Working Papers¶
Sourcing under Sanctions: Judicial Urgency and Pharmaceutical Procurement Costs¶
with Paulo Furquim de Azevedo (updated May, 2026 — v10, JPubE short paper) | PDF | Online Appendix
Abstract
Court mandates secure access to medicines, but they can also change how governments buy. We study São Paulo pharmaceutical procurement on BEC, covering 479,330 purchase-offer-item observations from 2009--2019. Higher prices under legal urgency can reflect two very different margins: incumbent suppliers charging more to sanctioned buyers, or fragmented sourcing that changes scale and supplier matching. Because court orders originate outside procurement offices, we interpret ordinary-versus-urgent estimates as the procurement effect of externally imposed legal urgency, conditional on item, time, and purchasing-unit fixed effects. A selected administrative urgent channel provides the closest feasible comparison without court sanctions; Lee bounds place the litigated-over-administrative price gap between 15.9% and 21.1%. The cost margin is not mainly a broad same-firm markup in deep repeated urgent markets: within the same firm, buyer, and item, prices are statistically indistinguishable across urgent regimes. Instead, judicial urgency operates through sourcing. Administrative orders are 3.3 times larger, and modal winners differ in 70.2% of item-buyer pairs. A residual within-firm price gap persists in thinner or earlier markets. The policy margin is not weaker access, but procurement capacity that preserves aggregation and supplier matching under legal urgency.
Media coverage
The Price of Exclusion: SME Set-Asides in Public Procurement¶
(updated June, 2026 — JPubE submission) | PDF | Online appendix | Highlights | Cover letter
Abstract
Set-asides expand small-firm access to public procurement by removing rival bidders from the auction. When the excluded bidders are the ones forming the price, the policy replaces competition with eligibility at a high static cost. I study São Paulo's centralized electronic procurement platform, where a legal reinterpretation expanded SME-only tendering into medical and hospital supplies. The platform's reverse auctions record losing bidders' drop-out prices, which reveal willingness to supply; I use them to split the set-aside price effect into the competitive discipline lost when non-SMEs are excluded and the offset from the post-policy SME pool. In standardized non-pharmaceutical markets the protected pool responds but does not replace the lost discipline: the full set-aside generates a static welfare loss of 28.9 percent of the open-regime price at \(\lambda = 0.30\). Pharmaceutical procurement shows larger but more model-sensitive losses. A 10 percent SME price preference keeps non-SMEs in the auction at near-zero price cost while delivering less redistribution. The relevant frontier is not SME support versus none; it runs between exclusionary redistribution and support that preserves the price-forming pool.
Media coverage
Cheap Signals, Costly Proof: The Reach and Limits of Award-Layer Screening in Cartel Enforcement¶
with Paulo Furquim de Azevedo (updated June 2026 — JLEO R&R version, two-platform extension: BEC + ComprasNet; three-document package: paper ~46 pp + online appendix ~28 pp + online supplement ~23 pp) | PDF | Online appendix
Abstract
Cartel enforcement must allocate costly proof-producing effort before legal proof exists. This paper asks how agencies should govern cheap administrative screens before opening bid-level records. Using São Paulo's BEC procurement platform and CADE adjudications, we audit a minimal award-layer construct---persistent zero-win participation---against a reproducible adjudication-anchored cobidder label that never uses the screen itself. The audit identifies an over-crediting problem: when both the screen and a contact-defined label load on participation volume, raw screening metrics overstate evidentiary value. In BEC, much of the raw reach is exposure rather than conduct, and strict prospective ranking fails outside incumbent pools. We then organize follow-up as a cost--recall frontier: one operating point opens only 12% of firms but roughly 67% of bid rows, showing why firm-count savings overstate forensic-cost savings. A second-platform stress test shows that the audit logic travels; the score does not. Liability remains in the richer record.