Refugee mobilities in East Africa: understanding secondary movements

ABSTRACT

There is significant policy interest in refugee migration, particularly in relation to ‘secondary movements’ – the movement of refugees from the first country in which they arrive. Yet, there is very little theoretical or empirical research on refugee mobilities in the Global South, where the overwhelming majority of refugees reside. Existing literature on refugee migration focuses mainly on people who have already selected onward migration to the Global North. This leaves a gap in terms of describing, understanding, and explaining refugee migration patterns within and from low and middle-income regions of the world. Drawing upon cross-sectional data for Kenya, Ethiopia, and Uganda, we describe aspirations relating to mobility; and drawing upon panel data for refugees based in Kenya, we describe actual patterns of mobility. While a majority of refugees ‘hope’ to migrate inter-regionally and a smaller majority ‘expect’ to migrate inter-regionally, actual mobility patterns are very different. Whereas refugees are highly mobile, the overwhelming majority of their mobility is internal and most international migration is intra-regional. By describing these patterns for one region, the article challenges policy assumptions relating to secondary movement and offers a starting point for further comparative research on refugee mobilities.

KEYWORDS:

• Refugees
• mobility
• migration
• East Africa
• secondary movement

Acknowledgments

We thank colleagues and participants at numerous seminars and conferences for helpful feedback and suggestions. We are especially grateful to Isabelle Aires, Madison Bakewell, Jordan Barnard, Raphael Bradenbrink, Imane Chaara, Antonia Delius, Eyoual Demeke, Maria Flinder Stierna, Leon Fryszer, Aregawi Gebremariam, Abis Getachew, Jonathan Greenland, Louise Guo, Jana Kuhnt, Hiwot Mekonen, Patrick Mutinda, Rashid Mwesigwa, Halefom Nigus, and Clarissa Tumwine for their assistance and inputs at various stages of the research. We thank the enumeration teams in each country for their fantastic work and express our deepest gratitude to all households who participated in our surveys. All errors and opinions expressed in the paper remain ours, the authors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 We recognise that refugee decision-making can be understood along a spectrum of extreme involuntariness to extreme voluntariness. See, for example, Erdal and Oeppen (2018).

2 For example, consent rates were 99% in Kakuma and 96% in Nairobi.

3 These questions are subjective in nature and hence more susceptible to mismeasurement problems. First, respondents may intentionally lie to the interviewer if the respondent does not trust the survey, especially if the respondent does not want to reveal an intention to migrate through illegal channels. To minimise this possibility, we recruited enumerators who were part of the community and worked with community leaders to establish trust between the respondents and the research team. The fact that a majority of respondents reported hoping and expecting to move abroad suggests that misreporting is not a major issue in our study. A second source of bias could come from misinterpreting the questions. We trained our enumerators such that they were fully equipped to handle such situations. There were occasions when respondents reported not being able to distinguish between ‘expectation’ and ‘hope’ due to fatalistic or religious beliefs (e.g. ‘as God wills’). In response, we would argue that these types of answers nevertheless reflect the respondent's real expectation and hope. The fact that we obtain notably different answers for the ‘expectation’ and ‘hope’ questions in Ethiopia suggests that at least some proportion of the respondents were able to distinguish between expectation and hope. We so, however, recognise that these concepts are fuzzy, implying that mismeasurement is possible.

4 Enumerators were trained to administer these surveys following a protocol written before the launch of the survey. The steps included first calling the surveyed households if a phone number was provided, and second surveying neighbours, caretakers, community leaders, and nearby shop keepers to obtain information about the household. Following these two steps, households were re-interviewed in person if they had moved within the survey areas. In Nairobi, we defined research areas as Eastleigh, Kasarani, Githurai, Kayole, and Umoja. In Kakuma, the research area was the whole Kakuma refugee camp. Because of budget constraints, we did not seek to directly re-interview respondents who moved outside of these research areas. For these attrited households, we only collected data on where they moved to and reasons behind the relocation by asking people living in the neighbourhood.

5 These individuals were either re-interviewed or we had obtained information about their new locations.

6 We note that the first-difference and difference-in-differences estimators (and more generally all estimators requiring the measurement of $X_t$) cannot be used in our analysis because explanatory variables are unobserved at time t for individuals and households who moved.

7 For all outcome variables, values are coded as missing if the household cannot be found.

8 Control variables include age and its square, a gender and marital status dummy variables, measures of poor physical and mental health, a measure of trust in the host population, the dependency ratio, and a dummy identifying individuals originating from an urban background.

9 We emphasise that not all questions and themes can be equally studied with quantitative and qualitative methods. For example, qualitative data shows that the requirement to obtain a movement pass is a strong barrier to refugee mobility in Kenya, but this barrier cannot be studied using quantitative data because all refugees in Kenya face the same constraint. Similarly, increasing rents has been consistently reported as a key driver of mobility during qualitative fieldwork in Nairobi; yet quantitative data cannot be easily used to study this phenomenon because rental prices are only observed for refugees who do not move. In contrast, describing the prevalence of different patterns of mobility and studying the relationship between living standards and mobility is possible with quantitative methods, but challenging with qualitative data. Quantitative and qualitative analyses are therefore often supplementary rather than complementary.

10 This null result holds when disaggregating the data by survey location, except in Kakuma were we find weak evidence of a U-shaped relationship between living standard and expectation to migrate internationally.

11 This relationship appears to be driven by refugees living in camp-like settings (Table A.10 in Appendix).

12 We use the following formula to compute the average movement rate per year (r): $r=1-(\frac{100-r^{raw}}{100}{)}^{1/years}$ where $r^{raw}$ is the percentage of individuals in Wave 1 who were found to have left in Wave 2, and years is the number of years between the two waves, i.e. two for Nairobi and three for Kakuma.

13 Receiving remittances is positively correlated the living standard index. As regression results are to be interpreted ‘ceteris paribus’, the coefficient of remittances captures the specific effect of remittances that is not mediated through living standard. In other words, the regression coefficient captures whether there is something specific about remittances that is not already captured by the living standard index.

14 We computed average movement rate per year by using the following equation: $r=1-(\frac{100-r^{raw}}{100}{)}^{1/years}$ where r is the estimated average yearly migration rate, $r^{raw}$ is the percentage of individuals in Wave 1 who were found to have left in Wave 2, and years is the number of years between the survey waves.

15 Using regression analysis, we find a significant and positive relationship between expectation and actual migration status (Tables A.14 in Appendix).

16 Using official exchange rate in 2019, KES 18,000–25,000 was approximately USD 176–245 and KES 7000–8000 was approximately USD 69–78.

17 Using official exchange rate in 2017, KES 8000 was approximately USD 77, KES 16,000 was approximately USD 155, and KES 10,000 was approximately USD 97. Using official exchange rate in 2018, KES 16,000 was approximately USD 158. Using official exchange rate in 2019, KES 20,000 was approximately USD 196.

18 Using official exchange rate in 2019, KES 5000 was approximately USD 49 and KES 4000 was approximately USD 39.

19 Our data does not distinguish clearly between ‘regular’ and ‘irregular’ migration. However, based on qualitative research, we can assume that migration to North America and Australia is mainly regular and via resettlement or complementary visa pathways, and that migration to Europe is based on mixture of irregular and regular routes.

Additional information

Funding

We gratefully acknowledge financial support from the IKEA Foundation. This study received research ethics approval from the Departmental Research Ethics Committee (DREC) of the Oxford Department of International Development, under protocol CUREC 1A/ ODID C1A-17-094.