AN-012: 10% SME price preference benchmark

Intuition (plain-language)

What if, instead of banning non-SMEs, you give SMEs a 10% bid discount for winner selection? Everyone still competes, so the simulated price barely moves — yet SMEs win more often. The preference buys redistribution without sacrificing price discipline; it is a static design benchmark, not a forecast of a specific legal regime.

Question

A static price preference is a non-exclusionary design alternative to a set-aside: SMEs receive a percentage haircut to their bids for winner selection only, but all firms remain eligible to participate and the government pays the actual winning bid. What is the simulated price effect of a 10% SME preference, and how much SME win-rate gain does it deliver?

Design

• Sample: structural cells from AN-010.
• Variation: counterfactual policy regime \(V_3\) (10% SME price preference for selection) vs \(S_1\) (open).
• Specification: BNE simulation; SME bid is multiplied by 0.90 for selection only; the government pays the actual (unhaircut) winning bid.
• Outcomes: \(\Delta p = p^{V_3} - p^{S_1}\) (simulated price shift); SME win-rate gain.

Results

From v8-jpube/output/values.tex (10% SME price preference for selection only; government pays actual winning bid; entry held fixed):

Class	\(\Delta p\) vs \(S_1\)	SME win-rate gain
Non-pharma	−0.004	+4.3 pp
Pharma	+0.002	+1.4 pp

Both price shifts are essentially zero in reference-price units — non-SMEs continue to bid and discipline the price-forming order statistic. The SME win-rate gain is smaller than under full set-aside (which delivers full SME-winner concentration by construction); the preference trades redistribution for preserved competitive discipline.

Output: macros in v8-jpube/output/values.tex; paper table tab_welfare_policy_v8 columns 5–6 in §5.

Interpretation

The 10% preference preserves the price-forming pool: non-SMEs continue to bid and discipline the price-forming order statistic. The simulated price is essentially unchanged from \(S_1\). The preference is a redistribution device — SMEs win more often than under open competition — but the redistribution is smaller than under the full set-aside.

The reading is that the relevant policy frontier is not "support SMEs vs do nothing" but "exclusionary redistribution vs support that preserves the price-forming bidder pool". The 10% preference is a static design benchmark of the second branch, not a forecast of a legal preference regime.

Confidence: yellow. The simulation holds entry fixed at the pre-policy level by construction — an entry-elastic version would require a separate entry model. The 10% margin is the cleanest design benchmark; the diminishing-returns curve and the preference level at which non-SMEs start dropping out is in the optimal-preference search.

Follow-ups

• Optimal-preference search (v7-jpube-tight/scripts/61_optimal_preference.R) over margins \(\in \{0\%, 5\%, 10\%, 15\%, 20\%, 25\%\}\): deserves its own AN page to expose the design frontier.
• Distributional incidence (v7-jpube-tight/scripts/60_distributional_incidence.R): is the SME win-rate gain spread across SME firm sizes, or concentrated on marginal firms barely above the SME threshold? Important for evaluating the preference against alternative SME-support instruments.
• Entry-elastic preference benchmark: a version where preference shifts firms' decision to participate. The static-cost result may not survive endogenous entry; out of scope for this paper.