Paper: arXiv 2408.11257

Authors: Arun Kumar Polala, Bernhard Hientzsch

Abstract

In [“1”], we calibrated a one-factor Cheyette SLV model with a local volatility that is linear in the benchmark forward rate and an uncorrelated CIR stochastic variance to 3M caplets of various maturities. While caplet smiles for many maturities could be reasonably well calibrated across the range of strikes, for instance the 1Y maturity could not be calibrated well across that entire range of strikes. Here, we study whether models with alternative local volatility terms and/or alternative stochastic volatility or variance models can calibrate the 1Y caplet smile better across the strike range better than the model studied in [“1”]. This is made possible and feasible by the generic simulation, pricing, and calibration frameworks introduced in [“1”] and some new frameworks presented in this paper. We find that some model settings calibrate well to the 1Y smile across the strike range under study. In particular, a model setting with a local volatility that is piece-wise linear in the benchmark forward rate together with an uncorrelated CIR stochastic variance and one with a local volatility that is linear in the benchmark rate together with a correlated lognormal stochastic volatility with quadratic drift (QDLNSV) as in [“2”] calibrate well. We discuss why the later might be a preferable model. [“1”] Arun Kumar Polala and Bernhard Hientzsch. Parametric differential machine learning for pricing and calibration. arXiv preprint arXiv:2302.06682 , 2023. [“2”] Artur Sepp and Parviz Rakhmonov. A Robust Stochastic Volatility Model for Interest Rate Dynamics. Risk Magazine, 2023

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: The paper is mathematically dense with detailed SDEs for various Cheyette and stochastic volatility models, while also providing empirical calibration results with specific parameters and comparison plots for caplet smiles.

Research Flowchart

  flowchart TD
  Goal["Research Goal<br>Improve calibration of 1Y caplet smile<br>using alternative Cheyette SLV models"] --> Inputs["Data / Inputs<br>1Y Caplet Market Data<br>Existing models from [1]"]
  Inputs --> Method["Methodology<br>Implement generic simulation & pricing frameworks<br>from [1] + new frameworks from this paper"]
  Method --> Models["Calibration Process<br>Test various model configurations:<br>- Linear LV + CIR (Benchmark)<br>- Piece-wise Linear LV + CIR<br>- Linear LV + QDLNSV [2]"]
  Models --> Results["Results & Findings"]
  Results --> Result1["Calibration Success"]
  Results --> Result2["Discussion"]
  Result1 --> Details1["Models with Piece-wise Linear LV + CIR<br>and Linear LV + QDLNSV<br>calibrate well across strike range"]
  Result2 --> Details2["QDLNSV model preferred<br>due to robustness & properties"]

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 alternative one‑factor Cheyette model settings with different local volatility forms and stochastic volatility/variance specifications, and a generic simulation, pricing and calibration framework to achieve better 1‑year caplet smile calibration.

Source sentence
We find that some model settings calibrate well to the 1Y smile across the strike range under study. In particular, a model setting with a local volatility that is piece‑wise linear in the benchmark forward rate together with an uncorrelated CIR stochastic variance and one with a local volatility that is linear in the

Assumes 4

  • Mean reversion speed λ is set to 0.03 and is assumed not to impact results beyond numerical stability. model
    Source sentence
    In this paper, we use λ = 0.03, a common value from the literature. The choice of λ does not seem to impact the results too much beyond numerical
  • The Feller condition for CIR variance is enforced in all calibrations. model
    Source sentence
    In this paper, we enforce the Feller condition in the calibrations.
  • Stochastic volatility is used only to shape the distribution and does not affect expected volatility (θ = 1, ϑ0 = 1). model
    Source sentence
    As is common in rates and some applications in equity, we use the stochastic volatility only to shape the distribution but not to impact the expected volatility by choosing θ = 1 and ϑ0 = 1.
  • Piece‑wise linear local volatility requires anchor strikes to be chosen by trial and error. model
    Source sentence
    we do not need to find the appropriate anchor strikes for the piece‑wise linear local volatility by trial and error, instead of mostly automatically through some optimization.

Data

No dataset named in the text.

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Model settings with piece‑wise linear local volatility and uncorrelated CIR stochastic variance calibrate well to the 1‑year caplet smile across the strike range. pricing-calibration
    Source sentence
    a model setting with a local volatility that is piece‑wise linear in the benchmark forward rate together with an uncorrelated CIR stochastic variance
  • Model settings with linear benchmark‑rate local volatility and correlated lognormal stochastic volatility with quadratic drift (QDLNSV) also calibrate well to the 1‑year smile. pricing-calibration
    Source sentence
    with a local volatility that is linear in the benchmark rate together with a correlated lognormal stochastic volatility with quadratic drift (QDLNSV)
  • The generic simulation, pricing, and calibration frameworks make such calibration studies feasible, efficient, and easy. theoretical
    Source sentence
    We see that the presented generic simulation, pricing, and calibration frameworks make such calibration and modeling studies, feasible, efficient, and even easy.
  • Calibration using the code‑generated MC pricers or PDML approaches can be performed within a minute with relatively small computational effort. performance
    Source sentence
    the code generation and calibration set‑ups lead to calibrations that work well with relatively small computational effort and often return calibrated parameters within a minute.
  • The lognormal stochastic volatility with quadratic drift is preferred because it avoids the need for trial‑and‑error anchor strike selection for piece‑wise linear local volatility. theoretical
    Source sentence
    We prefer the lognormal stochastic volatility with quadratic drift since then we do not need to find the appropriate anchor strikes for the piece‑wise linear local volatility by trial and error.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 22 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.