Skip to content

Exclusion dominates the price decomposition

Intuition (plain-language)

Split the set-aside's price effect into two opposing forces. One: removing non-SMEs strips out the toughest price-setters (the lost-discipline channel). Two: with the door shut, more SMEs crowd in and partly compensate (the protected-pool offset). The first force is the lion's share — about 72% of the combined magnitude. Crowding more SMEs in helps, but it cannot manufacture the competitive pressure the excluded firms used to supply.

🟡 The structural decomposition of the SME set-aside price effect into the lost-discipline channel (non-SMEs removed, SME pool held fixed) and the protected-pool offset (post-policy SME pool replacing the fixed one) finds that the first channel accounts for ~72% of the sum of absolute component magnitudes in standardized non-pharmaceutical procurement (AN-010). The exclusion-to-net ratio likewise exceeds 1 — the protected-pool offset works in the right direction but cannot recreate the lost competitive discipline.

The decomposition rests on three counterfactual price-formation objects recovered from BEC Pregão drop-outs under the maintained IPV-clock interpretation: \(S_1\) (open pre-policy auction), \(S_2\) (SME-only with the pre-policy SME pool held fixed), \(S_3\) (SME-only with the observed post-policy SME pool). The jump \(S_1 \to S_2\) is the lost-discipline magnitude; the decline \(S_2 \to S_3\) is the protected-pool offset; the remaining gap \(S_1 \to S_3\) is the realized price cost of the set-aside.

See Figure 4 (Counterfactual price decomposition) in paper.pdf §4.

The simulated second-order statistic \(E[c_{(2)}]\) normalized by the buyer reference price, for \(S_1\), \(S_2\), \(S_3\). \(S_2 - S_1\) is the lost-discipline magnitude; \(S_3 - S_2\) is the protected-pool offset working in the opposite direction. The absolute exclusion share \(|S_2-S_1| / (|S_2-S_1| + |S_3-S_2|)\) is ~72% in non-pharma.

Caveat. The decomposition inherits the maintained IPV-clock interpretation of Pregão drop-outs (see H:ipv-clock-admissible). The diagnostic battery has now landed: AN-015 finds Conley close-pair shares stable in non-pharma (16.9% → 16.8%) and falling in pharma (27.6% → 24.4%); Bajari-Ye T1 ratios fall in both classes (NP 2.63 → 1.83; PH 1.29 → 1.11). The threat that the smaller eligible SME pool generated a new coordination shock is rejected. Strict invariance (AN-017) reinforces the dominance ordering, raising the exclusion share to 85% (NP) and 79% (PH) when composition is held fixed. The reading remains 🟡 because residual baseline clustering exists (realized close-pair shares are 1.6–1.9× the null mean) and the screens do not prove IPV-clock holds — they rule out the most obvious deviations. The pharma analog inherits additional model sensitivity (AN-016), which is why pharma is treated as a boundary case rather than a second headline.

Sources.

  • Own analysis: AN-010 (BNE simulation + decomposition table); supporting pieces AN-013 (drop-out extraction), AN-014 (Krasnokutskaya-style heterogeneity correction), AN-015 (collusion screens), AN-017 (strict-invariance sensitivity).
  • Reports: none direct.
  • News anchors: none direct.
  • Cross-refs: H:exclusion-dominates; H:protected-pool-responds; paper §4 (Exclusion and the Price-Forming Pool).
  • Validation: v7-jpube-tight/scripts/45_bne_simulation.R and 46_decomp_compare.R produce the decomposition table cells; the v8-jpube 46_bne_empirical_counts.R is the empirical bidder-count robustness used to discipline the simulation against observed entry.