{"id":"e9717328-94ab-43c9-9c79-0e8fdbfab32c","arxiv_id":"2412.14956","paper_version":3,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"For time-changed Markov processes built from the undershooting of a subordinator, expected payoffs solve a coupled non-local equation; for α-stable subordinators this yields a fractional Black-Scholes equation and a renewal formula for seasoned options.","lead":"The paper proves existence and uniqueness for a class of coupled non-local equations, where one operator acts on time and space together, and uses the result to derive a fractional Black-Scholes equation for option prices when jumps depend on the previous waiting time. A reader in finance or physics might care because it gives a rigorous pricing equation for models with dependent trade durations and returns.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (6.35) fails the terminal condition and the K=0 martingale check, so the seasoned-price renewal formula in Theorem 6.11 is not the conditional expectation it claims to be.","rationale":"The reader identified reliance on the Meerschaert-Straka Markov embedding as the weakest assumption, but the more serious problem is internal: the central renewal formula (6.35) contradicts the definition of q⋆ at maturity and the martingale property of the price process. The terminal check is immediate from the formula and requires no external results: at t=T the integral in (6.35) is empty, the kernel tail has mass one, and the formula returns a discounted payoff (x e^{-H(T)/2}-K)_+ instead of (x-K)_+. The K=0 check is equally direct and shows the state shift in the integrand has the wrong sign and magnitude. The existence and uniqueness proof for q in (6.36) may survive, but the theorem's applied content—the renewal formula for the seasoned price—is not correct as stated. This is a substantive mathematical error in the main result, not a gap in a cited theorem. No formal verification or reproducible code is provided to offset the failure. The paper's core claim therefore should not be accepted in its current form.","tokens_in":67805,"tokens_out":60776,"duration_ms":459138,"concrete_test":"Perform the symbolic check: substitute K=0 into (6.35) and compare with q⋆(t,x,w)=x, which must hold because X is an eP-martingale. At t=T the formula collapses to (x e^{-(T-v-w)/2})_+, not x, unless T-v-w=0. Run this for v=0, w>0, and any α∈(1/2,1); the identity fails, settling the concern.","verdict_should_be":"REJECT","load_bearing_attack":"Set t=T in (6.35). For w>0, the integral over [w,w+T-t)=[w,w) is empty and K_w(R×[w,∞))=1, so the formula gives q⋆(T,x,w)=(x e^{-(T-v-w)/2}-K)_+. But by definition q⋆(T,x,w)=eE[(X(T)-K)_+ | X(T)=x,γ(T)=w]=(x-K)_+. Since T-v-w=H(T)>0 a.s., the two disagree. Equivalently, with K=0 the option value must equal the underlying x for all t,w, yet the right-hand side of (6.35) contains the extra factor e^{-(t-v-w)/2} in the no-jump term and e^{y+t-w+τ-v} in the jump term; integrating the Gaussian e^y gives e^{τ/2}, leaving e^{t-v-w+3τ/2}, which is not 1. The fault is in the final step of the proof of Theorem 6.11: after a jump of size (τ,y), the P-representation state is not log x+y+t-w+τ-v; the Girsanov tilt shifts the log-price by -τ/2 (or equivalently the kernel must be tilted by e^{-y/2-τ/8}). Thus the central pricing formula is not the conditional expectation under eP.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper introduces a class of fully non-local, space-time coupled evolution equations of the form φ(∂t−G)q=νφ(t)q(0,·), where φ is a Bernstein function and G is the generator of a Feller semigroup. The main analytical result (Theorem 5.2) gives existence and uniqueness of solutions via a stochastic representation: q(t,x)=E[u(X(t))] for X(t)=M(H(t)), the undershooting time-change of a Feller process M by a subordinator inverse. A maximum principle yields uniqueness. The paper then applies the framework to option pricing: for φ(λ)=λ^α, α∈(1/2,1), it claims (Theorem 6.11) that the zero-age call price q(t,x) solves the coupled non-local Black-Scholes equation (∂t−G)^α q = t^{-α}/Γ(1−α) q(0,x), and provides a renewal formula (6.35) for the seasoned price q⋆(t,x,w) with positive sojourn time w.","tokens_in":68067,"tokens_out":19525,"duration_ms":121082,"significance":"If Theorem 5.2 and the zero-age pricing equation (6.36) are correct, the paper makes a valuable contribution: it gives a rigorous stochastic representation and uniqueness theory for a class of non-local operators with joint time-space non-locality, and connects them to physically motivated CTRW limits. The maximum-principle-based uniqueness proof is a strength, as is the detailed verification of the regularity conditions in Sections 5 and 6.2. The option-pricing section is ambitious and provides explicit formulas that could be useful. However, the renewal formula (6.35) for seasoned prices is not a valid conditional expectation, as the terminal-condition and K=0 martingale checks show; this undermines the applied claim and requires a substantial correction.","major_comments":[{"comment":"The renewal formula fails the terminal condition. Setting t=T and w>0, the integral over [w,w+T−t)=[w,w) is empty and K_w(R×[w,∞))=1, so (6.35) gives q⋆(T,x,w)=(x e^{-(T−v−w)/2}−K)_+. By definition, q⋆(T,x,w)=eE[(X(T)−K)_+ | X(T)=x, γ(T)=w]=(x−K)_+. Since H(T)=T−v−w>0 a.s. for the α-stable subordinator, the two expressions disagree. Equivalently, for K=0 the call price must equal the underlying x for all t,w by the eP-martingale property of X (Proposition 6.3), but the right-hand side of (6.35) contains the extra factor e^{-(t−v−w)/2} in the no-jump term, and the jump term does not compensate. The formula is therefore not the conditional expectation it claims to be.","section":"Section 6.2, Theorem 6.11, Eq. (6.35)"},{"comment":"The step eE[(X(T)−K)_+ | X(t), γ(t)] = E[(e^{Xe(T)−(T−v−γ(T))/2}−K)_+ | X(t), γ(t)] is not justified. Proposition 6.6 (Cameron-Martin formula) gives equality of the unconditional laws of (Xe+H/2,γ) under eP and (Xe,γ) under P; it does not give equality of conditional laws given X(t),γ(t), because the event {X(t)=x, γ(t)=w} under eP corresponds under the transformed representation to {e^{Xe(t)+H(t)/2−H(t)}=x} rather than {Xe(t)=log x}. Consequently, the subsequent use of the P-renewal equation for the semigroup Q_t with the payoff g_{T,v} does not yield the eP-conditional expectation; the correct renewal equation would require the eP jump kernel and an unshifted no-jump payoff. This is the root cause of the failure described in the previous comment.","section":"Proof of Theorem 6.11, final paragraph"}],"minor_comments":[{"comment":"The restriction α∈(1/2,1) for the option-pricing theorem is not mentioned; the abstract states the theory applies generally, while Theorem 6.11 holds only for this range.","section":"Abstract"},{"comment":"The formula φX(t)(z)=1F1(α, t/2) is incomplete as written; the hypergeometric parameters and argument should be specified accurately.","section":"Remark 6.7"},{"comment":"The notation for the Radon-Nikodym derivative omits the constant e^{x/2} in the displayed definition (deP_T/dP|_{N_T}=Z(T)), which may confuse readers; the factor is introduced a few lines earlier.","section":"Section 6.1, definition of eP"},{"comment":"The proof is only a reference to [45, Theorems 3.2 and 4.1] with a brief remark on the translation operator; since the setting is not identical, a more self-contained argument would help verify the Markov embedding.","section":"Theorem 6.1"},{"comment":"The phrase 'for tehcnical reasons' contains a typo and should read 'for technical reasons'.","section":"Section 6.2, line before Remark 6.7"}],"recommendation":"major_revision","confidential_remarks":"To the editor: The paper contains a substantial contribution in its general theory (Sections 1–5), and the zero-age pricing equation (6.36) may well be correct. However, the renewal formula (6.35) for seasoned prices is demonstrably incorrect, as shown by the terminal-condition and K=0 checks. This is not a minor gap; it invalidates the main applied theorem as stated. I recommend major revision rather than rejection because the error is localized to the derivation of the renewal formula and could in principle be fixed by reworking the eP-renewal equation. The authors should also address the heavy reliance on [45] in Theorem 6.1."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The first half of this paper is genuinely good. Theorem 5.2 gives existence and uniqueness for a class of coupled non-local equations via the undershooting of a subordinator, with a clean stochastic representation and a maximum principle. The proofs are detailed and the regularity estimates are handled carefully; the α>1/2 restriction is stated explicitly in Remark 6.16, even if the abstract omits it. I would be comfortable citing Theorem 5.2 in my own work on time-changed Feller processes.\n\nThe second half, though, has a load-bearing flaw. Formula (6.35) in Theorem 6.11 fails the terminal condition: set t=T and w<T. The integral over [w,w+T-t) is empty, and the formula gives q⋆(T,x,w)=(x e^{-(T-v-w)/2}-K)_+, which is not (x-K)_+. Equivalently, for K=0 the option value must equal the underlying x for all t,w because X is a martingale under eP; the right-hand side of (6.35) instead contains extra exponential factors that do not cancel. This is not a typo — it is a consequence of the proof. In the derivation of Theorem 6.11, the authors take the renewal formula from Meerschaert–Straka, which applies to the P-semigroup of (Xe,γ), and then substitute the Cameron-Martin transformed payoff into that same formula. But the conditional expectation defining q⋆ is under eP, not P; Girsanov changes the transition kernel and the state variable. The proof simply identifies eQ with Q, and that identification is false. The K=0 and t=T checks make the failure unmistakable.\n\nOther issues are minor: the abstract omits α>1/2, Remark 6.7 has an incomplete hypergeometric formula, and the Markov embedding argument relies on a modification of [45] rather than a full reproof. None of those matter next to the pricing section.\n\nWho is this for? People working on coupled non-local equations will get real value from the first half. The finance application, as written, is not reliable. I would send this to a serious referee because Theorem 5.2 is important and the flaw in Section 6 is contained and possibly repairable — but the referee should be told to check the Girsanov step carefully. In current form I would not accept it.","headline":"Theorem 5.2 is a solid contribution, but the seasoned-price formula in Theorem 6.11 fails elementary checks and is not the conditional expectation it claims to be.","tokens_in":68600,"tokens_out":5712,"would_cite":false,"duration_ms":49550,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60G53","60K50","60K15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The seasoned price of an intraday call option solves a coupled non-local Black-Scholes equation, and all positive-sojourn-time prices follow from it by a renewal formula.","keywords":["time-changed processes","Black-Scholes","undershooting","semi-Markov processes","subordinators","coupled non-local equations","anomalous diffusion","option pricing"],"falsifier":"Simulate the model with a concrete choice such as $\\alpha=0.75$, $T=1$, $K=1$, and compare Monte Carlo estimates of the conditional expectation $\\tilde{\\mathbb{E}}[(X(T)-K)_+\\mid X(t)=x,\\gamma(t)=w]$ against the right-hand side of the renewal formula evaluated with the numerically solved $q$ from (6.36); a systematic mismatch at any state would falsify the renewal representation.","tokens_in":2292,"feed_emoji":"📈","tokens_out":7566,"duration_ms":112160,"temperature":0.7,"pith_summary":"The paper establishes existence and uniqueness for a class of fully non-local Cauchy problems in which a single non-local operator acts jointly on time and space: $\\phi(\\partial_t-G)q(t,x)=\\nu_\\phi(t)q(0,x)$. The solution is shown to be a stochastic expectation: $q(t,x)=\\mathbb{E}[u(X(t))]$, where $X(t)=M(H(t))$ is a Feller process time-changed by the undershooting $H$ of an independent subordinator, so trajectories have trapping intervals of constancy followed by jumps coupled to the preceding wait. Uniqueness is proved via a positive maximum principle. In the financial application, when the subordinator is $\\alpha$-stable with $\\alpha\\in(1/2,1)$, the seasoned price of an intraday call option is shown to satisfy the coupled fractional Black-Scholes equation $(\\partial_t-G)^\\alpha q=t^{-\\alpha}q(0,x)/\\Gamma(1-\\alpha)$, with $G=\\frac12 x^2\\partial_x^2$, initial datum $(x-K)_+$, and a renewal formula expressing prices at positive sojourn times in terms of the renewal-state price. If correct, the paper turns a previously Fourier-and-simulation-based pricing problem into a PDE with a unique solution and an explicit renewal structure.","feed_headline":"Fractional Black-Scholes equation prices seasoned calls","feed_subtitle":"The paper proves a coupled non-local PDE gives the seasoned option price when returns and trade durations are dependent.","key_machinery":"The load-bearing object is the undershooting $H(t)=S(L(t)-)$: it is the last value of $S$ strictly before its inverse $L$ crosses level $t$, and it is the time variable seen by the parent process. Writing $X(t)=M(H(t))$ turns the pair $(M^\\phi,S)$ into a Markov-additive process, and adding the age variable $\\gamma(t)=t-S(0)-H(t)$ makes $(X,\\gamma)$ a time-homogeneous Markov process, a fact imported from the referenced Markov-embedding theory with a modified translation operator. The governing non-local operator is defined through the semigroup action of the generator $G$ as $-\\phi(\\partial_t-G)f(t,x)=\\int_0^\\infty(P_s f(t-s,\\cdot)1_{[0,t]}(s)-f(t,x))\\,\\nu_\\phi(ds)$; the Sonine pair of the special Bernstein function $\\phi$ supplies the density formulas that make the Laplace-transform proof and the renewal formula work. Uniqueness for the Cauchy problem is carried by a positive maximum principle, and for the $\\alpha$-stable case the undershooting has the explicit density $g_{H_0}(s;t)=s^{\\alpha-1}(t-s)^{-\\alpha}/(\\Gamma(\\alpha)\\Gamma(1-\\alpha))$.","core_discovery":"The central claim is that the governing equation of a Markov process time-changed by the undershooting of an independent subordinator is a coupled fully non-local equation, and that this equation has one solution. Let $M$ be a Feller process with generator $G$, let $S$ be a subordinator with Laplace exponent $\\phi\\in SB_0$ and $\\log(\\cdot)u_\\phi(\\cdot)\\in L^1[0,1]$, and set $X(t)=M(S(L(t)-))$ where $L$ is the inverse of $S$. Theorem 5.2 states that for every $u$ in the domain of $G$, the function $q(t,x)=\\mathbb{E}_{(x,0)}[u(X(t))]$ is the unique solution of $\\phi(\\partial_t-G)q(t,x)=\\nu_\\phi(t)q(0,x)$ with initial datum $q(0,\\cdot)=u$ and a local-uniform vanishing-at-infinity condition. In the option-pricing part, the authors specialise to $M(t)=e^{B(t)}$, so that $G=\\frac12 x^2\\partial_x^2$, and to $\\phi(\\lambda)=\\lambda^\\alpha$ for $\\alpha\\in(1/2,1)$; Theorem 6.11 then identifies the seasoned call price $q^\\star(t,x,w)=\\tilde{\\mathbb{E}}[(X(T)-K)_+\\mid X(t)=x,\\gamma(t)=w]$ with a renewal formula whose only input is the function $q(t,x)=q^\\star(T-t,x,0)$, the unique solution of the coupled non-local Black-Scholes equation (6.36). The price thus depends on the sojourn time $w$ only through the renewal kernel and the renewal-state solution.","pith_inferences":["For a general special subordinator, a renewal formula may still hold, but the explicit Beta density and the $\\alpha>1/2$ regularity threshold are special to stable subordinators; extending the pricing theorem would require a new argument.","At $\\alpha\\le 1/2$, the paper's own Remark 6.16 shows the solution is not twice differentiable at the strike, so a weak or viscosity formulation of the coupled equation would be needed there.","Because the seasoned price depends only on the renewal kernel and the renewal-state price, time-since-last-trade quotes could be used to infer the subordinator exponent from observed option prices."],"forward_implications":["The coupled non-local equation (5.1) becomes a governing equation for any Feller process time-changed by an undershooting, placing the trapping intervals of continuous-time random walk limits into an analytic partial differential equation framework.","For the dependent-returns-and-durations option model, valuation reduces to solving a single deterministic equation, (6.36); the renewal formula then yields the seasoned price for any positive sojourn time without simulation.","The seasoned price is a function of the current log-price and the time since the last trade, so it is determined by a finite state that is observable in tick data.","The time-fractional structure implies a correction to ordinary Black-Scholes prices, controlled by the trade-duration distribution and the stability index $\\alpha$."],"supporting_citations":[{"why":"Supplies the Markov embedding and transition semigroup of the age-augmented CTRW limit process, invoked with a modified translation operator in Theorem 6.1.","marker":"[45]"},{"why":"Defines the dependent-returns-and-durations model whose option price this paper characterises by a PDE, extending its Fourier-based pricing formula.","marker":"[34]"},{"why":"Provides the semi-Markov multiplicative CTRW limit and its governing equation, which the present theory unifies with the dependent model.","marker":"[58]"},{"why":"Introduces coupled CTRW limits and the undershooting construction, including the heuristic version of the coupled fractional heat equation.","marker":"[10]"},{"why":"Supplies the potential-measure and density formulas for the undershooting process, including the law used for the stable subordinator.","marker":"[12]"},{"why":"Provides the theory of Bernstein functions, special Bernstein functions, potential densities, and Sonine pairs used throughout the proofs.","marker":"[59]"},{"why":"Supplies the Levy-process and compensation-formula facts used in the Laplace-transform proof of Theorem 5.2.","marker":"[11]"},{"why":"Gives Phillips' theorem and subordination results used to identify the generator of the subordinated pair and the density of its range.","marker":"[54]"}],"fun_headline_variants":["Coupled nonlocal PDE prices options with sticky jumps","Time-change theory yields unique nonlocal Black-Scholes","Fractional Black-Scholes from Markov time-changes","Option pricing with jump-dependent trade durations","Nonlocal equations solve seasoned option pricing"],"cache_read_input_tokens":70784,"weakest_assumption_plain":"The formula for seasoned prices assumes that the pair 'log-price plus time since last trade' remains a Markov process after the change to the pricing measure; this is imported from a cited embedding theorem with modifications rather than fully reproved, so if that Markov property fails the renewal formula lacks rigorous support.","fun_headline_variants_meta":{"raw":{"variants":["Coupled nonlocal PDE prices options with sticky jumps","Time-change theory yields unique nonlocal Black-Scholes","Fractional Black-Scholes from Markov time-changes","Option pricing with jump-dependent trade durations","Nonlocal equations solve seasoned option pricing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000631,"raw_usage":{"total_tokens":2976,"prompt_tokens":1072,"completion_tokens":1904,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":688,"completion_tokens_details":{"reasoning_tokens":1834}},"tokens_in":688,"tokens_out":1904,"duration_ms":11638,"temperature":1.0,"reasoning_tokens":1834,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:46:25.226450+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the model with a concrete choice such as $\\alpha=0.75$, $T=1$, $K=1$, and compare Monte Carlo estimates of the conditional expectation $\\tilde{\\mathbb{E}}[(X(T)-K)_+\\mid X(t)=x,\\gamma(t)=w]$ against the right-hand side of the renewal formula evaluated with the numerically solved $q$ from (6.36); a systematic mismatch at any state would falsify the renewal representation.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Markov embedding and transition semigroup of the age-augmented CTRW limit process, invoked with a modified translation operator in Theorem 6.1."},{"cited_title":"Jacquier and L","cited_arxiv_id":null,"evidence_quote":"Defines the dependent-returns-and-durations model whose option price this paper characterises by a PDE, extending its Fourier-based pricing formula."},{"cited_title":"Scalas and B","cited_arxiv_id":null,"evidence_quote":"Provides the semi-Markov multiplicative CTRW limit and its governing equation, which the present theory unifies with the dependent model."},{"cited_title":"Becker-Kern, M","cited_arxiv_id":null,"evidence_quote":"Introduces coupled CTRW limits and the undershooting construction, including the heuristic version of the coupled fractional heat equation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the potential-measure and density formulas for the undershooting process, including the law used for the stable subordinator."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the theory of Bernstein functions, special Bernstein functions, potential densities, and Sonine pairs used throughout the proofs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Levy-process and compensation-formula facts used in the Laplace-transform proof of Theorem 5.2."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives Phillips' theorem and subordination results used to identify the generator of the subordinated pair and the density of its range."}],"review_version":1}