Rainfall, Drought, and Emigration in Guatemala, 2015-2018

By Sam Colvett

Introduction

Climate change and migration are some of the most hotly contested subjects in the United States today. This contention is especially evident with the various debates on both sides of the political aisle during the 2020 Presidential election. According to the Pew Research Center, 52% of Americans said that immigration was important to them in the 2020 election, and 42% of Americans indicated that climate change was important to them. Despite these being controversial issues, research into whether and to what extent these two phenomena are related to each other has only been addressed in recent years (e.g., Hunter et. al. 2015; Moran-Taylor and Taylor 2010, 2021). In the analysis that follows, I first explore the literature on how migration and environmental issues are interconnected. Following this discussion, I then examine such linkages in Guatemala, a country characterized by agricultural employment, changing climactic events, and food insecurity. In doing so, I model a statistical investigation into the lagged effects of rainfall variation on migration in Guatemala. My methods in this investigation are based heavily on a similar study that Puente et al. conducted in 2015 in Mexico. I focus this analysis on municipal emigration in the year 2018 and explain the causality of the rainfall variation by linking it to food insecurity throughout the country. I then explore the potential sustained impact of the severe 2015 drought by relating it to emigration rates in 2016 and 2017. I conclude by highlighting the findings of this investigation and the ways in which it could potentially be modified and extended in future studies.

Literature Review: Climate Change and Migration

While the effect of climate change on migration is now widely acknowledged, there is still some debate as to whether climate events contribute more towards short-range or long-range migration and whether that migration will be permanent or temporary. Nawrotzki et al. conducted a study in Mexico in which they considered the effects of climate change variables on international versus domestic migration and found that it more strongly affected the former. In these findings, they particularly note that both abnormally high temperatures and abnormally high precipitation affect what they refer to as the “probability of migration,” which is simply the likelihood that an individual will emigrate. Similarly, Bohra-Mishra et al. conducted a study in 2014 in Indonesia where they looked at the effects of rapid-onset disasters, rainfall, and temperature on the likelihood to migrate in households. In this study, they conclude that “some 95% of the migrant households end up migrating only once and do not return to their original province during the entire 15-year period” when slow onset disasters are the primary driving factors. However, they also found that “temporary and short distance moves rather than permanent moves” are more common with rapid-onset migration events. This migration pattern tends to predominate as migration due to slow-onset events can be more properly planned than spontaneous migration from natural disasters.

Bohra Mishra et. al. also find that “agriculture-related outmigration is mostly sensitive to temperature rather than precipitation.” This is re-affirmed by Thiede et al.’s 2016 study on climate effects on migration in South America (specifically Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, and Uruguay). However, in contrast, Puente et. al.’s study highlights different findings and uses data from the Mexican census to study how precipitation affects Mexican migration to the United States. In this study, the authors conclude that rainfall levels compared to the historical average at the municipal level in Mexico has a statistically significant impact on migration with “higher than normal levels of rainfall reducing migration.” Though they do not control for temperature in their study, which could be correlated closely with rainfall according to Bohra-Mishra et al., however, Puente et al. do control for factors such as education, marriage rates, and the number of emigrants 15 years prior on a municipal level. Here, then, I draw on Puente et al.’s study and explore a similar linear regression analysis of the effect of rainfall in Guatemala in 2015, 2016, and 2017 compared to the historical average on the number of migrants to the United States from a given municipality during 2018.

Lastly, there are limited studies on the effects of climate change on food security in heavily agricultural areas in places like Guatemala. The International Organization for Migration and the World Food Programme developed a ground-breaking study addressing Central American migration and food insecurity and state that “there is extremely limited academic research into how they relate to each other” . They find that in Latin America and the Caribbean, “Guatemala has the highest rate of chronic undernutrition, at nearly 50 percent nationally, with significantly higher levels in some parts of the country.” This is extremely significant to this study, as the report concludes that “there is a positive correlation between food insecurity and migration” in this region of Central America. Because crops are so important to nutrition in many areas of Guatemala and farmers rely on predictable levels of precipitation to maximize crop yield, studying the relationship between rainfall and migration is important because it considers food insecurity as a driving factor. Therefore, this is the causal mechanism that I explore in this study: that climate change affects crop yields, which affects people’s livelihoods, food security, and, ultimately, may play a role in their decision to migrate.

The Guatemalan Context

Building on previous studies that acknowledge Guatemala’s challenges with migration, climate change, human development, and food insecurity (e.g., Cohn et al. 2017; Magrin et al. 2014; Anderson et al. 2019; IOM 2016), I first outline specific country conditions related to these topics before beginning my analysis. Guatemala is a Central American country characterized by high levels of emigration. The Pew Research Center identifies Guatemala as an increasing migrant population in the United States relative to other Latin American migrant populations. They also affirm that work, family reunification, and flight from conflict are common causes for this migration trend. Besides being a large source of international migration to the United States, Guatemala also has several characteristics that make it optimal for studying the effects of environmental change. Anderson et al. point out that Guatemala experiences what is known as canícula or midsummer drought (MSD), which is a period of relatively little rainfall in the middle of the usually wet summer and typically occurs in July and August. Below is a graph of the average total monthly rainfall in Guatemala from 2007 to 2018 that shows the general dip in rainfall for the months of July and August.

Figure 1: Monthly Rainfall from CHIRPS data.

Anderson et al. confirm that the midsummer drought, on which “is bound to the livelihoods and food security of many rural communities… is expected to significantly shift by the end of the 21st century in response to anthropogenic modification of the climate system.” It is important that they explicitly link the MSD to food security in their study, as the IOM report also asserts that food security can be a significant factor in migrants’ decisions to move. To further illustrate this point, a policy brief from the Berkley Center for Religion, Peace, and World Affairs in Georgetown University acknowledges that “recurring drought(s)… have strained agricultural production, with an estimated crop loss of up to 60 percent” in Guatemala. Lastly, a report from the Scientific American drew a causal connection between drought and food insecurity in Central America and increased migration rates. They cite that “30% of the households with migrants in the drought affected areas cited climate-induced lack of food as the main reason for leaving their homes and becoming migrants,” reinforcing the findings from the IOM study on how food scarcity can lead to migration.

Guatemala conducted a census in 2018 for their entire population, and many of the indicators that I use in this analysis come from its results. There are several relevant findings from this census for the study of rainfall, food production, and migration in Guatemala. For one thing, the Guatemalan Census Report finds that between 2002 and 2018, the “urban population grew 7.7 percentage points” and that “in the history of Guatemala, this was the first measurement that reported a greater number of people in urban areas than in rural areas.” Thiede et al. “observed climate effects on migration to urban areas” in their study in South America meaning that climate change could partially explain this shift in Guatemala.

Figure 2: Percentage of people living in an urban area in Guatemala, 2002 vs 2018 (2002, 2018 censuses)

From the literature involving climate change, food security, and migration and the various studies on demographic and environmental changes in Guatemala, it is already clear that these phenomena are connected. In the sections that follow, I attempt to partially follow Puente et. al.’s 2015 methodology in Mexico to conduct a municipality-level analysis of the effects of rainfall levels in Guatemala on migration in the subsequent years to examine the linkages between climate change and migration.

Data and Methods

I chose migration data for this investigation from the National Institute of Statistics in Guatemala (2018 census and environmental statistics databases). There are several important factors to consider with this data source. First, Guatemala’s statistical capacity is not as sophisticated as in other Latin American countries. For example, the World Bank gave Guatemala a 71.1% rating for its overall statistical capacity score in 2018, a measure that considers “a nation’s ability to collect, analyze, and disseminate high-quality data about its population and economy.” In my analysis, however, I exclusively used data from the INE, potentially giving my analysis a higher internal reliability. Similarly to Puente et al., my dependent variable from INE is the number of migrants recorded per municipality in 2018. I control for the number of migrants in 2003 alone, as data for marriages and education was difficult to process due to the large size of the files. Further controls for marriage and education could be included in future investigations. Based on Puente et al., my reasoning for including the number of emigrants in 15 years prior is that higher previous levels of emigration could “reduce the information costs associated with migration” and encourage others in the municipality to “view migration as a potential opportunity” in the future.

As part of the 2018 census conducted by INE, individuals were asked to list the members of their household that had left since 2002 and not returned to Guatemala since. This question did not include internal migration, and only asked participants to list individuals who had gone to another country. This data source could potentially be skewed because of how hard it is to reach certain individuals (especially in rural areas) or inaccurate reporting. However, it is still beneficial because it was collected at such localized and individual level. While many of the entries left out the year of departure, I only utilized the number of emigrants for which this was recorded, meaning the number of migrants could be significantly higher, but certainly not lower, than the number listed. I chose precipitation data from the Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS) database, developed by the US Geological Survey and the Climate Hazards Center in conjunction with the University of California, Santa Barbara. I downloaded data for the Central America and Caribbean region (camer-carib) at the 0.05 degree resolution every month since 1981 to 2018 to have a solid base with which to compare recent precipitation trends in Guatemala. I then used ArcGIS, a geospatial processing software, to develop the CHIRPS raster files into tables usable in Excel. I took a yearly average of the rainfall across every month for each municipality and for Guatemala as a whole from this data. As Puente et al. point out, it is not only the amount of rainfall but also the difference in rainfall from the historical average that is important to consider with regards to migration. Therefore, I proceeded to subtract this 38 year average from the average monthly rainfall in 2015, 2016, and 2017 for each municipality. So, a data point of -66.777 for a given municipality in 2015 means that it received 66.777 millimeters less of rainfall in 2015 than the historical average from 1981 to 2018.

Figure 3: Rainfall totals in Guatemala, 2015-2017, and the 38 year average (CHIRPS).

Table 1: Variable and descriptive statistics.

To compare these variables, I ran a regression model between the deviance from the 38-year average rainfall in 2015 and the number of emigrants in 2018, with the number of emigrants in 2003 as a control variable. In this model, I control for emigration in 2003 alone, as data for educational attainment and marriage rates (both of which were used in Puente et al.’s analysis) were too large to process using the software I had available. There were 64 municipalities that lacked data for migration or rainfall, so they are excluded from this analysis. Nonetheless, the 267 municipality sample size still provides a varied and representative sample for municipalities in Guatemala.

Results and Discussion

In my regression analysis, I found that the rainfall in 2015 was significantly and negatively correlated with the number of migrants in 2018 at the p=.001 level. Migration 2003, the control variable, was also extremely positively significant, giving further proof for the migrant network theories that assert that community connections impact migration patterns.

Table 2: Models with Migration 2018 as the dependent variable (*** = Significance at the P=.01 level).

Figure 4: Scatterplot of Rain Levels, 2015 and Number of Emigrants.

Upon further investigation, it became abundantly clear that the reasons for the significant relationship between rainfall and emigration in 2015 was the severe drought that Guatemala suffered in 2015. The regional office of the United Nations Office for the Coordination of Humanitarian Affairs reports that this drought, caused in part by El Niño, was “one of the most severe in the history of Central America." They further note that, according to various research groups, “severe acute malnutrition oscillates from 3.3% in Polochic… to 5.7% in the east of the country” and cite the IOM report on food insecurity and migration to claim that this could be “a trigger for the migration of people living in the Northern Triangle of Central America.”

Figure 5: Monthly rainfall in Guatemala compared to the historical monthly average (CHIRPS).

The impact of the drought on crop yields can be evidenced by looking closer into the maize planting season, a crop that “occupies 40 percent of total agricultural land-use in Guatemala.” This crop is especially dependent on regular rainfall patterns and variations from the required moisture amounts can have drastic effects on corn yields. A corn crop handbook produced by Kansas State University claims that “one day of moisture stress… within a week after silking (1,400 growing degree days) can result in up to 8 percent yield losses” while “a period of 3 to 4 days of severe moisture stress at this time can easily reduce final grain yields by 30 percent.” Assuming many farmers began their planting between April and June, the “1,400 growing degree days,” a term incorporating both temperature and days since planting, would be reached sometime in the summer months, which were obviously significantly drier than average in 2015. The OCHA report further evidences this point in stating that subsistence farmers “are dependent on rainfall as they work in farming without irrigation, have limited access to basic health services and education, and face difficulties accessing the basic food basket.”

            With the drastically lower levels of precipitation in Guatemala in 2015 during the most vulnerable growing periods of maize, a staple crop, it seems clear that the yield from this year would be much lower than average. Therefore, it is worth considering how the 2015 drought affected emigration rates in 2016 and 2017 as well due to food insecurity. To test this, I ran a regression analysis with the difference in average monthly rainfall from 2015 as the independent variable, the migration in 2003 as the control, and emigration in 2016 and 2017 as the dependent variable. In these analyses, the deviance in monthly rainfall was even more significantly associated with out-migration, showing that the drought of 2015 had a sustained impact on emigration in the subsequent years.

Table 3: Results with Migration 2016 as dependent variable. (*** = Significance at the 0.001 level.)

Table 4: Results with Migration 2017 as dependent variable. (*** = Significance at the 0.001 level.)

Conclusion

The presence of the 2015 drought is clearly significantly related to migration levels in subsequent years. While future studies could examine whether the significant relationship between rainfall and emigration is sustained in the absence of such an extreme weather event, this study clearly shows that long-onset weather events are of importance in examining climate and migration relationships. Specifically, changing precipitation patterns affect crop yield, which then leads to increased food insecurity and consequently may affect migration decision-making processes. However, scholars highlight that even when environmental factors (like changing precipitation) are the proximate cause of migration, there are often several other population pressures at work that likely play a role in migrants’ decisions to move (e.g., Suhrke 1994; Hugo 1996). While this impact on subsistence farming and food crops certainly seems acute, future studies could further examine the relationships between precipitation and the loss of cash crops, which could prompt migration through imposing financial losses on families that depend on their sale.

Regardless, I believe this analysis and others like it (e.g., Puente et al. 2015; Thiede et al. 2016) show that migration and slow-onset climate events (like changing precipitation and drought) are clearly connected to one another. While the data I explore in this study might not be sufficient to show true causation in and of itself, it still further evidences the increasingly-acknowledged relationship between environmental change and migration. There are two lessons resulting from this statement: 1) Combatting climate change can have positive impacts for food security and responsible migration policy, and 2) The concept of a climate migrant or a climate refugee should be incorporated into existing migration law and policy. Climate change is already impacting people’s lives, and with the knowledge of just how extensive its effects can be, influential leaders could be uniquely positioned to make a greater impact in this area through more informed policymaking.