Benford's Law (Accounting Fraud Detection)

Frank Benford, a physicist at General Electric, documented the phenomenon in 1938 after observing that logarithm tables were more worn at the beginning — where numbers starting with 1 were listed — than at the end. The law states that the probability of a first digit being d equals log10(1 + 1/d). This produces probabilities of approximately 30.1% for digit 1, 17.6% for digit 2, 12.5% for digit 3, declining to 4.6% for digit 9. The law holds across an astonishing range of naturally occurring datasets including population figures, stock prices, physical constants, and financial accounts.

The forensic application of Benford's Law to financial statements rested on a straightforward insight: naturally generated accounting data — revenue figures, expense line items, account balances — should follow the Benford distribution because they arise from multiplicative growth processes and span several orders of magnitude. When a company or individual fabricates numbers, they tend to use intuitive but statistically unnatural distributions — for example, starting many invented figures with 5 or 7 — producing a fingerprint of manipulation detectable through statistical testing.

In Indian forensic accounting practice, Benford's Law was used as a screening tool by statutory auditors, internal audit functions, and forensic accounting firms engaged in due diligence or fraud investigation. Auditors extracted large transaction datasets — vendor payment records, sales invoices, journal entry amounts — and computed the observed first-digit frequency distribution. A chi-squared test or a mean absolute deviation (MAD) test was then applied to measure how much the observed distribution deviated from the expected Benford distribution. High deviations in specific digit ranges triggered targeted investigation of the underlying transactions.

Several high-profile Indian corporate fraud cases that subsequently came to light through other means were examined retrospectively and found to exhibit Benford anomalies. Companies engaged in inflating revenues through fictitious invoices, overstating trade receivables, or manipulating expense figures to smooth earnings often left detectable first-digit distribution fingerprints. The Satyam Computer Services fraud of 2008-2009 — one of India's largest accounting scandals — was later cited in academic and practitioner studies as a case where retrospective Benford analysis would have flagged anomalies in cash and bank balance reporting years before the fraud was disclosed.

Practical limitations were important to understand. Benford's Law did not apply to all financial datasets. It was inappropriate for data with restricted ranges (telephone numbers, prices constrained by regulation), data with minimum threshold values (minimum transaction sizes), or datasets with fewer than 500 observations. Financial ratios — which were derived variables rather than raw transaction data — also did not follow Benford distributions. Applying Benford analysis to inappropriate datasets produced false signals that wasted investigative resources and could incorrectly flag legitimate companies.

For Indian equity investors conducting forensic pre-investment analysis, Benford's Law was most usefully applied to disaggregated data — when company-level transaction files were available, typically in the context of acquisition due diligence where management provided detailed records. At the published financial statement level, where only aggregated line items were available, Benford analysis had limited applicability but still served as a starting point for flagging unusual rounding patterns or suspiciously round number concentrations that warranted further investigation.