H:no-collusion-confound — The price drop is not produced by a post-policy shift in bidder coordination

The structural decomposition reads drop-out prices as willingness-to-supply observations under the maintained IPV-clock interpretation (H:ipv-clock-admissible). If bidder coordination loosened or tightened around the March 2018 policy break — for instance because the smaller eligible SME pool repeatedly meets the same competitors — the drop-out distribution no longer reads as cost; it reads as strategic bid behavior. The decomposition would then become a price-forming re-weighting exercise rather than a structural counterfactual. Classical bid-rigging screens (Bajari-Ye, Schurter, pair classification) test whether such a coordination shift is detectable in the data.

Intuition (plain-language)

If the smaller post-policy SME pool started colluding — the same few firms meeting repeatedly and softening their bids — the price rise could be a coordination story rather than a lost-competition story, and the drop-out-as-cost reading would break. Classical bid-rigging screens check for exactly that. They find baseline clustering but no post-policy intensification, so the price increase is not manufactured by a new collusion shock.

Evidence strength: Partial (strongly supported). Three classical screens land clean for the differential question. AN-015: Conley-Decarolis close-pair share flat in NP (16.9% → 16.8%) and falls in PH (27.6% → 24.4%); Bajari-Ye T1 ratios fall in both (NP 2.63 → 1.83; PH 1.29 → 1.11). AN-026 class-conditional pair test: persistent-pair signature \(T_3\) sits at or below the within-class null in all 4 cells (p ∈ [0.730, 1.000]) — the "persistent meeting" pattern is segmentation, not coordination, and it is stable across the cutoff. The post-policy price increase is not produced by a coordination shock; yellow because the screens detect first-moment clustering, not all conduct deviations.

Theory

Classical bid-rigging detection (Porter and Zona 1993, 1999; Bajari and Ye 2003; Schurter 2017; Conley and Decarolis 2016) tests for coordination signatures in observed bidding patterns: low bid dispersion within auctions, persistent firm-pair co-bidding clusters, deviations from independent rationality in residual bids. The screens were designed for first-price auctions but transfer to English-clock drop-out data under the IPV reading. A change in screen flags across the policy break would indicate the maintained IPV restriction is unstable.

Prediction

The screens should show no significant shift in bidder-coordination flags between the pre-policy and post-policy bidder pools. Specifically: - Bajari-Ye: residual bid correlation across firm pairs should not shift discontinuously around March 2018. - Schurter: the test-statistic distribution should be invariant pre-vs-post. - Pair classification: persistent co-bidding pairs should not exhibit systematically different coordination signatures pre-vs-post.

Competing prediction

Coordination shift drives the result. If the smaller post-policy SME pool exhibits more cover-bidding (perhaps because the eligible firms meet each other more often), drop-outs are strategically suppressed in the post-period. The decomposition's \(S_3\) then understates the true willingness-to-supply distribution, and the "protected-pool offset" reading collapses into a coordination artifact. This is the load-bearing alternative.

Setting evidence

Post-policy SME pool composition can be tracked at the CNPJ-pair level using the BEC participation logs. The same firms repeatedly meet across items; the question is whether their relative drop-out patterns shift across the break.

Empirical test

• Outcome variables: per-screen test statistic distributions (Bajari-Ye, Schurter, pair-classification), per period.
• Variation: pre-policy vs post-policy bidder pools in switched group 65.
• Specification: each screen applied separately; results pooled across screens.
• Sample: BEC Pregão bid logs spanning March 2018.

Data requirements and limitations

The screens require the full bid-event log (already in v7-jpube-tight/); the pair-classification screen also requires firm-pair historical co-bidding. The screens were designed for specific institutional contexts (Italian roads, US highway, etc.); the São Paulo BEC adaptation requires careful interpretation of test- statistic baselines.

Evidence

Analysis	Bearing	Key takeaway
AN-015	Supports	Conley realized share flat NP (16.9→16.8) / falls PH (27.6→24.4); Bajari-Ye T1 falls in both (NP 2.63→1.83; PH 1.29→1.11). Differential-coordination shock rejected.
AN-026	Supports	Class-conditional pair screen: \(T_3\) persistent-pair signature at/below within-class null in all 4 cells (p ∈ [0.730, 1.000]). Persistent meeting is segmentation, not coordination.

Open tests

Within-firm bid-pattern stability

If individual firms' drop-out behavior is strategic-coordination driven, within-firm bid-cost distributions should shift discontinuously across the policy break. A within-firm DiD on bid moments would be a clean falsification.

Buyer-level heterogeneity

Some PBUs are more sophisticated than others. A buyer-level cross-cut (v7-jpube-tight/scripts/62_buyer_heterogeneity.R) on the screens would identify whether coordination shifts (if any) are concentrated in specific buyer types.