Effects of out-word usage in colombian twitter data

In December 2023 I joined Professor Tomas Rodriguez’s project to replicate Rathje et al., (2021) work on the effects of out-group targeting terms in social media platforms. Using a dataset of all the tweets done by colombian congresspeople (that had twitter accounts) in 2020-2021 we built a set of dictionaries of the out-group targeting terms, and after validating it, explored the incidence of out-word usage in both the virality of the tweets as well as the engagement (number of likes). We extended the analysis by tagging each tweet using the BETO sentiment analysis model and executed the same analyses exploiting the sentiment categorization as well as the emotion variables. We’ve explored a variety of econometric specifications and have consistently found the same results as Rathje, et al. (2021) in the sense that the presence of out-words is positively related to tweet virality (see Figure 1).

The emotional tag of our tweets also yielded the larger presence of anger-tagged tweets during the 8 most heated weeks of the 2021 national strike in Colombia (see Figure 2). Other relevant results will be eventually shared in the working paper.