Eve is here. Readers will probably disagree, but I found this study infuriating. This paper makes two assumptions that undermine its validity. Its stated aim is to challenge long-standing research, such as that conducted by researchers at the US Army War College in the early 2000s, that climate change causes mass migration through flooding in areas particularly exposed to sea level rise, such as Bangladesh and Sri Lanka. From NASA:
The analysis misses the point. First, in contrast to the effects of climate change that threaten housing and livelihoods, the study focuses on gradual changes in temperature that may prompt people who migrated to the U.S. Sunbelt to return north. Although internally displaced, 3.5 million people emigrated from the Great Plains during the U.S. Dust Bowl in the 1930s. Climate change not only directly causes population outflows through severe floods and droughts, it can also cause resource conflicts over water and intensify ethnic conflicts.
Second, and this is not my point, they assume legal immigration. Did they lie dormant during war-induced migrations, such as the exodus from Syria in 2014 and 2015, or during migrations in Africa, where significant numbers are fleeing upheaval? As the exodus from Myanmar shows, crisis-based mass migration and low-level migration are often not legal.
Third, the fact that rich people are more mobile and thus able to avoid the negative impacts of climate change early on (when they are mild) does not prove that low-income people will not move out of necessity when the impacts become more severe.
Written by Céline Azemar, Rodolphe Debordes, Markus Eberhard, and Eric Neumeier. Original publisher is VoxEU
It is widely expected that millions of people will move across borders due to climate change displacement. This column, based on monthly bilateral inflows and outflows of people from 127 developing countries, found the opposite. The idea is that temperature shocks reduce migration rather than increase it, because climate stress reduces incomes in countries of origin and poor households are unable to finance expensive international migration. Evidence of this is that ‘climate refugees’ are an exception, with hardship often trapping poor people in place rather than displacing them.
Editor’s note: This column is based on CEPR Discussion Paper 21634, “Stuck at Home: Weather Shocks, Income, and International Migration.”
“Climate refugees” have become a staple of policymaking. The Institute for Economics and Peace (2020) estimates that ecological threats will put 1.2 billion people at risk of displacement by 2050, and security and immigration authorities are increasingly working with large-scale influxes across borders. Research records are more careful. In its dedicated assessment, the IPCC (2022) determined that the evidence on climate and international migration is mixed, with flows likely to rise or fall depending on circumstances.
Part of the disagreement is a question of data. A bilateral migration series long enough to study slow-moving climate trends built from census data in Ozden et al. (2011) are recorded at 5- or 10-year intervals, which are too coarse to distinguish temporary weather anomalies from all other phenomena that move along with them. The sparseness of observations precludes confounding factors correlated with a warming climate (among them origin incomes, migration policies, diaspora networks, and business cycles at both ends of migration corridors) from being fixed, a long-standing and well-known limitation in the climate and income literature (Dell et al. 2014, Desbordes and Eberhardt 2024).
A recent paper (Azémar et al. 2026) revisits this issue using monthly data. The migration flow is taken from Chi et al. (2025) transformed the location history of approximately 3 billion Meta/Facebook users into a monthly panel of cross-border movements. We limited our sample to 127 developing countries of origin and 180 destinations from 2019 to 2022. Monthly frequency is what makes our exercises possible. We compare each corridor to itself, removing everything specific to origin-destination pairs in a given year, everything tied to a country’s unique seasonal calendar, and global monthly shocks common to all migration corridors, including migration disruptions due to pandemics. What remains are weather anomalies within the corridor, i.e. deviations of the country’s temperature (or precipitation) in a particular month from the country’s seasonal norm. And it is this residual variation that specifies the transition response.
Warmer-than-normal weather leads to reduced cross-border movement
Our results run counter to the climate refugee narrative. (Developing countries) As the country of origin continues to experience warmer than usual years, cross-border movement will actually decrease. A 1 degree anomaly in source temperature reduces bilateral flow by approximately 5.7% over the subsequent 12 months, but has no detectable effect on source rainfall.
This estimate survives many robustness checks. These include alternative fixed effects, leaving the largest travel corridors, reclustering spatially correlated weather, alternative specifications for pandemic control or removing them altogether, exercises that exclude only one destination, and placebo tests that reshuffle the weather calendar timings for each departure location and see that the effect disappears when the true timings are scrambled. The same effect, and the same zero effect on rainfall, is reproduced with an entirely different data source. Annual arrival flows from 1990 to 2019 were constructed from 30 years of IPUMS census microdata covering 177 origins.
The average effects across countries hide significant asymmetries (Figure 1). When we split the origin by long-term baseline temperature, the response is close to zero in the colder half of the origin, and the response is strongly negative in the already warmer half, with a 1 degree anomaly reducing emigration by about 16%. The effects will be concentrated precisely where the relationship between temperature and income is steepest, and precisely where further warming is expected to be most rapid. If we split the sample by income instead, the response remains the same, so this is not just about the poorest origins. The effects apply across the income spectrum in developing countries.
Figure 1 Effect of a 1 degree anomaly in origin temperature on bidirectional migration
Note: Impact of a 1 degree anomaly in origin temperature on bilateral migration over 12 months for all developing countries of origin and countries of origin divided by the median baseline temperature over the long term (1990-2018). Estimated by Poisson pseudo-maximum likelihood for monthly bilateral flows from 2019 to 2022.
Source: Azémar et al. (2026).
Income falls due to temperature shock, making it impossible to move
Why does the heat keep people indoors? Traveling across borders is expensive. It involves visa fees, travel costs, recruitment costs, and increasingly the financial reserves that wealthy countries’ immigration systems require up front. Households who have lost income cannot simply substitute costly emigration abroad. You may lose all funding options. This is the resource-constrained immobility trap of Black et al. (2013) and Benveniste et al. (2022), and it has clear empirical implications: temperature shocks that suppress immigration should also suppress incomes in the country of origin.
We demonstrate doing this using two separate datasets. On a subnational grid of GDP per capita (Rossi-Hansberg and Zhang 2025), a 1 degree annual temperature anomaly reduces income by about 2%, consistent with the macroclimate literature (Burke et al. 2015). A monthly panel of satellite night lights built for the same country returns the same negative sign for income weather anomalies at the monthly frequency at which migration responses are identified. Reading the two elasticities together, we find that every 1% climate-induced decrease in home income is associated with an approximately 2.6% decrease in immigration. When your budget gets tight, moving becomes unaffordable.
If the binding force is not temperature per se, but the financing of migration, then shocks that erode the incomes and solvency of the country of origin should similarly immobilize migration. We will demonstrate that it is possible. When entering the monthly index of domestic conflict risk (Aizenman et al. 2026) into the regression model, an increase in origin conflict risk also reduces cross-border runoff and night illumination in the country of origin, while leaving the weather coefficient unchanged. The climate response is not a disguised conflict effect, and the constraints are general rather than climate-specific.
The bidirectional structure of the migration data allows one additional test. Routes to wealthy countries should be more budget constrained, as it costs more to travel to wealthier destinations. We confirmed this by finding a significantly stronger migration response in corridors to high-income destinations. Existing larger diasporas have been found to reduce migration costs and turn climate shocks into migration drivers for certain US-bound routes (Mahajan and Yang 2020), but no such amplification occurs here when destination income is taken into account. This suggests that financial constraints, rather than network effects, shape international migration due to climate change.
The premise of the “climate change refugee” policy is wrong
These findings have direct implications for how climate change is projected. A common predictive shortcut is to multiply the global population at climate risk by a positive migration elasticity. For poor, hot countries of origin that are subject to constraints, the calculation will have the wrong sign and therefore overestimate short-term cross-border migration. In the short term, climate stress is more likely to disrupt orderly migration from poorer countries of origin than to send waves of refugees to far-flung wealthy countries, as the people most exposed to the harms of climate change are typically those who cannot afford international migration.
These findings are troubling. Although the main benefits of immigration are provided to those who remain through remittances (di Giovanni et al. 2015), income shocks that prevent departure also impede remittances on which the origin country economy depends. Cross-border movement is itself a form of adaptation, and as conditions worsen, constraints that become more severe will inhibit adaptation from those least responsible for the warming that is driving it. Whether people who cannot leave the country will instead move within the country is a breadth that our cross-border data cannot confirm, and that is the obvious next question.
Current evidence suggests that climate refugees are the exception rather than the rule. These kinds of difficulties tend to keep poor people in place rather than mobilizing them, setting up political debates built on the opposite premise to get it wrong.
See original post for reference
