Theorem Strategy-Proofness of Gale-Shapley for the Proposing Side (Roth 1982)
theorem staged

Roth 1982: Strategy-Proofness for the Proposing Side

Theorem (Roth 1982). The men-proposing Gale-Shapley mechanism is dominant-strategy incentive-compatible (DSIC) for the men: for every man $i$, truthfully reporting his preference list $\succ_i$ weakly dominates every misreport $\succ_i'$, regardless of what the other men and women report.

Formally: let $\mathrm{GS}^M : \succ \mapsto \mu^M(\succ)$ denote the men-proposing matching function. For every $i \in M$, every $\succ_i'$, and every fixed report profile $\succ_{-i}$:

$$\mathrm{GS}^M(\succ_i, \succ_{-i})(i) \succeq_i \mathrm{GS}^M(\succ_i', \succ_{-i})(i)$$

where the LHS uses $i$'s true preferences $\succ_i$ for comparison.

Asymmetry: Not Strategy-Proof for Women

The women-proposing analogue holds for women, but in the men-proposing mechanism the women have profitable misreports (they can sometimes get a strictly preferred match by truncating their preference list). This asymmetry is fundamental: Roth showed that no stable matching mechanism is DSIC for both sides simultaneously (Roth 1982 impossibility).

Proof

Sketch

The key tool is the blocking lemma (Gale-Sotomayor 1985): if a man $i$ misreports and gets a strictly better partner under the mechanism, then in the resulting matching there must be a blocking pair involving $i$ and his true preferred achievable partner — contradicting stability.

More concretely, suppose $i$ misreports $\succ_i'$ and ends up with $j' = \mathrm{GS}^M(\succ_i', \succ_{-i})(i)$ such that $j' \succ_i \mathrm{GS}^M(\succ)(i)$. Truth-telling gives $i$ his best achievable partner under the true preferences (men-optimal property). So $j'$ must be unachievable under the true profile — meaning there is no stable matching in which $i$ gets $j'$ when truth is told. Combining with $j'$'s acceptance of $i$ under the misreport gives the contradiction.

Cross-Reference

Connects matching theory to mechanism design's general DSIC framework ([[mechanism_design.basic.dsic_predicate]]).

Lean Status

Tracked by MT-L3 (#205). Depends on: - [[gale_shapley_stable]] (stability of GS output), - [[proposing_optimal]] (men-optimality), and - the blocking lemma (Gale-Sotomayor 1985) which will need its own helper lemma in Lean.

References

  • [MSZ Ch.22, Thm 22.16] Maschler, Solan, Zamir, Game Theory.
  • Roth (1982). The Economics of Matching: Stability and Incentives. Math. Oper. Res. 7, 617–628.
  • Dubins & Freedman (1981). Machiavelli and the Gale-Shapley Algorithm. AMM 88.
  • Gale & Sotomayor (1985). Some Remarks on the Stable Matching Problem. Discrete Applied Math 11.
  • Roth & Sotomayor (1990), Ch. 4.

Also in