A estatística como ferramenta de apoio à auditoria: uma discussão a respeito da capacidade de identificação de exceções em dados de empresas de energia elétrica

Abstract

Objective: Assess the ability to use statistical techniques and methods to identify exceptions in support of audit work. Method: Type of research carried out, as to the objectives, descriptive, and, as to the procedures, documentary. Data from the companies CEMIG, COELBA and ELETROPAULO, published by ANEEL, from January 2017 to July 2020 were used. The techniques used were exploratory statistics, linear regression, ARIMA models for time series, Benford's law and Principal Component Analysis. Results: The exploratory analysis of the data, by calculating the mean, median, standard deviation and mode, proved to be effective in identifying exceptions. The application of the correlation in the data of Consumption, Revenue, Number of UC's and Tariffs made it possible to identify patterns that need an analysis of the microdata segregated by customers. The greatest benefit of Benford's Law was to give evidence of elements that are not necessarily extreme values, which can be very useful in directing samples to be performed. For Cook's distance, the most influential elements were more concentrated in the year 2020, due to the impact of the pandemic. In relation to Principal Component Analysis, the greatest benefit was to direct the analysis towards the variables that have the greatest capacity to explain the variability of the database. Conclusion: the statistical techniques used were able to identify exceptions and assist in directing further analysis.