A Structural Text-Based Scaling Model for Analyzing Political Discourse

TL;DR

The Structural Text-Based Scaling model is introduced to infer ideological positions of speakers for latent topics from text data and it is seen that a speaker's region of origin influences their ideological position more than their religious affiliation.

Abstract

Scaling political actors based on their individual characteristics and behavior helps profiling and grouping them as well as understanding changes in the political landscape. In this paper we introduce the Structural Text-Based Scaling (STBS) model to infer ideological positions of speakers for latent topics from text data. We expand the usual Poisson factorization specification for topic modeling of text data and use flexible shrinkage priors to induce sparsity and enhance interpretability. We also incorporate speaker-specific covariates to assess their association with ideological positions. Applying STBS to U.S. Senate speeches from Congress session 114, we identify immigration and gun violence as the most polarizing topics between the two major parties in Congress. Additionally, we find that, in discussions about abortion, the gender of the speaker significantly influences their position, with female speakers focusing more on women's health. We also see that a speaker's region of origin influences their ideological position more than their religious affiliation.

Figures (12)

• Figure 1: The STBS model. Observed data (rectangles), latent variables of interest (rounded corners), additional parameters (circles). Colored planes indicate the relationship to documents ($D$), authors ($A$), the vocabulary ($V$), topics ($K$) and covariates ($L$).
• Figure 2: Comparison of polarity induced by the variance of $\phi_{\eta kv}^\mathrm{loc} \cdot \phi_{\mathbbb{i} ak}^\mathrm{loc}$ under fixed and under topic-specific ideological positions.
• Figure 3: Regression \ref{['eq:example1']} summary plot of the model with fixed ideological positions across topics. CCP codes: $*** < 0.001$, $** < 0.01$, $* < 0.05$, $\cdot < 0.1$.
• Figure 4: Topic 9 (Veterans and Health Care). Regression \ref{['eq:example2']} summary plot for the model with topic-specific ideological positions. CCP codes: $*** < 0.001$, $** < 0.01$, $* < 0.05$, $\cdot < 0.1$.
• Figure 5: Topic 11 (Climate Change). Regression \ref{['eq:example2']} summary plot for the model with topic-specific ideological positions. CCP codes: $*** < 0.001$, $** < 0.01$, $* < 0.05$, $\cdot < 0.1$.

Theorems & Definitions (2)

• Lemma 1
• proof