Theorem IsMaxminShare ↔ IsAlphaMMS 1
theorem staged

IsMaxminShare ↔ IsAlphaMMS 1

Theorem. An allocation $A$ satisfies the exact MMS guarantee ([[social_choice.fair_division.indivisible.maximin_share]]) if and only if it is 1-MMS in the α-MMS sense ([[social_choice.fair_division.indivisible.is_alpha_mms]]): $$ \mathrm{IsMaxminShare}\ v\ \mathrm{allGoods}\ A \;\iff\; \mathrm{IsAlphaMMS}\ 1\ v\ \mathrm{allGoods}\ A. $$

In Lean: SocialChoice.FairDivision.Indivisible.isMaxminShare_iff_isAlphaMMS_one.

Proof

Unfold both predicates against the MMS value ([[social_choice.fair_division.indivisible.mms_value]]):

  • IsAlphaMMS 1 states $1 \cdot \mathrm{MMS}_i \le v_i(A(i))$, i.e.\ $\mathrm{MMS}_i \le v_i(A(i))$, i.e.\ agent $i$'s bundle value beats the maximin threshold.
  • IsMaxminShare states the worst-bundle-of-any-partition $B$ is bounded by $v_i(A(i))$: $\forall B,\ \exists j,\ v_i(B(j)) \le v_i(A(i))$.

The bridge: $\forall B,\ \exists j,\ v_i(B(j)) \le v_i(A(i))$ is exactly the statement that the maximin value $\max_B \min_j v_i(B(j))$ is bounded by $v_i(A(i))$, which is $\mathrm{MMS}_i \le v_i(A(i))$.

Both directions follow from the mmsValue_le_of_forall and iInf_partition_le_mmsValue building blocks ([[social_choice.fair_division.indivisible.mms_value_bounds]]).

Why both predicates exist

The two formulations are complementary:

  • IsMaxminShare avoids Finset.sup' / Finset.inf' — a slick definition for exact MMS that doesn't require the auxiliary mmsValue function. Useful for stating theorems compactly.
  • IsAlphaMMS parameterises smoothly over the approximation parameter — needed for $\alpha < 1$ approximation results.

This equivalence theorem lets call sites switch freely between the two.

References

  • Budish, E. (2011). "The Combinatorial Assignment Problem". J. Pol. Econ.
  • [AGT Chapter 11] Nisan, Roughgarden, Tardos, and Vazirani, Algorithmic Game Theory. MMS vs α-MMS equivalence at α = 1.

Used by

Also in