I ran a regression with the total amount of bilateral aid an origin country receives as a baseline. In this analysis, bilateral aid is the total amount of aid that an origin country receives from all of the OECD countries combined. The results reaffirm the literature’s consensus that aid, as a whole, increases migration. In 2007 to 2010, the model estimates that every one percent increase in total aid received results in a 0.106 percent change in migrant stock. The small coefficient confirms that there is an indirect link between aid and migration. From the time period 2011 to 2015, the relationship is the same with a slightly larger magnitude of 0.173. This positive relationship between migrant stock, as a dependent variable, and bilateral aid given is also seen in Barthelemy et al.’s paper in 2009, however their results show a larger coefficient of 0.31. The larger coefficient is due to a different methodology: they separate the data into different country income groups and use a three-stage least squares model using two equations describing wage differentials and one regarding the cost of migration.

The analysis of both periods show that Social Infrastructure and Services aid is not statistically significant. The subcategory data, however, show that there are aid types that fall under Social Infrastructure and Services that are statistically significant. From 2007 to 2010, Education aid is not statistically significant, but Water exhibits a statistically significant negative relationship with a coefficient of 0.014 with migrant stock. From 2011 to 2015, both Education and Water have negative relationships with migrant stock. Holding all other variables constant, a one percent increase in Education aid, on average, would decrease migrant stock by 0.019 percent. For Water aid, a one percent increase in aid would decrease migrant stock by 0.012 percent. Lanati and Thiele’s paper argues that better public services may outweigh aid’s reduction of migration costs, decreasing migration from the origin country. My analysis shows that this may not be the case, or at least aid directed at building public services has varying impacts. This provides us with mixed initial results of the relationship between public services aid and migrant stock.

I was curious whether the presence of humanitarian aid in conflict situations makes an impact. In period 1, humanitarian aid and the interaction of conflict and humanitarian aid were not statistically significant. In period 2, both are statistically significant, but show different relationships with migrant stock. Humanitarian aid in a conflict zone increases migrant stock in destination countries by 0.094 percent given a one percent increase in aid. In conflict zones, which are defined to have more than 25 battleground deaths and the state is a counterparty, it decreases migration by 0.092 percent — an opposite effect. In period 1, there were 1773 conflicts during this period, whereas in period 2, there were 361 conflicts. Donor countries gave a total of $4.4 bn of humanitarian aid in period 1, whereas in period 2 they gave $6.6 bn of aid and there were less conflicts. Just an indicator, assuming donors only gave humanitarian aid to conflict zones and donated aid evenly, in period 1 each conflict received roughly $2.48m in aid whereas in period 2 each conflict received almost $18.31mm in aid. Of course, equal distribution of aid is an oversimplifying assumption to have and this does not reflect the severity of conflict, but the existence of conflict. I suspect the differences in geopolitical events in these two periods may explain the relationships but cannot conclude from this data analysis. I suspect that humanitarian aid reduces the cost of migration in non-monetary ways and increases migration. Figure 1 shows that refugees, the subset of migrants that receive the most humanitarian aid, move the least distances to destination countries. Their needs and preferences are more about safety than economic opportunity, unlike high-skilled emigrants (World Bank 2018)

Humanitarian aid like food, emergency response and reconstruction relief decreases the short-term cost of rebuilding after conflict and reduces the cost of migration. In other words, if migrants are provided with basic services and care during instability, then they can pursue safety and a better life elsewhere. In connection with this finding, food aid is estimated to be statistically significant and inversely related to migrant stock in period 1 -- further supporting the hypothesis that aid decreases the cost of migration. In period 2, however, food aid is not statistically significant. This shows that aid can have differing effects on migration depending on factors like time, origin, destination and other factors that may not be shown in my analysis.

I will discuss Economic and Production Aid together. This is often done when analyzing aid data because these two categories of aid are usually given to well-governed origin countries (Akramov 2012). In the first period, both Economic and Production sector aid show a negative relationship with migrant stock. A one percent increase in Economic or Production aid, on average, all else held constant, will yield an approximately 0.04 percent decrease in migrant stock. The subcategories in period one loosely affirm this effect as both Transport and Communications and Agriculture, Forestry and Fishery aid show a negative relationship with coefficients of 0.012 and 0.038 respectively. These results are in line with the majority of empirical work that determine that economic opportunity is a large driver of migration (World Bank). Opportunity of more economic activity in origin countries may change individuals’ inclinations to migration.

In period 2, both Economic and Production Sector aid are not statistically significant. The majority of sub-categories with the exception of agriculture, forestry and mining aid shows no statistically significance. Agriculture, forestry and mining aid depicts an opposite relationship than expressed in period 1. As this aid type increases by one percent, my model estimates that migrant stock in destination countries increases 0.012 percent, holding all else constant. This estimate of a positive relationship between migrant stock shows an alternative hypothesis. Instead of a decrease in migration when economic opportunities are better at home, individuals may choose to migrate due to a relative decrease in migration cost to find better wages elsewhere. This scenario is what the majority of economists agree with, including Clemens and Postel.

Lanati and Thiele use a similar three stage least squares model as Barthelemy et al. in 2009, but use migrant flow as the dependent variable instead. They find a general negative relationship between emigration rates and each type of aid: Social, Economic and Production with coefficients of -0.119, -0.046, -0.065 respectively. This is not exactly an apples to apples comparison, but the differences estimate the relationship between aid and migration. In my analysis, Economic and Production aid also exhibit a negative relationship with migration, however Social aid’s impact is not statistically significant in both periods. In the subcategories of each category, the negative relationship is clearer, as discussed above. The amount of variability explained in their model is very close to mine, at around 90 percent of variability explained. They use residuals squared (r-squared) values to describe the variability explained, while I use deviance squared (d-squared), but the interpretations are the same.

A key assumption in this paper is that aid is given by donor countries completely exogenously, implying that there is nothing that origin countries can do to receive more aid. Immigrants in destination countries can lobby their governments to give out more aid in their origin countries. This presents a reverse causality problem where large migrant stock can actually increase foreign development aid, as opposed to developmental aid increasing migrant stock. My analysis does not control for this, so the results of the analysis may overstate the magnitude of migrant stock in destination countries. Another source of endogeneity are the omitted variables that are related to the error term. My analysis also assumes that the OECD countries allocate aid in countries and sectors where it is most needed. This is not necessarily the case as aid is often used as a foreign policy tool. The underlying incentives of the aid given is an omitted variable and one that is hard to measure. This creates a bias in my analysis by understating the relationship between migration and aid because the allocation of aid is imperfect.