H:implied-welfare-weight-large — A planner needs SME welfare weight ~2.42 to prefer the full set-aside over open auctions in non-pharma¶
The static welfare arithmetic produces a Saez-Stantcheva-style implied-weight statistic: the marginal welfare weight on SME producer surplus that a utilitarian planner would need to apply to be indifferent between the full set-aside and the open auction. In standardized non-pharma, the implied weight is ~2.42 — i.e., the planner must value a dollar of SME producer surplus at $2.42 of welfare to find the exclusionary regime optimal.
Intuition (plain-language)
Rather than assert the set-aside is bad, the paper asks: how much would a government have to favor SME profits — above ordinary citizens' money — to make full exclusion worth it? In non-pharma the answer is 2.42: the planner must value a dollar of SME producer surplus at $2.42 of social welfare. That is a high bar, and it makes the policy's cost transparent in welfare-weight terms rather than just reais.
Evidence strength: Partial (strongly supported). AN-011 reports the implied weight \(\omega = 2.42\) in non-pharma, 2.61 in pharma. AN-024 confirms the qualitative "substantially-above-unity" claim is stable across the λ grid [0.15, 0.45]: the implied \(\omega\) scales linearly with the welfare loss; even at λ=0.15 (the most conservative MCPF benchmark), \(\omega\) remains > 1.5. The welfare ranking \(V_3 \succ V_0\) stable across λ in non-pharma means no λ choice rescues the set-aside under utilitarian weighting.
Theory¶
The Saez-Stantcheva (2016) framework expresses social preferences as a weighted sum of agents' marginal welfare contributions, where the weights are recovered from policy choices. Applied to the SME set-aside, the implied weight on SME producer surplus is the welfare weight that rationalizes the full set-aside as optimal given the realized welfare loss. A weight > 1 indicates the set-aside is justified only by treating SME surplus as more socially valuable than dollar-for-dollar; a weight = 1 implies the set-aside is dominated by less-exclusionary designs at the planner's revealed preferences; a weight < 1 implies the set-aside is anti-redistributive under the revealed preferences.
Prediction¶
In standardized non-pharma, the implied SME welfare weight required for indifference between the full set-aside (\(V_0\)) and the open auction (\(S_1\)) should be substantially greater than 1 — paper headline ~2.42. In pharma, the implied weight should be even larger (paper §5 reports the corresponding pharma number).
Competing prediction¶
Conventional weighting justifies the set-aside. If standard distributional weights (e.g., Hendren 2020 MVPF benchmarks) routinely exceed 2 for low-income or small-firm beneficiaries, the implied weight of 2.42 is not out of the ordinary. The interpretation of "large" is therefore relative — the number is only meaningful by comparison to the planner's other implied weights revealed elsewhere in tax or transfer policy.
Setting evidence¶
The implied-weight statistic is a welfare-arithmetic output, not a revealed-preference recovery. It does not say what the São Paulo planner actually values; it says what they would have to value to make the policy optimal. The number is a comparison device, not a preference attribution.
Empirical test¶
- Outcome variable: implied SME welfare weight \(\omega\) such that \(\omega \cdot \text{SME transfer} = \text{DWL}_{\mathrm{alloc}} + \lambda \cdot \Delta_{\mathrm{gov}}\).
- Variation: solve for \(\omega\) at the recovered welfare-loss magnitudes from AN-011; λ = 0.30 baseline.
- Specification: closed-form inversion of the welfare-arithmetic identity; bootstrap CI from the welfare bootstrap.
- Sample: structural cells (non-pharma standardized; pharma boundary).
Data requirements and limitations¶
Inherits all the welfare arithmetic's limitations. The implied weight is calibration-sensitive to λ (a higher λ reduces the required weight by inflating the fiscal-distortion side of the ledger). The reading is non-comparative across studies unless the λ benchmark is aligned.
Evidence¶
| Analysis | Bearing | Key takeaway |
|---|---|---|
| AN-011 | Supports | Implied weight = 2.42 in non-pharma at λ=0.30; corresponding pharma number larger but model-sensitive. |
| AN-012 | Supports (indirectly) | Under the 10% preference benchmark, the implied weight required for indifference is smaller — quantifies the welfare gain from non-exclusionary design. |
| AN-024 | Supports | λ sensitivity: welfare ranking \(V_3 \succ V_0\) stable across λ ∈ [0.15, 0.45] in non-pharma; implied \(\omega\) > 1.5 even at λ=0.15. λ choice does not rescue the set-aside. |
Open tests¶
λ sensitivity¶
The implied weight scales with the λ baseline. A sweep over λ in {0.10, 0.20, 0.30, 0.50} would expose the calibration sensitivity. Partly done in paper §5; would benefit from an explicit AN page.
Comparison to literature welfare weights¶
Comparing the implied weight to those revealed in (i) corporate tax preferences, (ii) SME subsidy programs, or (iii) Saez-Stantcheva analyses of comparable redistribution would provide a calibration benchmark. Currently a discussion-section observation; a structured comparison would strengthen the policy interpretation.