Unconditionally positivity-preserving approximations of the Aït-Sahalia type model: Explicit Milstein-type schemes

Xiaojie Wang (王小捷), Yingsong Jiang, Ruishu Liu, Xiaojie Wang, Jinghua Zhuo

TL;DR

This paper introduces a novel class of explicit, unconditionally positivity-preserving Milstein-type schemes for a challenging financial model (Aït-Sahalia type) with a superlinear diffusion and a drift that blows up at the origin. The proposed schemes achieve a first-order mean-square convergence rate, offering a computationally efficient alternative to existing implicit methods while maintaining crucial positivity properties.

What Problem Does This Paper Address?

Many real-world financial models, particularly those describing interest rates, are governed by stochastic differential equations (SDEs). However, these models often feature coefficients that violate the standard Lipschitz continuity conditions, making traditional numerical approximation methods like the Euler-Maruyama scheme unreliable and prone to producing non-physical negative values. The Aït-Sahalia type model, a prominent example, exhibits a drift that blows up at zero and a diffusion term with superlinear growth, posing significant challenges for numerical analysis. Existing methods to address these issues often involve implicit schemes, which are computationally expensive, or explicit schemes with limited convergence rates or applicability to specific cases. This paper aims to bridge this gap by developing an explicit, unconditionally positivity-preserving scheme with a strong convergence guarantee.

Methods

The authors propose a novel explicit Milstein-type scheme that incorporates two key innovations:

  • Corrective Mapping ($\Phi_h$): A projection operator $\Phi_h$ is introduced to handle the problematic polynomially growing coefficients. This operator is designed to satisfy specific assumptions that ensure its effectiveness in maintaining positivity and facilitating error analysis.
  • Partial Implicitness: While aiming for an explicit scheme, a crucial term in the drift is implicitly handled by solving a quadratic equation. This partial implicitness is key to preserving the positivity of the numerical solutions.

The analysis of the mean-square error bounds is also novel, as it avoids the need for a priori high-order moment bounds of the numerical approximations, which is a common requirement in SDE numerical analysis.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and numerical experiments using simulated data.

Key Insights and Ideas

  • Explicit Schemes Can Be Positivity-Preserving: The paper demonstrates that it's possible to design explicit schemes that rigorously preserve the positivity of solutions for challenging SDEs, overcoming a significant limitation of standard explicit methods.
  • The Power of Corrective Mappings: The introduction of the $\Phi_h$ operator is a clever way to tame the problematic coefficients without resorting to computationally expensive implicit steps or complex taming strategies that require higher-order moment bounds.
  • Simplified Error Analysis: The novel approach to analyzing mean-square error, which bypasses the need for a priori high-order moment bounds, is a significant methodological contribution that simplifies the analysis of such schemes.
  • First-Order Convergence Achieved: The proposed explicit scheme achieves a first-order mean-square convergence rate, which is a desirable outcome for practical numerical simulations.

Key Results

  • The paper introduces an explicit, unconditionally positivity-preserving Milstein-type scheme for the Aït-Sahalia type model.
  • The proposed scheme achieves a mean-square convergence rate of order one (O(h)) for both non-critical and general critical cases of the model parameters.
  • Numerical experiments confirm the theoretical findings, showing that the proposed scheme (SIPMM) achieves convergence rates close to 1, outperforming existing methods like BEM and SITEM in terms of both accuracy and computational efficiency.
  • Table 1 shows that SIPMM significantly reduces root mean square errors (RMSE) and computational time compared to BEM.

Limitations and Future Work

  • Specific Model: The proposed scheme is tailored for the Aït-Sahalia type model. Extending it to other SDEs with similar challenging coefficient properties would be a valuable next step.
  • Higher-Order Schemes: While first-order convergence is achieved, the development of higher-order explicit positivity-preserving schemes for this model remains an open area for future research.
  • Parameter Sensitivity: While the paper demonstrates effectiveness across different parameter regimes, further investigation into the sensitivity of the scheme's performance to various parameter choices could be beneficial.

Summary updated

Abstract

The present article aims to design and analyze efficient first-order strong schemes for a generalized Aït-Sahalia type model arising in mathematical finance and evolving in a positive domain $(0, \infty)$, which possesses a diffusion term with superlinear growth and a highly nonlinear drift that blows up at the origin. Such a complicated structure of the model unavoidably causes essential difficulties in the construction and convergence analysis of time discretizations. By incorporating implicitness in the term $α_{-1} x^{-1}$ and a corrective mapping $Φ_h$ in the recursion, we develop a novel class of explicit and unconditionally positivity-preserving (i.e., for any step-size $h>0$) Milstein-type schemes for the underlying model. In both non-critical and general critical cases, we introduce a novel approach to analyze mean-square error bounds of the novel schemes, without relying on a priori high-order moment bounds of the numerical approximations. The expected order-one mean-square convergence is attained for the proposed scheme. The above theoretical guarantee can be used to justify the optimal complexity of the Multilevel Monte Carlo method. Numerical experiments are finally provided to verify the theoretical findings.
BibTeX
@article{2025unconditionally,
  author = {Xiaojie Wang (王小捷) and Yingsong Jiang and Ruishu Liu and Xiaojie Wang and Jinghua Zhuo},
  title = {{Unconditionally positivity-preserving approximations of the Aït-Sahalia type model: Explicit Milstein-type schemes}},
  journal = {Numer. Algorithms},
  year = {2025},
  doi = {10.1007/s11075-024-01861-5},
}
Figure