Skip to content

H:protected-pool-responds — Post-policy SME participation rises but does not recreate the lost competitive discipline

The set-aside protects SME firms from non-SME competition. If protected firms enter in numbers and on costs comparable to the excluded non-SMEs, the protected pool can recreate the price-forming discipline. The data say it does the first (entry rises sharply) but not the second (the realized post-policy price remains above the open benchmark).

Intuition (plain-language)

Defenders of set-asides expect the protected SMEs to step up and fill the gap left by excluded non-SMEs. They do step up — SME participation roughly doubles — but they do not fill the gap: the realized price stays above the open-competition benchmark. More SME bidders is not the same as more price discipline, because the entering SMEs are higher-cost on the margin that sets the price.

Evidence strength: Partial (strongly supported). AN-010 shows the post-policy SME bidder count roughly doubles (0.94 → 1.87 NP; 0.55 → 1.22 PH). AN-030 extends: (i) Pre→Post asymmetric — SME +99% NP but non-SME −44%, net loss of 0.25 bidders per auction; (ii) Calibrated entry costs 4.7× higher for non-SMEs (κ = R\(2.57 vs R\)0.54 NP) — explains why SMEs cannot structurally replace non-SMEs (different cost-vs-win-prob equilibrium); (iii) Dominance ordering survives 2×2 method grid: intensive share ∈ [67.9%, 99.7%] across 8 cells. The protected-pool offset is real, bounded, and well-characterized.

Theory

Entry models of procurement (Levin and Smith 1994; Athey, Levin and Seira 2011) predict that protected-pool participation responds to the reservation rents created by removing rival bidders. Under type-conditional cost dispersion, the protected pool's contribution to the price-forming order statistic depends on (i) how many additional SMEs enter and (ii) how their cost distribution compares to the excluded non-SMEs'. The data on the BEC structural sample inform both margins simultaneously: drop-outs identify the cost distribution; the bidder count identifies the participation response.

Prediction

(i) The number of SME bidders per auction in switched group 65 should rise sharply between the pre-period (open auctions) and the post-period (SME-only). (ii) Despite this rise, the realized post-policy price \(S_3\) should remain above the open benchmark \(S_1\) — the entry response is not enough to close the gap.

Competing prediction

Discouragement. If the policy reduced expected SME rents (e.g., through perceived implementation friction or stricter scrutiny), SMEs might not enter additional auctions. The data rule this out — SME participation rises rather than falls. The relevant comparison is then between the strength of the response and the size of the lost discipline.

Setting evidence

Pregão entry is observed at the firm-auction grain. SME participation rises from 0.94 → 1.87 SME bidders per auction in non-pharma (roughly doubles), and from 0.55 → 1.22 in pharma. Non-SME participation moves in the opposite direction but does not vanish: 2.68 → 1.50 in non-pharma, 2.61 → 1.66 in pharma — a residual non-SME presence persists because many auctions are above the R$80K SME-eligibility ceiling and remain open. The composition shift on the SME side is large and mechanical; the open question is whether the additional SMEs are cost-equivalent to the excluded non-SMEs.

Empirical test

  • Outcome variables: (i) average number of SME bidders per auction pre vs post; (ii) simulated \(E[c_{(2)}]\) under \(S_3\) relative to \(S_1\).
  • Variation: pre vs post the March 2018 cutoff in switched group 65, with class-level (non-pharma standardized, pharma) decomposition.
  • Specification: descriptive bidder-count tabulation + BNE simulation with post-policy SME cost distribution.
  • Sample: BEC Pregão drop-outs, structural cells.

Data requirements and limitations

The interpretation of the protected-pool response combines additional participation with changes in the active SME pool composition (selection into the protected market, sourcing/technology changes, product-mix changes). The \(S_3 - S_2\) term is therefore not a pure entry parameter; it is the net contribution of the post-policy SME pool to the price-forming statistic.

Evidence

Analysis Bearing Key takeaway
AN-010 Supports SME bidder count rises sharply post-cutoff in both non-pharma and pharma cells; \(S_3 - S_1 > 0\) in both. The protected-pool response is real but does not close the price gap.
AN-016 Pending Pharma protected pool is thinner; composition changes more under the policy, making this hypothesis more fragile in pharma.
AN-030 Supports Per-cell entry rates: NP Pregão SME +99%, non-SME −44%, net −0.25 bidders/auction. Entry cost 4.7× higher for non-SMEs. Decomposition grid: intensive share ∈ [67.9%, 99.7%] across 8 cells, all > 50%.

Open tests

Mechanism decomposition of \(S_3 - S_2\)

Split the offset into a pure participation channel and a pure composition-of-active-SMEs channel. The strict-invariance specification (AN-017) sets one of the two to zero by construction; cross-comparing the two would identify which margin drives the offset.

Phased adoption sensitivity

v7-jpube-tight/scripts/70_phased_adoption.R simulates a counterfactual phased SME-only rollout. The protected-pool response should be smaller under phased rollout (less concentrated entry incentive); the simulation would test whether the dominance ordering of H:exclusion-dominates is robust to that implementation choice.