Skip to content

H:price-discipline-loss — Open auctions yield lower procurement prices than SME-only auctions

Removing non-SMEs from a competitive bidding pool weakens the price-forming order statistic. If the excluded bidders were on the price-forming margin, opening the auction back up to them should lower the negotiated price. The March 2018 PGE-SP reinterpretation that re-imposed SME-only status on group 65 (medical/hospital supplies) flipped the regime, providing a DiDiR testbed: comparing group 65 (switched) to other groups (always treated) yields the open-vs-SME-only price effect.

Intuition (plain-language)

Procurement prices are set by the most competitive bidders, and non-SMEs are often those price-setters. Kick them out and prices should rise; let them back in and prices should fall. The March 2018 PGE-SP reinterpretation that re-imposed SME-only status on medical supplies flips the regime cleanly, letting us read off the open-vs-set-aside price gap directly: about 10–11% lower under open competition.

Evidence strength: Partial (strongly supported). AN-001 reports the v8 reduced-form benchmark β = −0.113 (18m, ~10–11%, p<0.01; the earlier v1–v4 pipeline gave −0.131 to −0.133); reading survives placebos (AN-004), Lee bounds (AN-005), HonestDiD (AN-006), Sun-Abraham (0.108 ≈ DDR 0.113, AN-018), Goodman-Bacon (100% weight on clean 2×2, 🟢 AN-020), and synthetic control (AN-021). Concentrated at low quantiles (AN-007). Structural confirmation in AN-010. Falsification against CMED pharmaceutical-inflation alternative: price effect persists in both medications (β = −0.174) and non-medication medical supplies (β = −0.097) within Group 65 (AN-027). Phased-adoption robust (AN-023).

Theory

Set-aside policies that remove a subset of bidders from the auction must either rely on the protected pool's response or accept higher prices. In independent private values English auctions, the expected payment is the second-order statistic of the bidder cost distribution (Vickrey 1961; Milgrom and Weber 1982; Haile and Tamer 2003). Removing the lowest-cost subset of bidders raises the expected second-lowest cost; the question is how much the protected pool's entry response offsets the removal. Empirical work on SME procurement preferences (Marion 2007; Reis and Cabral 2015; Nakabayashi 2013) finds positive price effects of exclusionary preferences whose magnitude depends on the strength of the disciplinary margin removed.

Prediction

Negotiated log prices in switched group 65 should be lower in the pre-period (when the group was open to all bidders) than in the post-period (when the SME-only rule applied), relative to the always-treated control groups. Operationalized as DiDiR: the coefficient on \(g65 \times \text{Pre}\) in the price equation should be negative and statistically significant.

Competing prediction

Cost-shock alternative. A common cost shock that hit group 65 differentially around March 2018 (e.g., the BNDES medical-supply credit line, COVID-19 anticipations, currency moves in dollarized imports) would produce a level shift in group-65 prices that the DiDiR would mechanically attribute to the policy. The placebo test (AN-004) and HonestDiD sensitivity (AN-006) are the load-bearing pieces ruling this out: cost shocks should produce a positive placebo coefficient or a binding M̄ in HonestDiD; both come back null.

Setting evidence

Before 2018, Group 65 stayed open because São Paulo applied the R$80,000 SME-only threshold to the total value of the purchase notice, not item-by-item; medical notices bundle many small items inside large orders, keeping them outside the SME-only default. PGE-SP Parecer 151/2017 (December 2017) held that the threshold applies item-by-item, and Group-65 buyers reached mass take-up around the March 2018 operational cutoff. From that point Group 65 joined the other groups under SME-only tendering by default — a clean switching event with an open pre-period (controls already SME-only) and an SME-only post-period. See docs/paper.md for the institutional account.

Empirical test

  • Outcome variable: log negotiated price (per item).
  • Variation: DiDiR — group 65 is the switched group; other groups are always treated. The coefficient on \(g65 \times \text{Pre}\) is the open-vs-SME-only effect for group 65.
  • Specification: \(y_{pigt} = \eta_i + \gamma \text{Pre}_t + \beta\,(g65_{pgt} \times \text{Pre}_t) + x'\delta + \varepsilon_{pigt}\) with item FE \(\eta_i\), controls \(x\) (log quantity, tender type), item-clustered SEs.
  • Fixed effects: item; PBU FE in second specification.
  • Sample: completed items (oc_item_status == 1), 6/12/18-month windows centered on March 2018.

Data requirements and limitations

Requires the BEC parquet cache (data/processed/paper2_me_epp.parquet) covering Jan 2016–Dec 2019. The DiDiR identifies the open-vs-SME-only effect in group 65 — not a treatment effect on the universally-treated groups. The estimate is a reduced-form policy comparison; it does not decompose the price difference into the lost-discipline channel and the protected-pool offset (see H:exclusion-dominates for that decomposition).

Evidence

Analysis Bearing Key takeaway
AN-001 Supports v8 reduced-form benchmark β −0.113 (18m, ~10–11%), p<0.01, item-clustered; stable across 6/12/18-month windows. Earlier v1–v4 pipeline gave −0.131 to −0.133.
AN-004 Supports (via rule-out) Pre-reform price placebos small (β = −0.013/−0.030/−0.034 at Sep/Mar/Jun 2017) vs real-cutoff −0.108 to −0.142 — rules out the cost-shock confound.
AN-005 Supports Lee bounds tight; completion-selection has negligible impact.
AN-006 Supports Price effect survives substantial M̄ violations in HonestDiD.
AN-007 Mixed Effect concentrated at τ ≤ 0.50 (strongly negative); turns positive at τ = 0.90. Mean DiDiR masks heterogeneous distributional impact.
AN-010 Supports Structural counterfactual confirms direction; \(S_1 - S_3 > 0\) in standardized non-pharma.
AN-018 Supports Sun-Abraham ATT 0.108 matches DDR 0.113 within 0.005; CS2021 same direction.
AN-020 Supports (🟢) Weight = 1.000 on clean Treated-vs-Untreated 2×2; no contamination from heterogeneous-timing.
AN-021 Mixed Synth pre-gap exact 0.0000; post ATT 0.171 directionally consistent; placebo p=0.103.
AN-027 Supports Within-Group-65 split: medications −0.174, non-medications −0.097. CMED inflation alternative rejected.
AN-023 Supports DiD survives dropping BEC enablement window (β −0.109 → −0.087, both p<0.01).

Open tests

Cross-modality validation

Re-estimate the price effect on Convite first-price bids using a GPV (Guerre-Perrigne-Vuong 2000) recovery, and compare with the Pregão drop-out recovery. Convergence would upgrade this hypothesis to 🟢-eligible. Lives in v7-jpube-tight/scripts/38_cross_modality.R; documented as part of AN-014 robustness battery.

Adherence-rate sensitivity for fiscal translation

The structural welfare loss translates to ~R$38–89M/year on group 65 alone across the 30–70% adherence range (see docs/extensions.md). A first-stage-by-PBU adherence calibration (scripts/35_pref_sdid_class.R) would tighten the fiscal interpretation but is not load-bearing for the welfare arithmetic, which uses simulated rather than realized prices.