Introduction

Since the Mexican Drug War was started in December 2006 by the Mexican government, there has been an increase in violence and public insecurity, as a result of the confrontations between the armed forces and criminal groups, the struggle between drug cartels for drug production and distribution, and the increase in forced displacement due to insecurity and violence in different regions of the country (^{CMDPDH, 2014}; ^{CNDH, 2016}; ^{CONAPO, 2019a}; ^{IDMC, 2010}). However, few empirical studies have tried to relate the increase in violence with the rising levels of forced internal displacement, and there are even fewer analyzes that attempt to quantify and characterize displaced people in recent years (^{Centro de Estudios Migratorios, 2020}; ^{CMDPDH, 2019}; ^{CONAPO, 2019b}).

Due to the lack of official and statistical information on the magnitude and characteristics of forced internal displacement in the country, this study explores the relationship of violence with changes in municipal internal emigration for the 1995-2015 period. First, the existence of socio-economic and demographic differences of emigrants from municipalities with high levels of violence is determined, compared to the rest of the internal migrants in Mexico. Subsequently, the effects of violence on female and male municipal internal emigration are analyzed before, during, and after the so-called “Mexican Drug War.”

Forced displacements in mexico and the world

Worldwide internal and external forced displacements have increased in the last two decades by 89.8 percent, from 37.3 million people in 1996 to 70.8 million in 2018, while the population grew in the same period has been by 31.6 percent (^{ACNUR, 2016}, ²⁰¹⁹; ^{World Bank, 2020}). Despite this, most of the migration research in the world continues to focus on labor and /or economic migration, leaving out forced displacements caused by different kinds of violence, natural disasters, or mega construction projects that force individuals and families to leave their homes and move to other places within or outside their countries (^{ACNUR, 2019}, ^{ONU, 1998}).

Depending on the international borders crossed by people, studies that address the relationship between forced displacement and different types of violence can be divided into forced internal displacement and forced external displacement (refugees or asylum seekers). Although there can be many and varied types of violence and perpetrators, most of the empirical research has focused on collective violence inflicted by different criminal, communal, and guerrilla groups, or by the State itself against certain groups of individuals, or the population at large (^{ACNUR, 2016}; ^{OMS, 2002}).

There are several reasons why people are forced to leave their homes due to the increase in violence; however, research conducted in America (^{Alvarado & Massey, 2010}; ^{Morrison, 1993}; ^{Ramírez & Meza, 2012}) and Africa (^{Bariagaber, 1997}) concur that individuals and/or families decide to move to other regions or countries when a violence threshold is exceeded; in other words, until the costs of staying in their homes are greater than the costs of displacement. In addition to the economic costs of traveling, displacement also implies the loss, at least temporarily, of their assets and sources of income. For this reason, leaving their homes is a drastic action people take when they face imminent life and safety-threatening risk.

As ^{Morrison (1993}) and ^{Bariagaber (1997}) note, while some of the displaced persons can leave their homes in a matter of hours or days -particularly as a result of direct threats or violent actions- other displaced people make this decision after months and even years of violence, reaching a point where they can no longer continue their productive activities or where the safety of a member or the whole family is imminently jeopardized.

The socio-economic and demographic characteristics of the forcibly displaced people can be very diverse, particularly in contexts of generalized violence within a country or region. However, studies such as those of ^{Pareja and Domínguez (2014}), of the National Center for Historical Memory (^{CNMH, 2015}), and that of ^{Hennion (2014}), show that women and children are among the most common displaced persons because they are more vulnerable to violent events that may occur in their home region.

On the other hand, as revealed by the research of ^{Bariagaber (1997}), in Africa, and of ^{Sayago (2011)}, in Latin America, most of these movements occur at a short distance, particularly because of the displaced people’s desire to return to their home region in the short or medium term. In addition to this, longer travel distances mean higher travel costs, which makes it impossible for most people to travel to places further away. In this regard, ^{Bariagaber’s study (1997)} demonstrates that internal or external forced displacements are conditioned by the size of the violent conflict, the distance between home regions and international borders, as well as by the social and family networks that can help people reduce several economic and social costs of moving within or outside their countries.

In the case of Mexico, forced displacement dates back to the 1970s and 1990s, particularly due to communal, agrarian, and megaproject conflicts in Guerrero, Oaxaca, and Chiapas; however, this phenomenon grew substantially in the 1990s as a result of the uprising of the Zapatista Army of National Liberation (EZLN), displacing up to 30,000 people in Chiapas between 1994 and 2011 (^{CMDPDH, 2014}; ^{Rubio Díaz-Leal, 2014}). Nevertheless, as documented by the Mexican Commission for the Defense and Promotion of Human Rights (^{CMDPDH, 2014}), the National Human Rights Commission (^{CNDH, 2016}), and the Internal Displacement Monitoring Center (^{IDMC, 2020}), forced displacements caused by violence apparently increased in Mexico after a change in the security strategy in 2007.

In this regard, in a first effort to relate violence to international migration to the United States, ^{Alvarado and Massey (2010}) analyze how violence impacts the probability of emigration in four Latin American countries between 1979-2003. The authors find that for the specific case of Mexico, there is an inverse relationship between the increase in violence and the probability of people emigrating to the United States; the number of emigrants reduces when violence escalates. They conclude that the increase in violence has a greater impact on lower social classes, which limits their international migration, as they do not have sufficient economic resources to migrate. Furthermore, the country’s social and political status has prevented violence from driving out the middle and upper social classes.

In a subsequent investigation by ^{Ramírez and Meza (2012}), the authors analyze the relationship of criminal violence in Mexico with international emigration to the United States for the 2005-2010 period, given the increase in violence recorded after a change in the security strategy in 2007. The authors conduct an analysis at a municipal level where they find evidence of an increase in migration to the United States as a result of an escalation of violence in the municipality. However, they also mention that this is not a linear relationship because until violence exceeds a certain level (or threshold), it has an impact on emigration rates, but before exceeding this threshold, it reduces the levels of international emigration in Mexico.

For their part, ^{Acosta and Cruz (2016}) analyze the economic and social factors associated with internal migration in Mexico for the 1995-2010 period, determining that during the 2005-2010 period, violence had a negative and significant effect on a state’s probability of having a positive net balance, increasing internal emigration flows and reducing internal immigration flows in states with high rates of violence; for instance, Chihuahua, Durango, and Tamaulipas, all highly violent states, saw a reduction in their net migratory balances during this period. Therefore, the authors conclude that, for the 2005-2010 period, variables related to social conditions and quality of life have a significant impact on internal migration in Mexico.

On the other hand, to show the causal effect of violence on emigration rates in regions and municipalities with higher levels of violence, ^{Gutiérrez and Rivero (2012}) measure the violence effect on municipal internal migration based on counterfactual models, showing that during 2010 violence increased internal migration in municipalities with high levels of violence. However, this analysis excludes the period before the implementation of the 2007 security strategy; therefore, it does not show the possible changes in forced internal displacement as a result of this security strategy implementation.

In short, it can be stated that forced displacements, both internal and external, show differences in relation to economic or labor migration since it is a phenomenon in which people are forced to migrate, leaving behind their possessions and patrimony to try to safeguard their lives. Although this may be a planned event that results from evaluating the pros and cons of leaving or staying, the outcome is the loss of social and economic well- being caused by leaving their homes, which affects women and children the most.

However, in Mexico, most of the research has focused on international migration (^{Alvarado & Massey, 2010}; ^{Ramírez & Meza, 2012}), leaving out forced internal displacement, ignoring if the socio-economic and demographic characteristics are different from the rest of the migrants, during a time of change where violence escalated in northern Mexico and municipalities of Guerrero and Michoacán, as a result of the federal security strategy (^{Arteaga-Botello, Dávila-Cervantes, & Pardo-Montaño, 2019}; ^{CMDPDH, 2019}; ^{CONAPO, 2019a}).

This article contributes to the violence effect evaluation on female and male municipal internal emigration rates for three different periods, trying to find differences before, during, and after the change in the 2007 government security strategy. Furthermore, it analyzes whether the impact of violence between 1995 and 2015 is the same at different levels of internal emigration, since, as mentioned above, the violence effect may be greater in municipalities with higher emigration rates if violence exceeds the costs of staying in their home regions, leading to the departure of hundreds and even thousands of people and/or families from some communities and municipalities of the country.

This article is based on the hypothesis that there is a significant effect of violence on the increase in female and male municipal internal emigration rates, which has escalated after the implementation of the 2007 security strategy. However, regional differences in the rising level of violence have had a differentiated impact on emigration rates, where the greatest effects have been on municipalities with higher rates of internal emigration and higher rates of violence; therefore, it is expected that the effect of violence will be different at distinct levels of municipal internal emigration.

Data

This article uses microdata from the Population and Housing Censuses (CPV in Spanish) of 2000 and 2010 (^{INEGI, 2018a}) and the data from the Intercensal Survey (EI in Spanish) 2015 prepared by the National Institute of Statistics and Geography (^{INEGI, 2018b}). Census data provide information about people’s home municipalities at the time of the census and five previous years for the 1995-2000, 2005-2010, and 2010-2015 periods, calculating municipal internal migration indicators for the total number of municipalities at the beginning of each period (2,428 municipalities in 1995; 2,454 in 2005; and 2,456 in 2010).²

These indicators are complemented by other control indicators for the economic, social, and demographic determinants of internal migration in Mexico proposed by ^{Acosta and Cruz (2016}), ^{Massey et al. (1993}), and ^{Gutiérrez and Rivero (2012}). Given the availability and periodicity of the data, this analysis shows variations according to the information obtained for each five-year period, which represents a limitation of the data sources and the findings obtained (see results tables).

Variables

Due to the lack of statistics on forced internal displacement at the municipal level in Mexico, the dependent variables are the male, female, and total emigration rates per municipality for the periods 1995-2000, 2005-2010, and 2010-2015 estimated from the question of the place of residence five years earlier, which is defined as:

Where:

TEImij: Intermunicipal emigration rate of municipality i from year t to t+4 MIi: Total emigrants from municipality i from year t to t+4

POBi: Total population of the municipality i half of the period t to t+4

The analysis of internal emigration rates is limited to the municipalities at the beginning of each period because including recently created municipalities would result in miscalculations of their emigration rates, as well as the lack of data available for the analysis.

Violence, an independent variable of interest, is measured from the five-year homicide rate per 1000 inhabitants at the municipal level, which is constructed with the administrative records of deaths from ^{INEGI (2020)}, particularly data on homicides per municipality. The World Health Organization (^{OMS, 2002}) points out that this variable provides a measure of the degree of violence and insecurity that occurs in a region or country as it is a comparable indicator and with fewer underreporting problems than other crimes. Furthermore, it has been used by other studies on violence in Mexico and Latin America, such as those by ^{Alvarado and Massey (2010}), ^{Enamorado, López-Calva, and Rodríguez-Castelán (2014}), and ^{Gutiérrez and Rivero (2012}).

Social, economic, and demographic indicators obtained from ^{INEGI (2018a}, ^2018b), the National Population Council (^{CONAPO, 2013}, ²⁰¹⁵), and the National Council for the Evaluation of Social Development Policy (^{CONEVAL, 2015}) are considered independent control variables. They are presented in Table 1, which shows its definition, type of indicator, source of collection, and availability over time. The Absolute Migration Intensity Index (IAIM in Spanish) and the Absolute Municipal Marginalization Degree (GAMM in Spanish) are used because they allow making comparisons between the municipalities over different years (2000 and 2010), unlike the Migratory Intensity Index (IIM in Spanish) and of the Marginalization Index (IM in Spanish), which only allow making comparisons for the same year (^{CONAPO, 2015}).

Methods

First, the socio-economic and demographic characteristics of emigrants from municipalities with very high violence rates are explored with regard to the characteristics of emigrants from municipalities with very low violence rates. Emigrant’s home municipalities are divided into five equal-sized groups according to their levels of violence: very high, high, medium, low, and very low violence using their five-year homicide rates per 1,000 inhabitants as an indicator.

Subsequently, for the analysis of the economic, demographic, and social factors associated with the municipal internal emigration rates (female, male and total), we start from log-linear models (linear regression of the logarithm of the dependent variable) to determine if, in particular, violence effect is significant on municipal emigration rates for the 1995-2000, 2005-2010, and 2010-2015 periods. This type of model is used because the internal emigration rate (dependent variable) cannot take negative values, so simple linear regressions cannot be used, and logarithmic transformations must be applied to municipal internal emigration rates (^{Gordon, 2015}). This type of model makes it possible to determine which of the proposed variables in the models have had a significant effect on the increase or reduction of municipal internal emigration rates in the selected periods.

Finally, to test whether the effect of violence on emigration rates is significant for different levels of emigration, multinomial logistic regression models are developed, in which the dependent variable is defined from the classification of municipalities into five equal-sized groups according to their internal emigration rates: very low, low, medium, high and very high internal emigration, taking the very low internal emigration as a reference category of the dependent variable. These models are used to compare the effect of the factors on different degrees of emigration, and particularly, to compare whether the effect of violence, if any, shows variations between different levels of emigration (^{Gordon, 2015}).

Socio-economic and demographic characteristics of migrants from municipalities with high levels of violence

This section analyzes the socio-economic and demographic characteristics of migrants from municipalities with very high levels of violence (quintile V), which are compared with data for emigrants from municipalities with very low levels of violence (quintile I) to identify the differences that may result from different levels of violence in their home municipalities. As can be observed in Table 2, emigrants from municipalities with very high and very low violence tend to be mostly working-age women between 25 and 28 years old. However, as of 2010, the age of emigrants from municipalities with very high violence increases, and a growth of 9.7% is observed when comparing their average age between 2000 and 2015.

In general, the average schooling of migrants from municipalities in quintile I and quintile V tends to increase over time (from 9 to 11 years), but as of 2010, there is a significant increase in the schooling of migrants from municipalities in quintile V (very high violence), from 8.6 years in 2000 to 10.3 years in 2010 (an increase of 19.8%), while there is a decrease of 2.2% for migrants from municipalities with very low violence in the same period.

Most of the emigrants in the three analyzed periods are married or live together; however, there is a higher percentage of widowers or separated emigrants from municipalities with very high violence compared to migrants from municipalities with very low violence, particularly in 2010 when the difference between the two groups was of 2.6% points, that is, 38.2% more widowers or separated emigrants from municipalities with high violence.

Throughout the three periods, emigrants from quintile I show higher percentages of indigenous language speakers compared to emigrants from quintile V. While emigrants from quintile V tended to show higher percentages of female headship in 1995 and 2010, this changed in 2015 because quintile V had a lower percentage of female-headed households, with a difference of 1.6% points between quintiles.

In the three analyzed periods, it is observed that emigrants from municipalities with high violence show higher percentages of unemployment after migration than migrants from municipalities with low violence. Migrants from municipalities with high violence show the highest level of unemployment, registering 3% of their population aged 15 to 64 years, particularly in 2010. That said, the previous data does not provide information to learn if they were also unemployed before the migration; therefore, surveys are necessary to determine the characteristics of people before and after their displacement.

Although most migrants from highly violent municipalities are employed, there has been a percentage increase in unemployed migrants after their emigration in recent years; this may have an even more negative effect on their vulnerability condition compared to other groups of migrants, particularly those who were displaced due to violence and insecurity.

In short, migrants from municipalities with very high violence (quintile V) are mostly working-age women with an education that in recent years has tended to increase to little more than secondary education. Even though most of these migrants are married or living in consensual unions, there is a high percentage of widowers and separated persons with respect to migrants from municipalities with very low violence during the three analyzed periods. This could be associated with other factors, such as the increase in homicides related to organized crime in these municipalities and family disintegration as a result of violence.

Conclusions

In this article, linear and multinomial models were developed to determine the effect of violence on municipal internal emigration rates in three different periods: before, during, and after the “Mexican Drug War,” the federal security strategy initiated in December 2006 by former President Felipe Calderón (^{IDMC, 2010}). The purpose of the analysis was to identify the effects of this strategy on male and female emigration rates at a municipal level, and indirectly, the effect of violence on the increase in forced internal displacement in Mexico in recent years. On the other hand, the effect of violence was analyzed to learn if there were differences at distinct degrees of internal emigration to examine the hypothesis of the violence threshold.

The empirical results show that the increase in violence has greater significant effects on female emigration rates in two of the three studied periods: 1995-2000 and 2010-2015. As for the 2005-2010 period, the effect of violence on male emigration rates was similar to that of female emigration rates when homicide rates, particularly male homicide rates, escalated as a result of the “Mexican Drug War” and the confrontations between criminal groups. The foregoing statement seems to be consistent with other qualitative studies, such as those conducted by the ^{CMDPDH (2019)}, that show that most of the forced internal displacement in Mexico in recent years has been of women and children.

Furthermore, the violence effect appears to be non-linear since the increase in violence had a significant effect on the relative risk of a municipality having high and very high emigration rates for the three analyzed periods, but not for other quintiles of lower emigration rates. Therefore, this article contributes to the hypothesis of the violence threshold; in other words, people decide to leave their home regions when the costs of violence exceed the costs of migrating, since before reaching this point, violence does not have significant effects on emigration rates.

Although the purpose of this analysis was not to determine the level or levels of violence at which people decide to move to safeguard their lives -since it is believed that they depend on the individual, family, and context characteristics-, the results showed that particularly 20% of the municipalities with the highest homicide rates present significant differences in their internal emigration rates, when compared with the rest of the municipalities, particularly when compared with 20 percent of the municipalities with the lowest homicide rates.

Therefore, this analysis contributes to highlighting the social effects of the increase in violence in Mexico as a result of the Mexican Drug War, and particularly of the increase in internal migration in Mexico caused by the increase in forced displacement. This article also shows the need for policies and care strategies differentiated by sex for victims of forced displacement in regions where confrontations between criminal groups and the armed forces have led to the highest homicide rates recorded in the country in the last twenty years, in addition to various social and economic problems, both for the population that has migrated and for the population that has remained in these communities.

However, this article leaves out the spatial analysis of the increase in violence and migration within the country that can help to focus economic and political efforts to assist the victims, as well as the possible spatial effect on the decisions of people who have been forced to displace in recent years. Nor does it analyze the effect of violence on people’s individual decisions to migrate within or outside Mexico. Nor the possible levels or thresholds of violence at which individuals and families decide to leave their home regions.

Future work could investigate the existence of communities’ “clusters” with high rates of violence and migration, as well as determine causal effects of violence on internal or international emigration. Furthermore, qualitative studies could explore how the individual and family decisions of displaced persons are reached in Mexico, as well as their social, economic, and demographic effects.

Translation: Bárbara Ramírez De Valdez