PolicyCost.org

Measuring the human cost of federal policy changes since January 2025

Our Research Methodology

Last updated: January 18, 2026

PolicyCost.org uses rigorous data collection and analysis techniques to quantify the human, economic, and social impacts of recent federal policy changes. Our goal is to make these impacts visible and understandable through real-time counters that show the scale and scope of these changes.

This page explains our overall research approach, data sources, and calculation methodologies. For detailed information about specific counters, please visit the individual methodology pages linked below.

Research Approach

Our research process combines multiple approaches to ensure accuracy and comprehensiveness:

Advanced AI Research

We use AI research tools including Perplexity AI and Claude 3.7 Sonnet to systematically gather and analyze large volumes of policy data, economic indicators, and impact assessments. These tools help us identify patterns, correlations, and potential impacts that might be missed in manual research.

Primary Source Analysis

All AI-gathered data is verified against primary sources including government publications, peer-reviewed research, official statistics, and direct reports from affected agencies. We prioritize original documentation over secondary reporting.

Expert Validation

Our methodologies draw on established economic, public health, and policy analysis frameworks developed by recognized authorities in these fields. Where possible, we use calculation methods consistent with those used by government agencies and research institutions.

Conservative Estimation

When faced with a range of potential impacts, we generally use the more conservative estimates to avoid overstating effects. We acknowledge uncertainty and limitations in our methodology pages.

Individual Counter Methodologies

Each counter on PolicyCost.org focuses on a different aspect of policy impacts. Click below for detailed explanations of each counter's specific data sources, calculation methods, and assumptions: