Weather Extremes in Indigenous Communities in Mexico and Undocumented US Migration and Its Duration

INDIGENEITY AS A FEATURE OF PLACE

Indigeneity—having origins in precolonial populations—is both an individual characteristic and a feature of place. Across Latin America, who “counts” as indigenous has long represented a political project, defined and redefined through state practices of enumeration (Loveman 2014). Mexico has an extensive history of the state pursuing the assimilation of its indigenous populations into mestizaje, an idealized blend of indigenous and European origins characterized by Spanish-language use and urban residence (Ruiz Medrano 2010). For their part, indigenous communities have organized to resist these efforts (Martínez Casas 2014; Postero et al. 2004). By the late 1980s and throughout the 1990s, Mexico’s gradual turn toward democratization coincided with an ostensible embrace of its indigenous population (Hale 2005; Loveman 2014), reflecting a larger turn in Latin America (Telles and PERLA 2014). Mexico committed to introducing self-identification for enumerating indigenous populations following the International Labour Organization’s 1989 Indigenous and Tribal Peoples Convention (No. 169) (Del Popolo 2017). A 1992 reform of Mexico’s constitution further signaled Mexico’s turn toward a “pluricultural” nation (Loveman 2014; Telles and PERLA 2014; Flores et al. 2023).

There are multiple ways to measure indigeneity. Mexico’s national census (INEGI, by its Spanish initials) has shifted its official criteria to combine language use, self- and external identification, and cultural traits in different ways over time (for a helpful summary, see Figueroa-Rodríguez 2020, table 9.1). Early censuses classified individuals by “habitual language” (Spanish or an indigenous language) and race (“pure indigenous,” “mixed,” or “white”), while later censuses experimented with markers such as clothing or diet. Self-identification was formally introduced in the Mexican Census in 2000.

Despite these myriad changes, stratification scholars generally treat indigenous-language proficiency as the most conservative—and consistent—indicator of the relationship between indigeneity and inequality (Loveman 2014; Telles and Torche 2019; Flores et al. 2023; Martínez et al. 2025). About 6 percent of Mexicans report speaking an indigenous language, while about twice that share self-identifies as indigenous (Martínez et al. 2025). While indigenous identity subsequently became a basis for political inclusion and cultural rights, this inclusion was most accessible to Spanish-speaking, upwardly mobile individuals in urban areas who self-identified as indigenous (Barbary and Martínez Casas 2015; Flores et al. 2023; Martínez Casas 2007). In contrast, indigenous-language speakers—those most connected to indigenous communities—remain disproportionately poor and fare worse along nearly every dimension of socioeconomic well-being (Villarreal 2014a; Telles and Torche 2019).

In addition to being an individual characteristic, indigeneity is spatially organized. Indigeneity, in other words, is a feature of place (Batalla 1996; de la Peña 2006; Martínez Novo 2006). The largest indigenous populations are concentrated in the central and southeastern regions of Mexico, particularly Chiapas, Oaxaca, Veracruz, Puebla, Yucatán, and Guerrero. At the municipal level, nearly 44 percent of Mexico’s 2,258 municipalities in 2010 had only a minimal indigenous presence (fewer than 1 percent of residents speak an indigenous language); about 23 percent had a moderate presence (between 1 and 10 percent); and about 33 percent were considered “indigenous” or “predominantly indigenous” (10 percent or more).¹ We follow this typology employed by Mexico’s National Office of Social Development (CONEVAL 2012) throughout this article to evaluate whether and how the context of indigeneity relates to multiple forms of inequality.

This spatial organization of indigeneity across municipalities matters for inequality, regardless of whether an individual speaks an indigenous language or identifies as indigenous (Batalla 1996; Friedlander 2006; Martínez Novo 2006). Table 1 shows the average municipal socioeconomic characteristics in the 2010 Mexican Census microdata accessed via IPUMS International (Ruggles et al. 2024) by context of indigeneity following this typology. Compared to low- and medium-indigenous municipalities, high-indigenous municipalities in 2010 had substantially lower literacy and education levels, higher shares of self-employment and labor force participation in the agricultural industry, and low wages. These municipalities also had less infrastructure on average, with lower shares of households with electricity, piped water, sewage disposal, and finished flooring.

The unequal distribution of socioeconomic resources by context of indigeneity reflects indigenous municipalities’ long history of subordination in Mexico (Ruiz Medrano 2010). Even as the country embraced pluriculturalism in 1992, another constitutional amendment that year permitted the privatization of communal farmlands (ejidos), undermining agricultural livelihoods in many indigenous municipalities (Jung 2003). The North American Free Trade Agreement (NAFTA) took effect two years later over strong opposition from indigenous communities—most notably, the Zapatistas in Chiapas (Harvey 1998; Kelly 2001; Stephen 2002). NAFTA had uneven effects that deepened existing inequalities by contexts of indigeneity: poverty rates increased in indigenous municipalities, while nonindigenous ones gained in income and wealth (Hamilton 2011; González Meza 2006; Fernández-Kelly and Massey 2007).

WEATHER EXTREMES IN INDIGENOUS CONTEXTS

The context of indigeneity thus patterns multiple forms of inequality in Mexico. We propose that one additional form of inequality that this context patterns is indigenous municipalities’ exposure to weather deviations or extremes. We define weather deviations as periods of unusually high or low rainfall and temperature. Although climate change has not substantially altered Mexico’s annual mean precipitation, its average temperature has increased by roughly 0.71°C between 1951 and 2017 (Murray-Tortarolo 2021). It is expected to experience an additional 1.1°C to 3°C average increase by 2060 (McSweeney et al. 2010). The seasonal distribution of rainfall has also shifted over this same period. Wet seasons (June– November) have become wetter, and dry seasons (December–May) drier (Murray-Tortarolo 2021).

These climatic changes exhibit an uneven regional patterning across Mexico. However, explicit attention to the context of indigeneity is rare in existing quantitative studies, in part because analyses typically occur at spatial (that is, regional or state) scales that obscure important municipal-level variation in indigeneity. Still, the available evidence suggests that regions and states with the largest shares of indigenous residents disproportionately experience these changes. Central and southern states, home to most of the country’s indigenous populations, show the largest fluctuations in both temperature and rainfall over time (Murray-Tortarolo 2021). Mexico’s recent multi-hazard climate risk assessment confirms a similar pattern (INECC 2021), identifying Oaxaca as the state with the highest number of highly climate-vulnerable municipalities, followed by Veracruz. Other states with many climate-vulnerable municipalities—namely, Yucatán, Chiapas, and Guerrero—also have high concentrations of indigenous-language speakers. Oaxaca has experienced a pronounced decline in precipitation over the last two decades (Cuervo-Robayo et al. 2020), while Yucatán has undergone a steady drying trend over the past half century (World Bank 2022). Simulations suggest that summer precipitation in southern Mexico could decline by 13 percent by century’s end (Colorado-Ruiz and Cavazos 2021).

Differential exposure to weather extremes is consequential for several reasons (see Klinenberg et al. 2020 for a review). Here we focus on its potential impact on agricultural production, the predominant industry in indigenous municipalities. Indigenous municipalities rely heavily on rain-fed subsistence farming, particularly through the traditional milpa system, which is a polycropping method combining maize, squash, and legumes (de Frece and Poole 2008). Maize, which occupies the largest cultivated area nationally (Bellon et al. 2011), is highly sensitive to deviations in temperature and rainfall (Ureta et al. 2020). This sensitivity is most acute for subsistence farmers (Bellon et al. 2011), who disproportionately reside in indigenous municipalities (Bada and Fox 2022). One study suggests that “states in which high levels of poverty and subsistence farming prevail, such as Chiapas (6.5 percent), Oaxaca (6.4 percent), and Guerrero (3.1 percent), account for a considerable fraction (16.0 percent) of the costs of climate change impacts on agricultural yields” (Estrada et al. 2022, 126). Although Francisco Estrada and colleagues (2022) do not explicitly consider indigeneity in their analysis, these states make up 25 percent of Mexico’s rural municipalities (Bada and Fox 2022) and have large indigenous populations relative to the national average, with at least 21 percent of residents classified as indigenous (CONEVAL 2012).

Taken together, indigenous municipalities are disproportionately rural, agricultural, poor, and vulnerable to weather extremes. These overlapping inequalities pose unique challenges for residents adapting to changing weather conditions. Rurality limits access to public infrastructure and essential services following extreme weather events (Eakin 2006). Dependence on subsistence agriculture heightens risks of crop failure and income volatility (Bellon et al. 2011). Meanwhile, high poverty rates among residents further constrain municipalities’ capacity to recover from extreme weather events (Estrada et al. 2022). This convergence of social and environmental inequalities underscores the plausibility that the context of indigeneity likely patterns responses to weather extremes.

CONTEXTS OF INDIGENEITY AND INTERNATIONAL MIGRATION AND RETURN AFTER WEATHER EXTREMES

We examine whether international migration to and return from the United States is one potential response to weather extremes that varies by the context of indigeneity. We do so for several reasons. First, research links weather deviations to human mobility within and across national borders (for example, Boustan et al. 2012; Thiede and Gray 2017; see Hunter et al. 2015; Garip and Reed 2025 for reviews). Second, while most weather-related migration occurs within countries (de Haas 2023), international migration—especially that which is undocumented or otherwise unauthorized by the receiving country—is more likely where transnational networks connect sending and receiving contexts and where return migration is feasible (Hunter et al. 2015). This is the case for Mexico-US migration, the world’s largest sustained binational flow. Finally, demographers emphasize that “the social vulnerability of households and places” mediates how communities respond to weather deviations (Riosmena et al. 2018, 471). Yet despite the rising frequency of both weather extremes and migration from indigenous municipalities since the 1980s (Asad and Hwang 2019a, 2019b; Murray-Tortarolo 2021), the context of indigeneity is largely absent from demographic analyses of weather-related migration and return (for example, Zhu et al. 2024).

We summarize key theories used to explain Mexico-US migration and return to situate our analysis (Massey et al. 1993; Garip and Reed 2025). Neoclassical and new economics approaches frame migration as a strategy to maximize individual income or minimize household risk, respectively (Harris and Todaro 1970; Stark and Bloom 1985). Once income or wealth goals are achieved, return to the origin community is likely (Garip 2016). Cumulative causation theory explains how migration flows, once initiated, become self-sustaining through social networks that reduce the costs and increase the benefits of migration (Massey 1990). As with neoclassical and new economics approaches, return to the origin community is likely once income or wealth goals are achieved, or when the benefits of migration no longer outweigh the costs. Finally, the aspirations-ability framework emphasizes that migration reflects both desires and constraints: some people who wish to migrate lack the means to do so, while others who can migrate may not (Carling 2002; de Haas 2021; Schewel 2020). Return to the origin community is likely if migrants aspire to do so and have the material and social resources necessary.

These theories each point to varied expectations about the relationship between undocumented migration to and return from the United States across contexts of indigeneity following weather extremes. First, migration entails substantial start-up costs that those with the greatest access to economic resources can afford (Massey et al. 1994). Such resources are unequally distributed across Mexico’s municipalities: individuals in nonindigenous municipalities generally have greater access to wage labor, savings, and credit than those in high-indigenous ones (Fierros-González and Mora-Rivera 2022). Economic restructuring during the 1980s and 1990s—including ejido privatization and NAFTA’s implementation—further marginalized indigenous municipalities (Jung 2003; Fernández-Kelly and Massey 2007; Hernández-León 2008). Weather extremes can compound these constraints, especially by eroding household income via reduced agricultural yields in municipalities dependent on rain-fed farming. This erosion is likely to be most acute in indigenous municipalities, where access to economic resources is scarcer on average. On the one hand, then, individuals in indigenous rather than nonindigenous municipalities may be less likely to migrate following weather deviations if they lack the resources needed to finance international movement; they may be “trapped” in place despite wanting to migrate (Black and Collyer 2014). On the other hand, individuals in indigenous municipalities may be more likely to migrate following weather deviations if declining agricultural incomes increase the need to maximize earnings through migration, especially when local livelihood options are limited.

Second, the costs of international migration are lower in places with longer migration histories and stronger networks linking origin and destination. Connections with prior migrants offer knowledge, financial assistance, and normative pressures that facilitate migration (Massey 1990; Garip and Asad 2016). While Mexico-US migration has persisted for more than a century, the earliest sending regions were disproportionately nonindigenous (Fox 2006; Asad and Hwang 2019a, 2019b). During the Bracero Program (1942–1964), over 4.6 million Mexican farmworkers—mostly from central-western states in Mexico—migrated to the United States (Cornelius 2001). After the program’s abrupt termination, US agricultural demand for Mexican farmworkers continued, and millions more entered without authorization at their encouragement (Massey et al. 2016). Indigenous municipalities were only modestly represented in these early flows; whereas migrants from indigenous municipalities comprised about 5 percent of all Mexican migrants in 1980, they accounted for roughly 20 percent by 2010 (Asad and Hwang 2019a, 1039). On the one hand, then, weaker and more recently established networks in indigenous municipalities may make undocumented migration less likely following weather deviations if individuals lack the information, financial support, or normative facilitation needed to undertake high-cost migration (Black and Collyer 2014). On the other hand, even less-established networks may provide sufficient footholds—through kin, friends, or co-ethnic ties—to render undocumented migration a viable or necessary response if weather deviations erode local livelihoods, especially when alternative adaptive strategies are limited.

Third, even among those with the economic and social resources to migrate, migration aspirations are not inevitable. Subsistence farming and communal land tenure often bind livelihoods and identities to place (de Frece and Poole 2008; Liffman 2014). Since individuals in indigenous municipalities are disproportionately reliant on these systems, stronger symbolic attachments to land may inhibit migration aspirations. In such contexts, weather deviations may produce “acquiescent immobility” (Schewel 2020)—remaining in place despite risk—because migration threatens not material well-being but rather a collective sense of belonging tethered to land (Eakin 2005). On the one hand, then, migration may be less likely after weather deviations for individuals in indigenous than nonindigenous municipalities if symbolic attachments outweigh material pressures. On the other hand, prolonged or severe weather deviations may erode the very livelihoods and identities tied to land, potentially heightening migration aspirations if individuals perceive remaining in place as no longer viable.

While we have thus far focused on undocumented migration, return is another aspect of the migration journey from Mexico to the United States. Between 2005 and 2014, about two million Mexican immigrants left the United States, some of whom were undocumented; during this period, the undocumented Mexican population declined from 6.3 to 5.6 million (Gonzalez-Barrera 2015). Scholars attribute this trend to economic growth in Mexico, as well as economic downturns and intensified immigration enforcement in the United States (Masferrer and Roberts 2012; Villarreal 2014b; Masferrer et al. 2024). Still, weather deviations may also relate to return migration, particularly when they persist in migrants’ origin communities (Zhu et al. 2024). Consistent with this theory and evidence, return may be less likely for migrants from high-indigenous municipalities relative to low-indigenous municipalities if sustained weather deviations continue to undermine agricultural livelihoods in origin communities, reducing the viability and desirability of returning home. Conversely, return may be more likely if undocumented migration enables migrants to offset weather-related losses—through accumulated earnings, savings, or investments—and to transition back into valued land-based livelihoods or symbolic obligations. Finally, because weather deviations are increasingly widespread and unpredictable across Mexico, they may suppress return similarly across contexts of indigeneity, producing no meaningful differences in the likelihood of return between migrants from high- and low-indigenous municipalities.

Despite varied expectations about how weather deviations relate to undocumented migration and return across contexts of indigeneity, direct empirical evidence remains limited. Most research relies on ethnographies of a few indigenous communities in a single period, drawing on retrospective accounts of weather. Some studies document rising migration pressures linked to declining agricultural yields (Ebel et al. 2018). In X-Pichil, Yucatán, for example, residents attribute increased emigration to the unreliability of milpa production amid prolonged droughts (Ebel and Castillo Cocom 2012). Other studies highlight adaptation rather than migration: farmers in Oaxaca’s Mixteca Alta region have delayed planting cycles to align with later rainy seasons (Rogé et al. 2014), and communities in Quintana Roo have adopted fertilizers and pesticides to sustain yields under drought (Ebel et al. 2018). While these studies offer valuable insights, a social demographic approach that examines multiple contexts of indigeneity over time—using high-resolution, daily weather data—can more systematically evaluate how overlapping social and environmental inequalities relate to the migration journey.

SUMMARY OF EXPECTATIONS

Indigenous municipalities in Mexico face overlapping forms of disadvantage that shape opportunities for international migration and return. These disadvantages—higher rates of poverty, rural residence, and dependence on subsistence agriculture—also correspond to greater exposure and sensitivity to weather extremes. Yet, given our theoretical framework and the mixed empirical evidence reviewed above, the expected direction of the relationship between origin-weather deviations and international migration and return remains uncertain. All else equal (including destination weather conditions), these deviations could either intensify, dampen, or be unassociated with migration pressures across contexts of indigeneity.

For undocumented migration, weather extremes in sending communities may be associated with changes in the probability of undocumented migration, but the direction of this relationship may vary by the context of indigeneity. On the one hand, undocumented migration may be more likely in indigenous municipalities, where livelihoods depend on rain-fed agriculture and climate shocks threaten subsistence. On the other hand, it may be less likely due to limited financial resources, shorter migration histories, and strong land attachments. Alternatively, origin-weather deviations may influence undocumented migration similarly across indigenous and nonindigenous contexts if their effects are sufficiently widespread.

Regarding return migration, origin-area weather extremes may also relate to the probability of return, but again, the direction of this relationship is theoretically ambiguous. Return may be more likely among migrants from indigenous municipalities if strong symbolic or economic ties to land draw them back despite weather-related uncertainty. Conversely, return may be less likely if sustained agricultural disruptions or local economic scarcities make reestablishment difficult. As with migration, origin-weather deviations may also constrain return similarly across contexts of indigeneity.

DATA AND METHODS

We assemble multiple data sources to examine the relationship between weather extremes and undocumented migration to and return from the US by origin contexts of indigeneity in Mexico.

Descriptive Statistics

Table 2 shows average sample characteristics across the context of indigeneity. Most of the person-year observations (N = 944,416) come from low-indigenous communities (N = 641,108). A community can move across indigeneity categories over time. As a result, the sum of communities across indigeneity categories (171) exceeds the total number of communities in our sample (N = 147). Undocumented migrants make up less than 1 percent of person-year observations, but about 5 to 8 percent of persons in our sample. Their share is higher among low- (7 percent) and high- (8 percent) indigenous communities. Return migration is common among undocumented migrants. About 18 percent in low-indigenous communities return after their first year in the US; 14 percent return after the second year, 10 percent after the third year, and 7 percent after the fourth year.⁵ Return rates are somewhat higher in medium-indigenous communities and, in high-indigenous communities, they are lower after the first year but higher in the following years. In terms of exposure to droughts, high-indigenous communities are at the top, with 23 percent of person-years experiencing very dry or dry weather (as opposed to 20 percent in low-indigenous communities and just 13 percent in medium-indigenous places). In terms of heat exposure, high- and medium-indigenous communities also rank highest, with 17 percent of person-years having hot or very hot conditions relative to 14 percent in low-indigenous communities. In terms of US destination weather exposure, migrants from low-indigenous communities are more likely to encounter very wet, very cool, or cool conditions and less likely to experience hot conditions than migrants from medium- or high-indigenous communities. The differences in means (determined with a Chi-square test for categorical indicators and with analysis of variance for continuous indicators) across community indigeneity groups are statistically significant (p < 0.05) for most indicators.

View this table:

• View inline
• View popup
• Download powerpoint

Table 2.

Average Sample Characteristics by Community Indigeneity Level

Figure 1 shows two maps: the share of indigenous-language speakers in Mexican municipalities according to the 1990 census (left) and rainfall deviation in Mexican municipalities according to 2000 Daymet data (right). The deviation equals the average rainfall in the same year minus the average rainfall in the baseline period (1980–1990) in a municipality, divided by the standard deviation of rainfall in the baseline period. Indigenous population shares are high in southeastern states, like Oaxaca, Yucatán, Chiapas, and Veracruz (see figure 1 for state names) and in central-western states, like Hidalgo, Puebla, and San Luis Potosí. Rainfall deficits (as indicated by the lighter shades) are also common in the same regions in 2000 (as they are in most other years in our data). This pattern aligns with national reports that indicate regions of indigenous concentration to be especially vulnerable to climate-related hazards (INECC 2021). The pattern is also consistent with descriptive statistics from our own data (table 2) that suggest higher exposure to rainfall and temperature extremes in high-indigenous communities.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 1.

Distribution of Indigenous Populations in 1990 and Precipitation Deviations in 2000 in Mexican Municipalities

Source: Authors’ calculations from the Mexican Census and Daymet data.

Note: Indigenous population is measured by the percent speaking an indigenous language in the 1990 Mexican Census. Precipitation deviation is computed with respect to the municipality baseline in 1980–1990; it equals the difference between the average precipitation in 2000 and the average in the baseline period, divided by the standard deviation in the baseline period. The optimal way to view this figure is in color. We refer readers of the print edition of this article to https://www.rsfjournal.org/content/12/4/102 to view the color version.

RESULTS

We first estimate a linear probability of first undocumented migration to the US. Our main indicators measure the presence of anomalies in rainfall (very dry, dry, wet, and very wet) and temperature (very cool, cool, hot, very hot) in the prior year relative to the community norm. Prior work establishes migration as a selective and network-driven process in the Mexico-US setting (Massey and Espinosa 1997; Massey et al. 1994). To capture these patterns, our models include individual characteristics such as age, sex, education, as well as household characteristics such as land, business, and property ownership and the presence of prior US migrants. They also include community characteristics, such as share in agriculture, share earning below the minimum wage, share earning more than twice the minimum wage, share with six or more years of schooling, share of ever migrants, homicide rate per million residents, and distance to the US border, along with state and year dummy variables. Table 3 shows results from models estimated on communities in municipalities with low (less than 1 percent), medium (greater than or equal to 1 and less than 10 percent), and high (10 percent or more) shares of indigenous-speaking populations. Recall that, for brevity, we refer to these communities as low-, medium-, or high-indigenous. Very dry weather is positively related to subsequent undocumented migration only in low-indigenous communities. Despite their higher exposure to weather extremes, people in high-indigenous communities migrate at a lower rate, possibly due to resource constraints (financial or network-based) that limit their ability to undertake a costly move (Adger 2006). Figure 2 offers a visual summary of origin and top US destination effects. Panel A shows coefficients for precipitation and temperature extremes in low-indigenous communities, with hollow circles for origin measures and gray circles for destination measures; horizontal lines show 95 percent confidence intervals, and coefficients whose intervals exclude zero are statistically significant. Panels B and C repeat the same results for medium- and high-indigenous communities. Rainfall in a community’s primary US destination is not associated with out-migration, while very hot destination conditions are negatively related to migration from medium-indigenous communities.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 2.

Coefficient Plots for Weather Indicators from Linear Probability Models of First Undocumented Migration

Source: Authors’ calculations from the MMP data.

Note: The three panels present the estimates from communities with (A) low, (B) medium, and (C) high levels of indigeneity, respectively. Each panel displays results for weather indicators in the origin community (hollow circle) and the top US destination for that community (full gray circle); the outcome is taking a first undocumented trip to the United States.

Social and economic factors also play a role in migration behaviors. Across all levels of indigeneity, the results show a higher likelihood of clandestine crossing among younger individuals, men relative to women, those with lower levels of education, and those living in households with prior migrants or in communities with higher prevalences of US migration. Households owning a business, and thus being tied to the origin community, are less likely to send migrants in all communities, but the p-value for the coefficient in low-indigenous communities (p = 0.09) is slightly below our threshold for statistical significance. Interestingly, in high-indigenous communities, undocumented migration is more likely from communities that are farther away from the US border, and homicide rates—a proxy for cartel presence—are negatively associated with migration, perhaps reflecting alternative income opportunities tied to trafficking or the deterrent effects of extortion and violence on would-be migrants. Availability of a water source (irrigation, dam, or reservoir) is also negatively associated with migration, but only in high-indigenous settings. This pattern suggests that factors mitigating potential agricultural loss (such as access to water during a drought) are especially important in less-resourced communities.

We next estimate a linear probability of return migration separately for low-, medium-, and high-indigenous communities. Our sample includes all undocumented migrants who have been in the US for at least a year. The outcome is whether a migrant returns sometime within their second year. We include all the control variables in the undocumented-migration model (except the top US destination weather measures), along with indicators for rainfall and temperature extremes in migrants’ actual US destination, migrants’ occupation (whether it is in agriculture or not), and destination state (California, Illinois, or Texas).⁷

In low-indigenous communities, extreme origin rainfall (very dry, wet, and very wet) is strongly associated with lower rates of return migration among the undocumented in the US. In medium-indigenous places, very dry conditions seem to repel return moves, and, in high-indigenous places, very wet conditions are linked to lower return rates. Two patterns stand out. First, return migration is more sensitive to weather extremes relative to undocumented migration. A range of extreme conditions (both wet and dry) are related to return to Mexico, while a single condition (very dry) is associated with undocumented migration to the US. Second, the link between weather extremes and return holds across all three levels of indigeneity, whereas that between weather and undocumented migration is present only in low-indigenous communities. Figure 3 presents a visual summary of these origin-weather effects alongside the corresponding conditions in migrants’ US destinations. Rainfall at the destination shows no association with return, whereas very cool destination conditions are positively related to return among migrants from high-indigenous communities.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 3.

Coefficient Plots for Weather Indicators from Linear Probability Models of Return Migration After One Year in the United States

Source: Authors’ calculations from the MMP data.

Note: The three panels present the estimates from communities with (A) low, (B) medium, and (C) high levels of indigeneity, respectively. Each panel displays results for weather indicators in the origin community (hollow circle) and the top US destination for that community (full gray circle); the outcome is returning to Mexico after remaining in the United States for one year.

The models so far considered return migration among migrants who have stayed a year in the US. We extend this analysis to consider the return among migrants who have been away from their home for two, three, and four years, respectively. In each case, we relate weather extremes in the origin community to return behavior in the following year. Figure 4 displays origin-weather coefficients for low-, medium-, and high-indigenous communities, and figure 5 shows the parallel coefficients for destination weather. Hollow circles show estimates for migrants in the US for two years, gray circles for three years, and black circles for four years.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 4.

Coefficient Plots for Origin Weather Indicators from Linear Probability Models of Return Migration After Two, Three, and Four Years in the United States

Source: Authors’ calculations from the MMP data.

Note: The three panels present the estimates from communities with (A) low, (B) medium, and (C) high levels of indigeneity, respectively. Each panel displays results for origin weather indicators from three different models, where the outcome is returning to Mexico after remaining in the United States for two (hollow circle), three (full gray circle), and four (black circle) years.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 5.

Coefficient Plots for US Destination Weather Indicators from Linear Probability Models of Return Migration After Two, Three, and Four Years in the United States

Source: Authors’ calculations from the MMP data.

Note: The three panels present the estimates from communities with (A) low, (B) medium, and (C) high levels of indigeneity, respectively. Each panel displays results for US destination weather indicators from three different models, where the outcome is returning to Mexico after remaining in the United States for two (hollow circle), three (full gray circle), and four (black circle) years.

In low-indigenous communities, wet origin conditions are associated with a lower likelihood of return migration among all migrants, regardless of whether they have spent two, three, or four years in the United States. In medium-indigenous communities, very dry weather is linked to a higher chance of return among all cohorts. Among migrants who have been in the US for four years, very wet and hot origin conditions correspond to lower return rates, whereas cool conditions show the opposite pattern. In high-indigenous communities, the very dry and very wet weather origin conditions are negatively associated with return patterns among migrants who have remained in the United States for four years. Figure 5 shows corresponding destination-weather patterns: very dry and very wet destination conditions are negatively associated with return among those from low-indigenous communities. For medium-indigenous communities, very cool and cool destination conditions are negatively related to return, while very wet conditions are positively associated with return.

Although the sample size drops across migrant cohorts, our models in figure 4 still detect signals that align with those from the larger sample of migrants observed after their first year at the destination (table 4). In medium- and high-indigenous communities, we have only 338 and 428 migrants, respectively, who have remained in the US for four years. Despite the small sample size, return behavior is linked to weather extremes in origin communities even among migrants who have left their communities several years ago; this link is strongest among low- and medium-indigenous communities.

DISCUSSION AND CONCLUSION

In Mexico, indigeneity—having origins in precolonial populations—is a central dimension of inequality. Although indigeneity is an individual characteristic, it is also a feature of place that patterns residents’ access to material and social resources, regardless of their own ethnicity. Exposure to weather extremes represents one additional disadvantage that we have examined in this article. Indigenous communities are not only more exposed to weather extremes but also face overlapping hardships associated with this exposure, given their rurality, dependence on rain-fed agriculture, and limited access to economic and social resources that might allow them to absorb the impacts of weather extremes. We considered whether weather deviations also shape opportunities for undocumented migration to and return from the United States by the context of indigeneity.

Drawing on large-scale survey data from the MMP, linked with contextual measures of indigeneity from the Mexican Census compiled by IPUMS International and high-resolution daily precipitation and temperature data from Daymet, we find asymmetries in how the context of indigeneity relates to undocumented migration and return following weather deviations. Results from linear probability models show that weather extremes are associated with an increased probability of undocumented migration from low-indigenous communities. By contrast, sustained weather extremes in origin communities are associated with a lower probability of return throughout migrants’ tenure in the United States, regardless of their origin communities’ context of indigeneity. We nonetheless find evidence that different weather deviations matter for return in different contexts of indigeneity: very dry weather is associated with delayed return to low- and medium-indigenous communities among more recent migrants, while very wet and cool conditions are associated with reduced return to high-indigenous communities among more established migrants. Thus, opportunities for undocumented migration as a response to weather deviations vary by the context of indigeneity; a range of weather extremes may work together to constrain return across contexts of indigeneity.

We propose that these associations operate through weather deviations’ impacts on agriculture. Our analysis focuses on 147 MMP communities that rely on maize production. By measuring weather extremes during the corn growing season, we assume—but cannot show directly—that weather-related migration operates through declining crop yields (see also Zhu et al. 2024). Ethnographic evidence supports this interpretation, illustrating how weather deviations threaten agricultural livelihoods (de Frece and Poole 2008; Ebel et al. 2018). Yet the relationship between weather and migration is more complex when filtered through the context of indigeneity. Our results indicate that undocumented migration increases primarily among low-indigenous communities. One potential explanation is that international migration is a preferred adaptation strategy in agricultural communities with longer migration histories and denser transnational networks (Hunter et al. 2015). Another is that low-indigenous communities possess greater access to financial and social resources—such as wage labor, savings, credit, and coyote networks—that lower the costs of migration independent of historical migration ties (Asad and Hwang 2019a, 2019b). In the latter case, individuals in high-indigenous communities may be “trapped” in place, wanting to move but lacking the resources to do so (Black and Collyer 2014); if so, high-indigenous communities could experience even greater deprivation in the future as the network effects of migration increase inequality over time (see, for example, Garip 2008). We emphasize, though, that a lack of statistical association does not imply that weather extremes are inconsequential in high-indigenous communities. Rather, their relationship with our proposed agricultural mechanism may operate differently. High-indigenous communities may have already adapted to weather variability in ways that reduce the need for international migration (Ebel and Castillo Cocom 2012; Ebel et al. 2018; Rogé et al. 2014). Residents may also remain rooted due to strong symbolic and economic attachments to land (Eakin 2005). We find support for both these possibilities in our regression models, though a more explicit consideration of each potential explanation remains to be conducted.

Future research should build on these findings in several ways. First, while our analysis reveals asymmetries in weather-related migration and return using municipal-level measures of indigeneity, the MMP lacks individual-level indicators of indigeneity. Our associations may be even stronger among indigenous-speaking or indigenous-identified people in each community, although our current data do not allow us to test this claim. Second, more granular data could better approximate the context of indigeneity at the community rather than municipal level. Third, exploring cross-level interactions between individual and contextual measures of indigeneity could clarify the mechanisms underlying these patterns. Finally, while our study focuses on Mexico-US migration, results may vary across contexts of indigeneity marked by particular histories of US-bound migration (Fox 2006). They may also differ in other migration corridors beyond Mexico and the United States.

Our findings nonetheless make three contributions. First, we show that, alongside being disproportionately rural, agricultural, and poor, one additional hardship correlated with the context of indigeneity is indigenous communities’ disproportionate exposure to weather extremes. Although many studies based in Mexico suggest this relationship using other proxies (for example, rurality or poverty), few explicitly consider the context of indigeneity, in part because most analyses occur at scales that obscure municipal-level variation. Our analysis of Mexican Census data demonstrates that indigenous municipalities, relative to nonindigenous municipalities, already experience greater weather-related threats and are also projected to face more in the future. This result reinforces calls from recent systematic reviews emphasizing the need for subnational and local analyses of climate vulnerability—including indigeneity—in diverse national contexts (Zahnow et al. 2025).

Second, we adopt a social demographic perspective to clarify insights from qualitative research on how the context of indigeneity relates to the migration journey—not just migration to but also return from the United States—following weather extremes. While ethnographic work has documented both migration and adaptation in response to weather deviations (Ebel and Castillo Cocom 2012; Rogé et al. 2014), it typically focuses on a handful of indigenous communities at a single point in time. By contrast, we use fine-grained daily weather data and municipal-level measures of indigeneity across Mexico. We find that, although there are certainly instances of migration from indigenous communities following weather deviations, these are not widespread within the study period. Instead, the relationship between weather deviations and migration is most evident in low-indigenous communities. Taken together, these results suggest that immobility—staying in place—may be a common response to weather shocks in the most indigenous communities (Eakin 2006; Schewel 2020).

Finally, by analyzing the full migration journey, we uncover an asymmetry in the relationship between weather extremes, migration, and return by the context of indigeneity. Migrants from low-indigenous communities are more likely to leave for the United States following weather deviations but, once abroad, migrants from all contexts of indigeneity are equally likely to remain as adverse conditions persist in their origin communities. Weather extremes, in other words, reproduce existing inequalities in opportunities for international migration while constraining return across the board. This latter finding remains relatively underexamined (but see Zhu et al. 2024). Since the 1980s, intensified border and interior immigration enforcement in the United States has made return migration to Mexico less feasible (Massey et al. 2016). Our results suggest that sustained weather deviations in origin communities may further deter return, particularly when these shocks disrupt agricultural livelihoods and make continued residence abroad—despite deportation risks—a preferred strategy. These dynamics may help explain stability in the size of the undocumented Mexican population in the United States despite declining inflows, as both enforcement and environmental change reduce the feasibility of return.

In this sense, weather extremes may represent a double disadvantage for undocumented immigrants who experience climate disruptions in their origin communities and face heightened environmental risks on arrival. Many undocumented migrants work in sectors such as agriculture, construction, and landscaping, where exposure to extreme environmental hazards is especially high (Orrenius and Zavodny 2009). If migrants from high-indigenous communities are disproportionately represented in these sectors in the United States, this represents an additional hardship associated with the origin context of indigeneity (Méndez et al. 2024). Moreover, by discouraging return, weather extremes may indirectly increase migrants’ exposure to US immigration enforcement, which lacks legal protections for individuals displaced by climate-related disasters (Waters 2025). Unauthorized entry—even when motivated by climate-related livelihood shocks—carries severe legal consequences, including visa bans and felony reentry charges (Asad 2023; Golash-Boza 2015; Tosh 2023). Our results thus underscore the urgency of integrating climate vulnerability into immigration policy (McLeman 2019).

As climate change intensifies, weather-related migration pressures are likely to grow. Yet current legal frameworks—both in the United States and under international refugee law—offer no recognition or protection for climate-displaced individuals (Garip and Reed 2025). Future policy discussions should consider humanitarian or targeted visa programs for those displaced by extreme weather events (Waters 2025). If our results are any indication, addressing these challenges requires understanding weather-related migration as shaped by overlapping systems of social and environmental inequality.