{"id":"326c6c53-e4e8-48ac-a914-410fb117d61b","arxiv_id":"2512.08177","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Under worst-case approval constraints, the robustly optimal procurement mechanism is Baron-Myerson with a quantity floor, and price regulation beats quantity regulation only when demand uncertainty is small at the top.","lead":"A buyer who distrusts her demand and cost estimates first picks mechanisms that maximize the worst-case outcome, then chooses among those using her best-guess model. The result is a procurement rule that buys more from high-cost sellers and less from mid-cost sellers than the classical Bayesian optimum, and a comparison of price vs quantity regulation under model uncertainty.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pointwise-smallest demand in §2.1 is the load-bearing primitive; without it, Lemma 1's reduction to a single V and the q_l-floor characterization do not follow.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing condition: existence of a pointwise smallest inverse demand. I agree. I looked for internal inconsistencies (e.g., the DWL sign in Lemma 4, feasibility of demand-contingent transfers in price regulation) but the supplied plain-text rendering of integral limits is ambiguous, and the proofs are internally consistent in the simple linear examples I checked. The pointwise-smallest assumption is explicitly stated, so the paper is not incoherent; however, it is the condition on which the worst-case short-list results rest, and it fails for natural families such as crossing demand curves. The proposed concrete test would show whether the theorem extends to that case or whether the results should be read as applying only to ordered demand sets. Because the assumption is transparent and standard in max-min design, this does not change the reader's ACCEPT verdict.","tokens_in":25705,"tokens_out":30509,"duration_ms":276198,"concrete_test":"Take a two-element demand family with crossing inverse demands, e.g. P1(q)=2-q and P2(q)=1.5-0.5q on q in [0,2], with V1,V2 their integrals and a conjectured uniform F* on [0,1]. Compute the two-step optimum directly: discretize weakly decreasing quantity schedules q, evaluate G(M)=min(W under V1, W under V2) over point-mass cost distributions, find the short list, then maximize conjectured welfare over it. Compare the argmax with Proposition 1/2's prediction (floor q_l and conditions (7)-(8)). If the selected schedule differs, or q_l is not well-defined, the pointwise-smallest assumption is consequential for the central claims.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2.1 assumes a pointwise smallest inverse demand P in the admissible set P. This is not a mere regularity condition: it makes V the worst-case demand for every quantity and every type, and it is used in Lemma 1 to collapse the guarantee to G(M)=inf_theta{V(q(theta))-theta q(theta)-u(theta)}, in Lemma 2 to obtain the short-list constraint (4), in the definition of q_l and G*, and in the Lemma 4 equivalence behind Proposition 1's conditions (7)-(8). If admissible demands cross, no pointwise smallest P exists. The lower envelope of the family is generally not concave and need not belong to P, so the worst-case demand depends on q(theta) and theta. The max-min problem then has no reduction to a single V, and the Baron-Myerson-with-floor characterization, including the floor q_l itself, has no stated analogue. The online supplement establishes existence of robustly optimal mechanisms but does not relax this maintained assumption. This is a domain restriction rather than an internal contradiction, but it is exactly the point where the central claim is least secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies robust procurement design when a buyer faces a monopolistic seller with privately known cost θ ∈ [θ, θ] and is uncertain about both the value of the good and the cost technology. The buyer has a conjectured model (V*, F*) but, fearing misspecification, first restricts attention to mechanisms maximizing the worst-case welfare over an admissible set V × F, and then selects among those the mechanism maximizing expected welfare under (V*, F*). The main results characterize robustly optimal quantity mechanisms: Proposition 1 gives necessary and sufficient conditions under which the Baron–Myerson schedule with a quantity floor q_l is robustly optimal; Proposition 2 describes the departures when it is not, with high-cost types receiving at least q_l, intermediate types receiving reduced quantities relative to Baron–Myerson, and low-cost types following the Bayesian benchmark. For market settings, Proposition 3 shows every robustly optimal price regulation uses the price schedule p(θ) = min{z*(θ), θ}, and Proposition 4 compares quantity and price regulation, giving primitive conditions for each to dominate. The analysis is supplemented by an online appendix covering existence, undominatedness, and extensions. Proofs are collected in appendices.","tokens_in":25992,"tokens_out":20887,"duration_ms":171149,"significance":"If the results hold, the paper makes a substantial contribution to robust mechanism design and regulation. It provides a tractable two-step criterion that nests the standard Bayesian benchmark and yields sharp, falsifiable characterizations: efficiency at both ends of the cost distribution, downward adjustments for intermediate costs, and a non-trivial ranking of price versus quantity regulation. The proof structure is careful and the main propositions are stated with explicit conditions. The paper also credits and distinguishes itself from recent related work (Dworczak–Pavan, Guo–Shmaya, Kambhampati, Bergemann et al.). The main limitation is the maintained assumption of a pointwise smallest inverse demand in the admissible set, which is not discussed as a restriction; this assumption is load-bearing for almost every result. A second concern is a technical step in the proof of Proposition 2 that is not correct as written, though it appears fixable. With appropriate revision, the contribution would be a solid addition to the literature.","major_comments":[{"comment":"The assumption that there exists a pointwise smallest inverse demand P in P is load-bearing far beyond a regularity condition. It is used to identify V as the worst-case demand at every quantity and type, to reduce the guarantee in Lemma 1 to G(M)=inf_θ{V(q(θ))-θq(θ)-u(θ)}, to derive the short-list constraint (4) in Lemma 2, to define q_l and G*, and to prove the equivalence in Lemma 4 (Appendix D) behind Conditions (7)-(8). For natural families of crossing demand curves, no pointwise smallest P exists; the lower envelope of the family need not be concave or belong to V, so the worst-case demand varies with θ and q(θ). The manuscript does not flag this as a scope restriction, nor does it provide primitives under which such a P exists (e.g., vertical shifts of a base demand). Please add a thorough discussion of this assumption, its role in each step, and, if possible, an extension or at l","section":"Section 2.1; Lemmas 1, 2, 4; Proposition 1"},{"comment":"The proof of Lemma 7 asserts that, for the constructed schedule q̃=min{q*, qOPT}, W(θ,q̃)>W(θ,qOPT) for θ in the relevant range. This inequality is not generally true. When q_BM(θ)<qOPT(θ)<D(θ), reducing output to q_BM(θ) lowers the static surplus V(q)-θq, and this loss can exceed the rent saving ∫_θ^θ(qOPT-q̃). A concrete linear-demand example with q_BM=2, qOPT=4, D(θ)=5.5 gives W(θ,q̃)<W(θ,qOPT). The conclusion that q̃ belongs to the short list can still be reached by using Lemma 4: q̃ satisfies the integral condition (22) for all θ, W(θ,q̃)=G* (since q̃(θ)=q_l), and W(θ,q̃)≥G* (since q̃≤qOPT and qOPT∈M_SL). Please rework the proof to avoid the incorrect pointwise comparison.","section":"Appendix B, proof of Lemma 7"}],"minor_comments":[{"comment":"The symbol V is used both for the set of value functions and for a generic element; this is occasionally confusing. Distinguish the set, e.g., as V_set or a calligraphic symbol.","section":"Section 2.1"},{"comment":"The deadweight-loss term is written as ∫_θ^{P(q(θ))}(D(y)-q(θ))dy. When q(θ)>D(θ), the upper limit is below the lower limit; the orientation is implicit but should be clarified (or the integral defined with absolute value orientation) so that DWL is manifestly nonnegative.","section":"Equation (5)"},{"comment":"In condition (11), the notation D is used both as a demand realization and as the set of demand functions; this is standard but could be clarified, e.g., by writing 'for every D ∈ D \\ {D, D*}'.","section":"Section 5.1, Definition 4"},{"comment":"The integration variable p in (14)-(16) is a price, while p(θ) is also the regulated price function. The notation is understandable but may be streamlined to avoid confusion.","section":"Corollary 6"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious contribution and the central claims are likely correct, but the revision needs to address two issues: (1) the unqualified reliance on a pointwise smallest inverse demand, which is a substantive restriction on the ambiguity set and should be discussed prominently; (2) a genuinely incorrect step in the proof of Lemma 7, which is fixable by a different argument but must be corrected. Neither issue seems to require rejection, but they are substantial enough that a major revision is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my take on Mishra-Patil-Pavan. The paper gives a genuine new twist on robust mechanism design: a two-step lexicographic criterion where the designer first maximizes the worst-case guarantee over all plausible models, then picks among those the mechanism maximizing expected welfare under a conjectured model. That order is not the standard min-max regret or max-min over beliefs, and it has teeth. The main results are real: Baron-Myerson with a quantity floor is characterized (Proposition 1), and when the floor is not robustly optimal, the general structure in Proposition 2 is a clean combination of floor at the top, Baron-Myerson at the bottom, and reduced quantities in between. The price-regulation result (Proposition 3) is also new, and the price-vs-quantity comparison in Proposition 4 gives a concrete policy-relevant ranking that goes against the usual Bayesian dominance of price regulation. The proofs look careful, and the appendix is detailed; I did not find an internal contradiction.\n\nThe soft spot is the one the stress-test flagged: Section 2.1 assumes a pointwise smallest inverse demand P in the admissible set. That is doing real work. Lemma 1 collapses the guarantee to a single worst-case V, and the entire quantity-floor and DWL decomposition in Lemma 4 relies on that. If admissible demand curves cross, the lower envelope need not be a member of the set, and there is no obvious analogue of q_l or the floor. The paper states the assumption clearly, so it is not a hidden error, but it is a domain restriction that is much stronger than a standard regularity condition. The online supplement apparently does not relax it. Anyone applying the results should know the characterization is conditional on this 'smallest demand' existence.\n\nAlso, the admissible set for cost distributions is all cdfs on the interval, which is broad; that is fine because it is the main text, and the supplement considers more general sets. The paper is careful about the conjectured-model selection step, and the self-citations to Dworczak-Pavan and Mishra-Patil are appropriate.\n\nBottom line: this is a serious paper, clearly thought through, with new results worth having. The main caveat is the pointwise-smallest-demand assumption; I would not sink the paper over it, but it needs to be stated as a limitation and ideally addressed with some examples or a partial relaxation. I would send it to a good theory referee and would cite it. Bring it to a reading group if you want a lively discussion about worst-case design.","headline":"Solid, careful robust-procurement paper with real new results; the pointwise-smallest-demand assumption is the load-bearing primitive and deserves an explicit caveat.","tokens_in":26403,"tokens_out":2037,"would_cite":true,"duration_ms":20901,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91B03","91B26"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper shows that robustly optimal procurement under demand and cost uncertainty is the Bayesian schedule with a worst-case quantity floor, and that when that floor mechanism fails, robustly optimal contracts distort intermediate costs d","keywords":["robust mechanism design","procurement","model uncertainty","worst-case guarantees","price regulation","quantity regulation","screening contracts","incentive compatibility"],"falsifier":"Let the admissible set contain two inverse demand curves that cross, so that no pointwise smallest P exists; then q_l and G* are undefined and Conditions (7)-(8) cannot even be stated. A more targeted falsifier: for a regular conjectured distribution with D*(θ̄)>D_(θ̄), compute the floor schedule; if it violates (7)-(8) yet an alternative schedule outside the paper's characterization yields higher conjectured welfare among worst-case-optimal mechanisms, the characterization is wrong.","tokens_in":25606,"feed_emoji":"🏛️","tokens_out":5793,"duration_ms":54988,"temperature":0.7,"pith_summary":"This paper asks what a procurement contract should look like when the buyer cannot fully trust her model of demand and seller costs. It proposes a two-step criterion: first keep only mechanisms that maximize the worst-case payoff over all plausible models, then among those choose the one with the highest expected payoff under the conjectured model. The central finding is that under this criterion the classical Bayesian-optimal quantity schedule survives almost unchanged, except for a floor: every type must supply at least the efficient quantity that would be optimal in the worst case (lowest demand, highest cost). When demand uncertainty is also present, the same floor mechanism is robustly optimal exactly when two integral conditions hold; otherwise robustness requires trimming the schedule for intermediate-cost sellers. In a downstream-market setting, the robustly optimal price contract is a simple cap on the Bayesian markup. These results imply that quantity regulation can outperform price regulation, contrary to the usual Bayesian ranking.","feed_headline":"Robust procurement: Bayesian schedule plus a worst-case floor","feed_subtitle":"Uncertainty about demand and cost pushes contracts to a worst-case floor; price regulation need not dominate quantity regulation.","key_machinery":"The engine is the short-list characterization: a mechanism attains the maximal welfare guarantee iff its highest-cost type receives zero rent and its ex-post welfare under the lowest demand and zero high-cost rent never falls below G*, the maximum possible guarantee. This turns the robustness requirement into a family of integral inequalities on the quantity schedule, from which the floor q_l follows as a necessary feature of every worst-case-optimal mechanism. The same characterization is then used to identify the unique worst-case-optimal price schedule and to compare quantity and price regulation.","core_discovery":"The paper's central claim is a full characterization of robustly optimal procurement mechanisms. If the floor schedule—the Bayesian-optimal quantity schedule truncated from below at q_l, the efficient quantity under the lowest plausible demand and highest cost—satisfies two integral conditions, it is the unique robustly optimal quantity schedule. If it fails, every robustly optimal mechanism still procures exactly q_l from all types above a threshold θ*, follows the Bayesian schedule below a lower threshold θ_m, and reduces output for the intermediate range [θ_m, θ*), with strict reductions somewhere in that range. For price regulation the analogous claim is sharper: every robustly optimal p","pith_inferences":["If the admissible demand family lacks a pointwise smallest inverse demand, the construction of q_l and G* breaks down; a natural extension would replace the pointwise lower bound with a partial order or an aggregate lower envelope, and the characterization would need reworking.","The floor logic suggests a testable prediction for procurement practice: contracts designed for approval under model uncertainty should exhibit a compressed quantity range, with high-cost suppliers guaranteed a minimum quantity close to the worst-case efficient level.","The comparison of price versus quantity regulation could be extended by allowing the regulator to mix both instruments or use more flexible ex-post contingent transfers; the guarantee-equivalence result suggests such mixtures would not raise the worst-case payoff but could improve expected payoff."],"forward_implications":["When only costs are uncertain, the floor schedule is always robustly optimal and restores efficiency at both the lowest and highest cost types.","When demand is also uncertain, robustness pushes output up for high-cost types (to q_l) and down for intermediate-cost types, compressing the menu of quantities relative to the Bayesian benchmark.","Any robustly optimal price regulation must use the price schedule p(θ)=min{z*(θ), θ̄}, independent of the conjectured demand and admissible demand set.","If the floor schedule is robustly optimal and the conjectured demand at the highest cost exceeds the lowest demand there, quantity regulation dominates price regulation; if the floor fails and those two quantities coincide, price regulation strictly dominates.","Both regulatory forms achieve the same maximal welfare guarantee, so the choice between them is made under the conjectured model."],"fun_headline_variants":["Worst-case floor reshapes robust procurement","Bayesian schedule with worst-case floor: unique robust design","Truncate at q_l: robust procurement's floor schedule","Price vs quantity regulation: robust procurement insights","Two integral conditions decide robust procurement"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The existence of a pointwise smallest admissible demand curve—one below all others at every quantity—is what defines the floor quantity and the maximal guarantee; if the plausible demand curves cross so no such lower envelope exists, the whole characterization has no starting point.","fun_headline_variants_meta":{"raw":{"variants":["Worst-case floor reshapes robust procurement","Bayesian schedule with worst-case floor: unique robust design","Truncate at q_l: robust procurement's floor schedule","Price vs quantity regulation: robust procurement insights","Two integral conditions decide robust procurement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000237,"raw_usage":{"total_tokens":1285,"prompt_tokens":628,"completion_tokens":657,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":372,"completion_tokens_details":{"reasoning_tokens":586}},"tokens_in":372,"tokens_out":657,"duration_ms":6616,"temperature":1.0,"reasoning_tokens":586,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T17:44:01.227158+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Let the admissible set contain two inverse demand curves that cross, so that no pointwise smallest P exists; then q_l and G* are undefined and Conditions (7)-(8) cannot even be stated. A more targeted falsifier: for a regular conjectured distribution with D*(θ̄)>D_(θ̄), compute the floor schedule; if it violates (7)-(8) yet an alternative schedule outside the paper's characterization yields higher conjectured welfare among worst-case-optimal mechanisms, the characterization is wrong.","supporting_citations":[],"review_version":1}