Deterring Emigration with Foreign Aid: An Overview of Evidence from Low-Income Countries

Targeting aid toward the “root causes” of migration

The idea that development assistance can reduce migration pressures in developing countries has been widespread in policy circles for the past few decades. One of the largest expected benefits from the North American Free Trade Agreement (NAFTA) was to be an accompanying decrease in migration as economic conditions in Mexico improved. More recently, the U.S. State Department requested additional funds to contend with a surge of Central American child migrants, to “address the underlying factors driving migration or be prepared for what is likely to be an ongoing cyclical phenomenon—with significant impact and cost to the United States” (Jacobson 2015).

Addressing the “root causes” of migration has also been a main feature of development agencies across Europe since the 1990s. It continues to shape aid policy today (Knoll and Sherriff 2017). The new European Union Emergency Trust Fund for Africa is a cornerstone of the European response to the recent influx of asylum-seekers from Africa and the Middle East. Such programming often involves significant funding to facilitate return migration flows and build “migration management” capacity in origin countries. This article focuses solely on development aid interventions. It takes as given that the goals of this assistance are as stated: to foster development that renders migration less needed. Other interpretations of this rhetoric are possible.²

What “root causes” and what types of interventions are used?

What exactly are these “root causes of migration” and what sorts of aid programming are expected to mitigate them? The EU Trust Fund focuses on four key policy areas: employment creation, specifically for women and youth; basic local-level service provision; migration management; and governance, especially in regard to conflict prevention and including border management. Some public program documents mention specific interventions, such as vocational training, food security, and rule of law capacity building, but many rely on broader terms like “resilience building” and “migration management.”

It can therefore be difficult to ascertain the types and mandates of specific projects that fall under these broad umbrellas. The EU Trust Fund Sahel and Lake Chad funding window has designed a “situation index” to characterize individual country contexts along five axes: food insecurity, conflict risk, irregular migrants, socioeconomic vulnerability, and displaced people (European Commission 2016). But it is unclear whether specific interventions are meant to be targeted along these lines or if the organization’s main role is building situational awareness. Other parts of the EU Trust Fund’s work, such as the North Africa window, focus almost exclusively on “migration management.”

The United States Agency for International Development’s Central American Regional Security Initiative (CARSI) has a similar mandate in El Salvador, Guatemala, and Honduras: to address the “key factors…contributing to the migration of unaccompanied, undocumented minors to the United States” (Meyer and Seelke 2015). Economic development projects are focused on education, vocational training, and employment. Law-and-order programming is targeted to border security, antitrafficking and antigang programs, and counter-narcotics activities. As with the EU Trust Fund, more granular information is difficult to find. In general, publicly available information on these projects is relatively high level; it is likely only funders and implementing partners are aware of specific on-the-ground programming.

The effect of aid on conditions in migrant-origin countries

To deter migration, aid must not only target development outcomes that could shape migration in principle. Aid must also substantially affect those outcomes (Gregl and Logožar 2017). These effects could include contributing to the long-run growth of the economy, creating youth employment, and reducing conflict. The literature assessing aid’s effects on these outcomes is extensive. We will not summarize it. Rather, we will ask: Does the aid-effectiveness literature offer clear evidence that aid (if disbursed accordingly) can affect conditions in migrant-origin countries to a degree that is sufficient to deter migration? Here we focus on three plausibly desired outcomes: overall economic growth, job creation for youth, and conflict resolution. We expect that other aid sectors would demonstrate similar trends.

First, simply assume that aid can increase economic growth. This allows us to ask how long it would likely take for aid to deter emigration by this channel. The exercise is shown in Figure 2.⁴ As we will examine in the next section, various studies have found little systematic deterrent effect of economic growth on emigration until poor countries reach roughly PPP$8,000–10,000 in GDP per capita, such as Paraguay and the Philippines (Berthélemy, Beuran, and Maurel 2009; Clemens 2014; Dao et al. 2018; OECD 2017). The figure shows that today’s poorest quintile of countries, if they continued to grow at their historical rate of growth (over the last 24 years), would only reach PPP$8,000 in the year 2198. If development aid could systematically raise their per capita economic growth by one percentage point every year—more than doubling the historical rate—it would take until the year 2097. If aid could raise growth by two percentage points—a tripling—it would take until the year 2067.

Also, can development aid typically raise growth by these one or two percentage points, in a large group of the poorest countries over generations? This is far from well established. The research literature has not settled on a consensus that past development aid has had any positive growth impact on average (Qian 2015). Many rigorous studies fail to detect the slightest growth effect of aid. Ignore this controversy for a moment and consider the magnitude of the effect, even in the literature that does find a positive impact. For example, Clemens et al. (2012) find that raising growth by 1 percentage point per year in the average recipient would require on the order of 10 percent of recipient’s GDP in aid. That is much higher than any amounts of aid currently contemplated with the goal of deterring migration; for example, USAID’s Central American Regional Security Initiative dedicated 0.2 percent of GDP to economic growth programming explicitly intended to reduce violence driving migration in El Salvador in 2015 (Washington Office on Latin America 2017). And according to the exercise above, even that high level of growth would need to be sustained for three generations (until 2097) before it would begin to deter emigration from the average poor country.⁵

The research literature similarly invites skepticism about the ability of development aid to effect large changes in youth employment. McKenzie (2017: 20) reviews existing research on poor-country governments’ active labor market policies—efforts to increase both the number and quality of jobs. He finds that this research “shows these policies to generally be far less effective than policy makers, program participants, and economists typically expect.” There is no indication that active labor market policies supported by aid are more effective than those that are not. The least successful programs have focused on job training and matching; the most successful programs have assisted firms in overcoming regulatory barriers to hiring and assisted workers in leaving geographic areas where employment is scarce for more promising domestic destinations. There are exceptional cases where intensive job training through the best training centers has been found to substantially reduce youth unemployment (e.g., Alfonsi et al. 2017). In no case, however, has active labor market policy been shown capable of large reductions in youth unemployment at scale (Fox and Kaul 2017).

Moreover, the literature offers scant evidence that aid projects have been a systematically effective tool to mitigate civil conflict that could spark emigration. Zürcher (2017: 506) reviews all 19 existing studies on this topic that deploy some strategy for causal identification—including singlecountry studies (in such varied contexts as Afghanistan and Colombia) and cross-country studies. He concludes, “The evidence for a violencedampening effect of aid in conflict zones is not strong. Aid in conflict zones is more likely to exacerbate violence than to dampen violence. A violencedampening effect of aid appears to be conditional on a relatively secure environment for aid projects to be implemented.” This is true of all six types of aid-funded interventions he investigates—including conditional cash transfers, humanitarian assistance, and employment promotion. More research is needed; it may be that aid protects relatively stable regions from slipping into conflict, even if it is not frequently successful in mitigating ongoing conflict.

None of this implies that aid cannot affect conditions in poor countries that spur migration. Evaluations of aid effectiveness are retrospective, their results conditional on how aid was given in the past. Chandy, Seidel, and Zhang (2017) identify several ways for aid in fragile states to be more effective, including shifting from bilateral funding and fragmented projects toward multilateral agreements and programmatic assistance. But the evidence to date does carry one clear implication. The underlying rationale of “root causes” aid programming relies on claims that the magnitude of aid’s effect in specific sectors is sufficient to greatly reduce emigration. Such claims have not met this burden of proof. The evidence we have implies that aid would need to operate in unprecedented ways, at much higher levels of funding, over generations, to greatly affect some of the most important plausible drivers of emigration. That implies a case for experimentation and patience, but not confidence in a surge of aid to end a crisis.

The effect of development on migration

Even if we knew that aid greatly affected outcomes that theoretically drive increased emigration, we would still need evidence that those outcomes do in fact drive migration behavior. A helpful starting point is to understand two critical ways that many people in poor countries use migration to improve their economic lives: investment and insurance (Banerjee and Duflo 2011).

Sustained economic development tends to encourage emigration

First, poor families use migration as an investment. Among other things, migration is an exchange of substantial up-front costs for a stream of future benefits (Schultz 1972; Burda 1995; de Haas 2010). The costs include not just the direct costs of moving but foregone earnings at home and missed interactions with family. The benefits typically include large but delayed increases in income both for migrants (Clemens, Montenegro, and Pritchett 2016) and their families (Gibson and McKenzie 2014; Mergo 2016).

This immediately suggests that the effect of development on migration can be complex, for the same reasons that development has complex effects on other kinds of investment. For example, greater economic opportunity for people without a college degree might reduce the need for workers to invest in college education, but also increases their means to invest in college education. For similar reasons, greater economic opportunity at home might reduce the incentive for workers to invest in migrating abroad for work, but also increases their ability to make that investment. Which of these forces dominates is an empirical question.

It is now clear that emigration rates in middle-income countries are typically much higher than in poor countries (de Haas 2007; Clemens 2014; Carling and Talleraas 2016; OECD 2017; Dao et al. 2018). Economic development has also been positively associated with major asylum-seeker outflows over the medium and long term (Rotte, Vogler, and Zimmermann 1997; Vogler and Rotte 2000). Additional disposable income causes many poor families to invest it in more migration (e.g., Angelucci 2015; Bazzi 2017).

This suggests that in poor countries, development does more to encourage migration than to deter it. Figure 3a shows this relationship for the year 2013.⁶ Countries with GDP per capita of PPP$5,000–10,000 have, on average, roughly triple the emigrant stock of countries below PPP$2,000. Here, “emigrant stock” refers to the number of people born in a country but living outside it, divided by the population of the origin country. Only in countries above PPP$10,000 is there a clear negative relationship between real GDP per capita and emigrant stocks. This tendency for emigration to first rise and then fall with rising GDP per capita was first termed the “mobility transition” (Zelinsky 1971).

This pattern is counterintuitive for many who first encounter the evidence. The most basic economic theories of migration imply that greater economic opportunity at home will reduce the incentive and thus the tendency to emigrate, all else being equal (Sjaastad 1962). But all else is not equal. As development proceeds, human capital accumulates, connections to international networks increase, fertility shifts, aspirations rise, and credit constraintsareeased.Allofthesechangestendtoraiseemigration.Themost important of these factors appears to be rising education levels and international connections, which both inspire and facilitate emigration (Dao et al. 2018). Economic development is also often accompanied by falling child mortality rates while fertility rates are still high, leading to a surge in the number of young workers who have a high tendency to migrate. Greater disposable income means greater ability to pay the direct costs of migration, but also greater ability to invest in things that both inspire and facilitate migration—such as internet access, language skills, and overseas tourism.

Does the positive relationship between development and migration in poor countries accurately describe the effect of development on migration? One possible concern about the evidence in Figure 3a is that it shows a cross-sectional relationship at a single moment in time. In principle, this correlation could conceal parts of the causal relationship. It could conceal long-term effect so economic growth on migration by only showing a shortterm relationship. It could also conceal unobserved traits of today’s low income countries that might cause them to follow a different path than today’s middle-income countries did.

But the data do not support either of these possibilities. Figure 3b shows the relationship between the change in real GDP per capita between 1960 and 2013 and the change in emigrant stock over the same period. It shows all countries on earth with comparable data from both years. Each arrow represents the path of one country, from 1960 to 2013. The swath of arrows pointing up and to the right, around the middle of the figure, shows that emigration rose in almost all countries that grew into middle-income in this period. In the 71 countries that grew to middle-income or higher between 1960 and 2013, 67 had a concurrent rise in the emigrant share.⁷ For poor countries that remained poor, there was no clear rise.

In other words, the path followed by individual countries over time has resembled the cross-sectional relationship. As poor countries have become rich, almost all have experienced large migration outflows. Rising real incomes at home have gone hand in hand with higher levels of emigration.

A second possible concern about this evidence is that even if emigration rises with future economic development, emigration might fall with future improvements in noneconomic development indicators. For example, better health conditions in poor countries or greater opportunities to internally migrate to urban areas could in principle substitute for international migration. However, the evidence reveals no such patterns. Figure 4a shows the cross-country relationship between emigration and child survival—the probability that any given newborn lives to age five.⁸ People have a sharply greater tendency to emigrate from countries where children are less likely to die. This is likely because child mortality rates typically fall well before fertility rates fall. So typical countries must proceed quite far through the demographic transition before migration pressure from young workers abates (Figure 4b). And there is no statistically significant difference between emigration rates in high-urbanization countries versus low-urbanization countries (Figure 4c). In short, there is little sign that emigration is lower in poor countries with generally better public health or generally greater access to urban living, and little reason to believe that changes in such indicators will broadly and systematically deter emigration.

This evidence strongly suggests that in poor countries, development facilitates investment in emigration much more than it deters investment in emigration. But investment is only one way in which poor families use migration.

Greater short-term income diversification tends to discourage emigration

Poor families also use migration as a form of insurance. It helps them diversify income across members of the household, making the family less vulnerable to shocks like job loss or sudden sickness. Migration also helps diversify income over time, reducing the risk of dire lean periods for the whole household (Chen, Chiang, and Leung 2003; Giesbert 2007; Shonchoy 2011; Bignebat and Sakho-Jimbira 2013; Marchetta 2013; Bryan, Chowdhury, and Mobarak 2014).

This too suggests that the effect of greater economic development on migration decisions can be complex. More stable and predictable economies can reduce the need to ensure against negative shocks, but as aspirations rise, so can the need for insurance. For example, a family of subsistence agriculturists may perceive less harm from pulling a child out of 6th grade to work due to a parent’s job loss than would a middle-class family forced to withdraw their most capable child from 12th grade for the same reason. Beyond this, greater prosperity can also raise poor families’ ability to access the insurance benefits of migration. The net effect of greater prosperity on migration by this channel, too, is ambiguous in principle.

The youth employment rate is a sound, general guide to the probability that any given young person can supplement household income through wage work, and the data show us an unmistakable negative relationship between emigration rates and youth employment. In countries where youth employment exceeds 90 percent, the emigration rate is half as large as in countries where youth employment is just 70 percent (Figure 5a).⁹ This trend is evident not just among poor countries but across all countries. When youth employment is very low (40–70 percent), countries with marginally higher youth employment do not exhibit lower emigration rates. In countries where youth employment stagnates below 70 percent, marginal increases in job opportunities may both stimulate demand for the insurance of international migration and reduce that demand, to offsetting degrees. At higher rates of youth employment the migrationdeterring effect appears, on average, to win out.

It would be incorrect to infer, however, that successful assistance for youth job creation will clearly deter emigration. This is because the greatest engine of youth job creation is economic growth, which in poor countries tends to raise emigration. Figure 5b shows the same relationship in Figure 5a separately for countries in the two lowest quartiles of GDP per capita. For both the poorest quartile (below PPP$3,954) and the second quartile (between PPP$3,954 and 11,717), there is a clear negative relationship between youth employment and emigration. However, at any given level of youth employment, emigration is much higher for relatively richer countries. Figure 5b shows that countries in the richer quartile with 90 percent youth employment (e.g., Paraguay) have similar emigration rates to countries in the poorer quartile with 60 percent youth employment (e.g., Mozambique).

In other words, the evidence suggests that emigration is lower at higher levels of youth employment for any given level of overall economic development. Assisting families to diversify their incomes with domestic wage work may substantially reduce migration rates in poor countries that remain poor, but it may not do so in robustly growing and diversifying economies. This has important implications for policymakers, to be discussed in the final section.

The overall effect of aid on emigration

The evidence so far urges caution in interpreting any retrospective empirical estimate of the effect of aid on migration. The evidence leaves unclear the mechanism by which aid would systematically reduce migration. We do not have strong evidence that aid has typically been targeted to the sectors or places that most obviously would affect emigration. The ability of aid to cause large, short-term changes in national income, employment, or security is not independently demonstrated. And overall development—better incomes, health, and education—is, in fact, strongly associated with rising emigration.

All of these factors suggest that, as we review the limited literature directly testing the overall effect of aid on migration, we should not have a strong a priori view that it will detect substantial deterrence of migration by aid. Theoretical models have suggested that aid could deter migration by raising incomes at home (Myers and Papageorgiou 2000; Dula, Kahana, and Lecker 2006), but theory also suggests that any aid-produced rise in incomes at home could facilitate investment in migration by the poor (Gaytán-Fregoso and Lahiri 2000; Kahana 2007).

The first systematic quantitative assessment of the global average effect of aid on emigration is the gravity model in Berthélemy, Beuran, and Maurel (2009). They find that aid raises net emigration from the average poor country to high-income OECD countries: When aid rises by 10 percent of GDP, this raises the average emigrant stock as a share of population by 1.5 percentage points. They also find that aid shifts the composition of emigration toward low-skill migrants, and that the share of bilateral aid raises emigration about twice as much as aggregate aid. Other research has similarly found aid inflows to correlate positively with emigration from sub-Saharan Africa (Mugh 2011) and illegal emigration from Latin America (Bandyopadhyay, Chambers, and Munemo 2014). Recent and ongoing research has found no robust evidence that total aid inflows to a country reduce refugee outflows from that country (Dreher, Fuchs, and Langlotz 2017).¹⁰

One complication in interpreting these results, a complication common to many cross-country findings, is the possibility of overcontrolling— that is, holding portions of the relevant causal pathway constant. The regressions used by Berthélemy, Beuran, and Maurel (2009) control for the aid recipient’s GDP per capita, population, and trade with the migrant destination country. This is a sensible empirical choice because all of these factors can affect migration independently from aid, but it has the drawback that all of these factors can likewise form part of the causal pathway from aid to migration. Thus the coefficient estimates on aid itself show the relationship between aid and migration other than any effects that aid might have via any effects on economic growth, population growth, or trade. In principle, aid could affect these other factors in ways that reduce migration, or increase it even more.

The broad finding of Berthélemy, Beuran, and Maurel (2009) has been challenged by a small, recent literature. Lanati and Thiele (2018), also in a gravity model, find no effect of bilateral aid on migration, and a negative effect of aggregate aid on migration.¹¹ The principal innovation is the use of panel data, allowing fixed effects by origin (and destination). These results are difficult to compare with those of Berthélemy, Beuran, and Maurel (2009), because Lanati and Thiele (2018) use a different migration variable (gross migration flows annually 1995–2014 rather than net migration stocks in 2000), a different treatment variable (net aid flows rather than gross aid), a different set of countries (28 rather than 22 destination countries), and greatly different regression specifications. One example of how this complicates any comparison: aid could decrease outward migration but also decrease return migration, such that a negative effect on gross migration is not incompatible with a positive effect on net migration. Given that Lanati and Thiele (2018) theorize any positive effect of aid on gross emigration as the alleviation of credit constraints—a channel that would not affect return migration—it is difficult to understand why the two sets of results would differ so much unless aid does affect return migration.

Finally, Gamso and Yuldashev (2018) find, in a 15-year panel with country fixed-effects, that aid supporting rural development reduces emigration—but no other type of aid does so. The magnitude of the estimated effect is extremely small: It implies that a tripling of rural development aid would produce less than a one percent decline in the number of people emigrating from the average aid recipient country.¹² In other words, this study suggests that overall aid has no effect on emigration, and rural aid hasalmostnoeffectonemigration.Moreover,coefficientestimatesobtained from slicing aid into small subsets require further investigation when numerous such subsets are tested (Gamso and Yuldashev 2017). In such cases the null hypothesis of no effect can arise for some subsets of aid by chance alone if the number of subsets tested becomes large.

In sum, the few cross-country studies testing the overall relationship between aid and migration fail to offer clear evidence that aid has substantially deterred migration on average. Existing work either finds a positive relationship between aid and migration (Berthélemy, Beuran, and Maurel 2009), or finds no effect on overall migration from the large share of aid that is bilateral (Lanati and Thiele 2018), or finds extremely small negative effects of migration from a small subset of aid (Gamso and Yuldashev 2018). All of the studies face important challenges in interpreting the coefficient estimates on aid as the policy impact of an increase in aid, due to inherent limitations on identifying causal relationships in cross-country data.

The literature as it stands does not offer strong evidence that aid has systematically deterred migration, and rather implies that aid may have encouraged migration. As we emphasized at the beginning of this section, the literature testing for overall effects has a substantial burden of proof to bear because the evidence for each step in the underlying causal mechanism is itself weak. This literature does not yet offer lessons that would be programmatically useful for aid planners interested in deterring migration. A promising way forward is to seek large exogenous changes or discontinuities in aid exposure—such as across time, networks, cohorts, or areas—to look for well-identified effects on migration.

Four lessons

Today the evidence that aid can greatly and sustainably deter emigration from poor countries is weak at best. Aid tends to follow donors’ geopolitical concerns (Kuziemko and Werker 2006; Faye and Niehaus 2012) and there is minimal evidence that it has systematically targeted the geographic areas or sectors considered most influential for migration flows. Aid flows may have a positive effect on economic growth—though this remains controversial— but more importantly, economic growth has historically raised emigration in almost all developing countries. Large increases in youth employment may well deter emigration in the short term, in countries that remain poor, but the best evaluation evidence finds most donor projects have had little success creating youth employment on a large scale.

Aid practitioners’ most common questions of such research are operational: Where, when, and what kind of aid should they give to best serve their obvious implicit mandate to deter migration? However, the scant empirical research literature does not allow general operational lessons of this kind. We draw four lessons for aid policymakers and researchers with an interest in affecting migration flows from poor countries.

First, the evidence suggests that aid-supported programs to increase employment of young workers, in both rural and urban areas, can modestly reduce the potential for surges of emigration in the short term. We qualify this assessment with “modestly” because most such efforts have failed to achieve large changes in youth employment at scale. We qualify it with “rural and urban” because there is no clear evidence that poor countries with more urban or more rural populations have different emigration rates. And we qualify it with “short term” because the principal generator of youth employment—economic growth—tends to increase emigration in the medium- and long-terms. Among countries that remain poor but manage to get more youths into jobs, migration rates are likely to fall somewhat. But in countries that get more youths into jobs sustainably, by developing a dynamic and growing economy, the typical net medium- and long-term effect is likely to be an increase in migration.

This implies a reasonable aspiration for aid seeking to deter emigration: to help at the margin to get youths into predictable jobs, shielding their families from shocks and mitigating sudden surges of migration—during the decades that emigration steadily rises along with development. Although some aid projects have sought to go further, especially by mitigating migration pressures arising from conflict, the literature offers little evidence that aid to mitigate existing civil conflict has typically done so.

We clarify that these conclusions regard the ability of aid donors to deter emigration from the origin, not the ability of donors to deter immigration to the donor country. Donors seeking to use development assistance to deter immigration have another option: to divert emigrants rather than deter them, that is, to assist third countries in their absorption of emigrants from the origin. This is fundamentally the role of the aid-supported Migration Partnership Framework created by the EU in 2016. It is clear that immigration barriers in some countries can and do divert migration flows to other countries (Bertoli and Fernández-Huertas 2013; Elsner and Zimmermann 2016). It is plausible that these barriers, combined with development assistance to third countries, could tend to reduce immigration to donor countries even without reducing overall emigration from origin countries. We are not aware of direct empirical tests of this effect. We do note that development assistance to origin countries, to the extent that it is successful in fostering sustained development, is likely to create additional pressure on third-country hosting arrangements by encouraging greater overall emigration from origin countries. Moreover, most international migration— particularly from Africa, a major focus of “root causes” policy—remains intracontinental (Flahaux and de Haas 2016).

Second, greater transparency and reporting are necessary to understand this phenomenon. The public documentation available on aid programming is too high-level to help researchers and other policymakers understand how these projects actually play out on the ground. A good first step would be to follow the precedent set by OECD Creditor Monitoring System reporting of aid projects targeting environmental and gender inequality. An indicator on “facilitation of orderly, safe, regular, and responsible” migration has recently been added; however, root causes programming is specifically excluded. These programs will continue to be tracked under the broader aid sector of intervention, unfortunately precluding the information necessary for both impact evaluation and future programming.

Third, policy and research should focus on diverse experimentation in local context, continuously modified by feedback from rigorous evaluation. This approach has been called Problem-Driven Iterative Adaptation (Andrews, Pritchett, and Woolcock 2013). While more could be learned from refinements of cross-country regressions, that research program is unlikely to result in reliable guides to donors seeking to deter migration in a given setting. Carefully evaluating the impact of new aid efforts on international migration requires counterfactuals—places, cohorts, networks, or time periods with less or no exposure to the aid intervention, but otherwise similar, for comparison. It also requires tracking affected families to determine if they have gone abroad. Particularly important is testing a diverse range of new policies to raise youth employment, given the very poor track record of most past efforts. Research facilities created under the EU Trust Fund for Africa offer an excellent opportunity for new research of this kind—combining diverse experimentation with counterfactuals and migrant tracking—an opportunity that remains to be seized.

Fourth, aid seeking to shape migration must look far beyond efforts to deter migration. Demographic realities imply that large-scale migration will occur in some form. Most notably, these realities include estimates of the net increase of 800 million workers in sub-Saharan Africa by the year 2050, at a time when populations in many migrant-destination countries are aging and stagnating (Hanson and McIntosh 2016). The true demographic surge to come may be even larger (Bertoli 2017). The evidence reviewed here does not offer strong reasons to believe that a large portion of those future flows can be deterred with aid policy, and the literature suggests a limited role for deterring them with interdiction (e.g. Gathmann 2008; Czaika and de Haas 2017).

Instead, aid agencies seeking to shape future migration flows can focus on cooperation with migrant-origin countries that alters how migration occurs, maximizing its potential benefits for everyone involved. Aid agencies can work with migrant-origin countries to develop safe, lawful, and mutually beneficial channels for lower-skill labor mobility (Gibson and McKenzie 2014; Clemens and Postel 2017). Foreign assistance is often required for up-front costs like providing identity documents to potential migrants, developing systems to monitor and enforce labor recruitment laws, and agencies to monitor returns and prevent overstays. Such activities do not substitute for more traditional aid, but complement it. For example, the local economic stimulus of remittances would tend to raise demand for the produce of any aid-supported farms or small businesses (Minasyan and Nunnenkamp 2016).

Aid agencies can also shape higher-skill migration for mutual benefit, such as facilitating finance and technology transfers to equitably share the training costs of skilled migrants, and strengthening origin-country training institutions to give migrants precisely the skills they need for rapid employment and integration at the destination (Clemens 2015; Clemens, Graham, and Howes 2015). This field is wide open for innovation. For example, none of the projects currently supported by the EU Trust Fund for Africa have the goal of creating mutually beneficial labor mobility channels between Africa and the rest of the world. Current aid efforts around the world have devoted essentially none of their portfolio to supporting innovative ways to shape rather than deter migration. The first step in seizing this historic opportunity would be a substantial shift in aid agencies’ new mandate—away from an exclusive focus on deterring migration and toward shaping migration for mutual benefit.