Comparative research on populism faces a core difficulty: measuring degrees of populism across many parties and countries over time. Textual analysis helps with this task, and automated tools can make measurement scalable and timely. This article introduces a supervised machine-learning approach that converts national party manifestos into a continuous score of party-level populism.

🔎 Measuring Populism From National Manifestos

A supervised machine-learning model is applied to the text of national party manifestos to generate a continuous populism score for parties. This approach is designed to:

• Measure populism across a large number of parties and countries without resource-intensive human coding
• Produce updated, comparable information for temporal and spatial analyses
• Provide a continuous (rather than binary) indicator that allows fine-grained distinctions and reduces arbitrary classification decisions

🧠 How the Method Works

• Supervised machine learning is used to perform textual analysis on national manifestos
• The model outputs a continuous score that serves as a proxy for a party's level of populism
• The automated pipeline scales to many parties and years, enabling broad comparative work

📈 Illustration: Trends in Six European Countries

The resulting populism score is used as a proxy for party-level populism to analyze average trends in six European countries beginning in the early 2000s and spanning nearly two decades.

❗️ Why This Matters

This measurement strategy enables more fine-grained, scalable, and up-to-date comparisons of populism across parties, countries, and time—facilitating improved empirical tests of theories about the rise, diffusion, and consequences of populism.