H:entry-thickens-pool — Open auctions attract more firms and more valid bids than SME-only auctions¶
Eligibility expansion mechanically brings non-SME firms back into the bidding pool. If that translates into observed bid activity, open-auction procurements should record more participating firms and more valid bids than SME-only procurements. The reduced-form margin is mechanical (the non-SMEs were excluded before), but the magnitude is informative about the depth of the non-SME response — and about whether the protected SME pool fully replaces them under SME-only (it doesn't: see H:protected-pool-responds).
Intuition (plain-language)
A set-aside literally bars non-SMEs from bidding, so reopening the auction must add firms and bids — that part is mechanical. What is informative is the size of the response: how many non-SMEs actually show up measures how much competitive pressure the set-aside was suppressing, and whether the protected SME pool was ever a full substitute for them. It was not.
Evidence strength: Partial (strongly supported). Three converging channels documented: (i) Headcount — AN-002 DiDiR β on log firms +0.18 (6m) → +0.10 (18m), p<0.01. (ii) Extensive margin — AN-029 shows completion rate rises +10.7 pp at 18m (p<0.01). (iii) Composition / order statistic — same AN: the price effect survives fixing N at 2, 3, or ≥5 firms (β = −0.09 to −0.11***); the price-forming gap shrinks by 2.7 pp full-sample (p<0.01). Pure-headcount explanation rejected — holding N fixed, removing low-cost non-SMEs shifts the price-forming order statistic (a composition/admissibility effect, not strategic "less aggressive" bidding, per v8).
Theory¶
In English clock auctions with private values, more bidders shift the distribution of the second-order statistic downward (lower expected payment) and raise allocative efficiency (Milgrom and Weber 1982; Athey and Haile 2002). Entry into procurement auctions has been studied as endogenous to expected rents (Levin and Smith 1994; Athey, Levin and Seira 2011); SME-only rules act as an entry-eligibility shock by removing the non-SME bidder pool from the choice menu.
Prediction¶
Log number of bidder firms and log number of valid bids in switched group 65 should be higher in the pre-period (open auctions) than in the post-period (SME-only), relative to always-treated control groups. Operationalized as DiDiR: coefficients on \(g65 \times \text{Pre}\) in the firm-count and bid-count equations should be positive and significant.
Competing prediction¶
Concurrent demand expansion. A demand expansion specific to group 65 around March 2018 (e.g., hospital procurement bulk announcements, public-health emergency provisions) would attract more bidders without the price discipline interpretation. The placebo (AN-004) also shows significant pre-treatment shifts in firm and bid counts — these reflect the gradual SME restriction rollout across the control groups, not a group-65-specific demand shock; see the AN-004 discussion of the secular trend.
Setting evidence¶
BEC records every registered firm participation per item, including those that submit no winning bid. The administrative grain therefore includes both the extensive margin (firm registers for the item) and the intensive margin (firm submits a valid bid). Both are recorded in the parquet cache; both are DiDiR-able.
Empirical test¶
- Outcome variables: log number of bidder firms, log number of valid bids (per item).
- Variation: DiDiR — same as H:price-discipline-loss.
- Specification: same DiDiR equation; item-clustered SEs.
- Fixed effects: item; PBU FE in second specification.
- Sample: all items (not restricted to completed) — firm and bid counts are defined for non-completed items too.
Data requirements and limitations¶
The reduced-form firm-count effect attenuates with the window, which is not a defect — it reflects the convergence in the control groups as the SME restriction continued to be rolled out. The 18-month estimate (~10% more firms) is the conservative read; the 6-month estimate (~22%) is the cleanest "treatment effect" comparison but conflates more pre-treatment seasonality.
Evidence¶
| Analysis | Bearing | Key takeaway |
|---|---|---|
| AN-002 | Supports | DiDiR firms β +0.178 to +0.182 (6m), +0.149 to +0.154 (12m), +0.093 to +0.100 (18m); all p<0.01. Bid-count direction same. |
| AN-004 | Mixed | Pre-treatment placebo significant for firms (β ~ −0.10, p<0.01) — flagged in robustness as reflecting the secular SME-restriction rollout across control groups, not a group-65-specific cost shock. |
| AN-010 | Supports | Structural confirms: SME participation roughly doubles after the cutoff while non-SME participation falls — but the net post-policy bidder count is smaller than the pre-policy open count. |
| AN-029 | Supports | (a) Completion rate +10.7 pp at 18m; (b) Price β survives fixing N=2/3/≥5 at −0.09 to −0.11***; (c) price-forming gap shrinks −2.7 pp full sample. Two channels: extensive margin + composition/order-statistic (not behavioral aggressiveness). |
Open tests¶
Decompose the entry response by firm type¶
The reduced-form firm count combines the removal of non-SMEs and the entry response of SMEs. The structural decomposition in AN-010 separates them; the reduced-form DiDiR does not. A composition-by-type cross-cut on the DiDiR specification (planned, not run) would expose the SME-side entry-response margin directly.
Margin of intensive vs extensive¶
Are the bid-count gains driven by more firms registering (extensive) or by each firm submitting more valid bids on the same item (intensive)? Pregão re-bidding allows a single firm to submit many bids. A within-firm bid-count decomposition would isolate the strategic re-bidding margin.