{"id":"360e8f4e-4657-4929-a134-8cfd22912e60","arxiv_id":"1908.08777","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"In a continuous-time Kyle model, a non-fiduciary market maker who adds a fee proportional to order flow can earn expected profits exceeding those of a perfectly informed insider, while increasing price volatility.","lead":"A mathematical model of stock trading with a market maker who charges order-flow dependent fees shows the market maker can earn expected profits as large as or larger than a perfectly informed insider. The same model offers a mechanism for why speculative prices may be more volatile than fundamentals would predict.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The profit comparison is cap-selection rather than equilibrium: JM grows in κ, Appendix 2 gives only corner κ, and Table 2 shows the insider still wins at rv*=1.08.","rationale":"The reader's weakest-assumption analysis identified the exogenous price-setting rule and the regulator cap as load-bearing; my stress-test agrees, and the concrete evidence is Table 2 plus the unboundedness of JM in κ. The manuscript has real strengths: it provides a coherent continuous-time extension of Kyle, an explicit integral equation for the insider's trading intensity, a closed-form Kyle limit when κ=0, and an internally consistent accounting for the three agents' profits. The central weakness is that the headline profit comparison is not shown to be robust to the regulator's cap or to the market maker's own optimization. Since the reader already returned CONDITIONAL, my concern does not change the verdict, but it sharpens the condition: the profit-dominance claim should be presented as parameter- and cap-dependent, with a sensitivity analysis, rather than as a general result.","tokens_in":24551,"tokens_out":25303,"duration_ms":248978,"concrete_test":"Recompute the Table 2 cross-over with the same base-case parameters but with regulator caps rv*=1.08 and rv*=1.10, and iterate Eq. (4.3) until numerical convergence rather than the two rounds noted in Section 7.2. If pM(T,κ*) < pI(T,κ*) at either cap, then the headline 'market maker can beat a perfectly informed insider' is an artifact of the arbitrarily chosen 15% cap rather than a robust model implication.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that a modest fee lets the market maker beat a perfectly informed insider is not an equilibrium result of the model. In Section 3.3, Eq. (3.21) gives the market maker's expected profit as increasing in κ, and the paper states it grows without bound as κ→∞. An optimizing non-fiduciary market maker would therefore set κ at the regulator's upper bound. Appendix 2 confirms this: Theorem 11.2 shows that any Nash equilibrium has κ*=0 or κ*=K, never an interior 'modest' value. Thus the headline comparison is selected by an exogenous volatility cap, not derived from the strategic interaction. Table 2 makes the selection explicit: at rv*=1.08 the insider still earns pI=0.100 versus pM=0.060, while at rv*=1.15 the market maker dominates (pM=0.087 versus pI=0.055). The claimed crossover near κ=0.06 is therefore a consequence of choosing a 15% relative-volatility cap, not of the theory itself. If the regulator imposed a 10% cap, the central conclusion would fail. This does not invalidate the model as a conditional contribution, but it means the abstract's general statement overstates what has been established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper extends Kyle's (1985) continuous-time insider trading model by replacing the fiduciary market maker with a non-fiduciary market maker who sets prices according to pt = E[v|F_y] + (T−t)κ y_t. The insider's trading intensity β_t is characterized in Theorem 1 as a solution of an integral equation, and the paper derives closed-form expressions for the expected profits of the insider, the market maker, and the noise traders, as well as for the order-flow variance, relative volatility, and a price-informativeness measure. Numerical sections show that, as the fee parameter κ increases, the market maker's expected profit rises and eventually exceeds the insider's expected profit; the authors connect κ to a regulator-imposed cap on relative volatility rv*. The paper claims that setting a modest order-flow fee allows the market maker to earn profits of the same order as, or larger than, a perfectly informed insider, and that the model explains excess price volatility relative to fundamentals.","tokens_in":24824,"tokens_out":9016,"duration_ms":81917,"significance":"If the central claims were established, the paper would be a useful contribution to the literature on market power and intermediation in continuous-time insider trading: it provides explicit formulas for profits and price dynamics under a non-fiduciary price rule, and it frames a regulator's volatility cap as a constraint on the market maker's fee. The derivations in Section 3 and Appendix 1 are nontrivial, and the paper is transparent about several limitations, including the absence of an existence proof for the integral equation and the non-strategic treatment of the market maker. However, the headline profit comparison is currently presented as an unconditional result even though it is a consequence of an exogenously chosen regulatory threshold, and the technical optimality claim in Theorem 1 is not fully verified. With appropriate restatement and additional analysis, the continuous-time extension could be publishable.","major_comments":[{"comment":"Theorem 1 is derived as a first-order necessary condition, but the paper does not prove existence or uniqueness of a solution to the integral equation (4.3), nor does it show that the resulting β is a global maximizer of the insider's problem (4.2). The iterative procedure described in Section 7.2 is numerical, and no convergence argument is supplied, so the subsequent profit comparisons cannot be read as verified equilibrium outcomes. In addition, the displayed algebra is not internally consistent: the k^2 term appears with a plus sign in Eq. (10.3) and with a negative sign in Eq. (10.4), and the exponential factors in (10.4) do not obviously match the definition of V(t) used in the final expression (4.3). These points need to be fixed before the main theorem can be checked.","section":"Section 4 and Appendix 1, Eq. (4.3)"},{"comment":"Theorem 11.2 states that any Nash equilibrium has κ*=0 or κ*=K, never an interior 'modest' value. Combined with Eq. (3.21), which shows that JM is increasing in κ and grows without bound as κ→∞, an optimizing market maker would always choose the regulator's upper bound. The crossover near κ=0.06 reported in Section 7.4 is therefore not an equilibrium prediction of the model; it is an exogenous selection of the volatility cap. Table 2 confirms this cap-selection property: at rv*=1.08 the insider still earns pI=0.100 versus pM=0.060, while at rv*=1.15 the market maker earns pM=0.087 versus pI=0.055. The abstract and conclusion present the profit comparison unconditionally; this overstates what the model actually establishes.","section":"Appendix 2 and Section 3.3, Eq. (3.21)"},{"comment":"The claim that the model 'indicates why speculative prices are more volatile than predicted by fundamentals' is not supported by the analysis as written. The model compares the market price p_t to the fair conditional price m_t; it does not compare prices to any fundamental process such as dividends or earnings. The numerical finding in Table 1 that var(p_t) exceeds var(m_t) is a statement about the distortion caused by the fee schedule relative to a fiduciary benchmark, not about excess volatility relative to fundamentals. The conclusion in Section 9 should be reformulated as a conditional statement about fee-induced price distortion.","section":"Sections 6 and 9"}],"minor_comments":[{"comment":"The reference to 'Campbell and Schiller (1988)' should be 'Campbell and Shiller (1988)'.","section":"Section 1"},{"comment":"The phrase 'by by the above observation' contains a duplicated 'by'.","section":"Section 3.2"},{"comment":"In the second exponential in Eq. (3.23), 'ke' appears where 'kr' is intended.","section":"Section 3.3, Eq. (3.23)"},{"comment":"The phrase 'Vi have' should be 'We have'.","section":"Section 7.1"},{"comment":"The reference to 'Theorem 3.2' appears to be an error; the intended reference is likely Theorem 11.2 in Appendix 2.","section":"Appendix 3"},{"comment":"Table 2 reports profit values at t=9, while the text describes rv* as a supremum over t; the exact time convention used for the reported crossover should be stated explicitly.","section":"Section 5, Table 2"}],"recommendation":"major_revision","confidential_remarks":"This is a working-paper-style draft with a high number of typographical issues and a central claim that is currently overstated. The technical core may be salvageable, but the authors should either endogenize the market maker's choice of κ or explicitly frame all profit comparisons as conditional on an exogenously imposed volatility cap. They should also fix the algebraic inconsistency in Appendix 1 and either provide existence and uniqueness results for Eq. (4.3) or state clearly that the numerical solutions are only candidates. The paper's scope is appropriate for a quantitative finance journal, and the continuous-time extension is nontrivial, but the present version requires substantial revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick read on Aase-Øksendal. The genuinely new thing is the continuous-time version of their one-period non-fiduciary market maker model: the market maker sets price as filtered fair value plus k_t y_t, with k_t=(T-t)κ, and the insider's variational problem yields the integral equation (4.3) for trading intensity. That derivation is real work and its structure checks out; for κ=0 it reduces to Kyle/Back, so this is a proper extension, not a repackaging. The profit expressions in Section 3 are consistent with the price rule and the orthogonality properties, and the paper is honest about the role of the regulator.\n\nWhere it gets soft: the headline claim that a modest fee lets the market maker beat a perfectly informed insider is not an equilibrium result. Eq (3.21) shows JM increasing in κ, and the paper itself notes it grows without bound as κ→∞. Appendix 2, Theorem 11.2, confirms the Nash equilibrium κ* is only 0 or K, never an interior 'modest' value. The crossover in Table 2 is produced by the regulator's rv* = 1.15 threshold; at rv* = 1.08 the insider still wins. So the abstract overstates what is established. What is actually established is conditional: if the regulator permits roughly 15% extra relative volatility, then the fee-based market maker can dominate the insider. That is a real result, but it is not a robust 'modest fee' outcome.\n\nAlso, the paper does not prove existence or uniqueness of a solution to the integral equation (4.3). The numerical iterations look plausible, but a referee should ask about fixed-point properties. Minor but load-bearing: the assumption p_T = v kills the insider's terminal cross term; it is standard and stated, but without it the profit comparison shifts.\n\nThe reader's stress-test note holds up. I would push back only on calling this circular in a data-fitting sense: there are no data fitted, and the profit expressions follow from explicit assumptions. The issue is that the central comparison is selected by the regulatory cap, not that the paper sneaks in its conclusion. Self-citation is also not a problem here; the one-period predecessor is cited and the extension is real.\n\nBottom line: this is a serious theory paper worth refereeing. It should be accepted, if at all, conditional on rewriting the abstract to state the cap-dependence and on addressing the integral equation's existence/uniqueness. For a reading group it is a good discussion piece; cite it if you work on intermediary profits, but do not cite the headline as a general result.","headline":"A genuine but conditional continuous-time extension of Kyle with a fee-charging market maker; the headline profit comparison is selected by the regulator's volatility cap rather than derived as an equilibrium outcome, yet the integral-equation derivation deserves refereeing.","tokens_in":25339,"tokens_out":1915,"would_cite":true,"duration_ms":22369,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60G35","62M20","93E10","94Axx"],"pacs":[],"model":"deepseek-v4-flash","headline":"A non-fiduciary market maker who sets price by adding a fee proportional to cumulative order flow can earn expected profits comparable to or larger than a perfectly informed insider, while making prices more volatile than fundamentals…","keywords":["insider trading","market making","order-flow fee","asymmetric information","price volatility","non-fiduciary market maker","continuous-time auction","linear filtering"],"falsifier":"Let a second market maker enter and undercut the fee by any positive amount; if the only subgame-perfect outcome is $\\kappa=0$, then the profit-dominance result depends entirely on the assumed absence of fee competition and would not survive entry.","tokens_in":24351,"feed_emoji":"📈","tokens_out":9087,"duration_ms":84429,"temperature":0.7,"pith_summary":"This paper argues that the standard assumption of a fiduciary, zero-profit market maker is the wrong baseline. In a continuous-time version of the classic single-insider auction model, the market maker sets the price equal to the fair conditional expectation plus a fee proportional to cumulative order flow. With that fee schedule, the market maker's expected profit can match or exceed the profit of a perfectly informed insider, even when the fee is modest, and the resulting price is more volatile than fundamentals alone would produce. The paper derives the insider's optimal trading intensity from an integral equation and shows numerically how a regulator can cap the fee by monitoring relative price volatility.","feed_headline":"A modest order-flow fee lets market makers out-earn the insider","feed_subtitle":"In a continuous-time trading model, the fee also explains why speculative prices swing more than fundamentals imply.","key_machinery":"The central object is the fee schedule $k_t=(T-t)\\kappa$ embedded in the market maker's price rule $p_t=E[\\tilde v \\mid \\mathcal{F}^y_t]+k_t y_t$. Because $k_t$ is positive and linear in cumulative order flow, the price is pushed above the fair value after buying pressure and below it after selling pressure; since the order flow $y_t$ mean-reverts around zero, the market maker systematically buys low and sells high. The insider's trading intensity $\\beta_t$, defined by $dx_t=(\\tilde v-p_t)\\beta_t\\,dt$, is the other half of the mechanism: the insider's optimal $\\beta_t$ solves the integral equation in Theorem 1, and it declines as $\\kappa$ grows. The paper's main analytical tool is the linear-filtering representation of the conditional expectation $m_t$, which gives the variance $V(t)=E[y_t^2]$ and the mean-square error $\\gamma_t(\\beta)$ in closed form; these feed directly into both profit functions.","core_discovery":"The paper's central claim is that replacing the fiduciary price rule with a non-fiduciary one changes the economics of the single-insider auction model without destroying equilibrium. The market maker announces the price rule $p_t=E[\\tilde v \\mid \\mathcal{F}^y_t]+k_t y_t$, with $k_t=(T-t)\\kappa$; the insider still trades with intensity $\\beta_t$ against his information advantage $\\tilde v-p_t$. The expected profits are $J^M(\\kappa,\\beta)=w^M_0+\\int_0^T(k_s^2 V(s)\\beta_s+\\kappa V(s))\\,ds$ for the market maker and $J^I(\\kappa,\\beta)=w^I_0+\\int_0^T\\beta_s(\\gamma_s(\\beta)+k_s^2V(s))\\,ds$ for the insider, so the fee adds a positive, order-flow-driven income for the market maker. The insider optimally trades more softly when $\\kappa$ rises, but the market maker's fee income still dominates beyond a moderate threshold, which the paper's baseline calibration places around $\\kappa=0.06$. The paper also derives the price variance $\\operatorname{var}(p_t)=\\operatorname{var}(m_t)+k_t^2 V(t)+2k_t\\operatorname{cov}(m_t,y_t)$, which exceeds the fundamental-based variance, and constructs a regulator-observable relative-volatility measure $rv(t,\\kappa)$ that ties $\\kappa$ to a cap on price distortion.","pith_inferences":["If the market maker's fee-setting power were subject to competition from a second market maker who could undercut $\\kappa$ by a tiny amount, the profit-dominance result would likely shrink or vanish; the paper assumes a single market maker with regulatory protection, and testing that entry margin is a natural extension.","The linear fee $k_t y_t$ is what makes the profit formulas tractable; a nonlinear fee such as a fee on absolute order flow would change the mean-reversion argument, and the ranking of market-maker versus insider profits might not survive.","The model's volatility prediction is testable at the level of price impact: if the fee component is real, the coefficient of signed order flow on subsequent price changes should be positive and increasing in the market maker's market power, and the residual variance should exceed the variance implied by a fiduciary price.","Because the fee is paid by both insider and noise traders, the model implies that noise traders bear the cost of intermediation; in an overlapping-generations or entry version, this could generate persistent wealth transfer to intermediaries, connecting to the paper's motivation."],"forward_implications":["A modest order-flow fee can make the market maker's expected profit exceed that of a perfectly informed insider; in the paper's baseline calibration the crossover occurs around $\\kappa=0.06$.","Speculative price volatility contains a fee-driven component $k_t^2 V(t)+2k_t\\operatorname{cov}(m_t,y_t)$, so prices can be more volatile than fundamentals imply even without changes in dividend information.","The insider trades less aggressively when the fee increases, but intensity rises again near the horizon as $k_t\\to 0$, so the fee mainly shifts the timing and size of informed trading.","A regulator can control the distortion by monitoring the relative volatility $rv(t,\\kappa)$; a 15% cap in the baseline example keeps the market open for $\\kappa$ up to about 0.07, and a 21% cap would allow $\\kappa=0.09$.","Price informativeness $\\iota(t,\\kappa)$ falls as $\\kappa$ rises, so fee-driven price distortion makes prices less informative about fundamentals at every horizon."],"supporting_citations":[{"why":"The classic continuous-auction insider-trading baseline with a fiduciary, zero-profit market maker; this paper modifies its price rule.","marker":"[19]"},{"why":"One-period model with a non-fiduciary market maker, where the profit-dominance of the fee-charging market maker was first shown; the present paper extends it to continuous time.","marker":"[1]"},{"why":"Companion continuous-time insider-trading analysis that supplies the variational method and the identity $E[(\\tilde v-m_t)(\\tilde v-m_s)]=\\gamma_t(\\beta)$ used in the profit computations.","marker":"[2]"},{"why":"Filter-theory derivation of the insider's equilibrium trading intensity for the zero-fee case, used as the trial solution and benchmark $\\beta^0_t$ in the iterative solution of the integral equation.","marker":"[4]"},{"why":"Continuous-time insider-trading equilibrium with finite-variation strategies; provides the setting into which the fee rule is inserted.","marker":"[8]"},{"why":"Innovations theorem underlying the linear filter that delivers the conditional expectation $m_t$ and the variance $V(t)$.","marker":"[7]"}],"fun_headline_variants":["Market maker fees can beat insider profits in Kyle model","Non-fiduciary market maker outearns insider with order-flow fee","Order-flow fee flips profits: market maker beats insider","How a fee on order flow gives market makers the edge","Fee-based pricing makes market maker more profitable than insider"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole result depends on the assumed price rule $p_t=E[\\tilde v \\mid \\mathcal{F}^y_t]+(T-t)\\kappa y_t$, with a single scalar $\\kappa$ chosen by the market maker under a regulatory volatility cap; if market makers could undercut the fee, if the fee were nonlinear in order flow, or if the regulator's cap were tighter, the stated profit comparison and volatility result could fail.","fun_headline_variants_meta":{"raw":{"variants":["Market maker fees can beat insider profits in Kyle model","Non-fiduciary market maker outearns insider with order-flow fee","Order-flow fee flips profits: market maker beats insider","How a fee on order flow gives market makers the edge","Fee-based pricing makes market maker more profitable than insider"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000209,"raw_usage":{"total_tokens":1400,"prompt_tokens":934,"completion_tokens":466,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":550,"completion_tokens_details":{"reasoning_tokens":384}},"tokens_in":550,"tokens_out":466,"duration_ms":5055,"temperature":1.0,"reasoning_tokens":384,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:29:44.921390+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Let a second market maker enter and undercut the fee by any positive amount; if the only subgame-perfect outcome is $\\kappa=0$, then the profit-dominance result depends entirely on the assumed absence of fee competition and would not survive entry.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The classic continuous-auction insider-trading baseline with a fiduciary, zero-profit market maker; this paper modifies its price rule."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"One-period model with a non-fiduciary market maker, where the profit-dominance of the fee-charging market maker was first shown; the present paper extends it to continuous time."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Companion continuous-time insider-trading analysis that supplies the variational method and the identity $E[(\\tilde v-m_t)(\\tilde v-m_s)]=\\gamma_t(\\beta)$ used in the profit computations."},{"cited_title":"and Øksendal, B","cited_arxiv_id":null,"evidence_quote":"Filter-theory derivation of the insider's equilibrium trading intensity for the zero-fee case, used as the trial solution and benchmark $\\beta^0_t$ in the iterative solution of the integral equation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Continuous-time insider-trading equilibrium with finite-variation strategies; provides the setting into which the fee rule is inserted."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Innovations theorem underlying the linear filter that delivers the conditional expectation $m_t$ and the variance $V(t)$."}],"review_version":1}