A computational politics model for assessing the IO (de)legitimation. A case study of twelve IOs in six post-socialist countries

The media channel is crucial in understanding the legitimation processes of International Organizations (IOs). This paper applies a computational politics framework to develop a new model of IO (de)legitimation. The model uses Natural Language Processing and multivariate linear regression analysis to quantify three critical dimensions of legitimation through media: intensity, tone, and narratives. The model is applied to a corpus on more than 1.3m newspaper articles from six Eurasian post-socialist countries in four languages and twelve IOs. It yields three key results. First, to control for the country and newspaper idiosyncrasies, the tone of IO legitimation discourse in multi-country studies should be measured by relative sentiment. Second, the legitimation discourse tone crucially depends on the topic of the narrative. Third, there are significant differences for countries and IOs analyzed. For example, even after controlling for the negative narrative related to the Covid-19 pandemic, WHO faces the strongest deligitimation in post-socialist media, closely related to China’s activity. The tone of articles featuring EBRD is positive, except for negative articles that mention influential Russian politicians. Russian media contribute to IOs delegitimation, while Polish media engage primarily in the IOs’ legitimation. The model can be used to test many legitimation-related hypotheses. We verify the symmetric “guilt by association” hypothesis and document that the world powers significantly contribute to IOs delegitimation in the post-socialist press. The developed model is easily scalable and can be applied to any number of countries and IOs, providing that sentiment lexicons for languages spoken in these countries exist. <br><br>Data files used in this research will be uploaded in October 2021.<br><br><br>