Hostname: page-component-7b7567c8c-7pwfn Total loading time: 0 Render date: 2026-11-15T23:02:04.815Z Has data issue: false hasContentIssue false

LOCALIZED RADIAL BASIS FUNCTIONS FOR NO-ARBITRAGE PRICING OF OPTIONS UNDER STOCHASTIC ALPHA–BETA–RHO DYNAMICS

Published online by Cambridge University Press:  19 August 2021

N. THAKOOR*
Affiliation:
Department of Mathematics, University of Mauritius, Reduit 80837, Mauritius
Rights & Permissions [Opens in a new window]

Abstract

Closed-form explicit formulas for implied Black–Scholes volatilities provide a rapid evaluation method for European options under the popular stochastic alpha–beta–rho (SABR) model. However, it is well known that computed prices using the implied volatilities are only accurate for short-term maturities, but, for longer maturities, a more accurate method is required. This work addresses this accuracy problem for long-term maturities by numerically solving the no-arbitrage partial differential equation with an absorbing boundary condition at zero. Localized radial basis functions in a finite-difference mode are employed for the development of a computational method for solving the resulting two-dimensional pricing equation. The proposed method can use either multiquadrics or inverse multiquadrics, which are shown to have comparable performances. Numerical results illustrate the accuracy of the proposed method and, more importantly, that the computed risk-neutral probability densities are nonnegative. These two key properties indicate that the method of solution using localized meshless methods is a viable and efficient means for price computations under SABR dynamics.

Information

Type
Research Article
Copyright
© Australian Mathematical Society 2021
Figure 0

Figure 1Implied volatility for low-strikes problem (colour available online).

Figure 1

Figure 2Positive probability density using the RBF-FD MQ method.

Figure 2

Figure 3Positive probability density using the RBF-FD IMQ method.

Figure 3

Figure 4Implied volatilities for different strike prices.

Figure 4

Table 1Uncorrelated case.

Figure 5

Table 2Comparison of option prices in the correlated case.

Figure 6

Figure 5Accuracy comparison of the different methods with the exact simulation method when the expansion formula [16] is unreliable.

Figure 7

Figure 6Dependence of the $\ell ^2$ error on the shape parameter c (colour available online).

Figure 8

Figure 7Contour plots for the $\ell ^2$ error norm against the parabolic mesh ratio $\gamma _F$ against the spatial mesh $h_F$ in the F-direction.

Figure 9

Figure 8Contour plots for the $\ell ^2$ error norm against the parabolic mesh ratio $\gamma _\alpha $ against the spatial mesh $h_\alpha $ in the $\alpha $-direction.

Figure 10

Figure 9Surface plot for the $\ell ^2$ error against parabolic mesh ratios $\gamma _F$ and $\gamma _\alpha $.

Figure 11

Table 3Down-and-out call option prices for varying values of $\rho $.

Figure 12

Figure 10Comparison of down-and-out call options for small- and large-maturity problems.

Figure 13

Table 4Double knock-out call options prices using the RBF-FD method.

Figure 14

Figure 11Halton grid [15] with $M=100$ interior points for $F\in [0,\,2]$ and $\alpha \in [0,\,1]$.

Figure 15

Figure 12Option price fitted surface for $F\in [0,\,2]$ and $\alpha \in [0,\,1]$.

Figure 16

Table 5Option prices obtained on scattered nodes and structured nodes.