Pricing Average and Spread Options Under Local-Stochastic Volatility Jump-Diffusion Models

Abstract

This paper presents a new approximation formula for pricing multi-dimensional discretely monitored average options in a local-stochastic volatility (LSV) model with jumps by applying an asymptotic expansion technique. Moreover, it provides a justification of the approximation method with some asymptotic error estimates for general payoff functions. Particularly, our model includes local volatility functions and jump components in the underlying asset price as well as its volatility processes. To the best of our knowledge, the proposed approximation is the first one which achieves analytic approximations for the average option prices in this environment.In numerical experiments, by employing several models, we provide approximate prices for the listed average and calendar spread options on the WTI futures based on the parameters through calibration to the listed (plain-vanilla) futures options prices. Then, we compare those with the CME settlement prices, which confirms the validity of the method.Moreover, we show that the LSV with jumps model is able to replicate consistently and precisely listed futures option, calendar spread option and average option prices with common parameters.

Incorporating Prior Financial Domain Knowledge into Neural Networks for Implied Volatility Surface Prediction

“…Volatility models in finance can be classified into two groups [22]. The first group is called indirect methods, in which an implied volatility is driven by another dynamic model such as local volatility models, stochastic volatility models and Lévy models [21,24,27,34,38]. Models in this group usually have a limited number of parameters, and the volatility term is fitted by the market data along with the asset dynamics such as the geometric Brownian motion and the mean-revision jump-diffusion process.…”

Forecasting mortality rates with a coherent ensemble averaging approach

“…To overcome such an outstanding issue, this paper aims to employ the ensemble averaging, also known as the model averaging (MA) approach, to combine the coherent mortality models. MA is a sophisticated and well-developed approach in machine learning (Ley and Steel, 2009;Amini and Parmeter, 2012;Lessmann et al, 2012;Bork et al, 2020;Bravo et al, 2021), which has been widely applied in recent economics and finance research and practices (see, for example, Eicher et al, 2011;Mirestean and Tsangarides, 2016;Shiraya and Takahashi, 2019;Baechle et al, 2020;du Jardin, 2021, among others). With respect to mortality data, Shang (2012) and Kontis et al (2017) have employed various MA strategies, such as the Bayesian model averaging (BMA).…”