The cost to Brazil · study
The Real Costs of Washing Away Corruption: Evidence from Brazil's Lava Jato Investigation
- Authorship
- Claudio Ferraz (UBC, PUC-Rio, NBER), Luiz C. Moura (UFRJ), Lars Norden (FGV EBAPE), Ricardo Schechtman (Banco Central do Brasil)
- Institution
- National Bureau of Economic Research (NBER)
Working paper of the National Bureau of Economic Research (USA), not peer-reviewed at the time; the authors state that the opinions do not reflect those of the Central Bank, FGV or NBER. - Year
- 2025
- Period analyzed
- 2011–2016 (RAIS 2012–2016; credit 2011Q1–2016Q2)
- Document
- The Real Costs of Washing Away Corruption: Evidence from Brazil's Lava Jato Investigation · original ↗
Method
Diferença em diferenças com dados pareados empresa-banco (RAIS, registro de crédito do Banco Central, dados do MPF e do TSE); 21 construtoras investigadas contra grandes empresas de controle (>250 empregados); controle sintético e amostra pareada como robustez; efeito indireto via exposição de 44 bancos privados.
Results
| Fall in the wage bill of investigated firms | -63 pct_massa_salarial | source |
| Fall in employment at investigated firms | -54 pct_emprego | source |
| Fall in the volume of new loans to investigated firms | -50 pct_credito | source |
| Aggregate reduction in new credit to non-investigated firms, firm level | -18 pct_credito | source |
| Fall in the wage bill of the more exposed non-investigated firms | -12 pct_massa_salarial | source |
| Fall in employment at the more exposed non-investigated firms | -10 pct_emprego | source |
Study results are estimates from the authors' model, not measurements.
Assumptions
- Treatment is the start of the investigations (2014).
- Parallel trends before 2014 (supported by the event study).
Limitations
- It measures effects on large firms; the sample ends in mid-2016.
- The effect on the investigated firms includes the loss of public contracts and the recession in the construction sector.
- It does not convert the effects into GDP, nor separate the costs of corruption from those of how the operation was conducted.
- The indirect effect is relative (more exposed versus less exposed firms), not a total aggregate effect.
Divergences and other studies
Robust microeconomic evidence of a real cost to the investigated firms and through bank contagion, but it cannot be added to a GDP estimate. Literature cited by the paper itself (Colonnelli and Prem, 2022) finds positive effects of anti-corruption audits on activity in municipalities.
Extracted evidence
Ferraz, Moura, Norden, and Schechtman (NBER, July 2025), using a difference-in-differences design with data from RAIS and the Central Bank's credit registry, estimated that the 21 investigated construction companies saw a 63% drop in wage bill and a 54% drop in employment after the investigations began, compared with control firms with more than 250 employees; the volume of new loans to them fell 50%.
Source: The Real Costs of Washing Away Corruption: Evidence from Brazil's Lava Jato Investigation — NBER (Working Paper 33991), July 2025 · Introduction, pp. 1–2; Table 2 and Figure 3 · open original ↗
Study (primary) · full text examined · accessed on Oct. 24, 2026
Limitations: Working paper (not peer-reviewed at the time), sample of large firms through mid-2016. It measures the effect on the investigated firms of the combination 'investigation, loss of access to public contracts and credit, recession'; the text itself notes that the loss of public contracts occurred alongside the collapse of construction and the 2015–2016 recession. It does not measure the effect on GDP.
The same study estimated that private banks with greater prior exposure to the investigated firms reduced new credit to non-investigated firms (an aggregate effect of -18% in new credit at the firm level) and that non-investigated firms more exposed through their banking relationships reduced their wage bill by 12% and employment by 10% after the investigations began.
Source: The Real Costs of Washing Away Corruption: Evidence from Brazil's Lava Jato Investigation — NBER (Working Paper 33991), July 2025 · Introduction, pp. 3–4; Tables 4 to 7 · open original ↗
Study (primary) · full text examined · accessed on Oct. 24, 2026
Limitations: Relative effect between more and less exposed firms, not the total effect on the economy; sample of large firms (>250 employees); public banks are outside the main analysis.
Among the 21 investigated construction firms that make up the NBER study's treatment group, two went bankrupt (GDK and Schahin) and twelve entered judicial reorganization during or after the sample period; in only seven of them were the main executives not arrested or sought by the Federal Police, and all were excluded from business with the government (Szerman, 2023, cited).
Source: The Real Costs of Washing Away Corruption: Evidence from Brazil's Lava Jato Investigation — NBER (Working Paper 33991), July 2025 · Section 2, p. 8 · open original ↗
Study (primary) · full text examined · accessed on Oct. 24, 2026
Limitations: Count from the study, based on Campos et al. (2021), Netto (2016) and Gonzalez-Ocantos et al. (2023); the list of the 21 companies was not checked in this version of the collection.
The NBER study describes Lava Jato as an investigation that resulted in more than 360 convictions, nearly 400 cooperation agreements, and about 30 corporate leniency agreements, and states that defendants paid about US$3 billion in fines, penalties, and voluntary restitution (citing Da Ros and Taylor, 2022).
Source: The Real Costs of Washing Away Corruption: Evidence from Brazil's Lava Jato Investigation — NBER (Working Paper 33991), July 2025 · Section 2, p. 7 · open original ↗
Study (primary) · full text examined · accessed on Oct. 24, 2026
Limitations: Third-party context (Da Ros and Taylor, 2022); the recovery amounts must be checked against another record in the archive with documents from the MPF, CGU, and AGU.