Large Language Models Can Be Used to Estimate the Latent Positions of Politicians
This addresses measurement challenges in social science by providing new tools for estimating political positions, though it is an incremental application of existing LLM technology.
The researchers tackled the problem of estimating politicians' latent positions when data is limited by using large language models (LLMs) to pairwise compare lawmakers and scale the results, creating measures for U.S. senators on liberal-conservative ideology, gun control, and abortion. Their gun control and abortion measures predicted interest group ratings and votes better than ideology alone.
Existing approaches to estimating politicians' latent positions along specific dimensions often fail when relevant data is limited. We leverage the embedded knowledge in generative large language models (LLMs) to address this challenge and measure lawmakers' positions along specific political or policy dimensions. We prompt an instruction/dialogue-tuned LLM to pairwise compare lawmakers and then scale the resulting graph using the Bradley-Terry model. We estimate novel measures of U.S. senators' positions on liberal-conservative ideology, gun control, and abortion. Our liberal-conservative scale, used to validate LLM-driven scaling, strongly correlates with existing measures and offsets interpretive gaps, suggesting LLMs synthesize relevant data from internet and digitized media rather than memorizing existing measures. Our gun control and abortion measures -- the first of their kind -- differ from the liberal-conservative scale in face-valid ways and predict interest group ratings and legislator votes better than ideology alone. Our findings suggest LLMs hold promise for solving complex social science measurement problems.