By Theresa Beltramo, Rossella Calvi, Giacomo De Giorgi, and Ibrahima Sarr [1]
Standard per-capita poverty measures mask individual poverty within refugee and host families. Measuring individual consumption within refugee households, which is not currently done, should be viewed with increasing importance, especially in the current context of unabating forced displacement and funding shortfall.
In Somalia, the worst drought in forty years is causing horrific death and suffering with more than 1 million people displaced internally by drought. Thousands of Somalis have fled across borders (UNHCR and NCR, 2022) since the drought began, with more than 20,000 arriving in and around the Dadaab refugee camp in Eastern Kenya bordering Somalia (Save the Children, 2022). This recent influx of refugees due to drought and famine adds to the large number of existing refugees in neighboring Ethiopia (200,000+), Kenya (250,000+), and Uganda (60,000+).
The Integrated Food Security Phase Classification (IPC) estimates that between Oct0ber–December 2022, some 6.7 million people across Somalia experienced food insecurity (IPC Phase 3 or above) and children suffered more: approximately 513,550 children are estimated to be severely malnourished (IPC, 2022). In times of drought and famine, it may not be surprising that displaced children can be the most severely affected due to malnutrition. However, what about in times of non-drought?
In non-refugee settings, researchers have long found that vulnerable household members experience intra-household inequality and poorer outcomes in food intake, body measurements, and non-food expenditures. It is time to apply this knowledge to forced displacement settings. We examine poverty data for refugees and surrounding host communities in Kenya and Uganda from the period right before the 2020 drought began and test whether the standard poverty calculations, which assume that all household members consume equally, are accurate (Beltramo et. al, 2023).
We find that refugee children – defined as individuals under the age of 18 – can be up to three times as likely to be poor compared to adults. Further, as many as two out of three refugee children are estimated to be extremely poor (living on less than $1.90 per day). Non-refugee children in surrounding host communities are similarly poor. Notably, these findings hold true when differences in needs by age (and gender) are taken into account.
Calculating who is the poorest in a community is an essential mapping exercise to inform social assistance programmes to cover the gap between basic needs. Every country has a norm for the cost of satisfying a minimum set of basic needs in terms of food and non-food (shelter, clothing, transport, among others) encompassed in its own national poverty line. Social assistance programmes aim at helping households and individuals meet these basic needs. We estimate it would take $732/day to eliminate child poverty among refugees in the sample we analyze from the Kalobeyei Settlement in Kenya, and $379/day for the surrounding host community. In the samples from South West Uganda, bringing each child up to their age and gender-adjusted poverty line would require $1,783/day for refugees and $377/day for non-refugees.
The finding that children are the poorest household members in refugee and host communities has important implications for targeting humanitarian and development programmes. We use machine learning to model the household and individual characteristics that best predict poverty, and find that a few observable traits, such as a child’s age, household composition, and access to sanitation and clean water, stand out as the most influential in predicting refugee and host community children’s poverty.
Half of the children who consume less than $1.90 a day live in non-poor families
Our analysis reveals a trend previously unknown to humanitarian and development actors – almost half of the children estimated to be extremely poor are part of households above the poverty line. Targeting programmes using headcount poverty ratios would miss these children, making them especially vulnerable.
Given the deep vulnerability of refugee children, when resources are limited, policymakers should preserve programmes that support the nutrition and well-being of children under five. Extending these programmes to pregnant women and lactating mothers will protect the critical periods of fetal and infant development.
The graph below shows the shares of poor children living in poor and non-poor families. A poor child is defined as a child with estimated individual consumption below their gender and age-adjusted poverty line. A household is defined poor if per-capita consumption is below $1.90 per day.
Child poverty by household poverty
Child poverty is predicted by a few common observable traits, including the number of children in the household, which increases the chance of a child being poor
Our analysis suggests that 8 out of the 15 top predictors of refugee child poverty are the same in all three sites. These include the child’s age and gender, the age and education of the household head, household size, the number of children living in the household, and the number of rooms in the household’s dwelling. Moreover, there is a positive association between the number of children in the household and the chance of there being a poor child. This means that in families with more children, the odds that one of the children is poor is higher than in families with fewer children.
Family composition differs among refugee households, making refugee children more vulnerable
Another vulnerability unique to refugee households stems from the fact that their composition is fundamentally different from that of host communities in Kenya and Uganda. Refugee households tend to be larger and have higher youth dependency ratios (relative share of children to adults). They are also more likely than host community households to be female-headed – 72% of refugee households in the Kalobeyei Settlement in Kenya are female-headed, while men are not at all present in about half of refugee families in settlements in South West Uganda. This figure is substantially lower (27%) in the surrounding host communities in the region.
Unique refugee family composition can be attributed to the reconfiguration of families as a result of the violence and war that they fled, which frequently leaves only a subset of the original members intact due to unfortunate events such as conscription in the military for male members, death, kidnapping or separation of family members during the displacement event. Given the fundamental differences in refugee household structure, there is reason to believe refugees may allocate resources within households differently than non-refugees.
Implications for policy and assistance targeting
Standard per-capita poverty measures mask individual poverty within refugee and host families. As such, our findings that children are the poorest household members in refugee and host communities has important implications for targeting humanitarian and development programmes.
First, given the deep vulnerability of refugee children, when resources are limited, policymakers should preserve programmes that support the nutrition and well-being of children under five, pregnant women, and lactating mothers to protect them during critical periods of development.
Second, calculating intra-household inequality in refugee families and host communities is of primary policy relevance and critical for poverty measurement. Without explicitly accounting for the high incidence of children who are poor in non-poor households, existing humanitarian and development programmes targeting assistance to poor households will miss these children, making them especially vulnerable.
In refugee settings, there is scope for improving the targeting of child poverty if data exercises measuring poverty inside and around refugee camps and hosting areas are explicitly designed to allow for intra-household poverty calculations. A supervised machine learning analysis finds that the top predictors of child poverty are closely linked in all sites and include easily observable traits such as a child’s age and gender, the age of the household head, household size, and the number of children in the household. Other top predictors of child poverty across sites include measures of household food insecurity, the head of household’s education and employment status, the number of rooms in the dwelling, and access to water and sanitation.
By adding a few survey fields, UNHCR could yield better measures of household welfare and improve targeting
For UNHCR and our partners in government, which track refugee households using the proGres registration database, many of these household characteristics are already collected. If UNHCR were to add a few survey fields, including basic data on household water and sanitation access, housing stock characteristics, and collect employment status universally, it could improve current measures of inequality. This would improve the accuracy of targeting with potentially substantial gains for children’s well-being and poverty alleviation in the short and long run. UNHCR should also aim to partner with others to address this challenge.
Finally, it is worth noting that Uganda, unlike most refugee-hosting countries, offers refugees a relatively open enabling environment where they are allowed the right to work, cultivate land, and freedom of movement. In Kenya, refugee movement and employment are restricted. Despite the quite different policy enabling environments with Uganda touted as one of the most welcoming and Kenya as more restrictive in terms of refugee’s rights, refugee children in both Kenya and Uganda are disproportionately poorer than surrounding nationals, underlining the inherent vulnerability and transversal unique aspects of fragility facing refugee households in both countries.
[1] This blog is based on the research paper, Child Poverty Among Refugees, authored by Theresa P. Beltramo (UNHCR), Rossella Calvi (Rice University), Giacomo De Giorgi (University of Geneva), and Ibrahima Sarr (UNHCR). The opinions expressed do not necessarily represent the positions of their organizations.