{"id":"b5eafdd3-b703-4b08-87f7-aec56be625f0","arxiv_id":"1908.08786","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"In a two-period spatial competition model, a government that can buy winning plans later is indifferent between funding all plans upfront and funding none, so the paper's diversity recommendation is an overclaim.","lead":"This economics paper builds a game-theory model of how governments fund research when they do not yet know which projects will succeed. It argues that spreading money evenly across all fields is the right strategy, but the model actually leaves the government indifferent between spreading and concentrating.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The model's pricing rule makes the government's expected expenditure independent of the ex-ante adoption set, so the paper's conclusion that widespread funding is better than concentration does not follow from its own equations.","rationale":"The reader's weakest_assumption focuses on the unverified ex-post subgames for arbitrary adopted sets. That is a real gap, but it concerns the proof of equilibrium existence and could in principle be repaired. The concern raised here is more load-bearing: the paper's own pricing rule (ex-ante price equal to expected ex-post profit) makes the government exactly indifferent among all ex-ante adoption sets, including the empty set. The identity Σ_{i∈I} E[p_i^a] + E[1_{best∉I} gap] = E[gap] follows immediately from Eq. (7) and linearity, and it is verified numerically by the paper's own equilibrium prices. Consequently, the abstract's normative claim that widespread expenditure is 'better' than concentrated expenditure is not a theorem of the model; it is an overclaim. The off-path subgame issue identified by the reader is secondary: even a perfectly rigorous SPE construction with all plans adopted would not rescue the policy conclusion. Therefore the REJECT verdict is appropriate, but for a different primary reason than the reader's stated weakest assumption.","tokens_in":10839,"tokens_out":22189,"duration_ms":214465,"concrete_test":"Using the equilibrium locations z_i^*=(2i-1)/(2n), compute the government's expected total expenditure for three ex-ante adoption sets: I=∅, I=M, and I={odd indices}, with p_i^b as in Eq. (15) and ex-post prices as in Eq. (7). Show that all three equal Σ_i p_i^b. If they are equal, the indifference result is confirmed and the 'better than' conclusion fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Under the paper's pricing rule, p_i^b is set equal to E[p_i^a] with p_i^a defined by Eq. (7). For any realized t, at most one researcher has positive p_i^a(t): the plan closest to t, and its value is the gap between the best and second-best squared distances. Hence for any ex-ante adoption set I, expected total expenditure is Σ_{i∈I} E[p_i^a] + E[1_{best∉I}·gap(t)] = E[gap(t)], independent of I. In particular, buying every plan ex ante and buying no plan ex ante yield exactly the same expected government utility. The abstract's claim that 'equally widespread expenditure across all research fields is better than the selection of and concentration in some specific fields' is therefore contradicted by the model's own equations (7)-(15). Section 7's statement that it is 'optimal' to invest in all plans is at best weak optimality; the paper never establishes a strict gain. This is an internal inconsistency in the central claim, not merely a gap in the equilibrium proof. Even if the location-game deviation checks were repaired, the headline policy conclusion would remain unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a two-period procurement game between a government and n researchers on a Hotelling line. Researchers first choose locations (research fields), then set ex-ante prices; after the government learns its ideal point t, researchers can set ex-post prices and the government can buy additional plans. The authors claim that in a subgame-perfect equilibrium the government funds as many plans as possible ex ante, locations are evenly spaced at (2i-1)/(2n), and this outcome implies that spreading R&D funds evenly is better than concentrating on selected fields. They also derive an equilibrium number of plans when entry requires a fixed cost F.","tokens_in":11068,"tokens_out":17117,"duration_ms":150317,"significance":"If the conclusions were correct, the paper would offer a formal argument against selection-and-concentration policies in research funding and would contribute to the literature on optimal diversity of research portfolios. The model is transparent and yields closed-form results, and the first-order conditions for the symmetric location equilibrium are standard. However, the central normative claim is contradicted by the model's own pricing equations, and the proof of the location equilibrium contains incorrect deviation checks. As it stands, the paper does not establish its headline result.","major_comments":[{"comment":"The pricing rule makes expected total expenditure independent of the ex-ante adoption set. Since p_i^b is set equal to E[p_i^{a*}] and p_i^{a*}(t) is positive only for the plan closest to t, with value equal to the gap between the best and second-best squared distances, for any adoption set I the expected ex-ante cost is E[1_{b in I} gap(t)] and the expected ex-post cost is E[1_{b not in I} gap(t)]; the sum is E[gap(t)], independent of I. Consequently the government is indifferent between funding all plans, funding no plan, and every intermediate set. The abstract's statement that widespread expenditure is \"better than\" selection and concentration, and Section 7's assertion that it is \"optimal\" to invest in all plans, are therefore not supported by the model; at best the full-adoption plan is one among many weakly optimal outcomes.","section":"Section 4 and Section 7, Eqs. (7)-(15)"},{"comment":"The displayed expressions for a deviating researcher's profit are incorrect. For an interior l who deviates to z'_l outside (z_{l-1}, z_{l+1}), the profit formula using the original neighbors (2i-1)/(2n) and (2i+1)/(2n) is not valid because the ordering of plans changes; the relevant neighbors become different plans. For the endpoint check, the expression uses 1/(2n) as the neighbor location, but the equilibrium neighbor is z_2 = 3/(2n). These mistakes leave the claimed sufficiency of the first-order conditions unproved, so the location equilibrium z_i^* = (2i-1)/(2n) is not established.","section":"Section 5, deviation checks after Eq. (20)"},{"comment":"The ex-post price subgame is only solved under the assumption that the government already holds the second-closest plan, and the extension to arbitrary ex-ante adoption sets is asserted rather than proved. The statement that \"every z_j != z_{i+1} is strictly far away from z_i\" is false in general (e.g., z_{i-1} can be closer to t than z_{i+1} when t is in the left half of i's Voronoi cell). Since off-path histories are essential for a subgame-perfect equilibrium, the formal claim of subgame perfection is incomplete.","section":"Section 3, Eq. (6)"},{"comment":"The claim that \"there are two subgame perfect equilibria\" is inaccurate. Given the indifference established above, any ex-ante adoption set I can be part of an equilibrium, since researchers earn their expected ex-post profit either through ex-ante funding or through ex-post sales. The paper should either prove uniqueness of the two described equilibria or qualify the claim.","section":"Section 7"}],"minor_comments":[{"comment":"The notation z_{i+1} in Eq. (6) denotes the second-closest plan, while elsewhere it denotes the right neighbor; this ambiguity makes the price formula difficult to interpret.","section":"Section 3, Eq. (6)"},{"comment":"The line \"2 <= i <= n - 1\" appears after the brace without a clear connection to the displayed formulas; this is a typesetting issue.","section":"Section 4, Eq. (15)"},{"comment":"There are several typos, including \"filed\" for \"field\" and \"the chosen plan\" repeated; the manuscript should be carefully proofread.","section":"Section 7"},{"comment":"The paper should define the government's tie-breaking rule when it is indifferent between adopting and not adopting a plan; the current text simply asserts that the government adopts all plans or a part of them.","section":"Section 4"},{"comment":"The reference list is somewhat dated; more recent work on research funding, procurement, and innovation could provide useful context.","section":"Introduction"}],"recommendation":"reject","confidential_remarks":"The central result is unsupported by the model's equations, and the equilibrium existence proof contains incorrect displayed expressions. The paper would need a fundamental reworking to be publishable; I recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper extends the authors' 2015 spatial-competition model to a two-period government procurement setting: researchers choose locations, offer ex-ante prices, and the government can pre-purchase a subset before its ideal point is realized, then buy the realized best plan ex post. The option-pricing idea (setting the ex-ante price equal to expected ex-post profit) is a genuine extension of the cited literature, and the derivation of the equal-spacing location equilibrium and the 1/n^3, 2/n^3 payoff formula is straightforward and internally consistent. That part is worth reading.\n\nThe problems come when the paper reaches for policy conclusions. The ex-post price subgame is solved only under histories where the government already holds the second-closest plan; arbitrary adoption sets are never characterized, so the subgame-perfect equilibrium is incomplete. The location-game deviation checks contain undefined symbols (z_k in Eq. 16) and the displayed inequalities are not actually derived; at best they are sketches. More seriously, the central claim that widespread expenditure is 'better' than concentration does not follow from the model's own equations. Because p_i^b equals E[p_i^a], the government's expected total expenditure is E[gap(t)] whether it adopts all plans ex ante or none. The model yields indifference, not a strict gain. Strict losses arise only for partial adoption sets that sometimes exclude the two plans closest to the realized ideal point, but that is not the 'equally widespread' conclusion the abstract advertises. Sections 1 and 7 overstate what the model proves.\n\nI would not desk-reject this. The modeling idea is legitimate, the spatial-competition core is standard and cleanly presented, and the policy question is relevant. A conscientious referee could write a productive report: fix the deviation checks, solve the ex-post subgame for general adoption sets, and either prove a strict welfare advantage for full adoption or reframe the result as an indifference/neutrality theorem. As it stands, I would not cite it because the headline result is not established. But I would send it out for peer review, expecting major revision or, more likely, rejection. It could also be a useful reading-group example of a model overclaiming its conclusions.","headline":"A clean spatial-competition extension whose policy punchline is undone by the model's own indifference condition.","tokens_in":11582,"tokens_out":7934,"would_cite":false,"duration_ms":79836,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"In the model's subgame-perfect equilibrium, the government purchases every active research plan and the chosen plans are evenly spaced, so R&D funding is symmetric and unbiased.","keywords":["government R&D funding","research plan selection","spatial competition","subgame perfect equilibrium","research diversity","fixed entry cost","science and technology policy","quadratic location model"],"falsifier":"Solve the ex-post price subgame for an off-path history, such as three researchers at $0.1,0.5,0.9$ with the government already holding the middle plan when the ideal point is $0.2$. The equal-spacing equilibrium requires the left researcher's price to be $(0.2-0.5)^2-(0.2-0.1)^2$; if the actual equilibrium price differs, the ex-ante price schedule and the derived equilibrium locations collapse.","tokens_in":10595,"feed_emoji":"🔬","tokens_out":12774,"duration_ms":120981,"temperature":0.7,"pith_summary":"This paper asks how a government should buy research plans when it does not yet know which plan will prove most valuable. The authors model researchers who choose a location on the line of possible fields and set prices, and a government that buys plans before and after its ideal field is revealed. In the subgame-perfect equilibrium the government buys as many active plans as possible, and the plans are spaced evenly across the field space, giving endpoint researchers $1/n^3$ and interior researchers $2/n^3$. The result matters because it gives a formal argument for broad, diversified R&D spending rather than concentrated bets on a few fields, and it ties the number of fundable projects to the fixed cost of entry.","feed_headline":"Government should spread R&D funds evenly, not pick winners","feed_subtitle":"A two-stage game shows the government equilibrium is to buy every plan, spaced evenly over research space.","key_machinery":"The central mechanism is a one-dimensional location game on $[0,1]$ with quadratic loss. Each researcher picks a location; after the government's ideal point $t$ is revealed, the closest plan charges an ex-post price equal to the difference in squared distances to the next-closest plan, $p_i^a = (t-z_{i+1})^2 - (t-z_i)^2$. Expected ex-post profits determine the ex-ante price the government is willing to pay, and profit maximization over locations yields the equal-spacing condition $z_i^*=(2i-1)/(2n)$. A zero-profit entry condition then fixes the number of plans from the fixed cost $F$, giving the diversity result.","core_discovery":"The paper demonstrates that, in a two-period model in which researchers set locations and prices and the government buys plans before and after its ideal plan is revealed, the subgame-perfect equilibrium has the government purchasing all active plans whose ex ante price equals the researcher's expected ex post profit. Equilibrium locations are $z_i^* = (2i-1)/(2n)$ on $[0,1]$: the plans are equally spaced, the two end researchers earn $1/n^3$, and every interior researcher earns $2/n^3$. With a positive fixed cost $F$ of entering, the equilibrium number of operating researchers is the largest integer not exceeding $(1/F)^{1/3}$, with the usual adjustment when that number is an integer. From this the authors conclude that an equal spread of government expenditure across all research fields is better than selection and concentration in specific fields.","pith_inferences":["If the government's ideal point is drawn from a non-uniform distribution, the equal-spacing result should tilt toward high-probability fields; the model's proportionality is tied to the uniform-expectation calculation, so the diversity conclusion likely needs a density-weighted version.","The same option-value mechanism could apply whenever a buyer purchases a portfolio of variants before discovering its favorite, such as a firm prototyping several designs or a fund backing several start-ups; the model only studies a government and a line of fields.","A testable empirical consequence, which the paper does not run, is that portfolios of R&D grants spread evenly across fields should yield higher expected payoff per public dollar than portfolios concentrated on a few fields, holding fixed costs and uncertainty constant."],"forward_implications":["In the full-adoption equilibrium, the government takes every active plan whose ex-ante price equals its expected ex-post profit, so spreading funds over all fields is an equilibrium behavior, not a subsidy.","Because equilibrium locations are $z_i^*=(2i-1)/(2n)$, research plans are distributed proportionally across the field space, and no interior field is favored; the endpoint fields earn less than interior fields.","The number of supported plans falls with the fixed entry cost $F$ at approximately a cube-root rate, so cutting administrative and start-up costs increases research diversity.","Compared with a policy of concentrating funds on preselected fields, broad ex-ante adoption lowers the government's expected distance loss, since every extra plan is an option on the unknown ideal point."],"supporting_citations":[{"why":"Supplies the two-period game structure and the quadratic-loss utility that the government-purchase model extends.","marker":"[13]"},{"why":"Provides the zero-profit entry condition used to derive the optimal number of research plans from the fixed cost $F$.","marker":"[20]"},{"why":"Introduces the quadratic transportation-cost specification that lets the location subgame have a pure-strategy equilibrium.","marker":"[4]"}],"fun_headline_variants":["Even R&D funding beats picking winners","Spread research money evenly across fields","Diversify government research spending","Fund research plans evenly, not selectively","Equally spaced research plans win"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that in every subgame the price a researcher can charge after the ideal point is realized is governed by competition with the adjacent plan priced at zero; if that benchmark changes with different sets of already-owned plans, the ex-ante prices and the evenly spaced locations no longer follow.","fun_headline_variants_meta":{"raw":{"variants":["Even R&D funding beats picking winners","Spread research money evenly across fields","Diversify government research spending","Fund research plans evenly, not selectively","Equally spaced research plans win"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001167,"raw_usage":{"total_tokens":4769,"prompt_tokens":828,"completion_tokens":3941,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":444,"completion_tokens_details":{"reasoning_tokens":3883}},"tokens_in":444,"tokens_out":3941,"duration_ms":34579,"temperature":1.0,"reasoning_tokens":3883,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:24:14.190449+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve the ex-post price subgame for an off-path history, such as three researchers at $0.1,0.5,0.9$ with the government already holding the middle plan when the ideal point is $0.2$. The equal-spacing equilibrium requires the left researcher's price to be $(0.2-0.5)^2-(0.2-0.1)^2$; if the actual equilibrium price differs, the ex-ante price schedule and the derived equilibrium locations collapse.","supporting_citations":[{"cited_title":"and Nakagawa, K., (2015) ”Early Competition on Discoun t Tickets,” Journal of Transport Economics and Policy, 49(2), 219- 235","cited_arxiv_id":null,"evidence_quote":"Supplies the two-period game structure and the quadratic-loss utility that the government-purchase model extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the zero-profit entry condition used to derive the optimal number of research plans from the fixed cost $F$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the quadratic transportation-cost specification that lets the location subgame have a pure-strategy equilibrium."}],"review_version":1}