Benford’s law and electoral integrity: A forensic analysis of African elections

This study examines the applicability of Benford’s Law (BL) as a forensic auditing tool for detecting anomalies in electoral results. BL, which predicts the expected distribution of leading digits in naturally occurring numerical datasets, has been widely employed in financial fraud detection and academic research validation. While prior studies have explored BL’s potential in election forensics, its reliability in distinguishing between genuine fraud and natural statistical deviations remains inconclusive. Using electoral data from Zimbabwe and Kenya, contrasted with benchmark cases from more stable democracies that are South Africa and Botswana, this research applies multiple BL digit tests (first-digit, second-digit, and first-two-digit analyses) alongside complementary statistical measures (chi-square, Kolmogorov-Smirnov, Mean Absolute Deviation, and p-value tests). The results indicate that while BL can flag irregularities in election data, its limitations as a standalone tool necessitate caution. False positives may arise due to legitimate data quirks, and contextual factors can distort digit distribution patterns. The study concludes that BL should serve as a preliminary screening mechanism rather than definitive proof of electoral manipulation. To enhance election integrity, future forensic audits should integrate BL with advanced statistical techniques or machine learning models. These findings contribute to methodological debates in election forensics and provide practical recommendations for strengthening post-election audit frameworks.

Grantor dc:publisher

University of Pretoria

Year dc:date.issued

2025

Advisor dc:contributor.advisor

• Saville, Adrian

• UCTD
• Benford's Law
• Elections
• Election Forensics

Statement dc:rights

• © 2024 University of Pretoria. All rights reserved. The copyright in this work vests in the University of Pretoria. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of the University of Pretoria.

Language dc:language.iso

en

Dc Identifier Other

S2025

OAI identifier oai:identifier

oai:repository.up.ac.za:2263/103081