How Do Extreme Temperatures Influence International Migration?

This study asks how extreme heat affects international migration from El Salvador. The authors use detailed agricultural production data from 2013–2018 and household survey data from 2009–2018, combined with high-resolution temperature records. They find that each additional week of extreme temperatures during the main corn-growing season reduces annual corn production by about 3 percent and lowers producers’ demand for hired workers by roughly 3.6 percent. These income losses increase international migration: one extra week of extreme heat raises the probability that a corn-producing household sends a migrant abroad by about 0.15 percentage points, or 23 percent relative to the mean.

Why This Article Was Selected for The Policy Scientist

This article addresses an increasingly important policy issue: how climate-driven shocks shape migration patterns in low-income agricultural settings. The question matters broadly because extreme heat is rising globally, and many countries share El Salvador’s combination of subsistence farming, limited insurance markets, and large migrant networks. The paper’s empirical strategy relies on observational data rather than causal or experimental methods, yet the data sources are detailed, high-quality, and well matched to the research question. The findings are likely generalizable to similar agricultural economies where remittances reduce liquidity constraints. While the methods are primarily multivariate regressions, the analysis is careful, transparent, and grounded in the main theoretical work in the field. It offers a meaningful contribution by linking temperature shocks, labor markets, and international migration in a timely way, given ongoing climatic changes.

Full Citation and Link to Article

Ibáñez, A. M., Quigua, J., Romero, M. J., & Velásquez, A. (Forthcoming 2025). Responses to extreme temperatures: Migrant networks and international migration from El Salvador. American Economic Journal: Economic Policy. https://www.aeaweb.org/articles?from=f&id=10.1257/pol.20230447 

Central Research Question

The article investigates whether and how exposure to extreme temperatures influences international migration from El Salvador. The authors focus specifically on whether negative agricultural income shocks—driven by unusually high temperatures during the primary corn-growing season—reduce local labor demand enough to trigger increased migration to the United States. They also ask whether established migrant networks condition this relationship by reducing the financial and informational barriers to moving abroad. A further component of the research question concerns the distinction between short-term “distress migration” caused by immediate income losses and longer-term “investment migration” facilitated by cumulative exposure to favorable or unfavorable shocks.

Previous Literature

This work builds on several strands of research. First, a large climate-economy literature shows that extreme temperatures suppress crop yields, reduce farm productivity, and increase volatility in agricultural income. Foundational studies such as Deschênes and Greenstone (2007), Schlenker and Roberts (2009), Schlenker and Lobell (2010), Feng, Krueger, and Oppenheimer (2010), and Burke and Emerick (2016) document sizable negative effects of high temperatures on agricultural output across multiple contexts. A second body of research addresses household coping responses to weather-driven income shocks. Classic work by Rosenzweig and Wolpin (1993) and later contributions by Jayachandran (2006), Jessoe, Manning, and Taylor (2016), and Aragón, Oteiza, and Rud (2021) show that rural households, when faced with income shortfalls and incomplete markets, adjust labor supply, alter input use, or resort to distress migration. A third branch of literature focuses on weather-induced migration specifically. Studies such as Cattaneo and Peri (2016), Mueller, Gray, and Kosec (2014), and Nawrotzki (2015) show that climate anomalies can shift migration patterns, though findings differ across countries and institutional settings. A fourth strand emphasizes migrant networks as a critical determinant of international migration flows, highlighting their roles in reducing risk, lowering costs, and facilitating information transfer (Massey et al., 1990; Munshi, 2003).

This article contributes by linking these literatures, focusing on an environment with both high climate vulnerability and an exceptionally strong U.S.-bound migrant network. It also extends Kleemans’s (2023) model of migration under risk and liquidity constraints, applying it to international rather than internal migration and examining the dual role of migrant networks as both insurance providers and facilitators of mobility.

Data

The authors assemble a multi-source dataset combining agricultural surveys, household surveys, and high-resolution temperature measures. Their primary agricultural data come from El Salvador’s Multiple Purpose National Agricultural Survey (ENAPM) for 2013–2018, which covers roughly 19,325 corn producers across 14 provinces and 169 municipalities. The survey contains information on yields, land size, input use, and labor (distinguishing between hired and household labor). The focus on corn is justified because it is El Salvador’s main staple crop, highly temperature-sensitive, and produced by the vast majority of small farmers.

Labor and migration outcomes are taken from the Multiple Purpose Household Survey (EHPM), a nationally representative annual cross-section conducted from 2009 to 2018. This dataset includes more than 200,000 households and provides information on employment, wages, income, and whether any household member migrated abroad in the previous year. It allows the authors to examine both agricultural and non-agricultural labor outcomes and to classify households according to their sectoral engagement.

Temperature measures combine NASA’s MODIS Land Surface Temperature product (1×1 km resolution; eight-day averages) and the ERA5-Land dataset (daily measures at ~9×9 km resolution). MODIS offers strong spatial precision but limited temporal depth, so the authors construct historical baselines by imputing pre-2000 MODIS-equivalent data using ERA5. Their main heat-shock measure is the number of “weeks” (eight-day periods) in the primary growing season (June–July) in which temperatures exceed two standard deviations above historical means (1980–2003). They also construct cumulative measures capturing exposure in the previous four years to distinguish short- and long-run effects.

Methods

The empirical strategy links municipal-level shocks in extreme temperatures to outcomes for agricultural producers, labor markets, and households. The authors estimate several regression models using within-municipality variation over time, incorporating municipality fixed effects, year fixed effects, and municipality-specific linear time trends interacted with baseline socioeconomic variables. This approach is intended to isolate the effect of temperature anomalies from long-run trends or unobserved municipal characteristics.

Agricultural outcomes (yields, production, total factor productivity, and labor demand) are estimated using log-linear models and Poisson fixed-effects models for labor counts. For input use, the authors construct principal component indices and examine specific input categories such as seeds, fertilizers, pesticides, and post-harvest chemicals.

Labor outcomes rely on the household survey and include employment probabilities, sector-specific employment (agricultural versus non-agricultural), and wages. Migration outcomes are modeled with a one-year lag in temperature exposure on the assumption that international migration requires time to plan and finance. For migration, the key dependent variable is whether at least one household member migrated abroad in the previous year.

The models incorporate controls for rainfall anomalies, droughts, soil moisture, and demographic characteristics. Standard errors are double-clustered at the municipality and year levels. The authors also perform robustness checks using alternative definitions of extreme heat, Conley spatial standard errors, multiple-hypothesis corrections, and controls for violence.

Although the methods rely on observational data and fixed-effects regression rather than causal designs such as instrumental variables or experiments, the specification is transparent, thorough, and consistent with the current empirical literature on climate shocks in developing countries.

Findings/Size Effects

The article reports several key empirical results. First, extreme temperatures significantly reduce agricultural productivity. One additional week of extreme temperatures during the primary corn-growing season reduces total annual corn production by approximately 3 percent and yield per hectare by roughly 3.9 percent. TFP falls by about 3.2 percent.

Second, producers respond to these negative shocks by reducing input use—particularly post-harvest chemicals—and by cutting labor demand. The number of hired workers falls by about 3.6 percent for each additional week of extreme heat, while total workers decline by about 2.5 percent. There is some evidence of substitution toward household labor, consistent with consumption-smoothing behavior under liquidity constraints.

Third, local labor markets exhibit reductions in the probability of being employed in the corn-growing agricultural sector, though evidence of reallocation to non-agricultural employment is limited. Wages in the agricultural sector also show negative but imprecisely estimated effects.

Fourth, international migration increases substantially in response to adverse temperature shocks. For agricultural households primarily engaged in corn, one additional week of extreme temperatures increases the probability of emigration by approximately 0.15 percentage points, equivalent to a 23 percent rise relative to the mean annual migration rate. Non-agricultural households show no significant response, except for “unemployed households” heavily reliant on remittances.

Finally, migrant networks strongly amplify migration responses. Households in municipalities with higher baseline shares of migrants or remittance recipients exhibit substantially larger migration effects, indicating that networks effectively lower the liquidity and information barriers to international movement.

Conclusion

The study demonstrates that extreme temperatures meaningfully reduce agricultural output and labor demand in El Salvador’s corn-producing sector, and that these effects translate directly into increased international migration. The presence of strong migrant networks intensifies these responses by making migration financially feasible when incomes fall. The findings also indicate a distinction between short-run distress migration generated by contemporaneous income losses and longer-run investment migration facilitated by accumulated exposure.

The results are especially relevant for countries with similar structural characteristics: heavy reliance on subsistence agriculture, exposure to climate volatility, limited insurance and credit markets, and strong transnational migrant networks. While the methods rely on regression-based inference rather than experimental identification, the data quality is strong, the empirical strategy is well executed, and the contribution is timely given accelerating climate pressures and growing migration flows.