The impact of violence during the Mexican Revolution on Migration to the United States

with Ed Kosack and Zach Ward

 

Project funded by the 2020 Carnevali Research Grant, Economic History Society.

Media coverage: The Long Run, Yale Economic Growth Center

​

The number of individuals forcibly displaced, for reasons such as conflict and violence, has more than doubled from around forty million in 2011 to nearly ninety million in 2021, and the trend is expected to continue to increase in the coming year. Temporary conflicts might fundamentally shape migration flows depending on the magnitude and persistence of the migration response.

​

In this paper, we examine the migration response to violence during The Mexican Revolution (1910-1920), one the deadliest conflicts in world history: it is estimated that 1.4 million died—about 10% of the population—and another 350,000 fled north across the border. To do so, we collect two novel data sources that are unique in terms of frequency and completeness. First, we collect daily data on migration flows between 1910 and 1920 from individual border crossings registered at 24 entry points along the US-Mexico border. Second, we digitize daily data on insurgency events from military reports compiled in the "Military History of the Mexican Revolution." We link these two sources together at the municipa lity level (roughly equivalent to a county in the United States) and construct a month-by-municipality panel spanning from November 1910 to December 1915. 

​

We employ a fully flexible event-study design to estimate the impact of conflict events on migration. More specifically, we compare monthly migration rates from municipalities that experienced a conflict event to municipalities that had not yet experienced or never experienced a conflict event. We find that conflict caused a significant short-run increase in migration rates of 60\% relative to pre-event levels. This large effect only lasted for a few months after an event. After five months, migration rates returned to pre-event levels, and we do not find any permanent impact of conflict on migration rates beyond this time window. As part of our analysis, we show that our baseline estimates are unlikely to be biased by confounders, anticipatory behavior, interference between units in treatment assignment, or treatment effects heterogeneity, and therefore uncover the true effect of conflict on migration for treated municipalities. While we do not detect a permanent increase in migration overall, we find marginal evidence of larger and more persistent effects for women and children, which are likely to have changed the sex composition of the migration flow between 1910 and 1915. In addition, we show that violence was the main treatment channel, and we present new insights into how migration costs, migrant networks, and land ownership shape the decision to migrate in contexts of conflict.

​

Our main contribution is to use high-frequency data, which allows us to more accurately pinpoint the dynamics (magnitude, timing, and persistence) of migration in response to conflict and violence. We show that no dynamics can be identified with data varying at lower frequencies. Most studies on forced migration or displaced people examine conflict episodes characterized by the persecution of specific ethnic groups, whose migration response may not be representative of the whole population. We show that in the absence of persecution, violent conflicts can induce increases in migration rates and refugee flows. We also improve our historical understanding of the Mexican Revolution and the Mexico-US migration history. We know from anecdotal and indirect evidence that migration flows from Mexico to the United States increased during the Mexican Revolution. However, we are the first to provide direct, quantitative evidence on the impact that this conflict had on the scale and composition of flow at the local level.