Statistical detection of fraud in the reporting of Croatian public companies

Statistical methods based on Benford’s distribution, Z- and I‡2-statistics are being successfully applied to detect likely accounting and reporting fraud, for example in the daily usage of the Internal Revenue Service in the USA, and in historical analysis of Greek macroeconomic reporting. We adapt and apply the methodology to the analysis of the reporting of some leading Croatian public companies. We find indications of reporting fraud in several of the companies analyzed. In particular we find correlation between the likelihood of reporting fraud, measured as a deviation from Benford’s law, and reported net income losses, for companies large enough (with a revenue of at least 1 billion kuna). Finally, we suggest application of the methodology to improve the internal processes, efficiency and effectiveness of the State Auditing Office.Data availability: The data used in the study are corporate data in the public domain. For legal reasons, however, the identities of the companies are disguised. Contact the first author for the sanitized data sets that can be used to verify and replicate the analysis.