Migration Policy in Developing Versus Developed Countries

Developing countries host the overwhelming majority of the world's FDP— 85 percent of the 79.5 million FDP worldwide in 2019.Footnote ¹¹ More strikingly, least developed countries, which account for 1.2 percent of global GDP, host 27 percent of FDP. Only one developed country, Germany, is among the top ten host countries in total FDP population, and no developed country is a top FDP host on a per capita basis.Footnote ¹² However, the literature on migration policymaking, and particularly forced displacement policymaking, relies overwhelmingly on evidence from a small number of rich consolidated democracies.Footnote ¹³ This bias represents a critical gap in the literature, limiting the generalizability of existing theories of migration policymaking, and our understanding of global forced displacement.

The politics of migration policy differ between developing countries and developed countries.Footnote ¹⁴ This is because developing (host) countries face a unique array of domestic and international pressures around migration and forced displacement.Footnote ¹⁵ These include larger FDP flows, increasingly protracted periods of displacement, stronger transnational kinship networks, weaker state institutions, greater budgetary pressures, and challenges of nation and state building, among others. At the same time, developed countries have significantly broader welfare provisions, parties that are organized on a left–right ideological spectrum (rather than nonprogrammatic parties), and democratic governments that are generally more attuned to public opinion. Thus, policymakers designing forced displacement policies in the global South confront a variety of distinct opportunities and constraints.

Building on a growing qualitative literature on migration policy in developing countries, we argue that three key assumptions implicit in canonical models of migration management are unlikely to hold in the global South. Unpacking these assumptions is essential because they reveal unique influences on migration policymaking in developing countries.

Global Versus Local Policy Impacts

A key assumption in the literature is that policymakers favor restriction due to concerns about the global implications of policy reforms. However, the effects of migration policies are expected to be significantly more localized in the global South. Advanced industrial countries are integrated into world markets, supply chains, and transportation networks. Correspondingly, though distance between origin and destination countries always raises the costs of movement, wage differentials make the former attractive to prospective migrants globally.Footnote ¹⁶ By contrast, FDP flows to hosts in the South come overwhelmingly from neighboring countries: in the past decade, between 73 and 82 percent of all FDP fled to a contiguous neighbor.Footnote ¹⁷

Consequently, while policymakers in developed countries consider how migration policies may attract migrants worldwide, policymakers in the global South generally consider how policies will affect migration flows from their neighborhoods only.Footnote ¹⁸

The regional nature of displacement in the global South means that factors like transnational kinship weigh more heavily on policymakers than they do in developed contexts.

Rich Versus Poor Economies

Second, existing theories of migration policymaking often assume that host countries are wealthy, industrial economies.Footnote ¹⁹ While developed countries fear political fallout from greater competition between migrants and local low-skill workers, low-income countries sometimes view FDP as an asset in themselves.Footnote ²⁰ This is partly because, unlike in OECD countries, the human capital of the median FDP in the global South is greater than the human capital of the median resident in a host community.Footnote ²¹ Thus, host countries in the developing world may seek to facilitate self-reliance and grow local economies in host communities by affording FDP rights to employment, land, education, and free movement, among others.

Developing countries also face unique economic pressures stemming from slow growth, large debt burdens, and aid dependence. Such dependency precipitated the emergence of the “grand compromise,” whereby developed countries offered aid and trade concessions to developing countries in exchange for refugee hosting.Footnote ²² The explicit aim of the compromise is to reduce the number of FDP in the global North by financing a larger share of the monetary burden of hosting FDP borne by developing countries. By making migration policies in the global South and North interdependent,Footnote ²³ the “compromise” incentivizes governments, especially in sub-Saharan Africa and the Middle East, to leverage FDP hosting in exchange for Western aid.Footnote ²⁴ These considerations around aid in migration policymaking are obviously unique to the global South.

Reliance on external economic assistance can constrain and not only facilitate policy liberality. For example, in the 1980s and 1990s, fiscal austerity reduced the ability of (some) hosting countries in the global South to extend relatively generous support to refugees.Footnote ²⁵ This has been the case in Tanzania, where structural adjustment reforms forced social spending to be scaled back, ultimately causing the collapse of a previously generous, integrative refugee policy in favor of encampment in rural settlements subsidized by external assistance from the UN and other aid agencies.Footnote ²⁶

Liberal Versus Illiberal Regimes

Third, prominent theories of migration management assume that host countries are liberal, consolidated democracies.Footnote ²⁷ Migration policy in these settings represents a balance between liberal values, public opinion (which commonly but not uniformly supports restriction), interest-group advocacy, and labor-market considerations. By contrast, most developing countries are governed by hybrid regimes with weak electoral institutions. In nondemocracies, governments tend to be more insulated from public pressure. Thus, these regimes can more easily alter migration policies to support broader strategic aims. In fact, adopting relatively generous migration policies may paradoxically be easier in illiberal states, where external pressure promotes openness while insulation from public opinion mitigates possible mass demands for greater restrictions.Footnote ²⁸

A Theory of Policymaking on Forced Displacement

What factors determine when and how countries in the global South reform their asylum and refugee policies? Building on a rich qualitative literature on developing world migration policy,Footnote ²⁹ our theoretical framework identifies different data-generating processes underpinning the decision to reform policy and the direction of reform. Reform occurs when major shocks, namely civil wars in neighboring countries, increase expectations of FDP inflows, forcing a policy response. When the (relatively rare) decision to reform is made, different factors influence the direction of reform— liberalization or restriction (though decisions about whether to reform and the direction of reform typically occur simultaneously). In this study we chiefly focus on correlates of policy liberalization because forced displacement policies are trending in a more liberal direction over time in the global South (see Figure 2).

We argue that civil war in a country's neighborhood is the primary impetus for policy change. By contrast, factors that are auxiliary to regional civil wars—such as status of co-ethnics abroad and (in some cases) the extent of aid dependence— determine the direction of policy reform. This theoretical approach challenges prominent models of migration policymaking that view violence in origin countries only as a catalyst for restrictive policies in destinations.Footnote ³⁰

The Decision to Reform Policy

Because they are politically sensitive, migration policies are typically slow to change.Footnote ³¹ As Beine and coauthors note, changes in policy occur “episodically … with longer periods of constancy punctuated by changes via legislative reforms.”Footnote ³² This stability is partly due to the relative difficulty of altering migration policies. Politicians face a host of barriers to policy reform, including intra-elite disagreements and bureaucratic hurdles.Footnote ³³ Another reason for infrequent policy change is that migration is often a lower-salience issue in the global South, which allows governments, especially in low-income countries, to pursue vague policies and delegate FDP management to international organizations and nongovernmental organizations (NGOs).Footnote ³⁴ In particular, governments delegate management responsibilities to the UNHCR, partly to insulate themselves from politically fraught decisions about FDP.Footnote ³⁵

The decision to reform asylum and refugee policies, then, must be precipitated by factors that foster elite consensus on the need for reform, and that raise the salience of migration as a national policy issue. Factors that increase policymakers’ expectations of FDP inflows provide just this spark.Footnote ³⁶ Anticipation of increased inflows raises concerns among prospective host countries of a concomitant increase in the (political and economic) costs of hosting. This is because larger displaced populations are costlier to manage, ceteris paribus.Footnote ³⁷ Greater flows of FDP also reduce the attractiveness of outsourcing migration management to the UN and other nongovernmental bodies.Footnote ³⁸ As forced displacement becomes more salient, governments in host countries prefer greater control, partly due to anxieties over national sovereignty.Footnote ³⁹ In short, when the expected magnitude of migrant flows into a country increases, expectations of costs increase, raising the salience of migration and spurring elite consensus on the need for policy reform.

We do not anticipate heterogeneity across regime types in responsiveness to large FDP influxes for two reasons. First, large migrant inflows generate expectations of greater hosting costs irrespective of hosts’ political institutions.Footnote ⁴⁰ Although certain regimes (e.g., personalist autocracies) may find it easier than others (e.g., democracies) to rapidly shift displacement policies, there are common incentives to react when expected hosting costs increase. Second, regardless of regime type, most developing governments are acutely sensitive about national sovereignty. Concerns that international actors will supplant national institutions in the absence of a policy response to refugee influxes create a strong incentive to reform de jure migration policies when the expected magnitude of FDP inflows increases.Footnote ⁴¹

We argue that intense civil wars in neighboring countries are the primary factor policymakers use to calibrate expectations of FDP inflows. Globally, civil wars are the principal cause of forced displacement (see Figure 4).Footnote ⁴² The greater the violence, the more people flee to neighboring countries,Footnote ⁴³ producing waves of FDP.Footnote ⁴⁴ In recent episodes, civil conflict forced 6.6 million people from Syria, 2.7 million from Afghanistan, 2.2 million from South Sudan, and 1.1 million from Myanmar. The growing use of forced displacement as a deliberate wartime strategyFootnote ⁴⁵ has only sharpened the link between civil conflict and FDP flows.

We therefore hypothesize that civil conflict in a country's neighborhood is the chief impetus for migration policy change in the global South.

This hypothesis dovetails with findings in both developedFootnote ⁴⁶ and developingFootnote ⁴⁷ contexts about how security threats can spur migration policy changes. The negative regional externalities of civil wars are well known,Footnote ⁴⁸ and of these, FDP flight is perhaps the most widespread.Footnote ⁴⁹ Evidence also suggests that FDP flows can spur civil conflict spillovers by facilitating transnational flows of arms and combatants.Footnote ⁵⁰ Displacement also risks regionalizing civil wars by triggering clashes between the neighboring origin and host countries.Footnote ⁵¹ Further, as costs mount, host countries may face financial crisisFootnote ⁵² and local backlash.Footnote ⁵³ These factors help explain why existing models explicitly assume that policy restriction is the likely response to conflict.

By contrast, we contend that increasing policy restrictiveness is but one possible response. In the global South, colonialism left artificial borders and numerous split ethnic groups.Footnote ⁵⁴ As a result, conflicts in one state often affect groups with kinship ties in neighboring states.Footnote ⁵⁵ Conflicts, then, may precipitate policy liberalization when the ethnic kin of groups in a potential host country are affected by war in neighboring origin countries. Because of shared identity ties, host countries may choose to respond generously to co-ethnic FDP.Footnote ⁵⁶ By way of illustration, Chaulia describes how transnational kinship ties motivated Tanzania's liberal response to FDP from Burundi, the Democratic Republic of the Congo, and Mozambique in the 1960s and 1970s.Footnote ⁵⁷ Countries also have incentives to liberalize policy as a means of integrating war-displaced migrants, attenuating the effect of FDP on strife in host communities.Footnote ⁵⁸

Transnational Ethnic Kin and Policy Liberalization

There is ample evidence that ethnic networks pull migrants to specific destinations.Footnote ⁵⁹ Before individuals migrate, kin groups can relay information about the conditions in prospective destinations, as well as risks along the way. In destination countries, kin networks ease integration,Footnote ⁶⁰ reduce the risk of xenophobic attacks,Footnote ⁶¹ and help secure higher-paying jobs and better housing.Footnote ⁶² The more localized nature of migration in the developing world reinforces the importance of transnational ethnic kinship. However, while kin are a known pull factor in models of migrant decision making, little work explores their relevance in migration policymaking.Footnote ⁶³ Because borders were often drawn arbitrarily in the global South, splitting ethnic groups between countries, ethnic kin are an important policy influence.

Cross-border kinship plays an especially critical role in policy responses to regional civil wars. Many conflicts in the global South, and particularly those that produce the most FDP, have an explicit ethnic dimension.Footnote ⁶⁴ In origin countries, the presence of ethnic minority groups with powerful kin in neighboring countries can spur nationalizing campaigns directed at expulsion.Footnote ⁶⁵ Forced displacement of minority groups during civil war aims at dampening violent group mobilization,Footnote ⁶⁶ removing potential “fifth columns,”Footnote ⁶⁷ and consolidating territorial control.Footnote ⁶⁸

When groups are subject to wartime violence and discrimination, co-ethnic elites in neighboring countries frequently pursue policies intended to support kin abroad. Policy responses include sponsoring co-ethnic rebel groups,Footnote ⁶⁹ giving humanitarian aid to kin in war-affected countries,Footnote ⁷⁰ and punitively sanctioning countries engaged in expulsion.Footnote ⁷¹ These policies aim to deter or limit civil war violence and displacement targeted at foreign kin.

We argue that countries may also respond to the outbreak of civil war in their neighborhood with measures intended to provide an exit option for targeted co-ethnics. By liberalizing asylum and refugee policies, prospective host countries can assist kin fleeing the conflict and create a hospitable and open environment to ease integration. Consistent with this view, recent work suggests that de facto migration policies favor co-ethnic FDP.Footnote ⁷²

Specifically, policy liberalization is more likely when the co-ethnics of national political elites are excluded from political power in neighboring countries. In the context of civil war, relative disparities between ethnic groups in access to state power portend violent mobilization.Footnote ⁷³ Under such conditions, asylum policy changes in prospective host countries are likely to affect individuals with whom policymakers share an identity. Liberalization, then, follows as a function of co-ethnic solidarity.Footnote ⁷⁴

When co-ethnics flee persecution, host state elites may seek to improve their conditions by liberalizing policy, especially if such assistance is supported by the public. Policy liberalizations undertaken in Armenia and Azerbaijan between 1992 and 1999 are instructive. As conflict raged over the disputed region of Nagorno-Karabakh, the expectation of co-ethnic refugee flows induced policymakers to institute extensive, liberal policies for FDP.Footnote ⁷⁵

Of course, not all regimes face equal incentives vis-à-vis co-ethnics.Footnote ⁷⁶ We anticipate that more democratic regimes are more likely to enact more liberal asylum policies when co-ethnics of their political elites are discriminated against abroad. Crucially, kinship serves as a heuristic for parties and voters in the information-poor environments of the global South.Footnote ⁷⁷ As a result, elites in electorally competitive regimes face incentives to wield policymaking with an eye toward ethnic kin. Welcoming coethnics can broaden an incumbent's power base and improve their electoral fortunes.Footnote ⁷⁸

A failure to assist ethnic kin in times of need may prove unpopular for incumbent politicians.Footnote ⁷⁹

External Economic Assistance and Policy Liberalization

Apart from transnational kinship ties, external economic assistance may also affect migration policymaking in the global South. In the context of civil war, external assistance to conflict-affected countries and neighboring states often rises significantly, both to suppress violence and to facilitate reconstruction.Footnote ⁸⁰ Aid also floods into war-wracked regions in direct response to forced displacement.Footnote ⁸¹ Since developed countries increasingly seek to restrict the flow of refugees from South to North, they have an interest in promoting more liberal policy regimes in developing countries, especially those proximate to ongoing conflicts. Policy liberalization in the global South helps absorb regional FDP flows, reducing the hosting burden of countries in the global North.Footnote ⁸² The incentive for developed countries to deflect FDP is especially great when flows originate from places experiencing civil war; publics perceive FDP fleeing conflicts as security risks, and often oppose their resettlement.Footnote ⁸³

Foreign aid is critical to many developing countries, so rich countries can use it to pressure them to reform their asylum and refugee policies. Of course, developed countries must offer additional assistance to offset the costs of hosting large displaced populations.Footnote ⁸⁴ Bubb, Kremer, and Levine show that foreign assistance is an efficient solution to the negative externalities of policy restrictions in the West, and thus Western actors should compensate developing countries for hosting forced migrants.Footnote ⁸⁵

There is mounting empirical evidence that such a dynamic is unfolding.Footnote ⁸⁶ Some donor countries now explicitly tie external assistance packages to cooperative agreements with developing countries over migration management.Footnote ⁸⁷ Correspondingly, countries such as Turkey,Footnote ⁸⁸ Jordan,Footnote ⁸⁹ Libya,Footnote ⁹⁰ and MoroccoFootnote ⁹¹ are known to engage in such “migration diplomacy.”

Not all aid-receiving countries are equally vulnerable to donor pressure.Footnote ⁹² We hypothesize that external economic assistance to developing countries will be most effective at buying policy liberalization when the recipients are both aid dependent and illiberal. These countries are vulnerable to pressure from developed countries keen to deflect FDP flows and they stand to curry international favor by adopting a more open displacement policy regime such as Western actors desire.

Illiberal countries are more likely to respond to external economic incentives for several reasons. For one, autocrats are more insulated from public opinion, reducing the risk of domestic backlash to policy liberalization. Second, for illiberal countries that depend on external assistance, liberalizing asylum policy may be a strategic move to deflect attention from ongoing abuses by demonstrating a (symbolic) commitment to liberal international norms.Footnote ⁹³

The benefits of strategic policy liberalization for repressive regimes, then, are threefold. First, by building international credibility through asylum policy liberalization, illiberal regimes can generate goodwill, which in turn serves as political cover for lingering abuses. Second, liberalization makes authoritarian regimes seem like pliant partners, facilitating durable relationships with Western donors. Third, autocrats often rely on patronage to remain in power. Cooperation with the West on asylum policy can be a lucrative source of external revenue that can be used to sustain patronage networks, as in Gaddafi's Libya.Footnote ⁹⁴ Kenya's liberalization under Moi and Mozambique's liberalization under Chissano are also illustrative.Footnote ⁹⁵ These leaders pursued displacement policy liberalization to garner sustained international economic assistance while engaging in continued domestic repression.

Measuring Asylum Policy in the Developing World

To test the hypotheses, we construct an original data set of all national laws pertinent to forcefully displaced populations in a sample of ninety-two African, Middle Eastern, and South Asian countries between 1951 and 2017. In total, the DWRAP data set includes 229 unique, national-level migration laws.Footnote ⁹⁶ Construction of DWRAP considerably expands the geographic and temporal scope of asylum policy indices. In Table 1, we outline the temporal and geographic coverage of existing migration policy data sets compared to DWRAP, demonstrating the extent to which existing data are OECD-centric. Because there has been no systematic mapping of domestic laws and policies on forced displacement in the developing world, most of the world's asylum seekers and refugees reside in developing countries that are excluded from existing data sets. DWRAP thus represents the most expansive coding of asylum and refugee policies in the developing world to date.

Table 1. Country coverage of prominent migration policy indices

Notes: “Asylum specific” indicates whether the index includes indicators specific to displacement policy. Hatton's data represent the only other index we are aware of that is dedicated solely to mapping asylum and refugee policy. Other data sets capture asylum policy as one component in a broader tracking of immigration policy. Statistics for Beine et al. refer to planned coverage. The year range for de Haas, Natter, and Vezzoli is actually longer for some countries, but coding is most reliable after 1945. The single European country included in DWRAP is Cyprus, because it is in the Western Asia region of the UN geoscheme.

We conceptualize refugee and asylum policy as a combination of policy provisions regulating five core dimensions—access: the ease of entrance and security of status; services: provision of public services and welfare; livelihoods: the ability to work and own property; movement: encampment policies; and participation: citizenship and political rights. Consequently, for each law we code fifty-four provisions across these five policy fields. To allow fine-grained aggregation, we categorize the five policy dimensions into fourteen policy strands: status security, control measures, family unity, legal recourse, education, aid, health care, property, land, employment, settlement policy, document access, citizenship, and political rights. The provisions we code were selected after an exhaustive reading of migration policy coding guidelines employed in other key data sets, as well as UNHCR's reports on asylum policy. More details about the coding procedures are provided in the online supplement.Footnote ⁹⁷ The empirical tests suggest that our policy indices have construct validity by showing they associate with observed variables in ways consistent with theoretical expectations.Footnote ⁹⁸

Our data cover de jure policies on asylum and forced migration. Focusing on de jure (rather than de facto) policies has theoretical and empirical advantages. First, de jure policies offer a more objective measure because coding is based on legal texts, rather than subjective judgments about policy enforcement. Second, the international community can advocate adoption of certain policies, but enforcing implementation is much more difficult. By studying de jure policies, scholars can explore how policies, once implemented, create their own constituencies advocating implementation. For example, when Uganda mandated that refugee children have access to health and education, local communities near settlements pushed for the expansion of schools and clinics that serve both refugee and host populations.Footnote ⁹⁹ Third, charting the de jure environment in countries is important for understanding whether gaps, when they emerge, are a product of deficient policy frameworks or deficient enforcement of existing policies.

In developing DWRAP, we selected countries for inclusion according to their UN geoscheme region. Specifically, we selected all countries in Eastern Africa, Middle Africa, Northern Africa, Southern Africa, Western Africa, Western Asia, Central Asia, and Southern Asia. These regions were selected because they are underrepresented in existing migration policy indices, and because countries in these regions are large origin and destination states for externally displaced populations. In 2017, about 70 percent of all asylum seekers originated from and 35 percent sought refuge in DWRAP-covered countries; likewise, DWRAP countries produced 87 percent and hosted 81 percent of the world's refugees.Footnote ¹⁰⁰

We use 1951 as our start date because it marks the signing of the landmark Convention Relating to the Status of Refugees.Footnote ¹⁰¹ Before World War II, countries generally did not adopt formal asylum and refugee policies, and the announcement of the 1951 Convention was a watershed moment in the diffusion of national and international asylum policy. Since coding back sixty-six years raises measurement challenges, we evaluate the texts of laws individually, going systematically through national legal frameworks to code along the fifty-four provisions outlined in Table A.3. This approach facilitates reliable coding of historical policies.Footnote ¹⁰² By contrast, most existing migration policy indices rely on expert surveys. While this has the benefit of enabling an assessment of de facto policy provisions, it limits how far back in time one can code.Footnote ¹⁰³

The corpus of laws and policies pertinent to forced migration was identified chiefly using UNHCR submissions to the Universal Periodic Review, a mandated, cyclical review of UN member states organized by the Office of the High Commissioner for Human Rights. Individual countries, UN agencies, and third-party stakeholders may submit publicly available reports on human rights practices in individual countries under review to a Compilation Report. In practice, the UNHCR submits to the Universal Periodic Review process for virtually every country in a given cycle. UNHCR submissions detail the evolution of a state's forced displacement policies, or lack thereof, as well as states’ international legal obligations to FDP, and details of states’ de facto protection environments, including instances of refoulement and other violations of migrants’ rights. We use legal instruments referenced in UNHCR submissions to identify key laws and policies in individual states.Footnote ¹⁰⁴ We supplement information from UNHCR submissions with information from the UNHCR's Refworld database, the International Labour Organisation's NATLEX database, the International Organisation for Migration's Migration Law Database, the UN Office on Drugs and Crime's SHERLOC database, and the Law Library of Congress. Using these sources, we are able to locate full texts of more than 90 percent of the national laws in our data set. We rely on secondary sources, including historical reports and NGO assessments, to code provisions of the laws for which we cannot locate full texts.Footnote ¹⁰⁵

We use a straightforward aggregation procedure to transform the policy coding into a scale of displacement policy liberality. Specifically, we use a series of summary indices to aggregate from individual policy provisions to policy strands, policy strands to policy fields, and policy fields to policies. Each summary index is the mean of standardized outcomes weighted by the inverse of the covariance matrix, which maximizes the information captured in the index. To test the robustness of our results to the aggregation schema, we also show that principal component analyses give a similar decomposition of the data (Table A.5), and that identical results emerge when we aggregate indices using the equally weighted mean of standardized outcomes. Each index is further scaled to range from 0 to 1. By constructing index scores for policy strands, policy fields, and policies, we ensure comparability of policy regimes within and across countries over time.

Two Stylized Facts

Drawing on these new data, we present two stylized facts about displacement policies in developing countries. First, there is much diversity in asylum policy liberality, as shown in Figure 1, which plots asylum policy scores from DWRAP in 2017, the last year of our data set. This variability, which cannot be simply explained away by regional clustering, underscores the importance of our theory-building exercise.

Figure 1. Global South displacement policy in 2017

Second, our data suggest that developing countries have been gradually liberalizing their asylum and refugees policies at a time when developed countries have been moving in a more restrictive direction.Footnote ¹⁰⁶ Figure 2 depicts the sample average increase in our asylum policy measure over time.Footnote ¹⁰⁷ Policy liberalization has been especially pronounced with respect to the access and movement domains. Moves away from encampment and toward enhanced access to travel and identification documents are particularly important drivers of this trend. By comparison, liberalization of policy on livelihoods and participation has been more slight, on average. These policy dimensions, including employment and citizenship rights, are critical for refugee integration, but also politically sensitive, often sparking native opposition.Footnote ¹⁰⁸

Figure 2. Global South displacement policy over time

Importantly, the liberalizing trend in asylum policy across the global South is taking place at a time when refugee policies are becoming increasingly restrictive (Figure 3) in OECD countries. Divergent asylum policy trajectories across the global North and global South reinforce our study's assumption that a different set of analyses is needed to understand migration policymaking in the developing world.

Figure 3. Global North displacement policy over time

Empirical Strategy

To test the study's hypotheses, we leverage our fine-grained measure from DWRAP. We focus on when countries substantially reform their refugee and asylum policies. This focus reflects the fact that policies tend to change suddenly, not gradually—in fact, no change occurs in most country-years. In our primary analyses of policy determinants, therefore, the dependent variable is an indicator for policy change.

We define a policy reform as a one-standard-deviation change in a country's policy score from year t-1 to year t.Footnote ¹⁰⁹ We focus on one-standard-deviation changes because these reflect substantial, substantive modifications to countries’ asylum policy environment. Changes of smaller magnitude typically capture more limited legal and procedural modifications to policy frameworks, rather than substantive changes in policies themselves. In total, we record 71 changes of at least one standard deviation, including sixty-two liberalizations and nine restrictions, from 1951 to 2017.Footnote ¹¹⁰ Correspondingly, we use two binary dependent variables in our primary models: change, which refers to a one-standard-deviation shift in a country's policy score; and liberalization, which refers to a one-standard-deviation increase in a country's policy score. We focus on policy liberalization because we document many more liberalizing than restrictive changes. Nevertheless, we report results for restriction, a one-standard-deviation reduction in a country's policy score, in Table A.36.

We plot all policy changes of at least one standard deviation in the top panel of Figure 4, and the number of such changes over time in Figure A.14. All such changes are described in Table A.15. More than half of all asylum policy changes of at least one-quarter of a standard deviation (71 of 135 cases) are changes of at least one standard deviation. These reforms are temporally concentrated after the end of the Cold War, a period characterized by an increased incidence of civil war, and consequently, a dramatic increase in the global stock of refugees (bottom panel of Figure 4).

Figure 4. Policy reforms (top panel) and civil wars and forced migrant stocks (bottom panel) over time

Variable Description

We consider several independent variables corresponding to Hypotheses 1 through 3. To test our expectation about the relationship between conflict and policy change, we use data from the UCDP Armed Conflict data set.Footnote ¹¹¹ Usefully for our purposes, this data set disaggregates armed conflicts by intensity measured in battle deaths, which permits us to test whether the mere occurrence of conflict in a country's neighborhood is associated with asylum policy changes, or whether only intense conflicts (which are more likely to generate refugee flight) drive reforms. We use a binary measure of intense, proximate civil wars in our main analyses, which takes a value of 1 if a country has a contiguous neighbor experiencing a civil war causing 1,000 or more battle deaths in the prior year and 0 otherwise.

Hypotheses 2a and 2b refer to transnational ethnic kin. Data on politically relevant ethnic groups, their access to state power, and their transnational ties come from the Ethnic Power Relations (EPR) data set.Footnote ¹¹² Our measure of transnational ethnic power disparity is a binary indicator. Following EPR, we define elite ethnic groups as those that enjoy a status of at least the senior partner level. These groups exert influence on policymaking processes by virtue of the positions their members hold in government. Politically excluded ethnic groups are defined as those that are powerless or discriminated against. Excluded groups face tacit or overt barriers to accessing political institutions. Combining these definitions, we expect policy liberalization when transnational co-ethnics of a country's political elites are excluded abroad. Transnational kin tend to be regionally concentrated, and status differentials are most salient for groups in close proximity.Footnote ¹¹³ Thus our indicator takes a value of 1 when a country's political elites (ethnic groups at the senior partner level or higher) have politically excluded transnational kin (groups that are powerless or discriminated against) in another country in the region; and 0 otherwise.Footnote ¹¹⁴

Hypotheses 3a and 3b concern countries dependent on external economic assistance. To measure aid dependence, we take the five-year lagged moving average of net bilateral aid from donors on the Development Assistance Committee (DAC) of the OECD, divided by a country's GDP.Footnote ¹¹⁵ Because this measure is skewed, we transform it by taking the inverse hyperbolic sine. Our conditional hypotheses entail interactions between the level of democracy and elite kin discrimination (H2b) and aid dependence (H3b). To measure democracy, we use V-Dem's electoral democracy (polyarchy) index.

Our core specification also includes covariates that capture other factors relevant for migration policymaking. We measure wealth using PPP-adjusted GDP per capita. We also measure GDP shocks, defined as annual reductions in PPP-adjusted GDP in the top 15 percentiles of negative change. Other covariates include countries’ population, transnational terror incidents, and trade-to-GDP ratio. Time-variant covariates are lagged one year to mitigate temporal confounding. To conserve space, we present variable definitions and descriptive statistics in Tables A.16 and A.17 in the online supplement.Footnote ¹¹⁶

Estimation

Because we are interested in modeling when developing countries reform their asylum policies, our primary estimation is a series of hazard models.Footnote ¹¹⁷ The hazard is the conditional probability of failure. Repeated failures are a feature of the process we model: some countries reform their asylum policies several times over the course of the study period. For instance, Burkina Faso made substantial liberalizing changes in its policy framework in both 1988 and 2008. More broadly, repeated failures—in this case, asylum policy changes of at least one standard deviation—account for about one-quarter (17 of 71) of the changes we observe.

In traditional hazard modeling approaches, event times are assumed to be conditionally independent. This assumption is unlikely to hold in our setting for two reasons. First, event dependence means the risk of failure is correlated within units. The clustering of terminations among countries means that previous failures raise the risk of future failures. Second, there is likely unobserved heterogeneity across countries with respect to failure susceptibility. Covariates account for some of this heterogeneity, but it is unlikely that we can control for all factors that affect the risk of policy change, given the empirical complexity of migration policymaking. If unit heterogeneity in the propensity for policy change is also correlated with the likelihood of neighborhood armed conflict, our estimates will be biased.

Given the dual threats of event dependence and unobserved unit heterogeneity, we follow Box-Steffensmeier, De Boeuf, and Joyce and employ a conditional frailty gap-time Cox model.Footnote ¹¹⁸ This model is conditional because it stratifies the risk set by the number of failures a unit has experienced. By allowing the baseline hazard to vary by the number of failures, stratification helps control for event dependence. We use gamma-distributed frailties to account for unobserved heterogeneity in each country's propensity to fail.Footnote ¹¹⁹ This shared-frailty approach allows for partial pooling across countries while accounting for heterogeneity. Frailties are thus equivalent to random effects in other modeling contexts. Because we employ a Cox estimator, we make no additional parametric assumptions about the shape of the hazard function. Our choice of estimator follows other prominent analyses.Footnote ¹²⁰

The hazards in our framework can be written as

(1)$$\lambda _{u, s}( t) = \lambda _{0s}( t-t_{s-1}) e^{\kappa ( \hbox{intense, proximate civil conflict}_{u, s}) + \theta ( X_{u, s}) + \gamma _u}$$

(2)$$\lambda _{u, s}( t) = \lambda _{0s}( t-t_{s-1}) e^{\alpha ( \hbox{kin status}_{u, s}) + \beta ( \hbox{aid dependence}_{u, s}) + \theta ( X_{u, s}) + \gamma _u}$$

where (1) corresponds to H1 and (2) corresponds to H2a and H3a; u indexes countries, and s denotes the event number, which stratifies the risk set. λ₀s is the baseline hazard rate, and (t − t_{s −1)} specifies a gap-time formulation, where the hazard is the risk of failure for event s since the occurrence of event s − 1. k is the effect of intense civil conflict occurring in country u's neighborhood. α gives the effect when country u's political elites have excluded kin abroad. β is the effect of aid dependence. θ are estimates from a vector of covariates, and γ_u are gamma-distributed, unit-specific frailty terms. In all models, we cluster standard errors by country and use Efron's method for ties. For simplicity, we report standardized coefficients rather than hazard ratios.

Qualitative Data

Our empirical analysis also benefits from insights gleaned from 126 qualitative, semistructured interviews and focus group discussions with FDP, UN, and government officials, and NGO representatives, conducted in Uganda in June and July 2017.Footnote ¹²¹ We selected Uganda for our fieldwork because it hosts an exceptionally large forced migrant population—fourth-most globally in total stock and seventh-most on a per capita basisFootnote ¹²²—and because its asylum policy has drastically liberalized in the past decade. Uganda is a typical case in terms of responsiveness of inflows to policy.Footnote ¹²³

The evolution of Uganda's policy environment is depicted in Figure A.21. Between 2006 and 2017, Uganda's policy score increased 538 percent, and its annual number of forced migrant arrivals increased 2,811 percent. This dramatic policy liberalization reflected the passage of Uganda's 2006 Refugees Act and 2010 Refugees Regulations, which replaced the Control of Aliens Refugees Act of 1960. These policies are widely regarded as some of the most liberal in the developing world, affording refugees rights such as free movement, education, employment, and plots of land for cultivation. The case is not intended as a decisive test of the theory; rather, qualitative insights interspersed throughout help illustrate theoretical mechanisms and provide context for our quantitative results.

Civil Conflict and Asylum Policy Change

Turning to our empirical analysis, we find robust support for Hypothesis 1. Armed conflicts in neighboring countries are the primary impetus for asylum policy change in the developing world. We argue earlier that such conflicts shift expectations of FDP inflows, which in turn induce prospective host countries to reassess (and at times reform) their asylum policies. As a preliminary test, contingency tables (Tables A.23–A.25 in the online supplement) indicate that policy changes occur disproportionately when a country's neighbors are experiencing civil wars. The association is especially striking when, in year − 1, a neighboring country is experiencing an intense civil war, defined as causing at least 1,000 battle deaths in the prior calendar year. Such episodes account for nearly 40 percent of all one-standard-deviation policy changes.

Table 2 provides a more formal test of Hypothesis 1. In columns 1 through 7 we define a country's neighborhood in terms of contiguity, so coefficients refer to the occurrence of armed conflict in any of a state's territorial neighbors. In column 8 we define a country's neighborhood in terms of 1,500 kilometers of inter-capital distance, so coefficients refer to the occurrence of armed conflict in the broader set of countries surrounding a state; this column ensures that our core result is robust to how we define a country's neighborhood. The main dependent variable (used in columns 1, 2, 3, and 8) is an indicator for one-standard-deviation policy changes. Column 1 estimates a sparse model including only the coefficient of interest, our conflict indicator, and country-level frailty terms (i.e., random effects). In column 2 we introduce additional controls for political and socioeconomic conditions in a country.Footnote ¹²⁴ Columns 3, 5, and 7 use the equally weighted policy index to ensure that our results are robust to the aggregation scheme. To verify that our core finding is robust to the change threshold we choose, in columns 4 and 5 we study one-half-standard-deviation policy changes as the dependent variable, and in columns 6 and 7, one-and-one-half-standard-deviation policy changes.

Table 2. Conditional frailty models of policy change

Notes: Robust standard errors clustered by country in parentheses. The table displays standardized coefficients rather than hazard ratios. The models are stratified by the number of one-standard-deviation policy changes a country has made. Frailty terms are for country. Efron's method is used for ties. ICW refers to our main index, which is aggregated by inverse-covariance weighting. EW refers to our alternate, equally weighted index. *p < .10; **p < .05; ***p < .01.

Across all models, we find that countries are more likely to reform their asylum policies when intense civil wars are occurring in a neighboring state. Exponentiating the large, statistically significant coefficient from column 2, our best-fitting model, suggests that when one of a country's neighbors is experiencing an intense civil war, the target country is approximately 395 percent more likely to change its asylum policy by at least one standard deviation. The minimum estimated coefficient from column 8 still suggests that when one of a country's neighbors is experiencing an intense civil war, the target country is approximately 53 percent more likely to alter its displacement policy by at least one standard deviation. This result helps contextualize the importance of Fearon and Shaver's conclusion that battle deaths, and expectations thereof, drive refugee outflows during civil wars.Footnote ¹²⁵

As an additional check, in Figure A.26 we consider the relationship between neighborhood armed conflict and policy change at different levels of conflict intensity. We find that the policy-change-inducing effects of neighborhood armed conflicts are specific to intense civil war episodes rather than low-intensity conflicts. This suggests that conflict spurs policy reform by increasing neighboring countries’ expectations of FDP flows.Footnote ¹²⁶ More broadly, this finding lends confidence in our specification, and suggests that civil war battle deaths drive both forced displacement and displacement policy changes.

The Determinants of Policy Liberalization

Elites and Co-ethnics

Hypotheses 2a and 2b anticipate that countries, and particularly more democratic countries, enact liberalizing reforms when elites’ co-ethnics are discriminated against abroad. We assess these hypotheses in Table 3. Columns 1 and 5 represent our baseline, conditional frailty specifications. In columns 2 and 6 we add frailty terms for years. In columns 3 and 7 we introduce spatial lags to capture policy diffusion. These terms assess whether the probability of liberalization changes when a regional neighbor liberalizes, or when a country that shares the same legal origins liberalizes. To address the potential endogeneity of elite kin discrimination, we estimate instrumental variable (IV) models in columns 4 and 8. We instrument for elite kin discrimination using a country's number of contiguous neighbors and the fractal dimension of its borders.Footnote ¹²⁷ Countries with more contiguities have more split, cross-border ethnicities, raising the odds that an ethnic group is politically empowered in one country but excluded in another. Likewise, straighter borders reflect colonial-era boundary-making ignorant of local ethnic conditions. Tests suggest the proportional hazards assumption is satisfied (Figure A.27).

Table 3. Co-ethnicity, transnational kinship, and policy liberalization

Notes: Robust standard errors clustered by country in parentheses. The table displays standardized coefficients. The models are stratified by the number of liberalizing policy changes a country has made. Efron's method is used for ties. Democracy is a standardized index from V-Dem. Additional controls include neighborhood civil war, aid dependence, GDP per capita, negative GDP shocks, population, trade-to-GDP ratio, and transnational terrorism and civil conflict in the host country. IV Cox models are fit using the two-stage procedure described in Sjolander and Martinussen Reference Sjolander and Martinussen2019. *p < .10; **p < .05; ***p < .01.

We find robust support for Hypotheses 2a and 2b. Because elite kin discrimination is a binary variable, the substantive interpretation is straightforward. In line with Hypothesis 2a, the association between elite kin discrimination and policy liberality is large, positive, and statistically significant. Exponentiating the coefficient from our best-fitting specification, in column 2, suggests that the hazard of liberalization is 350 percent greater when national political elites’ co-ethnics in neighboring states are politically discriminated against. Elite kin discrimination increases the probability of policy liberalization by 818 percent in our IV model (column 4).

Interviews we conducted in Uganda corroborate the role of elite concerns about ethnic kin in neighboring states. For example, an international-NGO employee working on refugee outreach in Kampala told us that border artificiality in Uganda meant the logic of governance often made co-ethnicity a paramount regional policy consideration:

Some colonialists draw a line here, but the tribes are split on both sides, and that's really clear. It is that from northern Uganda to South Sudan, even DRC, they are brothers and sisters. Even literally, they have family on both sides … The reason is mainly ethnic contiguity … In Uganda it's all about your tribe, where your village is. And all of this—tribes and villages—which are then more connected to South Sudan or DRC than other places. So maybe these refugees are treated a little better.Footnote ¹²⁸

Other interviewees specifically highlighted Ugandan elites’ considerations about co-ethnic Bantu peoples in Rwanda and the Democratic Republic of the Congo as a cause of liberal policies toward FDP from these countries.Footnote ¹²⁹ During the 1994 Rwandan genocide, Ugandan president Yoweri Museveni welcomed Rwandan refugees as an extension of his campaign to support Rwandan rebel leader Paul Kagame. Kagame was raised in the same region of Uganda that Museveni was from, and served as a founding member of Museveni's National Resistance Army in the 1980s. An international aid worker in Uganda told us that ties between Museveni and Kagame were emblematic of the broader power of “Ugandan-Rwandese at the higher end of [Ugandan] society,” who promoted Ugandan openness toward co-ethnic refugees from Rwanda.Footnote ¹³⁰

In columns 5 to 8 we probe the conditional effect of elite kin discrimination across levels of democracy. We argue that incumbent elites have incentives to welcome co-ethnic FDPs as a means of bolstering their bases of support, particularly in electorally competitive environments. Elite kin discrimination significantly increases the probability of liberalization at all levels of democracy (H2a). In line with H2b, moreover, this effect is greater in more liberal countries. Coefficients on the interaction term are consistently positive, and achieve statistical significance in our best-fitting specification, in column 6. For ease of interpretation, we plot substantive effects in Figure 5. The effect of kin discrimination on the risk of policy liberalization increases about 335 percentage points as democracy increases from the twenty-fifth to the seventy-fifth percentile. This suggests that an electoral logic (at least partly) underlies elites’ considerations vis-à-vis kin.Footnote ¹³¹

Figure 5. Elite kin discrimination has a larger effect in more democratic settings

Importantly, more democratic countries may find it harder to liberalize migration policy because without incentives to welcome FDP owing to co-ethnicity, publics often oppose migrant influxes.Footnote ¹³² Estimates in Table 3 corroborate this intuition. Without liberalizing pressures tied to transnational kin, a one-standard-deviation increase in democracy reduces the probability of liberalization by 110 percent. Finally, asylum policy appears to diffuse spatially within regions. Regional diffusion makes sense in our theoretical framework because neighborhood civil wars and kin discrimination are regionally concentrated, cross-border phenomena.

Robustness of Liberalization Results

To probe the robustness of our core findings we conduct a host of additional tests. First, in Figure 6 we plot the estimated coefficients for our main variable, elite kin discrimination, at different thresholds of policy liberalization, and using a different aggregation scheme for our index. Elite kin discrimination explains large policy changes, but only weakly predicts smaller, procedural policy changes. This finding comports with our theoretical logic, because policymakers have incentives to make conditions as hospitable as possible when co-ethnic flows are anticipated.

Figure 6. Robustness of elite kin discrimination results

In Figure A.28 we re-estimate our core specification with a series of additional controls. The effect of elite kin discrimination on asylum policy liberalization remains large, positive, and significant, even after we control for a country's policy level, colonial history, unemployment, natural resource endowments, Chinese economic assistance, ruling party political orientation, strategic wartime displacement in the country's neighborhood, the Cold War and post-9/11 eras, and the number of laws and policies on displacement a country has enacted. The effect of elite kin discrimination also holds when we exclude country-years with no migration laws enacted, and exclude newly independent countries with an incentive to shift policies inherited from colonial powers.Footnote ¹³³

Figure A.29 verifies that our results are robust to the inter-capital distance threshold we use to define the region within which status differentials between kin groups are most salient. Figure A.30 tests for influential observations by iteratively dropping countries that experienced liberalizing policy changes above the ninetieth percentile of the annual difference in policy score. In Table A.31, we consider a competing-risks estimator. In Table A.32 we depart from a survival modeling approach and instead estimate two-stage Heckman and conditional mixed process models, along with a random effects OLS model. Here, we take the difference in a country's policy score from year t − 1 to year t as the main outcome of interest. Our main finding that the risk of policy liberalization increases when political elites’ co-ethnics are discriminated in neighboring countries holds.

Finally, in Figure A.33 we disaggregate our full index measure into its constituent policy fields: access, services, livelihoods, movement, and participation. Elite kin discrimination is associated with one-standard-deviation liberalizations of access, services, livelihoods, and participation, and the largest effect is on livelihoods. These results provide additional support for our theoretical logic. Given the centrality of employment rights for migrant integration and self-reliance,Footnote ¹³⁴ it makes sense that policy liberalization targeted at elites’ co-ethnics should prioritize the policy domains most consequential for migrant integration.

Transactional Aid

Moving to H3a and H3b, is it the case that countries dependent on external economic assistance engage in strategic asylum policy liberalization? We argue that aid-dependent countries are vulnerable to pressure from developed donors, which promote policy liberalization as a means of containing FDP in the global South. This incentive is particularly great for illiberal aid-dependent regimes, which can use strategic policy liberalization to cultivate goodwill and generate political cover. In the Ugandan case, these expectations seem to hold. Uganda is an authoritarian regime that engages in substantial violations of civil liberties but maintains a liberal asylum policy, which helps the regime retain generous external assistance packages and generates political cover for the government's illiberal activities. For instance, one activist told us:

For what illiberal practices does Uganda's liberal policy provide cover? Another humanitarian worker we spoke with pointed to Uganda's military engagements in East Africa, as well as Museveni's business interests in the region:

To explore whether these dynamics generalize, we conduct several additional tests (Table 4). As in Table 3, we introduce the main conditional frailty specification in columns 1 and 5, add year frailties in columns 2 and 6, and test for diffusion dynamics in columns 3 and 7. We include an indicator for whether a country in the top decile of DAC aid inflows liberalized its asylum policy in the prior three years. The logic is that countries should emulate the policies of top aid recipients, understanding that liberalization often accompanies increased aid flows.

Table 4. Aid dependence and policy liberalization

Notes: Robust standard errors clustered by country in parentheses. The table displays standardized coefficients. The models are stratified by the number of liberalizing policy changes a country has made. Efron's method is used for ties. DAC aid/GDP is the standardized, five-year lagged moving average of DAC aid receipts as a percentage of GDP. Democracy is a standardized index from V-Dem. Additional controls include neighborhood civil war, elite kin discrimination, GDP per capita, negative GDP shocks, population, trade-to-GDP ratio, and transnational terrorism and civil conflict in the host country. IV Cox models are fit using the two-stage procedure described in Sjolander and Martinussen Reference Sjolander and Martinussen2019. *p < .10; **p < .05; ***p < .01.

Caution is required when interpreting these results, however, because DAC aid inflows are endogenous. It is difficult to identify the causal effect of aid on asylum policy reform because reforms may follow aid, or aid may follow reforms.Footnote ¹³⁷ Aid is also endogenous to refugee flows.Footnote ¹³⁸ In columns 4 and 7 we therefore estimate an IV model, instrumenting for DAC aid inflows using the interaction of donor-government fractionalization and each recipient country's probability of receiving aid.Footnote ¹³⁹

Contrary to Hypotheses 3a and 3b, we find no evidence that aid-dependent countries are more likely to liberalize asylum policy, or that this effect is greater among more illiberal regimes. The evidence suggests that Uganda's case does not generalize widely.Footnote ¹⁴⁰ Coefficients are imprecisely estimated and occasionally significantly negative. Based on our best-fitting specification (column 2), increasing aid dependence by one standard deviation reduces the risk of liberalization by 89 percent. In Figure A.34 we plot substantive effects of the interaction term, which reveal little variation in the effect of aid dependence across levels of democracy. In Figure A.35, we show that the effect of aid dependence remains null or negative when we control for the strategic relevance of host countries, which may give them greater leverage over aid.Footnote ¹⁴¹

Given the qualitative evidence from Uganda, what explains the absence of a more general relationship between aid and liberalization? Several possibilities stand out. It could be that highly aid-dependent countries have low capacity, and merely liberalize their de facto environments, such as by relaxing enforcement, without changing their de jure frameworks. We cannot test this proposition, but it is a valuable avenue for future work. Two other possibilities also stand out. In columns 3 and 7 of Table 4 we include an indicator for whether a country in the top decile of DAC aid inflows liberalized its asylum policy in the prior three years. We expected that aid-dependent countries would emulate top aid recipients. Instead, we find that the risk of policy liberalization declines 386 percent when top aid recipients have recently liberalized.

This suggests two potential phenomena. First, it may be that countries observe liberalization by top aid recipients and free ride. If the fiscal fruits of liberal policy reforms (e.g., increased aid) are generally insufficient to cover increased expenditures stemming from a concomitant rise in FDP arrivals, developing countries considering liberalization might refrain from doing so when top aid recipients—which can better afford the increased hosting burden likely to accompany liberalization—make liberal reforms.Footnote ¹⁴² Second, it could be that developing countries learn about donor commitment and demands by observing liberalizations undertaken by top aid recipients. In particular, other states may observe the extent to which donors durably increase funding to top, liberalizing recipients, and the extent to which donors demand increased accountability over aid given in return for policy liberalization. If donor commitment to top aid recipients wanes quickly after the latter liberalize, or if donors demand intrusive accountability mechanisms over transactional aid-for-liberalization, countries may refrain from liberalizing when top recipients do because of “negative” learning.

Our qualitative evidence comports with this explanation. While they did suggest that Uganda engaged in transactional liberalization, our interviewees also expressed concern about the failure of the Uganda Solidarity Summit on Refugees, held in Kampala on 22 and 23 June 2017. The summit, which sought to raise USD 2 billion, raised only USD 385 million to fund Uganda's liberal policy. The head of a UNHCR implementing partner organization told us:

Nor is this problem unique to Uganda. A Tanzanian aid worker also spoke about Tanzania's failed plan to naturalize Burundian refugees:

This evidence suggests that the problem of insufficient aid is ubiquitous, and implies that one reason for the null results on H3a and H3b is that developing countries are justifiably skeptical of aid-for-liberalization schemes. But built-in donor accountability mechanisms may also deter aid-dependent regimes from accepting aid in exchange for policy reform. In Uganda, a damning UN audit released in November 2018 uncovered massive fraud, including inflated FDP figures, waste, and undocumented payments to government officials from the refugee aid budget. Four government officials, including the commissioner for refugees, were sacked as a result, and in February 2019 the UK and Germany suspended refugee aid to Uganda until “the implementation of stringent integrity measures.”Footnote ¹⁴⁵ This episode highlights a potential cost of aid in return for liberalization—increased donor scrutiny—which could deter countries from pursuing transactional aid.

Do De Jure Policies Matter?

Our DWRAP data set measures the de jure rather than the de facto policy environment. An obvious question is whether de jure forced displacement policies really matter, given potential enforcement gaps. A standard approach to build confidence in migration policy indices is to regress policy scores on migrant flows.Footnote ¹⁴⁹ Establishing that more liberal policies attract more FDP flows increases confidence in our policy measure. To the best of our knowledge, no study outside the OECD context has tested whether the target country's policies influence asylum seekers’ decision making. Encouragingly, in a companion paper we demonstrate that more liberal policies are associated with greater FDP flows, conditional on facilitating factors that enable the diffusion of information about policies in host countries, such as transnational kin and mobile penetration.Footnote ¹⁵⁰