Analyzing Brazil's economic subsidy program: the main instrument of non-reimbursable subsidies for R&D

Using the full contract-level census of Brazil's Economic Subsidy program–the country's main instrument for non-reimbursable R&D grants–we identify the determinants of resource allocation across firms and regions and of project execution. We analyse 2069 contracts (R$ 7.9 billion, 2007–2025) with OLS, Tobit, probit, negative binomial, Cox, quantile regression, TWFE panel, and event-study models, with IV/2SLS and Arellano–Bond GMM as robustness checks. Southeast-based firms receive grants approximately 31% larger than comparable firms elsewhere, a premium that rises to 65% at the 90th percentile; a phase-by-phase analysis shows this premium is a post-2020 phenomenon, absent in 2007–2019, and it is absorbed once state GDP per capita is controlled, indicating it reflects the Southeast's economic scale rather than a pure regional bias. Execution rates fall as grants grow larger (a 2.2-percentage-point drop per doubling of grant value), and larger counterpart (co-financing) commitments predict a higher probability of contract termination. Geographic concentration is high (Gini=0.751) but exhibits no significant trend over 2007–2025. Because the data cover awarded contracts only, allocation estimates are interpreted as conditional associations rather than causal additionality effects.