世界银行-MiDES：巴西当地采购和预算执行的新数据和事实（英）-2023.11-58页_5mb

Summary of MiDES Dataset Paper

Introduction and Methodology

This paper presents the MiDES (Microdados de Despesas de Entes Subnacionais) dataset, a harmonized and disaggregated collection of data on local procurement and budget execution in Brazil. It covers approximately half of Brazilian municipalities, spanning from 2003 to 2021, and provides granular details such as supplier identities, tender specifics, and payment timelines, unavailable in previous aggregate datasets. The data is sourced from State Audit Courts and validated against SICONFI, a primary public finance dataset in Brazil. Researchers can access and analyze the data using platforms like Data Basis and Google BigQuery.

Key Findings

• Local Procurement: About 25% of municipal purchases are from suppliers in the same municipality, with variations based on municipality size and no significant difference between competitive and non-competitive tenders.
• Payment Timeliness: Approximately 15% of payments to suppliers are delayed beyond the 30-day legal limit, and payment speed correlates positively with local per capita GDP. Average delays are around 21 days, but delays can reach up to 60 days in some cases.
• Data Validation: MiDES aggregates align closely with SICONFI at the municipality-year level, with discrepancies partially explained by factors like carryover payments and reporting differences.

Applications

• Local Supplier Contracting: The dataset reveals geographical variations in supplier locations, with wealthier municipalities showing higher local procurement shares, potentially linked to supplier availability and competition.
• Payment Delays: Analysis shows systematic delays correlated with income levels, impacting supplier liquidity and firm performance, with implications for policy reforms to expedite payments.

Conclusion

MiDES offers unprecedented insights into Brazilian local public finance, enabling rigorous analysis of efficiency, competition, and policy impacts. It opens avenues for further research on corruption prediction, business cycles, and budget execution quality, supported by high-quality, publicly available data.