Book Chapter
Understanding opinions towards migrants in transit: an analysis of tweets on migrant caravans in the US and Mexico
In: Aiello, L. M., Mejova, Y., Seneviratne, O., Sun, J., Kaiser, S., Staab, S. (Eds.): WEBSCI '24: proceedings of the 16th ACM Web Science Conference, Stuttgart, Germany, 22-24 May 2024, 1–10
New York, NY, Association for Computing Machinery (ACM) (2024)
Abstract
The study of opinions towards migrants is profoundly important to understanding migration as well as to politics. Previous research has contributed to understanding anti-immigrant attitudes using social media data. However, there is still a need for a better understanding of opinions towards migrants in transit. We study the case of Central American migrant caravans from 2018 to 2021 by looking at the opinions in both the US, the destination country, and Mexico, the transit country. Media highly covered these events, and an online debate about them started on social media. Our research aims to understand how migrant caravans are discussed online. We are particularly interested in how media salience and geographical variables are associated with the sentiment intensity of the opinions. We combine geolocated data from Twitter, GDELT (Global Database of Events, Language, and Tone), and Survey and Census data for the US and Mexico. We use topic modeling to find the latent topics within the online Twitter discussion, and VADER sentiment analysis to quantify tweets' sentiments to calculate the sentiment intensity score that is used as the dependent variable of our OLS regression models. For both countries, we found that similar topics were discussed with a more political discussion in the US. Our analysis of the sentiment score revealed that sentiment does not reflect stance adequately, which led us to analyze the sentiment intensity score (absolute value of sentiment). We found that, for Mexico, when the media generated a higher number of news articles about migrant caravans, the sentiment intensity was higher. For the geographical variables, we found no significant association in the US; however, for Mexico, tweets in bordering states had a lower sentiment intensity. These results shed light on the differences in the determinants of sentiment intensity in opinions between the two countries.
Keywords: Mexico, USA