The disease of corruption: Missing funds and health conditions in Brazilian municipalities

Abstract

The article analyzes how local corruption affects health indicators. Using a random sample of audited Brazilian municipal governments, an objective indicator of corruption was constructed and utilized as an explanatory variable for health indicators involving different dimensions of performance. Using instrumental variables and a series of controls to assure robustness of the results, a causal relationship was found between corruption and the performance of public services, confirming a chain reaction. The study confirms that additional resources are necessary, but not sufficient, to guarantee improvements in the performance of local government services in Brazil.

View correction statement:

Correction

Notes

1 Further information on the Family Health Program (in Portuguese): https://www:saude.gov.br/acoes-e-programas/saude-da-familia/sobre-o-programa

2 Program reliability is guaranteed by the security of the federal lottery draw carried out by the federally-owned savings bank, Caixa Econômica Federal. It has existed since the 1960s, and has been audited by independent bodies. Despite the existence of lawsuits which question the program, with regards to its frequency and possible problems in findings of irregularities, there has been no decision to support these issues, and together with media coverage, demonstrates the dependability of the program. Since 2015, CGU has incorporated the vulnerability matrix as criteria in the definition of the municipalities to be audited. During the period analyzed in our study, the draw was the exclusive definition criterion.

3 The irregularities found by auditors can be questioned by the municipal authority. However, only those identified by auditors were confirmed for the purposes of the calculation.

4 One of the most commonly-used variables to measure corruption is the corruption perception index (CPI). Among other problems found is the bias arising from tolerance of corruption. Therefore, the indicator is not adequate to determine the level of corruption in a locality (Svensson 2005).

5 In this case, funds were released for the purchase of ambulances, often more than necessary, through fraudulently overpriced processes. Further information (in Portuguese): https://www1.folha.uol.com.br/folha/brasil/ult96u400354.shtml.

6 The differences between the education and health sectors may explain the non-significance found in the relationship between the instrument and corruption in the health sector. Ferraz, Finan, and Moreira (2012) argue that there is a significant relationship between the percentage of resources that originate from FUNDEB, and corruption in the education sector. In order to prove the difference between the factors associated with corruption in the two sectors, I analyzed the relationship which exists between the percentage of resources from the Unified Health System (SUS) and corruption in health, controlled by a series of municipal characteristics, and a statistically significant relationship was not verified. Further potential explanations can be found in the other factors associated with corruption mentioned by Rose-Ackerman (1997). However, a detailed analysis of the characteristics of the education sector is beyond the scope of this study.

7 For a detailed account of IV and 2SLS estimation, see Angrist and Pischke (2009).

8 Other studies use corruption data in Brazilian municipalities and the provision of health sector indicators. However, these studies aim to understand the relationship between the auditing program and municipal performance (Lichand, Lopes, and Medeiros 2016; Funk and Owen 2020). In addition to different research problems, the empirical strategy is divergent, using the differences-in-differences strategy. The studies do not consider aggregate measures of municipal corruption, using service data as the unit of analysis.

9 Considering the high percentage of deaths from external causes in Brazil, such as violence, traffic accidents and suicide, this was also tested, including a control of the percentage of deaths from external causes. There was no change for the sign and significance of the relationship.

Additional information

Funding

This work was supported by CAPES.