Análise de risco em busca de não conformidades de empresas do Simples Nacional: uma análise empírica para os anos de 2021 a 2023 no Ceará, utilizando o método da fronteira estocástica

Abstract

This dissertation analyzes the risk of tax noncompliance among firms under the Simples Nacional regime in the state of Ceará, covering the period from 2021 to 2023, using the Stochastic Frontier Analysis (SFA) model. The main objective was to measure taxpayers’ efficiency in fulfilling their tax obligations and to propose a Tax Compliance Risk Index (IRCT), capable of ranking firms according to the likelihood of inconsistencies in their declarations. The database was constructed using information from the Cadastro Geral da Fazenda (CGF), the Simples Nacional Revenue Declaration System (PGDAS-D), the Payment Means Information Return (DIMP), and electronic tax documents. The descriptive analysis revealed strong heterogeneity among taxpayers, with significant variations in reported sales, purchases, and expenses. The SFA results indicated that firm size is a determining factor for compliance: small businesses (EPP) exhibited greater declarative efficiency, while microenterprises (ME) showed a higher propensity for deviations from the estimated frontier. From a sectoral perspective, industrial firms displayed lower levels of compliance compared to retail businesses, which demonstrated greater consistency between inputs and reported output. The geographic dimension also proved significant, with notable differences between the Metropolitan Region of Fortaleza and inland regions, suggesting that local factors influence tax behavior. The findings reinforce the usefulness of the IRCT as a practical tool for tax administration, enabling audits and monitoring to be targeted at higher-risk groups, thereby contributing to the reduction of the tax gap. The study also advances the national literature by applying the stochastic frontier methodology to the Simples Nacional in a recent state-level context, offering empirical evidence to support the design of more segmented and effective tax enforcement policies. Finally, the study acknowledges limitations such as the absence of institutional and behavioral variables, as well as the reliance on administrative records, suggesting the need for future research that incorporates qualitative dimensions, inter-state comparisons, and complementary machine learning methodologies.