Replication package: Political Instability and Bureaucratic Resilience — The Protective Role of Performance-Based Transfers in Brazilian Social Assistance

Analysis code and processed municipality-year datasets needed to reproduce every table and figure in the article Political Instability and Bureaucratic Resilience: The Protective Role of Performance-Based Transfers in Brazilian Social Assistance, and in its supplementary appendices. The study uses the removal of Brazilian mayors from office by the electoral courts, which triggers an off-cycle special election, as a quasi-experimental shock, and a staggered difference-in-differences design across all 5,570 municipalities (2012–2019), to test whether tying a federal transfer to the measured performance of an administrative routine (the IGD-M) shields that routine against political instability. Contents. The pipeline is written in R and runs from the repository root. Collectors download data from the official sources — VIS DATA 3/MDS for the registry update rate, the SUAS Census for the reference centres, and Base dos Dados for Cadastro Único, Bolsa Família and RAIS; treatment cohorts and the municipality crosswalk come from the Superior Electoral Court (TSE), and structural controls from the 2022 IBGE Census. Scripts are numbered 00–23 for the main pipeline and 26–29 for additional checks, each carrying a header explaining what it does and what it produces. The bootstrap seed is fixed in 00_setup.R. What is not included. Raw microdata are not redistributed, because of their volume and because some sources were originally credential-gated. Every step that builds the processed panel from them is scripted and documented, so the chain from official source to reported estimate can be inspected end to end. All data are public and aggregated at the municipality-year level, with no personal identifiers.