Mass ideology-based voting model

摘要

As one of the powerful tools in political science, ideal point estimation is always used to study the pattern behind the senators' voting behavior. In order to give a comprehensive estimation of senators' political positions, some researchers estimated the ideal points on different topics. However, for those senators who are not that polarized, their ideal points are so sensitive to the voting records that even a small change will make a big difference, which may mislead the readers. In this paper, we propose a mass ideology-based voting model taking the senators' latent ideology into consideration. Firstly, we model the senators' general ideal points by using the following links on Twitter due to the reason that we have homophily in social networks. Secondly, we use the roll call data of different bills, which can be decomposed as a combination of different topics, to estimate the senator's adjustment on different topics. Finally, we combine the general ideal points and the adjustments together to analyze the senator's political positions. Additionally, two-stage learning algorithms are also shown in the following section. Compared with the Issued-Adjusted model, our model has an edge on classifying the senators on different topics. This model can also be used to predict the voting behavior. Then, we show a case study of a moderate senator and try to explain her voting behavior for some bills according to our research.