{"id":"0979ad70-fbff-4c30-96c5-17f6dec5bc27","arxiv_id":"2501.11427","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A look-back option framework for bond illiquidity is extended to defaultable coupon bonds and used to estimate a 23-27 bps liquidity spread for an unquoted Republic of Italy bond.","lead":"This paper extends an option-based model to estimate the liquidity discount on bonds that trade rarely, adding credit risk and coupon payments to a known look-back option framework. It applies the model to an unquoted Italian government bond and finds a liquidity spread of 23 to 27 basis points.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (12) is an assumed equality, and since the unknown gamma enters the illiquid option payoff, solving it makes the estimated liquidity spread self-referential; the 23–27 bps result is conditional on this unvalidated identification.","rationale":"The reader's CONDITIONAL verdict is appropriate, and my concern is the same load-bearing one. The paper is a coherent extension of the Longstaff and Koziol-Sauerbier framework, and it is honest about the assumption in Eq. (12) and about the difficulty of separating liquidity from credit-curve shifts (footnote 8). Those caveats support a conditional verdict rather than a rejection. However, the central quantitative claim—that gamma is the theoretical liquidity spread—rests entirely on the second equality in Eq. (12) and its coupon-bond analogue Eq. (15). Because gamma appears inside the option payoff used to define it, the estimation routine is close to self-referential: it finds the gamma that makes an assumed identity hold. The market application does not break the circularity, since the probing frequency is calibrated to match the very spreads the model is supposed to explain. The proposed ground-truth simulation would settle whether the option-ratio relation recovers a known liquidity discount; until that is done, the 23–27 bps figure should be read as conditional, not as a validated estimate. My verdict remains UNCHANGED because the reader already marked it CONDITIONAL and my concern is the same one.","tokens_in":15093,"tokens_out":11070,"duration_ms":127566,"concrete_test":"Construct a ground-truth simulation: pick \\gamma_0=25 bps, simulate the G2++ short rate and CIR credit spread used in §3, and define the illiquid zero-coupon bond price as \\bar P(t,T)e^{-\\gamma_0(T-t)}. Compute O_i(\\gamma) and O_c on a common set of at least 10^6 Monte Carlo paths, with the 'continuous' grid at hourly frequency and the illiquid grid at the frequencies of §2.3, and apply the paper's Newton equation to estimate \\gamma. Repeat for 1y, 5y, and 10y maturities, and for a 3.5y coupon bond version to test Eq. (15). If the recovered \\gamma is not within Monte Carlo error of \\gamma_0, the identification in Eq. (12)/(15) is biased; report the bias in bps.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the second equality in Eq. (12). In the illiquid option value O_i, the payoff already contains the unknown gamma through \\tilde P(t,T)=e^{-\\gamma(T-t)}\\bar P(t,T), while the liquid value O_c uses \\gamma=0. Equating the ratio O_i/O_c to e^{-\\gamma T} is not a derived no-arbitrage relation; it is a nonlinear equation that the Newton routine in §2.3 solves by construction, so the gamma reported by the model is the number that makes the assumed identity true. The text itself calls the second equality 'the underlying assumption' and concedes the real discount is only a fraction c·\\gamma, so the estimate is an upper-bound parameter, not a measured spread. For coupon bonds, Eq. (15) posits the same proportionality between option-value ratios and bond-price ratios, but the coupon option payoff in Eq. (14) includes coupons and recovery terms, and no argument is given for why that ratio should equal VB(0,T)/VB_{\\gamma=0}(0,T). The Section 4 application does not resolve this: Table 10 calibrates the probing frequency to reproduce the yield-curve-based market spread, so the 23–27 bps for the unquoted bond is an interpolation from a fitted frequency distribution rather than an out-of-sample verification. If Eq. (12)/(15) is not a genuine equality, the estimated gamma is not the theoretical liquidity spread.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the option-theoretic approach to bond liquidity pricing (Longstaff 1995; Koziol and Sauerbier 2003) to defaultable coupon bonds. Illiquidity is captured by a look-back option whose payoff is evaluated under two trading schedules: continuous (liquid) and discrete (illiquid). The liquidity spread gamma is defined via the ratio of the option values under these schedules (Eq. 12), and the model is solved numerically using G2++ short-rate and CIR credit-spread processes. Numerical experiments in Section 3 study the effect of default events and credit-spread volatility. Section 4 calibrates the model's probing frequency to yield-curve-based liquidity spreads of quoted Italian sovereign bonds and uses the resulting distribution to estimate a 23-27 bps liquidity spread for an unquoted 3.5-year Republic of Italy bond.","tokens_in":15453,"tokens_out":10818,"duration_ms":89169,"significance":"If the identification in Eq. (12) held, the paper would provide a tractable framework for estimating liquidity spreads in defaultable bond markets, extending the prior literature by adding credit risk and coupons. The numerical experiments reveal interesting interactions between default events, credit-spread volatility, and the estimated spread, e.g., the non-monotone behavior for short maturities at low ratings. The paper is transparent about the underlying assumption and the upper-bound interpretation, and the market application demonstrates a practical calibration pipeline. However, the central equality between option-value ratios and bond-price ratios is assumed rather than derived, and the empirical calibration in Section 4 makes the reported fits in Table 10 close by construction. The contribution is therefore best read as a model-based upper-bound measure with an explicit calibration procedure, rather than as a validated estimator of market liquidity spreads.","major_comments":[{"comment":"The second equality in Eq. (12) is stated as the 'underlying assumption', but the illiquid option value O(0,T) on the right-hand side already depends on gamma through \\tilde P(t,T)=e^{-\\gamma(T-t)}\\bar P(t,T). Consequently, Eq. (12) does not derive gamma from observable quantities; it is a nonlinear equation that the Newton routine in Section 2.3 solves by construction, so the reported gamma is the value that forces the assumed identity to hold. Since the paper also concedes that the true discount is a fraction c·gamma, gamma is an upper-bound parameter rather than a measured market spread. Please clarify the status of gamma and discuss what empirical evidence would be needed to support the proportionality assumption.","section":"2.2, Eq. (12)"},{"comment":"Equation (15) assumes that the ratio of look-back option values for coupon bonds equals VB(0,T)/VB_{\\gamma=0}(0,T), but the coupon option payoff in Eq. (14) includes accumulated coupon payments, survival indicators, and recovery terms. No argument is provided for why such a ratio should equal the bond-price ratio; without it, the extension to coupon bonds in Section 4 lacks a theoretical foundation. This is the same identification problem as in Eq. (12) and should be either derived or explicitly labeled as an additional assumption.","section":"2.3, Eq. (15)"},{"comment":"The calibration in Section 4 solves for the probing frequency that reproduces the market liquidity spread, defined as the yield difference between each illiquid bond and the fitted liquid yield curve. The close agreement between Model Liquidity Spread and Market Liquidity Spread in Table 10 is therefore by construction and does not validate the model. The final estimate for the unquoted bond (Table 3) is an interpolation from the probing-frequency distribution fitted to these same quoted bonds, not an out-of-sample verification. Please provide an out-of-sample check (e.g., holding out some quoted bonds) or explicitly frame the procedure as a fitting exercise.","section":"4, Table 10"},{"comment":"Footnote 8 acknowledges that the yield difference between an illiquid bond and the liquid yield curve is hard to disentangle from credit-curve shifts. Since this yield difference is the target for calibrating the probing frequency, the estimated spread for the unquoted bond is conditional on this identification. If part of the difference is due to credit migration or term-structure misalignment, the calibrated probing frequencies and the resulting 23-27 bps range will be biased. The paper should discuss the direction and magnitude of this potential bias.","section":"4, footnote 8"},{"comment":"Even if Eq. (12) is accepted as an assumption, the model's practical output is not the actual liquidity spread but an upper bound c·gamma, with c unspecified. The paper does not calibrate c or provide bounds for its magnitude; the market application reports gamma values (23-27 bps) without adjusting for c, so it is unclear what economic quantity these numbers represent. The paper should clarify the relationship between the reported figures and the actual discount that would be applied to an illiquid bond.","section":"2.2 and 4"}],"minor_comments":[{"comment":"There are numerous typos throughout, such as 'liqudity' in the Introduction, 'substitutibg' before Eq. (14), and 'yeas' in Section 4; the manuscript needs careful proofreading.","section":"General"},{"comment":"In Eq. (12), the left-hand side is written as e^{-\\gamma(T-t)} while the right-hand side is evaluated at time 0; if t denotes the valuation date, the exponent should be e^{-\\gamma T} or the right-hand side should be time-t quantities.","section":"2.2, Eq. (12)"},{"comment":"The caption of Figure 1 is confusing: it refers to red as case 2 and green as case 3 on the left, while the text discusses default events (case 1 vs case 3) on the left; please clarify the panel labels.","section":"3, Figure 1"},{"comment":"The description of the probing-frequency distribution in bucket G is incomplete: the paper states that outliers #64 and #65 are removed and then uses 14, 17, and 19 days, but it does not report the mean and standard deviation of the distribution or the remaining sample size.","section":"4, Table 10"},{"comment":"In Table 10, for a few bonds the Model Liquidity Spread differs from the Market Liquidity Spread by 1-2 bps even after calibration; this should be explained, e.g., by the discreteness of the probing-frequency grid.","section":"4, Table 10"}],"recommendation":"major_revision","confidential_remarks":"The paper may be more appropriately framed as proposing an upper-bound liquidity measure rather than a true estimator. The empirical section is essentially a fitting exercise, and the paper would be strengthened by an out-of-sample validation or by comparing the estimated spreads with alternative liquidity proxies. The role of the parameter c should also be clarified, as the reported 23-27 bps are not the actual discount applied to the bond."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a competent, honest extension of the Longstaff/Koziol-Sauerbier look-back option framework to defaultable coupon bonds, with a useful numerical sensitivity study and a practical calibration workflow for unquoted bonds. The central identifying equation, however, is an assumption rather than a derivation, and the Section 4 empirical fit is partly constructed.\n\nWhat's genuinely new: adding credit risk (CIR spread plus default events) to the option payoff, and showing how default events raise liquidity spreads for short-maturity BB bonds while credit spread volatility can offset risk-free rate volatility. The Section 4 exercise—classifying Italian sovereign bonds by liquidity, calibrating a per-bond probing frequency, and using the resulting distribution to price an unquoted bond—is a sensible practitioner-oriented procedure.\n\nThe soft spots are real. Eq. (12) defines gamma via a ratio of look-back option values, but the second equality, tying that ratio to the bond's liquidity discount, is explicitly labeled as the underlying assumption. Gamma also appears inside the illiquid option payoff, so the Newton solve is finding the value that makes the assumption true. The authors acknowledge it's an upper bound and that the true discount is c·gamma, but c is never calibrated. The 23–27 bps for the Italian bond is thus a conditional upper-bound estimate, not a measured spread. In Section 4, the probing frequency is calibrated to reproduce the yield-curve-based market spread, so the close match in Table 10 is by construction; the few failures are attributed to non-uniform probing dates, which is another free parameter. No error bars, no out-of-sample test, no code or data.\n\nTo the authors' credit, none of this is hidden. They flag the disentangling problem in footnote 8 and the c-factor issue in the text. So the paper is a framework with a worked example, not a validation. The abstract's 'robust tool' overstates what is demonstrated.\n\nThis is worth a serious referee: the extension is natural, the numerics are careful, and the limitations are discussable. I'd push the authors to calibrate c, add an out-of-sample check (e.g., hold out some liquid bonds and see if model spreads match), and provide error bands on gamma. With those, it could be a solid applied piece.","headline":"Honest extension of the look-back option framework to defaultable coupon bonds, but the central identification is assumed and the empirical fit is partly by construction.","tokens_in":15957,"tokens_out":3662,"would_cite":true,"duration_ms":33773,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91G20","91G40","91G30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that the liquidity spread of a defaultable coupon bond can be estimated from the ratio of two look-back option values, one evaluated on discrete trading dates and one on continuous trading.","keywords":["liquidity spread","look-back option","defaultable bond","coupon bond","credit risk","probing frequency","Monte Carlo simulation","unquoted bonds"],"falsifier":"Run the model with a known $\\gamma$ and check whether the simulated ratio $O_{\\text{discrete}}/O_{\\text{continuous}}$ equals $e^{-\\gamma(T-t)}$; if it does not, the second equality of Eq. (12) is false, and the same test applied to the quoted illiquid bonds in Table 10 would reveal whether the market discount is a fraction of $\\gamma$.","tokens_in":14839,"feed_emoji":"📈","tokens_out":9656,"duration_ms":81039,"temperature":0.7,"pith_summary":"The paper claims that the liquidity spread of a defaultable coupon bond can be estimated by pricing liquidity as a look-back option: an ideal investor with perfect timing captures a payoff that depends on how often the bond can be sold. The central quantity is the ratio of the option's value when trading is restricted to discrete dates to its value when trading is continuous, and the model equates this ratio with the bond's liquidity discount factor $e^{-\\gamma(T-t)}$. This extends the equity-marketability framework [11] and the default-free zero-coupon version [12] by adding default events and credit-spread volatility, then solves the resulting implicit equation numerically. If the identification holds, the method produces a concrete upper-bound estimate for unquoted bonds; the Republic of Italy case yields a 23–27 basis point spread for a 3.5-year bond.","feed_headline":"Liquidity spread traced to look-back option ratio","feed_subtitle":"Method prices unquoted bonds by comparing continuous vs discrete trading dates.","key_machinery":"The central object is the look-back option $O(0,T)$, whose payoff is the maximum over allowed trading dates of the present value of the bond plus already-received coupons and recovery in default, with proceeds reinvested at the risk-free rate. Liquidity is priced through the ratio of $O$ evaluated on a discrete set of trading dates to $O$ evaluated on the liquid continuous market (approximated by hourly probing), and this ratio is set equal to the liquidity discount factor $e^{-\\gamma(T-t)}$. The numerical machinery is a G2++ short-rate model calibrated to swaptions, a CIR credit-spread model calibrated to the issuer's spread curve, Monte Carlo simulation with 10,000 paths, and a numerical root-finder that exploits the monotonicity of the option value in $\\gamma$.","core_discovery":"The central claim is that the loss of marketability of a defaultable coupon bond can be measured by the ratio of two look-back option values, $$$e^{{-\\gamma(T-t)}}$ = \\frac{O_{\\text{discrete}}(0,T)}{O_{\\text{continuous}}(0,T)} = \\frac{\\tilde P(0,T)}{\\bar P(0,T)},$$ where $\\gamma$ is the theoretical liquidity spread and $\\tilde P/\\bar P$ is the liquidity discount factor of the risky zero-coupon bond. The first equality is a definition; the second equality is the paper's underlying assumption, and the paper states that the true discount is a fraction $c\\gamma$ of this upper bound. For coupon bonds, the option payoff includes coupons already received and recovery in default, leading to the implicit equation (15), which is solved numerically. The paper argues that credit risk enters twice: credit-spread volatility changes the bond values feeding the option, and default events truncate the set of admissible probing dates, with the largest effect on short-maturity, lower-rated bonds.","pith_inferences":["Editorial extension: the identity in Eq. (12) can be tested directly on quotes of the same bond class by comparing model-implied $O_{\\text{discrete}}/O_{\\text{continuous}}$ with observed price discounts; a mismatch would separate the model's definitional part from its assumption.","Editorial extension: because the paper leaves the fraction $c$ undetermined, the 23–27 bps figures are best read as upper bounds; calibrating $c$ on liquid-versus-illiquid quoted pairs of the same issuer would convert the method into a point estimate.","Editorial extension: replacing the bucket-wise bid-ask/volume classification with a continuous liquidity score could absorb the negative-spread outliers (bonds #8 and #12) and the extreme probing-frequency outliers (#64 and #65) rather than discarding them."],"forward_implications":["Any coupon-bearing defaultable bond with a specified illiquid trading frequency gets a finite, positive liquidity spread, so unquoted bonds can be priced without a quoted market price.","Default events increase the estimated spread, especially at short maturities and lower ratings, so omitting credit risk understates the illiquidity cost for BB-rated issuers.","The estimated spread rises monotonically as the illiquid probing frequency increases, giving a calibration handle from the quoted spreads of illiquid bonds.","For the Republic of Italy bond, the calibrated probing frequencies of 14, 17, and 19 days produce liquidity spreads of 23, 24, and 27 basis points, respectively."],"supporting_citations":[{"why":"Supplies the original option-theoretic view of marketability as a look-back option, which the paper extends from equity to defaultable bonds.","marker":"[11]"},{"why":"Provides the default-free zero-coupon bond version of the look-back option framework and the upper-bound/c-factor discussion that this paper generalizes to coupon bonds with credit risk.","marker":"[12]"},{"why":"Gives an alternative closed-form pricing of illiquid corporate defaultable bonds; the paper distinguishes its liquidity-spread estimate from that 'sheer liquidity' approach.","marker":"[13]"},{"why":"Supplies the G2++ and CIR model definitions and calibration framework used in the numerical Monte Carlo analysis.","marker":"[24]"},{"why":"Provides the Svensson yield-curve fitting used to construct the liquid yield curve from selected liquid Italian bond prices.","marker":"[25]"}],"fun_headline_variants":["Liquidity spread as look-back option ratio","Option model prices unquoted bond liquidity","Discrete vs continuous dates set bond spread","Look-back options value defaultable bond marketability","Unquoted bonds priced via look-back option ratio"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole estimate rests on the assumption that the ratio of look-back option values under discrete vs continuous trading equals the bond's liquidity discount, and, in the market application, that the yield gap between an illiquid bond and the fitted liquid curve is entirely a liquidity premium.","fun_headline_variants_meta":{"raw":{"variants":["Liquidity spread as look-back option ratio","Option model prices unquoted bond liquidity","Discrete vs continuous dates set bond spread","Look-back options value defaultable bond marketability","Unquoted bonds priced via look-back option ratio"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000182,"raw_usage":{"total_tokens":1275,"prompt_tokens":876,"completion_tokens":399,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":343}},"tokens_in":492,"tokens_out":399,"duration_ms":4607,"temperature":1.0,"reasoning_tokens":343,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:17:31.877630+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the model with a known $\\gamma$ and check whether the simulated ratio $O_{\\text{discrete}}/O_{\\text{continuous}}$ equals $e^{-\\gamma(T-t)}$; if it does not, the second equality of Eq. (12) is false, and the same test applied to the quoted illiquid bonds in Table 10 would reveal whether the market discount is a fraction of $\\gamma$.","supporting_citations":[{"cited_title":"How Much Can Marketability Affect Security Values?","cited_arxiv_id":null,"evidence_quote":"Supplies the original option-theoretic view of marketability as a look-back option, which the paper extends from equity to defaultable bonds."},{"cited_title":"Valuation of Bond Illiquidity: An Option- theoretical Approach","cited_arxiv_id":null,"evidence_quote":"Provides the default-free zero-coupon bond version of the look-back option framework and the upper-bound/c-factor discussion that this paper generalizes to coupon bonds with credit risk."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives an alternative closed-form pricing of illiquid corporate defaultable bonds; the paper distinguishes its liquidity-spread estimate from that 'sheer liquidity' approach."},{"cited_title":"isbn:978-3-540-22149-4","cited_arxiv_id":null,"evidence_quote":"Supplies the G2++ and CIR model definitions and calibration framework used in the numerical Monte Carlo analysis."},{"cited_title":"Estimating and Interpreting Forward Interest Rates: Sweden 1992- 1994","cited_arxiv_id":null,"evidence_quote":"Provides the Svensson yield-curve fitting used to construct the liquid yield curve from selected liquid Italian bond prices."}],"review_version":1}