B1 — diagonal allocation of a symmetric IC mechanism is monotone and [0,1]-valued
ProvedMultiItemRev.TwoIID.diagonal_allocation_monotoneB1 — diagonal-allocation monotonicity (Hart–Nisan, Approximate Revenue Maximization with Multiple Items, arXiv:1204.1846v3, pp. 32–33, proof of Theorem B, Eq. (12) setup).
For a symmetric incentive-compatible mechanism M for two goods, the diagonal allocation t ↦ q(t, t)_0 — the allocation of good 0 when both reported values equal t — is monotone nondecreasing and takes values in [0, 1].
This is the missing bridge node in the Theorem B chain: display_12 bounds a symmetric mechanism's revenue by an analytic expression in its diagonal allocation, and rhs12_le bounds that analytic expression for any monotone [0, 1]-valued phi. Setting phi := fun t => M.q (diag t) 0, both of rhs12_le's hypotheses follow from this node.
Proof idea: two IC comparisons (true type (t, t) reporting (t', t') and vice versa) cancel the payments and give (t − t') · (Q(t) − Q(t')) ≥ 0 for the total diagonal allocation Q(t) = q(t, t)_0 + q(t, t)_1. Symmetry gives q(t, t)_0 = q(t, t)_1 (the swap fixes the diagonal), so Q(t) = 2 · q(t, t)_0 and the coordinate-0 allocation is monotone. The [0, 1] bounds are feasibility. Symmetry is genuinely needed: IC alone yields monotonicity of the sum Q, not of coordinate 0 alone.
Only the minimal hypotheses are assumed (feasibility + IC + symmetry); individual rationality, no-positive-transfer, and measurability are not used. display_12 (live, Open) is NOT re-published here — this is B1 only.
import Mathlib import Definitions.Def_MultiItemRev_TwoIID_Model open MeasureTheory open scoped NNReal ENNReal
namespace MultiItemRev.TwoIID
/-- B1: diagonal-allocation monotonicity (Hart–Nisan, arXiv:1204.1846v3,
pp. 32–33, proof of Theorem B, Eq. (12) setup).
For a symmetric incentive-compatible mechanism, the diagonal allocation
`t ↦ M.q (diag t) 0` — the allocation of good 0 when both reported values
are `t` — is monotone and `[0,1]`-valued. This is the bridge that lets the
diagonal allocation feed `φ` into `rhs12_le` via `display_12`: set
`φ := fun t => M.q (diag t) 0` and both of `rhs12_le`'s hypotheses
(`Monotone φ`, `0 ≤ φ t ∧ φ t ≤ 1`) follow from this node.
Proof idea: two IC comparisons — true type `(t,t)` reporting `(t′,t′)`
and vice versa — cancel the payments and give
`(t − t′) · (Q(t) − Q(t′)) ≥ 0` for the total diagonal allocation
`Q(t) = q(t,t) 0 + q(t,t) 1`. Symmetry gives `q(t,t) 0 = q(t,t) 1`
(`swap2` fixes the diagonal), so `Q(t) = 2 · q(t,t) 0` and the
coordinate-0 allocation is monotone. The `[0,1]` bounds are feasibility.
Symmetry is genuinely needed: IC alone yields monotonicity of the sum `Q`,
not of coordinate 0 alone. -/
theorem diagonal_allocation_monotone
(M : MultiItemRev.Decomp.Mechanism (Fin 2))
(hFeas : MultiItemRev.Decomp.IsFeasible M)
(hIC : MultiItemRev.Decomp.IsIC M)
(hSym : IsSymmetric M) :
Monotone (fun t => M.q (diag t) 0) ∧
∀ t, 0 ≤ M.q (diag t) 0 ∧ M.q (diag t) 0 ≤ 1 := by sorry
end MultiItemRev.TwoIID