https://orcid.org/0000-0001-6533-5352
![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
This paper analyses poverty dynamics and checks for the existence of poverty traps among refugee and host communities living close to each other in Uganda. Although some non-linearities emerge in asset dynamics, there is convergence towards one stable equilibrium for the whole sample that suggests the existence of a structural poverty trap. However, households are quite heterogeneous: when analysing refugees and hosts separately, refugees converge to a lower own-group equilibrium than hosts. The household size and education are asset growth enablers for both communities. Noticeably, access to land, past history and social cohesion are also significant correlates of refugees’ asset dynamics. From a policy perspective, structural poverty traps are bad news, because standard anti-poverty interventions would not unlock the trap. Our results stress the need of more structural approaches aimed at promoting economic growth in the whole area where refugee and host communities live, targeting both communities.
Acknowledgements
This paper is the result of the collaboration between the University of Florence and the Food and Agriculture Organization of the United Nations (FAO). We thank participants and discussants to: XVI EAAE Congress, 1st DevEconMeet workshop, 15th RGS Doctoral Conference in Economics, and ICDE 2022, who provided useful insights on an earlier draft of the paper. We are also grateful to the two anonymous reviewers and the editor. Any remaining errors are ours. Codes available on request.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Uganda’s refugee policy has been praised worldwide as it grants wide rights and freedom to the refugees, including the rights to access land and work.
2 The empirical literature on poverty traps has been fast-growing over the last years (Barrett et al., Citation2016; Barrett & Carter, Citation2013), but at the best of our knowledge there is no previous study assessing whether there is a poverty trap among refugees and hosting communities.
3 A fourth order is preferable to a third order polynomial as it does not impose the stable equilibria to be in the tails of the distribution (Naschold, Citation2013).
4 Plus, the Kampala district, where refugees do not reside in settlements and have a different profile from settlement-based refugees. Therefore, they are not included in the FAO-RIMA survey.
5 Other location-related variables, such as living in rural areas and the agroecological zone, were constructed with other data sources exploiting the household georeferenced coordinates.
6 SPEI stands for Standardized Precipitation Evapotranspiration Index (Beguería, Vicente-Serrano, Reig, & Latorre, Citation2014; Vicente-Serrano, Beguería, & López-Moreno, Citation2010), downloaded from SPEI Global Drought Monitor (https://spei.csic.es/) (cf. the Appendix for details).
7 The literature considers as flood and drought deviations that are ±1.5 (or ±2) standard deviations from the long-term average. This was not the case in the areas of analysis in the surveyed period, hence we speak of abundant and scarce rainfalls.
8 Human capital and social capital are not included because of the imperfect transferability of such assets. However, we controlled for these capitals in the main specifications.
9 This is partly due to how the questions are framed. The 2017 and 2018 questionnaires included many more expenditure items for both food and non-food categories. Generally, more detailed questions result in higher expenditure (Comerford, Delaney, & Harmon, Citation2009; Jansen, Verhoeven, Robert, & Dessens, Citation2013). Our approach identifies an expenditure lower bound by using only the categories included in all waves.
10 This can be inferred from a transition matrix across asset quintiles between the first and the fourth wave. In general, many households improved their position, though higher stability is found for better-off host households (55% of those that started in the richest quintile ended in the richest quintile) and worse-off refugee (57% of those that started in the poorest quintile remained in the same quintile).
11 Other functional forms confirm these results (cf. Figure A.1 in the Supplementary materials).
12 Refugees in our sample have on average about one year less of schooling than hosts. This is consistent with other surveys on refugees in Uganda (Development Pathways, Citation2020).
13 Social cohesion variables have been surveyed only from wave 2 onwards. Therefore, the analysis of these variables is restricted to wave 2 – wave 4 period. The whole sample’s wave 2 – wave 4 equilibria are not statistically different from the whole sample’s wave 1 – wave 4 equilibria.
14 FE estimator is preferred as it accounts for the unobserved time-invariant household heterogeneity. RE coefficients are shown to support the FE results (see Section 5.4 where this is essential for testing attrition).
15 Being female-heading is indeed a proxy means for targeting interventions and the refugees mostly dependent on assistance.
16 Note that most households live in tropic warm humid zones and no extreme weather event was reported in the period of analysis. The coefficient indicates that wetter seasons are negatively associated with asset growth. The coefficient for dry season (the variable is reversed, i.e. higher values mean drier conditions) is positive, which means that drier seasons favour assets accumulation. This result is confirmed if we replace it with a dummy variable or a self-reported measure for a dry season.
17 Özler et al., (Citation2021) found similar attrition rates in another panel of refugees. Nonetheless, overall absorbing attrition, i.e., the households at the first wave, which do not enter the balanced panel, is rather high accounting for 63% of the whole sample (68% of refugees and 57% of hosts).
18 As emphasized by Jacobsen (Citation2012), this may be the result of a location-selection strategy according to which households split with the better social and human capital endowed members leaving the camps and the others living on humanitarian assistance and remittances in the camps.
19 Specifically, whether the household belongs to the balanced panel (cf. column 3 and 8 of Table A.5 for OLS and RE, respectively) and a count of the waves each household is included in the survey (cf. column 9 for the RE) (Nijman & Verbeek, Citation1992).
20 We consider the distance to the closest border crossing point and the distance to the closest settlement, the month of interview dummies, and granular rural categories.
21 The auxiliary variables are socio-demographics, income variables, other resources, shock indicators, infrastructural variables, as well as first wave year of the interview, and month of the interview (see Table A.6 in the Supplementary materials for the complete list of variables). Baulch and Quisumbing (Citation2011) argue that a Pseudo R2 of 13% can be considered a relatively high explanatory power. We obtain values between 8% and 15%.
22 Not shown, but available upon request to the authors.
23 We use the expenditure median as poverty line (i.e., 0.10 US$ per day per capita).
24 Negative values are the results of the linear prediction model and are also found in other studies (Giesbert & Schindler, Citation2012; Walelign et al., Citation2021;), although they find equilibria close or above to the poverty line.
25 Available upon request.
26 This can give rise to the so-called ‘recall decay bias’ (Beegle, Carletto, & Himelein, Citation2012; Sawada, Nakata, & Tanaka, Citation2019). However, considering that the FAO-RIMA survey collected data on refugees and hosts using the same instrument and the same modalities, we can assume that this issue would be the same for refugees and hosts, making meaningful the comparison between the two communities.
27 Almost half of refugees in Uganda live in poverty, a proportion more than double than the national poverty headcount which was 21.4% in 2016–17 (World Bank, Citation2019). Hosts are generally better off, although in the Northern region both refugees and hosts’ poverty rates are above the national average, at 57 and 28%, respectively.
28 In presence of a single equilibrium, a one-time asset (in the specific case, livestock) transfer coupled with training has proven effective to improve household resilience and consumption in Zambia (Phadera, Michelson, Winter-Nelson, & Goldsmith, Citation2019). A similar success-case in presence of a multiple-equilibria poverty trap is a large asset transfer (specifically, one cow) coupled with training in Bangladesh (Balboni, Bandiera, Burgess, Ghatak, & Heil, Citation2022).