Skip to content

Changelog

v8 — May 2026 (JPubE submission)

  • Title is The Price of Exclusion: SME Set-Asides in Public Procurement. The framing moves from "sheltered bidding" (a within-auction conduct claim) to price formation under exclusion (the order statistic that prices the contract changes when non-SMEs are removed).
  • Decomposition rebuilt as \(S_3 - S_1 = (S_2 - S_1) + (S_3 - S_2)\): a lost competitive discipline component (exclusion of non-SMEs, SME pool held fixed) and a protected-pool offset (post-policy SME pool replacing pre-policy). The two components are reported as counterfactual price-formation objects, not reduced-form treatment effects.
  • Headline numbers at λ=0.30: non-pharma static welfare loss 28.9% of the open-regime price; exclusion accounts for 72% of the price decomposition (absolute share). Implied SME welfare weight to prefer full exclusion to a 10% preference: 2.42 (non-pharma) and 2.61 (pharma). Pharmaceuticals reported as a boundary case rather than a second headline.
  • Empirical bidder-count robustness (Online Appendix OA-D.2): replacing the Poisson bidder-count draws with sampling from the empirical class-period-type count distributions attenuates net effects by roughly a quarter but leaves the exclusion-dominant decomposition intact (exclusion shares 69.4% non-pharma, 63.1% pharma).
  • Threat-assessment table (Section 6, Table 5): eight identification and sensitivity threats mapped to diagnostic, finding, and remaining limitation. Closes the most predictable referee-level rejection channels.
  • JPubE compliance: Elsevier Harvard bibliography style, abstract under 250 words, 4 keywords + 5 JEL codes, highlights file, generative-AI declaration, acknowledgements section, data-availability statement.
  • 45 cited references covering procurement-policy classics (Marion 2007; Nakabayashi 2013; Krasnokutskaya–Seim 2011; Athey–Coey–Levin 2013), modern procurement empirical (Coviello–Mariniello 2014; Bosio et al. 2022; Best–Hjort–Szakonyi 2023; Decarolis et al. 2025; Coviello et al. 2026), Brazilian setting (Ferraz–Finan–Szerman 2016; Szerman 2023), auction theory canon (Vickrey 1961; Milgrom–Weber 1982; Haile–Tamer 2003; Athey–Haile 2002, 2007), recovery (GPV 2000; Bajari–Hortaçsu 2005; Krasnokutskaya 2011), welfare framework (Saez–Stantcheva 2016; Ballard et al. 1985; Hendren 2020; Finkelstein–Hendren 2020), and modern DiD literature.

v6 — May 2026 (Sheltered Bidding framing — superseded)

  • Title rebranded to Sheltered Bidding: The Within-Auction Cost of SME Set-Asides (commit 5d83c28).
  • Abstract pivoted to question-first structure with three headline numbers: ~10% DiD effect on winning prices; sheltered bidding = 2/3 to 3/4 of the simulated effect; R\(55M/yr welfare cost on Group 65 alone (out of São Paulo's R\)13B platform).
  • Sheltered bidding elevated from descriptive label to characterized property; new policy-design framing.
  • Aggressive compression (body 64→54pp; appendix 24→23pp) to fit JPubE conventions.
  • Macros (v6-jpube/output/values.tex) become the single source of truth for every headline number in the manuscript.
  • Final pre-submission cleanup (commit d21754d) + footnote translation of the PGE-SP Parecer 151/2017 ementa into English.

v5 — April 2026 (Submission-ready, structural decomposition)

Structural decomposition replaces the v1--v3 reduced-form headline: asymmetric IPV reading of Pregão, type-specific cost distributions point-identified from drop-out bids (Haile and Tamer 2003), Krasnokutskaya (2011) auction-level UH correction, Athey--Seira (2011) treatment of equilibrium SME entry, BNE Monte Carlo, cluster bootstrap, Saez--Stantcheva (2016) welfare weights. Headline magnitudes: within-auction share ~74% in non-pharma, ~66% in pharma. Welfare cost 28.9%/44.8% of pS1 at λ = 0.30.

March 2026 (v4, reduced-form layer)

  • Fiscal Cost Quantification (v4 reduced-form): subsection added with R\(84.5--85.8 million headline. **Superseded in v5/v6** by the structural welfare arithmetic (R\)55--128M/yr, λ-grid, MCPF, Saez--Stantcheva weights).
  • 5 advanced econometric methods: parallel trends sensitivity (HonestDiD), Lee bounds, causal forest, quantile DiD, Gelbach decomposition
  • 4 new tables + 5 new figures (18 tables / 15 figures total)
  • Expanded appendix with organized subsections and explanatory paragraphs
  • Polished main body text (~10% expansion with references to advanced methods)
  • New MkDocs page for advanced methods

March 2025

  • Full R/fixest pipeline with 14 tables and 10 figures
  • Robustness checks: placebo tests, alternative clustering, winsorization, randomization inference
  • Extensions: real prices, extensive margin, efficiency, winner composition, heterogeneity by item value and PBU type
  • MkDocs documentation site

January 2021 (First Version)

  • Initial manuscript with core DiDiR results
  • 4 main outcome variables: prices, firms, bids, distance
  • 3 time windows with baseline and PBU-controlled specifications