Anomalies Detection Using the Benford's Law: Application to the Kenyan Presidential Elections of 2017

Abstract

In the modern times, the populace in most African countries are left wondering whether the declared election winner actually got the most votes. The validity of the declared election results in most cases remain questionable. In order to determine the validity of the declared results, an empirical statistical methodology could be used to give some hint and or evidence of anomalies in the declared election count data. This paper therefore considers a statistical method based on the pattern of digits in vote counts known as 2 digit Benfords Law (2BL) that is useful for detecting fraud or other anomalies. The 2BL methodology and other extensions are applied to detect the possible anomalies and fraud in the 2017 Kenyan presidential elections results data. The analysis show that the data for the top two presidential candidates: Uhuru Kenyatta and Raila Odinga do not follow the 2BL distribution. The digits are signi cantly di erent at 5% signi cance level when tested using the chi-square and the Euclidean tests. The mean absolute deviation (M.A.D) also con rms the non-conformity of the data to the 2BL distributions test. Further tests namely,the second order test, the summation test and the duplication test are utilized in order to detected any possible anomalies and fraud that could be present. All the three additional tests con rm the presence of fraud and anomalies in the data. These are red ags on the credibility of the presidential election results data published by the Independent Electoral and Boundaries Commission (IEBC).

Keywords: Anomalies, Benford's Law, Kenyan Presidential Elections 2017