Climate change and international migration: Exploring the macroeconomic channel

from climate impacts (decades). Despite the improvements, these methods follow the majority of recent studies (e.g. [40–45]), focusing only on the temperature variable and its effects on the economic activity. Recently it has been shown that this channel is not the only one through which climate impacts the economy [46]. Therefore our estimates, using two different methods and one climate variable to calculate the GDPc impacts, give an indication of the possible range of outcomes, but not a complete one. The migration model, on the other hand, includes several important features that set it apart from previous studies [47–51]: (i) it considers bilateral flows between more than 170 countries globally, resulting in an unprecedented spatial coverage, (ii) it considers separately return and transit migration flows, augmenting the resolution on the type of bilateral migra- tion flows with respect to previous works, (iii) it represents the empirically observed non-lin- ear relationship between origin-country incomes and emigration rates, through the “migration hump” function. Nonetheless, our modeling approach also involves important simplifications and limita- tions. For instance, our projections of baseline population distribution, broken down by place of birth, rely on country-level natural population change rates applied to all migrant stocks uniformly. They ignore the effects of transient migration that would alter the migrant stock distribution between now and the future time period analyzed; effects that would need to be

taken into account for a quantitative prediction exercise. However, for our present analysis, we consider this simplification useful because it facilitates comparison between future and present migration patterns, both assuming the same initial distribution of relative migrant stocks. The same goes, importantly, for feedbacks from altered migration patterns on economic develop- ment and GDP [8], which we ignore for the sake of simplicity and transparency in our analysis of climate change effects. It is important to stress out that our modeling study focuses only on international migration. Internal, i.e. within-country, migration is an important phenomenon that would likely face impacts due to future climate change. Our global approach, and coun- try-level population and GDPc data do not allow for carrying out the same study at a higher spatial resolution. International migration models often include dyadic variables such as geographical dis- tance, colonial links, or common language, between origin and destination country. However, these can become insignificant once bilateral migrant networks are explicitly accounted for by including a diaspora variable ([52], p. 508). This is also the case for the present model specifica- tion, where these dyadic variables were estimated to be insignificant, and therefore neglected [26]. Moreover, the geographical distance in particular shows up more significantly in log- transformed models, while untransformed models such as ours predict more limited effects of distance [53, 54].

We acknowledge that international migration might be driven also by other macroeco- nomic factors as wage differentials and employment rates. We have opted for using only the GDPc because of the global coverage of GDPc projections data and the fact that we want to maintain the model parsimonious in the number of covariates. Moreover, by including the migration “hump function” we manage to capture a more complex dynamics of migration than that suggested by neoclassical economic theory [55] which would project decreasing migration flows as the wage differential decreases [8]. Indeed, observed migration flows have shown a more complex dynamics, with emigration rates from developing countries increasing during a first stage of economic development [15, 56]. The model also does not account for differences in within-country income distributions, nor for changes in the shape of these distributions; i.e. changes in within-country inequality. These distributions clearly differ between countries, and how they will change is likely going to depend on climate change impacts too. Another important effect comes from immigration PLOS ONE Climate change and international migration: Exploring the macroeconomic channel PLOS ONE | https://doi.org/10.1371/journal.pone.0276764 November 16, 2022 21 / 25