Extensions

Where this sits in v8

The first five extensions below are reduced-form-layer checks on the broader panel — they support the timing-and-sign benchmark, not the headline. The canonical v8 policy result is the structural decomposition and welfare comparison (Results); the Welfare cost section at the bottom reports the v8 annualized figures.

Intuition (plain-language)

These extensions poke at the same mechanism from different angles. Is the effect real or just inflation? (Deflated prices look identical.) Does opening tenders change who wins? (Fewer SMEs win — exactly what a competition story predicts.) Does the government get better deals against its own reference price? (Yes.) And the welfare arithmetic puts a number on it: about 29% of the open price per auction, R$38–89 million a year on this one product group. None of the first five is the headline; the welfare cost is the substantive payoff.

IPCA-Deflated Real Prices

Using IPCA-deflated (real) prices yields patterns virtually identical to the nominal results, confirming that the price effects reflect real differences in procurement costs rather than differential inflation across product groups.

---

Extensive Margin: Tender Completion

The probability that a procurement item is successfully completed may also be affected by the policy. Point estimates suggest that open tenders may improve completion rates, though the effects are less precisely estimated due to the use of group-level (rather than item-level) fixed effects.

---

Procurement Efficiency

Procurement efficiency, measured as the ratio of the final negotiated price to the government's reference price, improves under open tenders. Lower ratios indicate that the government obtained better deals relative to its own expectations.

---

Winner Composition

The probability that the winning firm is an SME decreases under open tenders, directly supporting the competition mechanism: open tenders attract larger firms that outbid SMEs on price.

---

Heterogeneity by PBU Type

Interacting the treatment indicator with a direct administration dummy reveals broadly similar effects across buyer types (direct administration vs. indirect administration entities), suggesting that the competition channel operates consistently across the institutional landscape.

---

Welfare Cost

The structural welfare arithmetic combines the allocative wedge \(\text{DWL}_{\text{alloc}} = c_{(1)}^{S_3} - c_{(1)}^{S_1}\) with an MCPF distortion \(\lambda \cdot (p_{S_3} - p_{S_1})\) at \(\lambda = 0.30\) (Ballard–Shoven–Whalley benchmark; updated MCPF frameworks in Hendren 2020 and Finkelstein–Hendren 2020).

Class	Loss / pS1 at λ = 0.30
Non-pharma	28.9%
Pharma (boundary)	44.8%

Translated to annual public-finance units on Group 65 alone, the welfare cost spans R\(38–89 million per year** (US\)11–25M) across the 30–70%** adherence range, against a Group-65 annual reference outlay of roughly R\(345M (non-pharma) and R\)363M (pharma). This is one product group of São Paulo's R$13 billion procurement platform.

R$ 38–89 M / yr v8 annual static welfare cost of bidder exclusion on Group 65, at λ = 0.30, across the 30–70% adherence range (US$11–25M). The headline per-auction loss is 28.9% of the open-regime price in standardized non-pharmaceutical procurement.