Quantifying polarization across political groups on key policy issues using sentiment analysis
Dennies Bor,Benjamin Seiyon Lee,Edward J. Oughton
DOI: https://doi.org/10.48550/arXiv.2302.07775
2023-02-15
Applications
Abstract:There is growing concern that over the past decade, industrialized democratic nations are becoming increasingly politically polarized. Indeed, elections in the US, UK, France, and Germany have all seen tightly won races, with notable examples including the 2016 Trump vs. Clinton presidential election and the UK's Brexit referendum. However, while there has been much qualitative discussion of polarization on key issues, there are few examples of formal quantitative assessments examining this topic. Therefore, in this paper, we undertake a statistical evaluation of political polarization for representatives elected to the US congress on key policy issues between 2021-2022. The method is based on applying sentiment analysis to Twitter data and developing quantitative analysis for six political groupings defined based on voting records. Two sets of policy groups are explored, including geopolitical policies (e.g., Ukraine-Russia, China, Taiwan, etc.) and domestic policies (e.g., abortion, climate change, LGBTQ, immigration, etc.). We find that out of the twelve policies explored here, gun control was the most politically polarizing, with significant polarization results found for all groups (four of which were P < 0.001). The next most polarizing issues include immigration and border control, fossil fuels, and Ukraine-Russia. Interestingly, the least polarized policy topics were Taiwan, LGBTQ, and the Chinese Communist Party, potentially demonstrating the highest degree of bipartisanship on these issues. The results can be used to guide future policy making, by helping to identify areas of common ground across political groups.