Detecting anomalous data in household surveys: Evidence for Argentina

Abstract

This paper advances in the detection of anomalous data in income reports of Argentina.In particular, income declared by households surveyed in the Encuesta Permanente deHogares (EPH, Permanent Household Survey in English) -for the period 2003-2017-and in the Encuesta Anual de Hogares Urbanos (EAHU, Annual Urban HouseholdSurvey in English) -for the period 2010-2014- are analyzed.A widely known technique in forensic accounting and auditing, such as Benford´s law-also known as the first digit law- is used. If the analyzed data were generated naturally-free of manipulation- it should follow the logarithmic distribution of Benford. The Chisquare test and the absolute mean deviation (MAD) are used for verification.The results suggest that the income reported in the EPH does not follow the Benforddistribution and the degree of compliance with this law decreases significantly between2007-2015 coinciding with the intervention period of the Instituto Nacional deEstadísticas y Censos (INDEC, National Institute of Statistics and Censuses inEnglish).