience.

Introduction 1
Recent figures from the Internal Displacement Monitoring Centre (IDMC) estimate that in 2019, 24.9 million people across the world were newly displaced due to natural disasters, with another 8.5 million people forcibly displaced due to conflict and violence (IDMC 2020). Currently close to 46 million people are internally displaced in their home country, with 26 million living as refugees 1 in a host country (UNHCR 2020). Some have moved for short periods, while others have been forcibly displaced for years or decades.
In 2019, fewer than half a million forcibly displaced individuals had returned to their homes -signaling the protracted nature of displacement that has increased throughout the last decade, partly due to ongoing political instability in North Africa and the Middle East, and Sub-Saharan Africa (IDMC 2020). And while the displaced population amounts to only 1% of the global population, they are disproportionately affected by deprivation, often chronic. For instance, over 80% of all displaced people reside in countries and areas affected by acute food insecurity and malnutrition (UNHCR 2020), and many of the internally displaced and refugee families live in temporary housing or camps with basic living conditions and have limited access to services and employment.
Existing gendered analyses of forcibly displaced persons tend to center gender-based violence as the organizing vulnerability in people's lives, rather than poverty status (Fiddian-Qasmiyeh 2014). Admasu et al.'s (2021) inclusion of the MPI as an element of examining gender difference added a new dimension to this field, finding that women were more likely to head multidimensionally poor households than men in most of the five countries studied here -Ethiopia, Nigeria, Somalia, South Sudan, and Sudan. As the MPI is a household-level measure of multidimensional poverty, strictly reporting the MPI headline results obscures the divergent experience of individuals within the household. This paper builds on the work of Alkire, Ul Haq, and Alim (2019) with its focus on individual deprivations and intrahousehold analysis using the gender-sensitive refugee [ †] World Bank, Washington, D.C, USA [ ‡] Oxford Poverty & Human Development Initiative, University of Oxford, Oxford, UK MPI, constructed in Admasu et al. (2021), to capture the deprivations of forcibly displaced individuals and their gendered lives.
The paper proceeds as follows. Section 2 reviews the existing literature to provide the background and motivation for the analysis, including a summary of the different country contexts covered by the case studies in this paper. Section 3 outlines the measurement strategy for deconstructing the MPI used for analysis and its limitations, followed by Section 4, which introduces the data. Section 5 presents the findings, first for deprivation results at the individual level and then results evaluating intrahousehold inequalities.
Concluding remarks are discussed in Section 6.

Individual-level measures of gender and multidimensional poverty
Individual-level analyses of multidimensional poverty have mostly centered around children, with various studies analyzing the relevance of indicators for children (aged 0-17 years), 2 as well as other age ranges. The MPI has also been used to better understand gender issues, for example, Batana (2008) implemented a women's MPI in Sub-Saharan Africa. Bhutan's Gross National Happiness measures (2010,2015), Vijaya et al. (2014), and Klasen and Lahoti (2016) are implemented at the individual level. The Women's Empowerment in Agriculture Index (WEAI) ) uses individual-level data, and the linked Gender Parity Index reflects inequalities across women and men's deprivation scores within the same household. Alkire, Apablaza and Jung (2014) design and implement an exploratory individual-level MPI for 31 European countries over six waves of data using EU-SILC data sets, finding no cases in which are women significantly less poor than men, and in many cases, they are significantly poorer.
Espinoza-Delgado and Klasen (2018) create an individual-level MPI to understand differences in poverty between women and men in Nicaragua, finding similar overall incidence, but much higher intensity of poverty among women. Bessell (2015) and Pogge and Wisor (2016) explore deeply contextual gendered poverty measures and elucidate the ways that participatory consultations can inform the design and uses of gendered measures. Rogan (2016) uses the global MPI to analyze the gender poverty gap in South Africa. Alkire, Ul Haq, and Alim (2019) use individual-level data alongside MPI data to expose gendered and intrahousehold differences among MPI poor and non-poor children.
In this study, we apply their techniques to understand gender differences among adults as well.

Intrahousehold analyses of multidimensional poverty
The literature on multidimensional poverty measurement and intrahousehold analysis is limited. Espinoza-Delgado and Klasen (2018) propose an individual-based multidimensional poverty measure for Nicaragua and estimate gender gaps in headline statistics. Klasen and Lahoti (2016) question the neglect of intrahousehold inequality in multidimensional poverty indices by comparing a standard household-level MPI and an individual-level MPI to the MPIs proposed by Alkire and Santos (2014) and UNDP (2014), finding that females recorded a far higher poverty rate when using the individual measure and that age differentials in poverty were also larger. We follow their work of investigating poverty in the indicators for which individual data is available and compare the achievements of men and women and boys and girls living together. This allows us to avoid relying on gender of the head to derive conclusions about gender gaps, while also recognizing the high prevalence of female-headed households among the displaced and large differences across countries between households based on the gender of the head (Admasu et al. 2021).

Country contexts
Sub-Saharan Africa is one of the most conflict-affected regions in the world, with onethird of the total number of conflicts taking place in the region (Pape et al. 2018). At the end of 2018, IDMC estimated that 16.8 million people in Africa were internally displaced because of conflicts and violence, which amounts to 40% of the global number. All five countries included in this study are among the most conflict-affected in the region and rank in the top 12 countries with the highest number of conflict-induced displaced population (Pape et al. 2018). East Africa (which includes four of the five countries considered in this study) is the sub-region with the largest number of internally displaced people and refugees, accounting for 22% of the global number. Ethiopia, Somalia, and South Sudan are three of the top five countries in the region with the greatest number of new displacements since 2019 (IDMC 2020).
Ethiopia is known for its open-door policy towards refugees, and thus is one of the largest refugee-hosting countries in the world. According to UNHCR, at the start of 2020, the country hosted 735,204 refugees. Most come from Eritrea, Somalia, South Sudan, and Sudan, driven by conflict, drought, political instability, and military conscription.
Ethiopia also had more than 1.8 million IDPs in 2020 (IDMC 2020, IOM 2020), due to ethnic and border-based disputes, and drought and floods. More recently, the conflict in the Northern part of Ethiopia has produced many IDPs and refugees, and a full-scale humanitarian crisis is unfolding at the time of writing in 2021.
Nigeria had more than 2.7 million displaced people at the end of 2019 (IDMC 2020), many fleeing the Boko Haram insurgency in the Northeast region, which has compounded climate change and complex ethnic and political factors, leading to conflicts between pastoralists and local farmers (Pape et al. 2018).
Somalia is one of the poorest countries in Sub-Saharan Africa, with more than 2.7 million IDPs -about one-fifth of the population -in 2019 (IDMC 2020), of a total population of 14.32 million. There are many interlinked drivers of displacement in Somalia, including Al-Shabaab attacks, prolonged conflicts and violence, drought, food insecurity, and human rights violations. Recent crop losses due to locusts and flooding have triggered further large-scale displacement. The overlapping crises have created a massive humanitarian and development challenge (IOM 2020).
In South Sudan there is widespread internal displacement after a major conflict in 2013, as well as intercommunal conflicts, drought, and floods. In 2019, there were more than 1.6 million IDPs -more than 10 percent of the population (IDMC 2020) -while many have also fled to neighboring countries. The combined effect of conflicts, flooding, and economic crises leaves many displaced South Sudanese in protracted situations.
Sudan has a long history of internal displacement, following the longest conflict on the continent, the conflict in Darfur, as well as floods and droughts. There were about 2.3 million IDPs in 2019 (IDMC 2020), many of whom live in protracted situations. Sudan also hosts refugees from neighboring countries like South Sudan and Eritrea.

The A-F method and individual deprivations
The Multidimensional Poverty Index (MPI) used in this paper was first presented, with a full methodological discussion, in Admasu et al. (2021). Here we present a general overview of the measure for the individual-level and intrahousehold analyses.
The MPI is constructed based on the Alkire-Foster (AF) method of multidimensional poverty measurement (Alkire and Foster 2011). Three key statistics characterize any MPI: incidence or headcount ratio (H), which is the proportion of the population who are multidimensionally poor; intensity (A), which is the average share of weighted indicators in which multidimensionally poor people are deprived; and adjusted headcount ratio (M0 or MPI), which is the product of the incidence and intensity (MPI = H × A). The AF method uses a dual-cutoff counting approach to poverty measurement. Having fixed relative weights across indicators that sum to 100%, it first identifies who is deprived in each indicator, then sums up the weighted deprivations each person experiences into a deprivation score. A person is identified as poor if their deprivation score meets or exceeds a cross-dimensional poverty cutoff that is greater than 0 and less than or equal to 100%. It then aggregates this information to compute society-level MPI, incidence, and intensity. The MPI can be decomposed by any groups for which the data are Consider two households, each of which has five eligible members with data on nutrition.
The aggregation rule in this example is that if any household member is undernourished then the household is undernourished. So, both households are deprived in terms of nutrition. But in Household A, 2/5 eligible persons are deprived, whereas in Household B, it is 4/5. Furthermore, in terms of gender, in Household A both deprived persons are female, and no males are deprived. But in Household B, the only non-deprived person is female. Household-level information would code both situations as identical, but we can use the underlying microdata to go further. the matrix below illustrates this example.

The MPI structure
The AF method allows flexibility around the selection of dimensions, indicators, deprivation cut-offs, weights, and poverty cut-off to reflect different contexts and priorities. After discussion with experts in both multidimensional poverty and forced displacement and testing of several trial measures, the structure shown in Table 1 was selected. This MPI measures deprivations in education, health, living standards, and financial security, with 15 indicators across these four dimensions. All dimensions are weighted equally, receiving a quarter of the total weight, with indicators within each dimension also receiving equal weights. The poverty cut-off is defined as k=50%, with those deprived in at least half of the weighted indicators considered multidimensionally poor. We focus on six of these 15 indicators that use individual-level data -viz years of schooling, school attendance, pregnancy care, early marriage, legal identification, and unemployment. Our intrahousehold analysis drops the two health indicators due to data limitations; the question about age at marriage was only asked to the household head in Ethiopia, Nigeria, and South Sudan, whereas in Somalia and Sudan, it was applied to more members than the head. The pregnancy care indicator is also excluded from the intrahousehold inequalities analysis as the reference populations for the analysis did not permit rigorous statistical testing.

Limitations
With a few exceptions, gendered MPIs have been designed using indicators that are present in standard survey instruments, which themselves struggle with normative challenges (Alkire 2018

Data sources
The five household surveys conducted in Sub-Saharan Africa and selected for the purpose of analyzing the MPI constructed in Admasu et al. (2021) are presented in Table 2. 8 These surveys differ in population and geographic coverage, as well as sample design and size, limiting the possibilities for comparisons across countries. A full analysis of these data sources can be found in Admasu et al. (2021).

Description of the data
The relative sizes of different comparator groups -displaced and non-displaced populations, adults and children, and displacement status groups by gender -are shown in Figure 1 and Table 3.  Source: Authors' calculations using the Ethiopia SPS (2017)  In Ethiopia, Northeast Nigeria, South Sudan, and Sudan, the dependency ratio is somewhat higher among the forcibly displaced population, whereas in Somalia, the reverse is true.  (2017), and Sudan IDP Profiling Survey (2018). Significance at 1% ***, 5% **, and 10% * levels.
Overall, while statistically significant differences do exist in the household composition of forcibly displaced and non-displaced households, the magnitude of these differences is mostly minor. The fact that the household composition is similar across the populations suggests that any relationships we observe regarding gender and displacement at the individual level are not confounded by individuals living in different household types.

Hypotheses
Several hypotheses interrogate the intersection of gender, displacement, and multidimensional poverty to extend our household-level analysis to the individual arena.
The first set of hypotheses are among women only:

H1:
Forcibly displaced women and girls are deprived more often than nondisplaced women and girls.
H2: Forcibly displaced women and girls are deprived and MPI poor more often than non-displaced women and girls.
The second set of hypotheses concern gender gaps: H3: Gender gaps between men and women are larger for forcibly displaced persons than non-displaced persons.
H4: Gender gaps between boys and girls are larger for forcibly displaced children than non-displaced children.
The final hypothesis examines intrahousehold inequality by displacement status: H5: Intrahousehold inequalities will be more common for forcibly displaced individuals than non-displaced individuals.

Findings
We present the key multidimensional poverty statistics, individual-level deprivations, and intrahousehold inequalities of the MPI poor for forcibly displaced populations and host communities in turn. The final section presents the relationships among gender, displacement, and multidimensional poverty for intrahousehold inequalities. Admasu et al. (2021) found that displaced households were on average MPI poorer in the same set of countries, with large variation in the poverty gaps between the displaced and non-displaced populations. We take these overall findings a step further by zooming into the individual deprivations experienced by men and women, and boys and girls, living in displaced and non-displaced households.

Headline MPI results
The MPI headcount and intensity by displacement status, by gender, are presented in  There is large variation in the MPI incidence and intensity within and across the five countries, both in terms of gender and displacement status. In Sudan, for example, male and female IDPs report an incidence of poverty nearly five times higher than their nondisplaced counterparts, while in Nigeria, the differences are far smaller in magnitude.
Still, the displaced populations always experience higher incidence of multidimensional poverty, although in Nigeria, Somalia, and South Sudan, the intensity of poverty is higher among the non-displaced population. Table 6 checks the significance of the differences using an adjusted Wald test of means on the female populations by displacement status (Table 6). In both Ethiopia and Somalia, women and girls who are forcibly displaced experience a significantly higher rate of multidimensional poverty, on average, than their nondisplaced counterparts, both in MPI and incidence; however, intensity of poverty is not significantly different between these two populations in either country. In Sudan, forcibly displaced women and girls see significantly higher rates of multidimensional poverty -both in incidence and intensity -than their non-displaced counterparts in all key statistics.
By way of contrast, in Nigeria, the differences in MPI and H are insignificant for the two female populations, although non-displaced women and girls face significantly higher rates of intensity of poverty, at 63.0%, compared to their displaced peers at 59.5%. South Sudan sees no significant differences between the displaced and non-displaced female populations, which may be partly due to the small sample size.
Looking at gender differences among the displaced population, women and girls in Ethiopia and Nigeria experience higher rates of multidimensional poverty compared to men and boys, although this is not the case in Somalia, South Sudan, and Sudan. Nor are there significant gender differences in multidimensional poverty rates among the nondisplaced population.
Further disaggregating by age shows that in Ethiopia, Nigeria, and Somalia, forcibly displaced children do not see a significant difference in multidimensional poverty rates according to gender, but gender is a major disadvantage for women (Tables A.1

and A.2).
In Ethiopia, nearly half of women refugees face multidimensional poverty compared to only a quarter of men refugees. We do not see gender gaps among the non-displaced children and adults in Ethiopia. In South Sudan, the intersection of age, gender, and displacement status does not reveal any significant differences except for the nondisplaced populations, where boys are poorer than girls and women are poorer than men.
Gender clearly plays a role in the experience of poverty among the host community of South Sudan. In Sudan, displacement appears to be the larger driver of poverty differences than gender, as significant differences are observed among women and girls when comparing displaced and non-displaced individuals, but gender gaps are not evident in the IDP population.
Overall, the gendered breakdown of the key statistics of the MPI exposes important, varied stories of displacement and age at the household level that we will now explore by explicating the results of individual-level deprivations.

Analysis of individual-level deprivations
We now turn to compare gendered differences in deprivations, as well as multidimensional poverty, by displacement status. These indicators span three of the four dimensions: education, health, and financial security.

Ethiopia
In Ethiopia (Table 7), deprivations in education are high, with around eight in ten poor children not attending school, regardless of displacement status (Table 7). Gender gaps in school attendance are significant only for the refugee community, where girls are more deprived than boys (3-percentage points among the wider population, and 5-percentage points among the poor). Meanwhile there are no significant gender differences in years of schooling, by displacement or poverty status. Most refugee in Ethiopia children attend NGO schools, although some attend public schools, and they also have higher enrollment rates than the national average of their country of origin (Pape et al. 2018). The only significant gender gap in legal identification is in the wider population (an 11percentage points gap), where women are more deprived than men in the host community.
Meanwhile, the only significant gender gap in unemployment is in the poor population (a 12-percentage points gap), where men are more deprived than women in the refugee community, although this may reflect women's overall inactive labor status. Early marriage rates are significantly higher among women across the board.

Nigeria
In Nigeria (Table 8), while there are no significant gender gaps in education, among the poor or the wider population, the deprivation levels do vary by displacement status. This may be because many IDP children have been out of school for at least three years since initial displacement (Pape et. al 2018). For more than half of displaced households, lack of financial resources is a main reason for not sending children to school, despite the provision for universal basic education in Nigeria and free primary schooling in the Northeast region. Among the wider population, women are more likely to be deprived in employment regardless of their displacement status, but among the poor, this gender gap only holds for displaced women. Women are more deprived in legal identification only among the wider population, although it is a serious issue among the poor, too, with eight in ten individuals lacking identification across the board. Furthermore, women report being married early more often than men in all permutations, although host women see smaller gender gaps than displaced women.

Somalia
Like Nigeria, there are no significant gender gaps in education in Somalia, but the differences in deprivation levels by displacement status are stark (Table 9). For example, over 95% of displaced individuals, poor or otherwise, are deprived in the years of schooling indicator, compared to six out of ten non-displaced individuals in the wider population. Meanwhile, women are more deprived in employment only among the nondisplaced population, and the gap is larger among the poor. The gender gaps in legal identification are significant only among the wider population, where women are more deprived overall. Early marriage rates are significantly higher among women than men in both displaced and non-displaced communities, although the gender gaps are larger among the nondisplaced. Among the MPI poor, non-displaced women also see higher levels of deprivation than their displaced peers.

South Sudan
In South Sudan (Table 10), the gender gaps in education are significant only for the host community, when comparing men and women's rates of primary school completion among the wider population and school attendance among poor children. Females are more deprived in both instances. The employment indicator, meanwhile, sees no significant gender differences, while the gender gaps in legal identification are significant in all cases. The gender gaps are also larger among the displaced community, with the gap nearly five times as large among the displaced than non-displaced poor (24percentage point and 5-percentage point differences, respectively). South Sudan sees the largest gender gap of all countries for displaced women, among the poor and population-wide (70-and 67-percentage points, respectively). Displaced women are also more deprived than host women, although the differences are larger among the wider population.

Sudan
In Sudan (table 11), the gender gaps in education and financial security are significant only among the wider population. Deprivations are higher for male hosts in primary school completion and female hosts in legal identification, whereas females face higher deprivations in employment in both displaced and host communities. Deprivation levels in education are low given the displacement situation. This may be due to good access to education in camps, where more than 80% of IDPs state that better access to education is what makes them want to stay in camps. Early marriage 0.09 0.64 -0.56 *** 0.14 0.63 -0.50 *** Source: Authors' calculations using the Sudan IDP Profiling Survey (2018). Significance at 1% ***, 5% **, and 10% * levels.
Early marriage remains an important issue for both displaced and non-displaced females, as it displays the highest deprivation rate among women and exhibits the lowest deprivation rate among men. The results show a substantial gender gap in early marriage among the displaced population (43-percentage points) and hosts (35-percentage points).
Among MPI poor households, the highest gender gap is also observed in early marriage, for both groups.

Intrahousehold inequalities
Our final hypothesis investigates whether forcibly displaced households have higher rates of intrahousehold inequality than their non-displaced peers, and whether girls are more likely to be disadvantaged. We focus on four individual-level indicators -school attendance, primary school completion, unemployment, and legal identification.
Displaced people experience multidimensional deprivations that result both from their current circumstances and the disadvantages they may have accumulated in their home country. For example, refugees fleeing long-term conflicts may not have completed primary education due to chronic disruptions, reducing years of schooling. Meanwhile, deprivations in the current school attendance indicator reflect the individual's present circumstances. Both accumulated and contemporaneous disadvantages shape the lived realities of forcibly displaced populations.
According to Table 12, all countries but Sudan show a significant relationship for schoolage children's experience of intrahousehold inequality and their displacement status, although sample sizes mean that only Nigeria and Somalia's results are robust. In Nigeria, most children experiencing intrahousehold inequality reside in non-displaced households -although it is crucial to note that these children constitute most of the school-age children (71.1%), and the levels of intrahousehold inequality are nonetheless far higher than anticipated if displacement status had no effect. In Somalia, displaced children are significantly more likely to experience intrahousehold inequality in school attendance, as they constitute 63.5% of school-age children experiencing intrahousehold inequality even though they only make up 35.3% of the school-age children population in the sample. For the years of schooling indicator, the overall lack of intrahousehold inequality among the MPI poor in years of schooling obscures meaningful or robust differences by displacement status.
Gender and displacement status appear to jointly have significant impacts in school attendance in Ethiopia, Somalia, and South Sudan. In Northeast Nigeria, it appears that displacement status has larger effects than gender. In Somalia, forcibly displaced school children experience intrahousehold inequality more often than non-displaced children, to the disadvantage of girls. This holds in both IDP and non-IDP households, although the gender gaps observed are larger for children in displaced households is some 15percentage points, compared to only a 1-percentage point difference among children in non-displaced households.
Gender affects intrahousehold inequality and primary school completion, and in all countries, males more often than females live in households where they have completed primary education and another household member has not -suggesting that men and boys' educational careers are the priority. However, the displacement differences are significant only in Somalia, where outcomes are worse for girls in displaced households.

Results of intrahousehold inequalities, financial security
Displacement can boost female employment rates if women in the household take on traditionally male tasks (e.g., as breadwinner), and/or as norms around women's paid work outside the home relax, although these opportunities may not arise.  Gender does not reveal any significant differences in intrahousehold inequality for employment in the countries, except for Nigeria, where men are more likely than women to live in MPI poor households where they are employed, and another adult is unemployed and looking for work. As Nigeria's results for differences by displacement were insignificant, we can rest assured that of the two traits, gender is the primary characteristic driving those differences. Moreover, in the legal identification indicator, the test for gender records robust significant differences in intrahousehold inequality only among the MPI poor of Somalia, where men more commonly have a form of legal identification when another household member does not.

Policy Implications and Conclusions
As leaders and policy makers reckon with the backwards slide of progress in the wake of the global pandemic, they must consider the disparate and varied experiences of boys and girls, and men and women, within the household, as questions of building back better turn into policies and programs directed at leaving no one behind.
The decomposability afforded by the MPI's individual-level indicators is a step towards informing better policies. The indicators analyzed in the individual and intrahousehold analysis -school attendance, primary school completion, unemployment, and legal identification -can be targeted by government policies and programs to improve the situation for those who are multidimensionally poor.
Our comparisons of the household and individual-level deprivations and intrahousehold analyses suggest that displacement status most affects household status, whereas gender drives differences within households. The individual deprivations reflect aspects of both displacement and gender. More broadly, the analysis underlines the importance of going behind household-level outcomes, especially in education.
Among the multidimensionally poor, children in displaced households, and especially girls, face greater barriers to educational equity than those of their host community peers.