Predicting Left-Right Positions from Hand-Coded Content Analysis using Machine Learning

Abstract

The Manifesto Project’s widely used left-right index of party policy positions (RILE), built from human-coded sentences from party manifestos, can be predicted using machine learning. We demonstrate this using some simple classifiers to show that using these conservative approaches as a baseline, performance is already as good as human coders. It works in multiple languages. Using transfer learning, we also show how a model trained on coded manifesto sentences can be used on new texts to predict left-right positions, and validate these with independent survey-based evidence.

Patrick J. Chester

Postdoctoral researcher

Patrick Chester is a postdoctoral researcher at the China Data Lab at UC, San Diego who received his PhD from New York University’s Politics Department.