Forced Displacement Policies and Migrant Decision Making

Canonical models view migrant decision making in terms of a rationalist, utility-maximizing framework (Hanson and McIntosh Reference Hanson and McIntosh2016). According to this view, individuals weigh the costs of leaving versus the prospective benefits of migrating to various destination countries before deciding whether and where to go, subject to uncertainty and budget constraints. Factors driving individuals to leave their home countries, such as civil conflict, are “push” factors, while factors inducing gravitation toward certain destinations, like contiguity and wage differentials, are “pull” factors. We draw on this framework but broaden the scope of many existing models by focusing on de jure policies as a pull factor. Theoretically, we follow Czaika’s (Reference Czaika2009) seminal model, which formalizes the intuition that asylum-seekers are attracted to more liberal policies.Footnote ³

In the context of forced displacement, FDP face severe constraints. The nature of FDP flight—from war and repression—means distance considerations (i.e., costs of travel) weigh paramount in their decisions (Moore and Shellman Reference Moore and Shellman2007). Precisely because most flee acute emergencies, more than three-quarters of all FDP are displaced to contiguous neighboring countries (United Nations High Commissioner for Refugees 2019). Still, among the limited set of proximate host countries FDP might flee to during crises, policies affording more expansive rights should be more attractive. This is because liberal de jure policies enhance migrant assimilation, security, and prosperity. This argument builds on evidence that prospective migrants seek information on destination policy environments (Holland and Peters Reference Holland and Peters2020) and prefer countries where they can access more and better services (Balcilar and Nugent Reference Balcilar and Nugent2019; Hatton Reference Hatton2016).

However, gravitation toward liberal policy environments is conditional, particularly in the Global South, where households have less knowledge about policy environments in other countries (Havinga and Böcker Reference Havinga and Böcker1999). We anticipate that two key factors—access to information communications technologies (ICT) and ethnic networks—facilitate the spread of information about policy environments. Opportunities to learn about host policies are greater in information-dense origins, where ICT penetration enables FDP to research potential host environments (Holland and Peters Reference Holland and Peters2020). Internet and mobile technologies have proven to be particularly important tools (Bacishoga, Hooper, and Johnston Reference Bacishoga, Hooper and Johnston2017). Similarly, ethnic kin play two roles.Footnote ⁴ First, coethnics in a destination ease integration by helping arriving migrant kin realize de jure rights (Rüegger and Bohnet Reference Rüegger and Bohnet2018). For instance, ethnic networks are vital for finding jobs where employment is allowed (Munshi Reference Munshi2003). Second, coethnics in host countries are a source of policy knowledge. Individuals in destinations can transmit information about de jure laws and de facto conditions to kin in origin countries (Helbling and Leblang Reference Helbling and Leblang2019). This networked policy communication is partly responsible for spurring migrant waves (Holland and Peters Reference Holland and Peters2020).

Data and Empirical Strategy

We test the effect of de jure migration policies on FDP flows using data from DWRAP, an original dataset of all national laws on forced displacement in a sample of 92 African, Middle Eastern, and South Asian countries (Blair, Grossman, and Weinstein Reference Blair, Grossman and WeinsteinForthcoming).Footnote ⁵ The dataset includes 229 national-level migration laws and represents the most expansive mapping of domestic laws and policies on forced displacement in the developing world. For each law in the data, we code 54 policy provisions, which we aggregate to five policy fields: (1) access—status and entry procedures; (2) services—access to education, aid, and health care; (3) livelihoods—property and employment rights; (4) movement—encampment requirements and identity documents; and (5) participation—citizenship and political rights. We use summary indices to aggregate from policy provisions to a policy score.

In our primary specifications, we estimate Poisson pseudo-maximum likelihood (PPML) gravity models (Silva and Tenreyro Reference Silva and Tenreyro2006). The dependent variable is the directed dyadic arrival rate, calculated as $ \frac{\left(\mathrm{Asylum}\ \mathrm{Applications}+\mathrm{Prima}\ \mathrm{Facie}\ \mathrm{Arrivals}\right)}{\mathrm{Country}\ \mathrm{of}\ \mathrm{Origin}\ \mathrm{Population}\ \mathrm{in}\;\mathrm{100,000}\mathrm{s}} $ .Footnote ⁶ By taking the arrival rate, we capture the magnitude and intensity of forced displacement between countries. Comparable rates are used in prominent gravity models of voluntary migration (Hanson and McIntosh Reference Hanson and McIntosh2016). Our main independent variable is the five-year lagged moving average of a destination country’s policy score.Footnote ⁷

To capture the conditional effects of liberal asylum policies, we interact our policy measure with three distinct variables. We evaluate the effect of host policy liberality conditional on origin ICT access by interacting our policy score with Dreher’s (Reference Dreher2006) index of information openness, which aggregates data on press freedom, Internet bandwidth, and television and Internet subscriptions.Footnote ⁸ To capture the influence of ethnic networks, we interact our policy index with measures of the presence and number of transnational ethnic kin (TEK) linkages between origin-destination dyads. Data on TEK are drawn from the Ethnic Power Relations dataset (Vogt et al. Reference Vogt, Bormann, Rüegger, Cederman, Hunziker and Girardin2015) and reflect politically relevant ethnic groups split between countries. These measures proxy for the social connections between migrants at origin and destination populations. We cannot directly observe dyadic ethnic networks between FDP and hosts, but TEK linkages increase the probability of shared ties ceteris paribus (Rüegger and Bohnet Reference Rüegger and Bohnet2018).

Our gravity model of forced displacement takes the following form:

$$ \begin{array}{c}{Y}_{o,a,t}=\mathit{\exp}({\alpha}_o+{\beta}_a+{\gamma}_t+\delta \left(\mathrm{Polic}{\mathrm{y}}_{a,t-1,5}\times \mathrm{Facilitator}{\mathrm{s}}_{o,a,t-1}\right)\\ {}\hskip-26pt + \unicode{x03BB} \left({G}_{s,t}\right)+\phi \left({X}_{o,t-1}\right)+\mu \left({Z}_{a,t-1}\right)+{\varepsilon}_{s,t}),\end{array} $$

where $ {Y}_{o,a,t} $ is the arrival rate of FDP from country of origin o to country of asylum a in year t and $ \delta $ captures the interactive effect of policies and facilitators, such as information openness and transnational kin, which are key to realizing the benefits of liberal policy provisions. $ {G}_{s,t} $ is a vector of controls (e.g., distance) for origin-destination dyad s, $ {X}_{o,t-1} $ is a vector of lagged covariates specific to origin o, $ {Z}_{a,t-1} $ is a vector of lagged covariates specific to country of asylum a, $ {\alpha}_o $ are origin fixed effects, $ {\beta}_a $ are destination fixed effects, $ {\gamma}_t $ are year fixed effects, and $ {\varepsilon}_{s,t} $ is the error term.Footnote ⁹ We cluster standard errors by dyad. Our quantitative results are bolstered by qualitative evidence from 126 interviews with FDP and other stakeholders on forced displacement in Uganda in summer 2017.Footnote ¹⁰

Disaggregating Policy Domains

Does the conditional positive effect of displacement policy liberality on FDP flows vary across policy provisions? Existing research points to employment rights (Holland, Peters, and Sánchez Reference Holland, Peters and Sánchez2019), welfare access (Hatton Reference Hatton2016), and citizenship (Alarian and Goodman Reference Alarian and Goodman2018; Fitzgerald, Leblang, and Teets Reference Fitzgerald, Leblang and Teets2014) as particularly attractive. To explore variation across policy domains, we repeat our core specifications from Table 1, substituting the full policy index for constituent subindices capturing access, services, livelihoods, movement, and participation. All effects are positively signed, but the largest and most robust associations emerge between the movement, access, services, and livelihoods indices and FDP flows.Footnote ¹² Liberal asylum policies related to free movement and identity documents (movement); welfare, education, and health care access (services); and employment and property rights (livelihoods) are critical for migrants’ integration prospects and economic opportunities in host countries (Betts et al. Reference Betts, Bloom, Kaplan and Omata2017; Hatton Reference Hatton2016). Thus, the disproportionate pull of liberal policies on movement, services, and livelihoods aligns with the neoclassical view of migrants’ decision making calculus as utility maximizing (Hanson and McIntosh Reference Hanson and McIntosh2016). More specifically, gravitation toward liberal welfare and employment provisions comports with recent evidence that migrants are drawn to destinations with right-to-work policies and social safety nets (Holland, Peters, and Sánchez Reference Holland, Peters and Sánchez2019). As in the Global North (Hatton Reference Hatton2016), we also find that FDP gravitate toward host countries where it is easier to gain recognition and reunify with migrant family members (access).

In developed countries, permanent residency and citizenship provisions draw migrants (Alarian and Goodman Reference Alarian and Goodman2018; Fitzgerald, Leblang, and Teets Reference Fitzgerald, Leblang and Teets2014). We find weak evidence of this dynamic in the Global South. That citizenship and political rights like voting (participation) only weakly influence refugees’ decisions about where to flee in the developing world has an intuitive explanation. South-South FDP typically want to return to their home countries or resettle in Western countries, not reside permanently in asylum countries in the region (Ghosn et al. Reference Ghosn, Chu, Simon, Braithwaite, Frith and Jandali2021; United Nations High Commissioner for Refugees 2019). As a result, considerations about permanent political membership in a host country are less salient.

Robustness

We conduct a host of additional tests to probe the robustness of these results. Our core finding—conditional gravitation toward liberal policies—holds with (a) alternative three- and five-year lags on the policy index (Tables A.10 and A.11), (b) a lower-limit Tobit estimator (Table A.12), and (c) and an alternative dependent variable that extends the analysis period back to 1992 (Table A.13). Evidence from alternative research designs and qualitative interviews supports a causal interpretation of the results.

Generalized Synthetic Control Method

Our primary analyses use gravity models to study FDP flows between country dyads. Gravity models are a well-known approach for estimating migrant flows (Hanson and McIntosh Reference Hanson and McIntosh2016), but they require strong identification assumptions. In particular, our PPML estimations (1) assume no unobserved, time-varying confounders and (2) estimate a single effect of policy liberality rather than an effect allowed to vary across countries. As an alternative, we estimate generalized synthetic control models (Xu Reference Xu2017).Footnote ¹³

In this framework, we fit a predictive model of flows to a destination country and compare the observed effect of displacement policy liberalization on flows to counterfactual flows absent liberalization. We define treatment as policy liberalization such that a country’s score is in the top quartile of all policy scores. As reflected in Figure 2, policy liberalization is associated with a large, statistically significant increase in total FDP arrivals (ATT: 35,364, 95% CI: 17,584-52,187) and FDP arrivals from TEK-linked origins (ATT: 36,204, 95% CI: 18,905-55,257). The increase in flows peaks between 9 and 14 years after liberalization. Preliberalization trends between treated and control units are parallel, assuaging concerns that liberalization is endogenous to prior flows. Results from a related approach, the panel event study, are substantively similar.

Figure 2.

Synthetic Control Estimates of the Effect of Displacement Policy Liberalization

Note: The plots depict observed FDP arrivals (solid line) versus imputed FDP arrivals in the absence of policy liberalization (dashed line). The left plot displays the results for all FDP arrivals, and the right plot displays the results for FDP arrivals from origins linked by TEK. The shaded region indicates the posttreatment period.