Benford's law

principle

In naturally generated datasets spanning multiple orders of magnitude, leading digits form a distinctive curve: about one-third begin with 1, while only small fractions begin with 8 or 9. Manipulation can disturb this fingerprint, making the law a screening tool for suspicious data—not proof of fraud.

A fifth-standard student could help uncover accounting or election fraud by ignoring the size of each number and looking only at its first digit. If the figures arose naturally, 1 should appear far more often than 9.

Growth leaves a first-digit fingerprint

When values grow across several orders of magnitude, they spend unequal amounts of that journey beginning with each digit. The resulting distribution is sharply lopsided: roughly one-third of values start with 1, then the frequency declines until 8 and 9 account for only about five and four percent. People fabricating figures tend not to reproduce that curve reliably, so altered data may reveal a different first-digit shape.

E1 R1

Where it shows up

Loan disbursements

An auditor can compare the first digits in bank-loan disbursement records with the expected curve. A mismatch identifies records or groups that deserve closer examination.

Election totals

The same screening idea has been applied to election data, including discussion of Iran’s 2009 election: inspect the aggregate digit pattern for anomalies before investigating particular results.

Office reimbursements

Reimbursement figures submitted by a manager or processed by HR offer another possible test: naturally varying claims should leave a different signature from numbers repeatedly chosen or manipulated by a person.

An anomaly is not a conviction

Benford’s law does not apply to every collection of numbers. Heights, for example, occupy a narrow range rather than spanning multiple orders of magnitude. Even in a suitable dataset, a deviation is a reason to investigate—not evidence that fraud occurred.

Run a first-digit triage

Take one naturally generated financial dataset that spans a wide range—such as disbursements or reimbursements—count its leading digits, and compare the percentages with the declining Benford curve. Use the largest departures to choose which records to audit first.