Large language models (LLMs) enable researchers to analyze text at unprecedented scale and minimal cost. Researchers can now revisit old questions and tackle novel ones with rich data. We provide an econometric framework for realizing this potential in two empirical uses. For prediction problems—forecasting outcomes from text—valid conclusions require “no training leakage” between the LLM's training data and the researcher's sample, which can be enforced through careful model choice and research design. For estimation problems—automating the measurement of economic concepts for downstream analysis—valid downstream inference requires combining LLM outputs with a small validation sample to deliver consistent and precise estimates. Absent a validation sample, researchers cannot assess possible errors in LLM outputs, and consequently seemingly innocuous choices (which model, which prompt) can produce dramatically different parameter estimates. When used appropriately, LLMs are powerful tools that can expand the frontier of empirical economics.

We propose a mechanism linking legislative gridlock to voters’ support for candidates who hold extreme policy positions: voters rationally discount policy proposals on gridlocked policy issues because on these issues policy change is unlikely. When voters have preferences that are moderate and broadly aligned with a single party across policy issues, gridlock increases support for extreme co-partisan candidates. We test our mechanism in a large-scale online experiment in which we randomly vary subjects’ perceptions of gridlock and measure subjects’ support for candidates in candidate-choice tasks. We verify that greater perception of gridlock on a specific issue increases moderate, self-identified partisan subjects’ propensity to vote for extreme co-partisan candidates on the gridlocked issue. We show that our experimental evidence is consistent with our mechanism and that other mechanisms are less likely to underlie our main result. We discuss and analyse additional predictions of our mechanism, including a possibly moderating effect of gridlock that occurs when voters have preferences that are extreme and do not align with a single party across issues. Our theory offers a possible causal connection from gridlock to elite polarisation that may inform further empirical work and suggests a novel tradeoff between elite polarisation and policy stability in constitutional design.