Isogeometric Analysis for the Pricing of Financial Derivatives with Nonlinear Models: Convertible Bonds and Options

Rakhymzhan Kazbek, Yogi Erlangga, Yerlan Amanbek, Dongming Wei · 2024

Read Isogeometric Analysis for the Pricing of Financial Derivatives with Nonlinear Models: Convertible Bonds and Options by Rakhymzhan Kazbek, Yogi Erlangga, Yerlan Amanbek, Dongming Wei (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-08987.txt.

Introduction

Rakhymzhan Kazbek, Yogi Erlangga, Yerlan Amanbek, and Dongming Wei test a fast math method for pricing complex deals. Their 2024 paper prices bonds that can turn into shares and call options with added costs. Both cases rest on hard nonlinear price equations.

The method, called isogeometric analysis, builds smooth curved pieces to solve the price equations. The team set its results next to two older methods that use grids or small blocks. Their method reached very high accuracy on far fewer grid points. Fewer points mean much less computing time. The gain matters for banks that price such deals many times a day under tight time limits.

Catalog record

CREATORRakhymzhan Kazbek, Yogi Erlangga, Yerlan Amanbek, Dongming Wei
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-08987

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Isogeometric Analysis for the Pricing of Financial Derivatives with Nonlinear Models: Convertible Bonds and Options Rakhymzhan Kazbek, Yogi Erlangga, Yerlan Amanbek, Dongming Wei Abstract Computational efficiency is essential for enhancing the accuracy and practicality of pricing complex financial derivatives. In this paper, we discuss Isogeometric Analysis (IGA) for valuing financial derivatives, modeled by two nonlinear Black-Scholes PDEs: the Leland model for European call with transaction costs and the AFV model for convertible bonds with default options. We compare the solutions of IGA with finite difference methods (FDM) and finite element methods (FEM). In particular, very accurate solutions can be numerically calculated on far less mesh (knots) than FDM or FEM, by using non-uniform knots and weighted cubic NURBS, which in turn reduces the computational time significantly. Keywords– Isogeometric Analysis, NURBS, Transaction costs, Options, Convertible bonds, Greeks. 1 Introduction Isogeometric analysis (IGA) is a finite-element method (FEM) that utilizes a special type of basis functions, known as non-uniform rational B-splines or NURBS [ 29 ] . First introduced by Hughes in [ 18 , 10 ] , IGA has since gained popularity as a robust FEM approach. Among its key strengths are the ability to deal with complex geometries and the natural link it creates between FEM and computer-aided design (CAD), a tool widely used in geometric design and mesh generation. Moreover, IGA’s use of continuous and sufficiently differentiable NURBS basis functions enables it to produce higher-order smooth solutions within the computational domain \Omega . In contrast, traditional FEM methods are limited in this regard, as they employ piecewise continuous basis functions. For instance, the linear P1-FEM solution achieves only C^{0} -continuous across the domain \Omega . In problems where derivatives of the solution are critical, such as in financial hedging strategies involving option prices (the Greeks), having only a C^{0}(\Omega) -continuous solution makes the calculation of derivatives conceptually not straight-forward. These derivatives can only be determined within the interior of elements, rendering their values non-existent at the boundaries. In option pricing, the finite difference method (FDM) is widely favored by financial practitioners due to its computational efficiency and straightforward implementation [ 15 , 9 , 26 ] . Moreover, FDM is well-suited to addressing non-linearities in the partial differential equations (PDEs) associated with option pricing. While higher-order FDMs can potentially enhance solution accuracy with fewer grid points, they often involve extended discretization stencils, which can complicate the treatment of boundary conditions. On the other hand, finite element methods (FEM) provide a more natural framework for achieving higher-order approximations and managing boundary conditions effectively. Despite these advantages, FEM has not gained the same level of popularity as FDM in option pricing [ 14 , 35 , 22 ] . Although FEM typically offers superior convergence rates compared to FDM [ 1 ] , the use of Lagrange polynomial basis functions in standard FEM formulations can result in highly oscillatory solutions. In contrast to the Lagrange polynomial basis functions used in the standard FEM, IGA utilizes NURBS, a class of functions that satisfy certain differentiability conditions across elements, as the basis functions. For instance, cubic NURBS basis functions are twice differentiable in the computational domain \Omega . Furthermore, the NURBS basis functions are nonnegative and non-oscillatory. By using proper weighting, NURBS basis functions can very accurately approximate (or fit) non-smooth functions, such as the pay-off function in option pricing, a feature not found in FDM or standard FEM. While the success of IGA has been demonstrated in a wide range of applications [ 28 , 33 , 17 , 8 , 5 ] , there is not so much work on using IGA for (nonlinear) option pricing. We should in this regard mention the work of [ 27 ] , which uses a special case of NURBS, called B-Splines, to solve the linear Black-Scholes equation, and the recent work on IGA for European option [ 28 ] . In this paper, we demonstrate IGA as an efficient alternative to the popular FDM and FEM for solving nonlinear partial differential equations (PDEs) arising in option pricing. In particular, with optimal weights and non-uniform knots, we show that NURBS IGA produces solutions that match accurately the solutions of FDM and FEM, but with low numbers of meshes/knots (hence, basis functions). The latter results in significant gain in computational time. For this purpose, out of many nonlinear models that are presently used for different types of options, we shall in particular consider two models: (i) the Leland model [ 25 ] , which arises in European call option with transaction costs; (ii) the AFV model [ 3 ] , used for convertible bond pricing with coupon payments. The AFV model is an example of exotic options due to its variable parameters and inequality constraints, which influence the solution at each time step. Additionally, it exhibits path-dependency because of its embedded bond and equity option properties, where the price depends not only on the final payoff value but also on the entire price path computed at each time step. Moreover, the model allows early exercise (American options), which is represented by free boundaries. All of these lead to a complex structure from a computational perspective due to the changing payoff values in the inequality constraints. One approach for solving free boundaries such as in the AFV model is by reformulating the problems as Linear Complementarity Problems (LCPs) [ 12 ] , which can be solved using, e.g., ADI-based predictor-corrector schemes [ 38 ] or through analytical approximations [ 23 , 4 , 7 , 6 ] . Alternatively, one can formulate the model as an optimal control problem [ 13 ] using discrete Hamilton-Jacobi-Bellman (HJB) PDEs. The latter is often solved by using policy iteration algorithms [ 9 ] , which eliminate the need to explicitly determine the optimal exercise boundary at each time step. In this study, we employ formulation of the AFV model into a penalty PDE [ 12 ] , resulting in a system of nonlinear PDEs and transforming the free-boundary problems into ones defined on a fixed domain. The paper is organized as follows. In Section 2 , we discuss briefly the IGA method, followed by the construction of an IGA method for the nonlinear Leland PDE and the nonlinear AFV model. In Section 3 , we discuss the time integration procedure, which is based on the \theta -scheme with linearization. Numerical results are presented and discussed in Section 4 . We shall draw conclusions and outlook in Section 5 . In the remaining part of this section, we briefly discuss the two PDEs used as the model problems in this paper. 1.1 European contract with transaction costs: The Leland model The European call option price with transaction costs and with the strike price \hat{K} at the terminal time T can be modeled by the nonlinear PDE \displaystyle\frac{\partial V}{\partial t}+\frac{1}{2}\sigma^{2}S^{2}\left(1+\textit{Le}\sign\left(\frac{\partial^{2}V}{\partial S^{2}}\right)\right)\frac{\partial^{2}V}{\partial S^{2}}+rS\frac{\partial V}{\partial S}-rV=0, (1) where V=V(S,t) is the option price at any time t\in(0,T) , S\geq 0 is the value of the underlying asset, r is risk-free interest rate, and \sigma is volatility. The constant Le=\sqrt{\frac{2}{\pi}}\frac{c}{\sigma\sqrt{\delta t}} is called the Leland number, with c be the round trip of transaction costs per currency limit. This model, introduced by Leland [ 25

Read the complete text