{"id":"f768e732-4686-4e77-8e5f-16e214efaa6c","arxiv_id":"2504.17467","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The FDA matching equals the DA matching when each hospital quota is its FDA fill count, which makes the DA outcome constrained efficient and weakly stable.","lead":"This paper proves that the flexible deferred acceptance (FDA) algorithm and the deferred acceptance (DA) algorithm, when the DA runs with per-hospital quotas drawn from the FDA's own output, always produce the same matching under regional caps. The result interprets the FDA as a way to endogenously design hospital-level quotas, so that DA can still deliver constrained efficiency.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 2 assumes without verification that Algorithm 2's target-capacity/order choice function satisfies the KK2018 substitutability and LAD conditions, so the doctor-optimality premise may not cover the FDA as defined.","rationale":"The reader's weakest-assumption analysis identifies exactly the load-bearing point: the proof is conditional on KK2018 Proposition 1, and the paper does not verify that the choice function generated by its own Algorithm 2 satisfies the required substitutability and law-of-aggregate-demand conditions. My independent reading confirms this is the most serious unresolved issue. The two-market optimality argument is otherwise plausible, and the suspicious Step 2(i) equality appears reparable by choosing x' at the same hospital as x~ or by observing that the case is vacuous. I therefore do not think the concern moves the verdict away from CONDITIONAL; it reinforces the reader's conditional recommendation. The proposed brute-force search is a concrete way to test whether the theorem itself is true; if it passes on many small instances, the remaining issue is proof rigor rather than correctness of the central claim. I agree with the reader that the paper is not fully self-contained and needs a clarification of the FDA definition and the delegated conditions.","tokens_in":11986,"tokens_out":28499,"duration_ms":284981,"concrete_test":"Run an exhaustive search over all preference profiles with up to 4 doctors and 3 hospitals (and 1-2 regions), all regional caps, target capacities, and hospital orders, comparing the outcome of Algorithm 2 with the DA outcome under adapted capacities q~_h=|muF_h|; any single mismatch would refute Theorem 2, and a clean pass would materially increase confidence that the asserted equivalence is correct despite the proof gaps.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is plausible and the two-market optimality strategy is attractive, but as written the proof of Theorem 2 hinges on an unverified identification. The paper invokes Kamada and Kojima (2018) Proposition 1 to assert that the FDA outcome X^F is a doctor-optimal stable allocation in the original market M_F. That proposition requires the aggregate hospital-side choice function to satisfy substitutability and the law of aggregate demand. The paper never checks these properties for the specific choice function induced by Algorithm 2, which includes a target-capacity filling phase and an order-based regional filling phase. If that choice function differs from the one covered by KK2018, or if the target-capacity phase is not part of the KK2015 FDA, then the premise 'X^F is doctor-optimal stable in M_F' is unsupported and the proof of Theorem 2 lacks its foundation. There is also a local gap in Step 2(i): the equality xi(X^F+x')=xi(X^D+x~) is not true for an arbitrary x' in X^D\\X^F; it holds only if x' is chosen at the same hospital as x~. The case can be made vacuous by transitivity, but as written it is an unexplained step. Neither issue proves the theorem false, but both need to be resolved before the proof can be accepted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies two-sided matching with regional caps. It claims that for any preference profile there exist allocations of the regional caps among hospitals (adapted capacities) such that the deferred acceptance (DA) algorithm produces a constrained efficient and weakly stable matching, and that this matching coincides exactly with the outcome of the flexible deferred acceptance (FDA) algorithm of Kamada and Kojima (2015). The proof embeds the original regional-constraint market and a shadow market with adapted capacities into the matching-with-contracts framework, invokes doctor-optimality of the generalized DA and the rural hospital theorem, and concludes that the FDA and DA outcomes coincide.","tokens_in":12239,"tokens_out":8095,"duration_ms":66796,"significance":"If the theorem holds, it provides a clean interpretation of the FDA algorithm as an endogenous capacity-design device for DA and offers a broadly applicable template for equivalence proofs among DA-based mechanisms, with natural extensions to hierarchical constraints. The paper builds on established results, and the two-market optimality argument is elegant and economical. However, the proof as written has gaps that prevent acceptance in its current form.","major_comments":[{"comment":"The proof relies on Kamada and Kojima (2018) Proposition 1 to assert that the FDA outcome X^F is a doctor-optimal stable allocation in the original market M_F, but it does not verify that the hospital-side choice function induced by Algorithm 2—with its target-capacity filling phase and order-based regional filling phase—satisfies substitutability and the law of aggregate demand, nor that Algorithm 2 is exactly the cumulative offer process of that contracts market. Because Step 1 of the proof depends entirely on this premise, the equivalence theorem is not yet established.","section":"Section 2.5, proof of Theorem 2; footnotes 7 and 10; Section 2.3, Algorithm 2"},{"comment":"The equality ξ(X^F + x') = ξ(X^D + ~x) is asserted for any x' ∈ X^D\\X^F with X^F + x' ≻_D X^F, but it holds only if x' and ~x involve the same hospital. This is a gap. The case appears to be vacuous because ~x ∈ X^F and ~x ∈ C_D(X^D + ~x) contradict X^D ≽_D X^F from (1), but the paper should state this explicitly or construct x' with the required hospital. As written, this step is unsupported.","section":"Section 2.5, Step 2(i)"},{"comment":"The definition of the FDA algorithm includes target capacities q̄_h, which are not part of the model primitives in Section 2.1 and are not part of the original Kamada and Kojima (2015) FDA as presented there. The paper should specify the domain of q̄_h and state whether Theorem 2 is meant to hold for all target capacities or for a particular choice; otherwise the statement of Theorem 2 is not fully well-defined.","section":"Section 2.3, Algorithm 2"}],"minor_comments":[{"comment":"The text says \"d4 and d5 remain unmatched,\" but in the displayed matching d5 is matched to h3; the intended statement is that d3 and d4 remain unmatched.","section":"Section 2.2, Example 1"},{"comment":"The phrase \"rejects the the lowest-ranking\" contains a duplicated article.","section":"Algorithm 1"},{"comment":"The definition of substitutability uses the variables x and x' in a way that could be confused with the surrounding text's use of x; consider relabeling.","section":"Footnote 7"},{"comment":"The word \"gerenralized\" is a typo for \"generalized.\"","section":"Footnote 9"},{"comment":"The statement that weak stability in the original market coincides with stability in the shadow market is asserted without proof; if this is intended as a direct consequence of Theorem 2, it should be made explicit.","section":"Section 3"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a short note with a promising idea. The main risk is whether the invoked Kamada-Kojima (2018) Proposition 1 covers the algorithm as defined; the author should verify the choice-function properties or prove them. If the gaps are filled, the result is publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nTwo things to know. The equivalence claim is genuinely new and the two-market optimality proof is an attractive idea. But the proof as written does not cover the algorithm it defines, and that is a real problem.\n\nWhat is good: the paper states a clean result — set each hospital's capacity in DA to the number of doctors the FDA placed there, and DA reproduces the FDA matching. That gives FDA a nice interpretation as endogenous capacity design and recasts weak stability as ordinary stability in a shadow market. The proof technique, using the Hatfield-Milgrom framework and KK2018's rationalization of choice functions, is compact and mostly fillable. The author is honest about relying on KK2018.\n\nThe soft spots. Algorithm 2 is presented as the Kamada-Kojima (2015) FDA, but it is a variant: the real KK2015 FDA has no target-capacity phase. Here the algorithm first fills target capacities, then lets the region fill remaining seats in hospital order. In Example 2 this produces a different outcome than the actual KK2015 FDA with the same order. The proof invokes KK2018 Proposition 1 to assert that the FDA outcome is doctor-optimal stable in the original market, but that proposition requires the hospital-side choice function to satisfy substitutability and the law of aggregate demand. The paper never checks these properties for the choice function induced by Algorithm 2, and the target-capacity phase makes it doubtful that the KK2018 result covers it. Without that check, the premise of Step 1 is unsupported. There is also a small gap in Step 2(i): the distribution equality holds because x' and the rejected contract must be at the same hospital; this follows from the rural hospital theorem but is not explained.\n\nSo the result is plausible and the proof strategy is a genuine contribution, but the paper is not ready as written. It needs either a direct proof of the choice-function conditions for Algorithm 2, or a redefinition of the algorithm to match the KK2015 FDA.\n\nWho is this for: people working on matching with distributional constraints, particularly those comparing DA-based mechanisms. It deserves referee time — the idea is worth taking seriously and the fix seems feasible.\n\nRecommendation: send to peer review, but make clear that the KK2018 identification and the target-capacity issue must be resolved before acceptance.","headline":"A plausible, novel equivalence between FDA and DA with adapted capacities, but the proof relies on an unverified identification of Algorithm 2 with the KK2015 FDA, so it needs revision before the claim is established.","tokens_in":12772,"tokens_out":7087,"would_cite":false,"duration_ms":62484,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91B68"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that deferred acceptance, run with per-hospital quotas taken from the FDA outcome's fill counts, reproduces the FDA matching exactly, so constrained-efficient allocations of regional caps always exist.","keywords":["FDA algorithm","DA algorithm","regional constraints","matching with contracts","weak stability","capacity design","distributional constraints","doctor-optimality"],"falsifier":"Run doctor-proposing DA on a small market (for instance, the three-hospital, five-doctor example in the paper) with each hospital's capacity set to its FDA fill count; if any FDA order produces a DA outcome different from the FDA matching, Theorem 2 is false. Alternatively, directly test the hospital-side choice rule of Algorithm 2 for substitutability and the law of aggregate demand, since the proof inherits doctor-optimality from a cited theorem only under those conditions.","tokens_in":11699,"feed_emoji":"🏥","tokens_out":11972,"duration_ms":90948,"temperature":0.7,"pith_summary":"The paper addresses a practical obstacle in applying the deferred acceptance (DA) algorithm to markets with regional caps: the caps must be split among hospitals, and a poor split can make DA inefficient. It proves that, for any fixed preferences, there exists a split under which DA produces a constrained efficient, weakly stable matching. The construction is not arbitrary—the right split is the doctor distribution produced by the flexible deferred acceptance (FDA) algorithm, and DA run with those exact per-hospital quotas returns precisely the FDA matching. This gives the FDA algorithm a new reading as an endogenous capacity-design tool for DA, and it offers a general way to prove equivalences among DA-based mechanisms.","feed_headline":"DA equals FDA when hospital quotas mimic FDA's fill counts","feed_subtitle":"Hospital quotas taken from the FDA outcome make deferred acceptance constrained-efficient and weakly stable.","key_machinery":"The key machinery is the matching-with-contracts model—contracts are doctor–hospital pairs, with all hospitals aggregated into one agent—together with rationalized hospital-side choice functions. For the original market the hospital side maximizes $g^F(\\xi(Y)) + f_R(\\xi(Y)) + \\epsilon f_H(Y)$; for the shadow market it maximizes $g^D(\\xi(Y)) + f_H(Y)$, where $\\xi(Y)$ is the vector of doctors per hospital. The first terms enforce feasibility (regional caps in one market, adapted capacities in the other), $f_R$ encodes the region-level preference, $f_H$ encodes hospital preferences, and the small $\\epsilon$ makes the region-level objective dominate. These rationalizations let the proof express stability in both markets as a choice-function condition, so that doctor-optimality of the generalized DA in each market forces the FDA outcome and the shadow-market DA outcome to be the same.","core_discovery":"The central result, Theorem 2, says that if $\\mu^F$ is the matching produced by the FDA algorithm and each hospital $h$ is assigned the adapted capacity $\\tilde q_h = |\\mu^F_h|$, then the DA matching $\\mu^D$ in the market with those capacities satisfies $\\mu^F = \\mu^D$. The proof constructs two markets: the original market $M^F$ with regional caps, where the FDA runs, and a shadow market $M^D$ without regional caps but with the adapted hospital capacities, where DA runs. Using the matching-with-contracts framework, the author shows that the FDA outcome is stable in both markets and that the DA outcome is stable in both markets. Since the generalized DA is doctor-optimal in each market, the two stable allocations must coincide. The author concludes that efficient allocations of regional caps therefore always exist, that they can be read off the FDA outcome, and that the same argument extends to hierarchical regional constraints and to any DA-based mechanism whose hospital-side choice function first picks a distribution and then assigns doctors.","pith_inferences":["A reverse reading is implicit: any DA run with adapted capacities can be viewed as an FDA run for some region-level preference order, whenever those capacities match the fill counts of some FDA outcome.","For applications such as residency matching, the equivalence suggests a regulator could compute fill counts once from FDA and then implement with plain DA, preserving strategy-proofness for doctors while avoiding the appearance of favoring one hospital order.","The proof is a transferable recipe for other constrained settings: model the hospital side as a distribution-picking choice function, verify the regularity conditions, and compare doctor-optimal stable allocations in original and shadow markets.","Because the theorem is stated for fixed preferences, it does not say how the adapted capacities should be revised when preferences change; a dynamic or preference-robust version of the equivalence remains open."],"forward_implications":["There always exists an allocation of regional caps under which the DA algorithm is constrained efficient and weakly stable (Corollary 1).","Starting from any weakly stable but inefficient DA matching, moving to the FDA-based allocation makes every doctor weakly better off (Theorem 1).","The FDA outcome can be implemented by a standard DA run in a shadow market with hospital-level quotas, so no region-level hospital order is needed at implementation time.","Weak stability in the original regional-cap market coincides with ordinary stability in the shadow market with adapted capacities.","The equivalence extends to hierarchical regional constraints and to any DA-based mechanism whose hospital-side choice first fixes a distribution and then assigns doctors, when the required regularity conditions hold."],"supporting_citations":[{"why":"Establishes that the FDA outcome is the doctor-optimal stable allocation in the matching-with-contracts market, the result the proof invokes as its foundation.","marker":"Kamada and Kojima (2018)"},{"why":"Introduces the FDA algorithm whose outcome the paper proves equals DA with adapted capacities.","marker":"Kamada and Kojima (2015)"},{"why":"Supplies the matching-with-contracts framework, the definition of stable allocation, and doctor-optimality of the generalized DA.","marker":"Hatfield and Milgrom (2005)"},{"why":"Introduces the deferred acceptance algorithm, whose doctor-optimal stable matching is the benchmark.","marker":"Gale and Shapley (1962)"},{"why":"Defines weak stability, the fairness notion the DA-with-adapted-capacities matching satisfies.","marker":"Kamada and Kojima (2017)"},{"why":"Provides the responsive-with-capacity hospital preference model used throughout.","marker":"Roth (1985)"}],"fun_headline_variants":["Hospital quotas from FDA output make DA match FDA exactly","FDA as capacity design: set caps to its fill counts","Adapted caps turn DA into FDA, solving cap-induced inefficiency","Efficient caps exist: set them to FDA's fill counts","DA and FDA coincide when quotas match FDA's fill"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the two-phase hospital-side choice rule in Algorithm 2 satisfies substitutability (a rejected offer stays rejected when more offers arrive) and the law of aggregate demand (more offers never reduce the number of doctors chosen), because the proof's equivalence argument inherits doctor-optimality from a cited theorem that requires those two conditions but does not itself verify them for this specific algorithm.","fun_headline_variants_meta":{"raw":{"variants":["Hospital quotas from FDA output make DA match FDA exactly","FDA as capacity design: set caps to its fill counts","Adapted caps turn DA into FDA, solving cap-induced inefficiency","Efficient caps exist: set them to FDA's fill counts","DA and FDA coincide when quotas match FDA's fill"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001105,"raw_usage":{"total_tokens":4559,"prompt_tokens":852,"completion_tokens":3707,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":468,"completion_tokens_details":{"reasoning_tokens":3625}},"tokens_in":468,"tokens_out":3707,"duration_ms":24221,"temperature":1.0,"reasoning_tokens":3625,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:45:15.511727+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run doctor-proposing DA on a small market (for instance, the three-hospital, five-doctor example in the paper) with each hospital's capacity set to its FDA fill count; if any FDA order produces a DA outcome different from the FDA matching, Theorem 2 is false. Alternatively, directly test the hospital-side choice rule of Algorithm 2 for substitutability and the law of aggregate demand, since the proof inherits doctor-optimality from a cited theorem only under those conditions.","supporting_citations":[{"cited_title":"and Kojima, F","cited_arxiv_id":null,"evidence_quote":"Establishes that the FDA outcome is the doctor-optimal stable allocation in the matching-with-contracts market, the result the proof invokes as its foundation."},{"cited_title":"and Kojima, F","cited_arxiv_id":null,"evidence_quote":"Introduces the FDA algorithm whose outcome the paper proves equals DA with adapted capacities."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the matching-with-contracts framework, the definition of stable allocation, and doctor-optimality of the generalized DA."},{"cited_title":"and Shapley, L","cited_arxiv_id":null,"evidence_quote":"Introduces the deferred acceptance algorithm, whose doctor-optimal stable matching is the benchmark."},{"cited_title":"and Kojima, F","cited_arxiv_id":null,"evidence_quote":"Defines weak stability, the fairness notion the DA-with-adapted-capacities matching satisfies."}],"review_version":1}