This paper reviews the main challenges faced by applied researchers to produce unbiased causal estimates of the effects of forced displacement on health. The authors sought to answer the following question: “What are the analytical challenges faced by current quantitative research examining the relationship between forced displacement and health, as well as the main methodological approaches employed to address those challenges?”
The review initially identified 1,454 studies from the health and social sciences disciplines published up to May 2021, of which 56 studies met the inclusion criteria. African countries were the most frequent origin of the forcibly displaced populations studied, while European countries were the most frequent destination, followed by African countries. The most frequent health outcomes reported in the relevant literature include: all-cause and cause-specific mortality; infant mortality; maternal and perinatal outcomes; child morbidity; and child growth. Less frequent are studies on: maternal mortality; self-perceived health; unmet health needs; access to services; mental health; fertility; and health-related behaviors. The most frequently used comparison group was the native population from the host community.
The authors classified the quality of the evidence according to a four-level categorization (very strongly credible, strongly credible, somewhat credible, and less credible) based on the capacity of the paper’s methodological (identification) strategy to produce a valid comparison group and to mitigate endogeneity concerns.
The authors discuss several analytical challenges associated with studies on the causal health impacts of forced displacement. These include:
- Non-random allocation. In most situations of war, conflict or natural disasters, there is some degree of agency in the decision to migrate and, therefore, no randomness in exposure to displacement. Also, the intensity of conflict violence can disproportionally affect certain individuals within a community (e.g., specific ethnic groups). Even in “natural experiments” where the entire population in a particular territory is homogenously exposed to an exogenous displacement event at the same time, it is nevertheless important to account for the influence of potential observable and unobservable differences between populations in the specific context.
- Limited comparability between displacement contexts. For example, the nature of the displacement shock (acute or protracted), income levels at origin, and the involvement of governments in forced resettlement schemes all vary across contexts. Contextual factors also affect individual characteristics that enable individuals to migrate, the selection of a migration destination, and experiences during the relocation process. Additionally, host territory characteristics can influence long-term outcomes in different ways.
- Disentangling the effects of violence from the impact of forced displacement. Displaced populations may have experiences that directly lead to, or compound, the impacts of forced displacement, e.g., direct exposure to human rights violations, violence, or disasters. To isolate the effect of forced displacement, researchers usually need to compare individuals with similar exposure to a conflict or emergency event, some of whom remained in their home territory while others migrated. However, duration/intensity of exposure to violence influences the probability of displacement, as well as health outcomes. The probability of displacement is also determined by often unobserved factors such as health status, wealth, and social connections.
- Data limitations. Due to the difficulty of predicting displacement events and constraints on collecting data in humanitarian crises, most studies use surveys or administrative data, covering short periods of time, for a few population groups, usually with no longitudinal follow-up allowing for panel data analyses.
- Difficulties of identifying control groups. In most studies, comparisons to the host population do not represent a valid control group for robust inference about the causal effect of forced displacement on health. Younger, healthier, and wealthier individuals are more likely to migrate, or migrate sooner, and host communities tend to be better off in many ways that influence health status. These contextual differences introduce endogeneity and have often confounded the reported links between forced displacement and health outcomes.
Main findings and recommendations:
- High-quality causal inference methods were rarely found in the literature analyzing the health effects of forced displacement.
- Most of the available empirical evidence for a wide range of health outcomes is prone to substantial bias, making it difficult to draw firm conclusions. This is due to issues around selection of valid control groups and the application of credible causal inference methods in several studies.
- Current research practice in the field could be strengthened through selection of valid control groups and application of more appropriate causal inference methods. In settings of internal displacement or displacement to neighboring countries, valid control groups may be found in geographically close communities with similar baseline characteristics and who were unaffected by the exogenous event that triggered displacement. In international displacement settings, comparisons with international (voluntary) migrants from a similar ethnic background (ideally from the same location), who relocated to the same host country as the displaced communities, is likely to be the best approach. More robust findings will require a wider/judicious use of non-experimental methods better suited for causal inference, such as instrumental variables, difference-in-differences, regression discontinuity and interrupted time series analyses.