Benford's law — methodology

This page explains why the Benford test exists, why we prefer the second-digit test (2BL) to the first-digit test for election data, what thresholds we apply, and how MAD and χ² should be read. The goal is to make transparent both what the test can and cannot say.

What you'll find

• Why the first-digit test often shows "deviation" on clean electoral data.
• Why 2BL (second digit) is preferred — Mebane's line of work.
• Thresholds: minimum 10 votes per section, minimum 30 sections per party.
• MAD bands: Close to Benford / Moderate / Strong deviation.

For the live screen see Benford's law by party.