Paper: arXiv 2410.14839

Authors: Adel Javanmard, Jingwei Ji, Renyuan Xu

Abstract

We study the dynamic pricing problem faced by a broker seeking to learn prices for a large number of credit market securities, such as corporate bonds, government bonds, loans, and other credit-related securities. A major challenge in pricing these securities stems from their infrequent trading and the lack of transparency in over-the-counter (OTC) markets, which leads to insufficient data for individual pricing. Nevertheless, many securities share structural similarities that can be exploited. Moreover, brokers often place small “probing” orders to infer competitors’ pricing behavior. Leveraging these insights, we propose a multi-task dynamic pricing framework that leverages the shared structure across securities to enhance pricing accuracy. In the OTC market, a broker wins a quote by offering a more competitive price than rivals. The broker’s goal is to learn winning prices while minimizing expected regret against a clairvoyant benchmark. We model each security using a \(d\)-dimensional feature vector and assume a linear contextual model for the competitor’s pricing of the yield, with parameters unknown a priori. We propose the Two-Stage Multi-Task (TSMT) algorithm: first, an unregularized MLE over pooled data to obtain a coarse parameter estimate; second, a regularized MLE on individual securities to refine the parameters. We show that the TSMT achieves a regret bounded by $\tilde{O} ( δ_{\max} \sqrt{T M d} + M d ) $, outperforming both fully individual and fully pooled baselines, where \(M\) is the number of securities and \(δ_{\max}\) quantifies their heterogeneity. Finally, our empirical experiment on U.S. corporate bonds shows promising performance of our TSMT algorithm and supports the theoretical findings.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper presents a sophisticated mathematical framework for dynamic pricing in credit markets, addressing a challenging problem with a novel multi-task learning approach. The empirical validation on U.S. corporate bonds supports the theoretical findings, demonstrating strong rigor. The combination of high mathematical depth and solid empirical work places it in the ‘Holy Grail’ quadrant.

Research Flowchart

  flowchart TD
    A["Research Goal: Learn Winning Prices for Credit Market Securities"] --> B{"Challenge: Infrequent Trading & Data Scarcity"};
    B --> C["Approach: Multi-Task Dynamic Pricing Framework"];
    C --> D{"Methodology: Two-Stage Multi-Task (TSMT) Algorithm"};
    D --> E["Data/Inputs: M Securities, d-dim Feature Vectors, Contextual Info"];
    E --> F["Computational Process: Stage 1 (Pooled MLE), Stage 2 (Regularized Individual MLE)"];
    F --> G["Outcomes: Regret Õ ( δ_ T M d + M d ), Outperforms Baselines, Empirical Validation on Corporate Bonds"];

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes the Two-Stage Multi-Task (TSMT) algorithm for dynamic pricing of many credit securities, leveraging shared structure across tasks.

Source sentence
We propose the Two-Stage Multi-Task (TSMT) algorithm: first, an unregularized MLE over pooled data to obtain a coarse parameter estimate; second, a regularized MLE on individual securities to refine the parameters.

Assumes 4

  • The context covariance matrices Σj are uniformly bounded away from zero and infinity across securities. statistical
    Source sentence
    Assumption 3.1. Assume that there exist 0 < λ < λ such that λ < min j∈[M] λmin Σj ≤ max j∈[M] λmax Σj < λ.
  • BCL yields are bounded above by a constant \bar r and the function A_t(r) ≤ 0 for r < \bar r. model
    Source sentence
    Assumption 3.2. There exists some ¯r > 0 such that for every t ∈[T], it holds that At(r) ≤0, ∀r < ¯r.
  • The noise in the yield model is i.i.d. Gaussian with finite variance and satisfies a technical inequality (3.6). statistical
    Source sentence
    ... (3.6) requires that the true σ in (2.2) lies in a confined range determined by bond primitives.
  • There are no strategic interactions between the broker and competitors (exogeneity of competitor pricing). market
    Source sentence
    means there are no strategic interactions between us and other competitors.

Data

Backtest sample: U.S. corporate bonds · fixed-income

  • TRACE transaction records · FINRA market-prices commercial
    Source sentence
    We construct the data set by combining transaction records from TRACE [42]...
  • OpenBond daily bond information · OpenBond fundamentals public
    Source sentence
    daily bond information from OpenBond [78]...
  • Mergent FISD contractual information · Mergent fundamentals commercial
    Source sentence
    contractual information from Mergent FISD [72]...
  • FRED interest-rate data · Federal Reserve Economic Data macro public
    Source sentence
    interest-rate data from FRED [40]...

Recognised sources: FRED

Costs

Unclear The classifier was not confident.

Finds 2

  • TSMT achieves a regret bound of \tilde O(δ_max \sqrt{T M d} + M d), improving over individual and pooled baselines. theoretical
    Source sentence
    We show that the TSMT achieves a regret bounded by \tilde{O} ( δ_{max} \sqrt{T M d} + M d ) , outperforming both fully individual and fully pooled baselines...
  • In the 2024 evaluation, TSMT reduces realized accumulated regret by 32.5% relative to pooling and 38.0% relative to individual estimation, and achieves the lowest MAE of 7.724 bps. performance
    Source sentence
    TSMT reduces realized accumulated regret by 32.5% relative to pooling and by 38.0% relative to individual estimation

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 9 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (106)

  1. Size matters for OTC market makers: General results and dimensionality reduction techniques (2021)
  2. Dynamics of market making algorithms in dealer markets: Learning and tacit collusion (2024)
  3. Inventory, market making, and liquidity in OTC markets (2024)
  4. Time series: Theory and methods (1991)
  5. Foundations of mathematical economics (2001)
  6. A first course in optimization theory (1996)
  7. Generalized envelope theorems: applications to dynamic programming (2018)
  8. Properties of reverse hazard functions (2011)
and 98 more
  1. Reinforcement Learning for Corporate Bond Trading: A Sell Side Perspective — → 2406.12983
  2. Market Making without Regret — → 2411.13993
  3. Dynamic Pricing in Securities Lending Market: Application in Revenue Optimization for an Agent Lender Portfolio — → 2407.13687
  4. Optimal policies for dynamic pricing and inventory control with nonparametric censored demands (2024)
  5. Optimal auction design (1981)
  6. Privacy-preserving dynamic personalized pricing with demand learning (2022)
  7. Structural models of corporate bond pricing: An empirical analysis (2004)
  8. The value of knowing a demand curve: Bounds on regret for online posted-price auctions (2003)
  9. Dynamic pricing in high-dimensions (2019)
  10. Perishability of data: dynamic pricing under varying-coefficient models (2017)
  11. Feature-based dynamic pricing (2020)
  12. Dynamic pricing with an unknown demand model: Asymptotically optimal semi-myopic policies (2014)
  13. Dynamic pricing without knowing the demand function: Risk bounds and near-optimal algorithms (2009)
  14. Dynamic pricing under a general parametric choice model (2012)
  15. Dynamic pricing with demand covariates (2016) — link
  16. On dynamic pricing with covariates (2025)
  17. Context-based dynamic pricing with partially linear demand model (2022)
  18. Pricing and positioning of horizontally differentiated products with incomplete demand information (2024)
  19. Robust dynamic pricing with demand learning in the presence of outlier customers (2023)
  20. Regression analysis when the dependent variable is truncated normal (1973)
  21. Provably optimal algorithms for generalized linear contextual bandits (2017)
  22. The fundamentals of commodity futures returns (2013)
  23. Facts and fantasies about commodity futures (2006)
  24. An anatomy of commodity futures risk premia (2014)
  25. Order flows and financial investor impacts in commodity futures markets (2022)
  26. Common risk factors in currency markets (2011)
  27. Deutsche mark–dollar volatility: intraday activity patterns, macroeconomic announcements, and longer run dependencies (1998)
  28. Order flow and exchange rate dynamics (2002)
  29. Efficient capital markets (1970)
  30. Common risk factors in the returns on stocks and bonds (1993)
  31. News and asset pricing: A high-frequency anatomy of the SDF (2025)
  32. Continuous-time fama-macbeth regressions (2025)
  33. The cross-section of expected corporate bond returns: Betas or characteristics? (2005)
  34. Corporate yield spreads and bond liquidity (2007)
  35. Understanding Fixed-Income Markets in 2023 (2023)
  36. Deep reinforcement learning for market making in corporate bonds: beating the curse of dimensionality (2019)
  37. The behavior of dealers and clients on the European corporate bond market: the case of Multi-Dealer-to-Client platforms (2016)
  38. Structural models of corporate bond pricing with maximum likelihood estimation (2008)
  39. Which factors affect corporate bonds pricing? Empirical evidence from eurobonds primary market spreads (2005)
  40. Deep learning in asset pricing (2024)
  41. The capital asset pricing model: Some empirical tests (1972)
  42. Empirical asset pricing via machine learning (2020)
  43. Machine learning in empirical asset pricing (2019)
  44. Financial machine learning (2023)
  45. Modeling corporate bond returns (2023)
  46. Characteristics are covariances: A unified model of risk and return (2019)
  47. Factor models, machine learning, and asset pricing (2022)
  48. Bond risk premiums with machine learning (2021)
  49. Are bond returns predictable with real-time macro data? (2023)
  50. Forest through the trees: Building cross-sections of stock returns (2019)
  51. Autoencoder asset pricing models (2021)
  52. Multitask learning (1997)
  53. Predicting multivariate responses in multiple linear regression (1997)
  54. Multilinear multitask learning (2013)
  55. An overview of multi-task learning (2018)
  56. Gradient surgery for multi-task learning (2020)
  57. How fine-tuning allows for effective meta-learning (2021)
  58. Predicting with proxies: Transfer learning in high dimension (2021)
  59. Robust angle-based transfer learning in high dimensions (2022) — link
  60. Transfer learning for high-dimensional linear regression: Prediction, estimation and minimax optimality (2022)
  61. Adaptive and robust multi-task learning (2023)
  62. Mean–Variance Portfolio Selection by Continuous-Time Reinforcement Learning: Algorithms, Regret Analysis, and Empirical Study — → 2412.16175
  63. Sublinear regret for a class of continuous-time linear-quadratic reinforcement learning problems (2025)
  64. Policy evaluation and temporal-difference learning in continuous time and space: A martingale approach (2022)
  65. Data-driven option pricing (2024) — link
  66. Generalised correlated batched bandits via the arc algorithm with application to dynamic pricing (2021) — link
  67. Learning from similar linear representations: Adaptivity, minimaxity, and robustness (2025)
  68. Model-agnostic meta-learning for fast adaptation of deep networks (2017)
  69. Learning across bandits in high dimension via robust statistics (2021) — link
  70. The double auction market institution: A survey (2018)
  71. Linear algorithms for online multitask classification (2010)
  72. Multitask quantile regression under the transnormal model (2016)
  73. Online meta-learning (2019)
  74. Meta-learning in neural networks: A survey (2021)
  75. Meta-thompson sampling (2021)
  76. Meta dynamic pricing: Transfer learning across experiments (2022)
  77. Transfer learning for reinforcement learning domains: A survey. (2009)
  78. A comprehensive survey on transfer learning (2020)
  79. Bandit algorithms (2020)
  80. Strategic Learning and Trading in Broker-Mediated Markets — → 2412.20847
  81. Market making with exogenous competition — → 2407.17393
  82. Broker-Trader Partial Information Nash Equilibria — → 2412.17712
  83. Risk-Aware Linear Bandits: Theory and Applications in Smart Order Routing (2022) — link
  84. Action elimination and stopping conditions for the multi-armed bandit and reinforcement learning problems. (2006)
  85. User-friendly tail bounds for matrix martingales (2011)
  86. Log-concave probability and its applications (2006)
  87. Robustness implies generalization via data-dependent generalization bounds (2022)
  88. How to clean enhanced TRACE data (2014)
  89. TRACE: The Source for Real-Time Bond Market Transaction Data (2024) — link
  90. LSEG Workspace (2024) — link
  91. Open Source Bond Asset Pricing: Replication Data and Code (2026) — link
  92. Mergent Fixed Income Securities Database (FISD) (2026) — link
  93. Federal Reserve Economic Data (FRED) (2026) — link
  94. “Fulfilled by amazon”: A strategic perspective of competition at the e-commerce platform (2022)
  95. Help grow your business with fulfillment by amazon (2018)
  96. NASDAQ ITCH Data (2022)
  97. Moody’s Seasoned Aaa Corporate Bond Yield (Accessed 2025) — link
  98. Moody’s Seasoned Baa Corporate Bond Yield (Accessed 2025) — link

7 of 106 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.