H:preference-near-zero-cost — A 10% SME price preference preserves the price-forming pool at near-zero static cost¶
A static price preference is a non-exclusionary design alternative to a set-aside. SMEs receive a percentage haircut applied to their bids for winner selection only; all firms remain eligible to participate, and the government pays the actual winning bid. Because non-SMEs continue to discipline the price-forming order statistic, the simulated price under a 10% SME price preference is essentially unchanged from the open benchmark \(S_1\), while SME win-rate rises. The preference is not a free version of the set-aside — it delivers less redistribution — but it shifts the static frontier.
Intuition (plain-language)
A price preference is the gentle alternative to a set-aside: SMEs get a discount applied only when picking the winner, but everyone still bids and the government pays the real winning price. Because the disciplining non-SMEs stay in the room, the simulated price barely budges — yet SMEs win more often. It is not a free set-aside (it redistributes less), but it buys SME support without throwing away the competition that holds prices down.
Evidence strength: Partial (strongly supported). AN-012 reports a price shift of −0.004 NP / +0.002 PH of the reference price under \(V_3\) (10% SME preference); SME win-rate gain +4.3 pp NP / +1.4 pp PH. AN-024 extends this: the welfare ranking \(V_3 \succ V_0\) holds across the entire λ grid [0.15, 0.45] in non-pharma under both main and strict-invariance specifications — the choice of λ does not flip the preference vs set-aside ordering. Pharma ranking is \(V_3 \succ V_0\) under main spec across all λ but reverses to \(V_0 \succ V_3\) under strict invariance (the boundary-case treatment).
Theory¶
In an English clock auction with bid-shading or scoring, a percentage preference applied to the SME bid for selection only (with the government paying the actual winning bid) does not distort the price-formation logic among non-SME bidders. The non-SMEs continue to drop out at their costs; the SME's effective bid is shaded by the preference for selection purposes. Provided the non-SMEs remain in the auction in numbers comparable to the open regime, the price-forming order statistic is essentially unchanged. The result is in the spirit of price-preference analyses in Marion (2007) and Krasnokutskaya and Seim (2011), where the design choice between exclusion and preference maps onto different bidder-pool retention.
Prediction¶
A 10% SME price preference simulated under the static BNE machinery should produce \(|\Delta p| \approx 0\) relative to \(S_1\) in standardized non-pharma. SME win-rate should rise by a meaningful but not enormous margin. The benchmark is a decomposition device, not a policy forecast.
Competing prediction¶
Non-SMEs exit when preferred against. If the preference makes non-SMEs systematically lose, they may stop participating altogether — in which case the preference behaves like a soft set-aside and the static-cost result collapses. The static analysis holds entry fixed by construction; an entry-elastic analysis would change the prediction.
Setting evidence¶
São Paulo never operationalized a price preference in BEC. The preference benchmark is therefore not identified from administrative data; it is a static design simulation using the recovered cost primitives. The reading is explicitly that of a decomposition device, not a policy forecast. See paper §5 for the scope statement.
Empirical test¶
- Outcome variables: \(\Delta p = p^{V_3} - p^{S_1}\) (simulated payment shift); SME win-rate gain.
- Variation: counterfactual policy regime \(V_3\) (10% SME price preference for selection) vs \(S_1\) (open).
- Specification: BNE simulation with the recovered cost primitives; the SME bid is haircut by 10% for selection only, with the government paying the actual winning bid.
- Sample: structural cells (non-pharma standardized; pharma boundary).
Data requirements and limitations¶
The preference benchmark inherits all maintained restrictions of the structural decomposition (H:ipv-clock-admissible). It additionally holds entry fixed at the pre-policy level by construction — an entry-elastic version would require a separate entry model. The reading is explicitly static and non-equilibrium: it asks the price-formation cost of preserving the non-SME bidder pool under a non-exclusionary preference.
Evidence¶
| Analysis | Bearing | Key takeaway |
|---|---|---|
| AN-012 | Supports | \(\Delta p\) = −0.004 NP / +0.002 PH at 10% preference; SME win-rate gain +4.3 pp NP / +1.4 pp PH. |
| AN-024 | Supports | Welfare ranking \(V_3 \succ V_0\) stable across λ ∈ [0.15, 0.45] in NP under both main and strict-inv specs. Ranking not λ-driven; determined by composition treatment in PH. |
Open tests¶
Optimal preference search¶
v7-jpube-tight/scripts/61_optimal_preference.R searches over preference
margins from 0% to 25%. The 10% headline is the cleanest design
benchmark; the search exposes the diminishing-returns curve and the
preference level at which non-SMEs start dropping out of the simulation.
Not yet documented as a standalone AN.
Distributional incidence of the preference¶
v7-jpube-tight/scripts/60_distributional_incidence.R distributes the
SME win-rate gain across the SME firm-size distribution. If the gain
concentrates on marginal SMEs barely above the size cutoff, the
redistribution is different than if it spreads to small SMEs.
Important for evaluating the design against alternative SME-support
instruments (subsidies, training, micro-finance).
Entry-elastic preference benchmark¶
The static reading holds entry fixed. A version with elastic entry — where the preference shifts firms' decision to participate — would test whether the near-zero price cost survives endogenous entry. Out of scope for this paper but a clean follow-up.