Scientific portfolio organised by area. Each page describes the question, the data, the method, the current stage and the limitations of the work.
Was the first official institutional adoption of AI followed by measurable aggregate productivity gains in the state courts?
Methodology: Court and year fixed effects; Staggered difference-in-differences (estimators robust to treatment heterogeneity)…
Latest public commit:
How much does public spending on education return in output, wages and tax revenue over time?
Methodology: Panel-data econometrics; Fiscal DSGE model with human capital and simulation of spending shocks
Latest public commit:
Did the installation of computer laboratories by ProInfo Integrado raise the Ideb index and Saeb scores of the public schools it reached?
Methodology: Staggered-adoption difference-in-differences (Callaway and Sant'Anna) with not-yet-treated schools as controls; Sensitivity analysis (HonestDiD), placebos and alternative treatment-timing definitions
Did participation in Prouca change the Ideb index, Saeb mathematics and Portuguese scores and the approval, retention and dropout rates of the schools involved?
Methodology: Difference-in-differences with school and year fixed effects, entropy balancing and matching; Pre-programme placebos, event study, multiple-testing adjustment and dose-response analysis
How do farm-gate prices transmit across Brazilian regions and which markets lead the adjustment process?
Methodology: Unit-root and cointegration tests; Vector error-correction model (VECM) and threshold version (TVECM) for asymmetries…
Latest public commit:
Which factors explain the value of moral damages awards and how much do decisions vary across similar cases?
Methodology: Text mining and variable extraction with manual validation; Regression models for award values and for dispersion
Latest public commit:
Can patterns of abusive litigation be identified in an explainable, auditable way that is compatible with human oversight and due process?
Methodology: Interpretable procedural-pattern indicators; Explainable machine learning with human-review safeguards
Latest public commit:
Do binding precedents actually diffuse and get applied by the courts of origin? How fast, and with which territorial differences?
Methodology: Natural language processing to detect application and citation; Diffusion models and time-to-adoption by court
Latest public commit:
When does the STJ agree to review the quantum of moral damages, and how does Súmula 7 act as a barrier to that review?
Methodology: Census and sampling of decisions; structured classification of grounds; Models for the probability of admitting the appeal and of revising the amount
Latest public commit:
Are drafts produced by retrieval-augmented language models legally reliable?
Methodology: Benchmark of language models with and without retrieval-augmented generation; Assessment across seven dimensions - legal correctness, reasoning, normative currency, citations, hallucinations, consistency and stability
Does case duration depend on where the case is processed, and which territorial and institutional characteristics explain that inequality?
Methodology: Reconstruction of procedural trajectories; Survival analysis and spatial models
Latest public commit:
Are the state courts adopting AI at the same pace at which they develop security, governance, data-protection and incident-response controls?
Methodology: Comparative document analysis across six dimensions: data governance; information-security policy; data-protection officer; impact assessment; incident response; AI regulation
Latest public commit:
Do global forecasting models, coherently reconciled across the municipality–state–region–Brazil hierarchy, forecast the geography of Pix better than local models — and in which territories is the forecast least reliable?
Methodology: Temporally stable geography and municipality-month panel with snapshot hashes and metadata; Global forecasting models evaluated by rolling origin against baselines…
Latest public commit:
Are Pix security interventions associated with changes in the relative incidence and financial severity of fraud, in fund recovery and in system use?
Methodology: Measurement audit of the public series; Interrupted time series and ARIMAX models, conditional on series length and comparability
What is the role of distance, air connectivity and price competitiveness in international tourism demand for Brazil?
Methodology: Gravity models estimated by PPML with fixed effects; Robustness checks with leave-one-out exclusion of origins
Latest public commit:
When a port suffers a physical shutdown, can the trade network absorb the shock or does regional economic activity fall?
Methodology: Event studies and difference-in-differences; Trade-network analysis and PPML models
Did the Porto Sem Papel programme reduce port times and trade costs? What informational transformation did digitalisation produce?
Methodology: Staggered difference-in-differences and event studies; Stochastic frontier of port efficiency…
Latest public commit:
How do marine heatwaves and changes in ocean productivity affect fishing effort along the Brazilian coast?
Methodology: Geospatial data cube (H3, Parquet/DuckDB); Cell-day panels with fixed effects and placebo tests
Latest public commit:
Do seasonal bans and marine protected areas reduce fishing pressure or displace effort to neighbouring areas?
Methodology: Spatial difference-in-differences and placebo tests; Vessel-level decomposition to separate coverage changes from real activity changes
Latest public commit:
No project matches the selected filters.
The projects listed are ongoing. Descriptions, methods and results may be updated after further validation.