# Large language model reveals an increase in climate contrarian speech in the United States Congress

## TL;DR

A large language model found that climate contrarian speech in the US Congress has increased, with Republicans making such claims much more frequently than Democrats.

## Contribution

The paper introduces a scalable large language model to classify climate contrarianism in Congressional speeches and analyzes specific claims made.

## Key findings

- Republicans are roughly 12 times more likely than Democrats to make climate contrarian claims.
- An increasing proportion of Congressional speeches critique the costs of climate solutions.
- Demographic factors and fossil fuel employment in districts predict contrarian claims in speeches.

## Abstract

Contrarian voices on climate change have reached the highest levels of government in the United States: the US Congress has been a focal point for climate obstruction, playing an important role in impeding US climate action and stymieing global negotiations. We take a closer look at these voices by examining Congressional speeches from 1994 to 2023, a period containing important moments in the history of climate (in)action in the US. We contribute to the literature by 1) developing a scalable large language model (LLM) to accurately classify climate contrarianism in Congressional speech and 2) offering a systematic analysis of the specific contrarian claims in Congress. We demonstrate that an increasing proportion of speeches are critiquing the costs of climate solutions and Republicans are roughly 12 times more likely than Democrats to make contrarian claims. Statistical analysis further suggests that demographic factors and district-level fossil fuel employment predict claims making in floor speeches.

Congressional opposition to climate policies has grown in the United States, and Republicans are approximately twelve times more likely than Democrats to make contrarian claims, according to an analysis of 30 years of speeches using a custom large language model.