Geometric Local Variance Gamma model

TL;DR

This paper introduces a geometric version of the Local Variance Gamma model with drift, extending previous models by allowing piecewise linear and quadratic local variance functions, and provides closed-form solutions for option pricing and surface recovery.

Contribution

It develops a geometric drift version of the Local Variance Gamma model with new piecewise linear and quadratic local variance functions, enabling closed-form solutions and efficient surface calibration.

Findings

Derived a differential equation analogous to Dupire's for the new model.

Achieved closed-form solutions for option prices under the new constructions.

Provided a fast, non-optimization method for local surface recovery.

Abstract

This paper describes another extension of the Local Variance Gamma model originally proposed by P. Carr in 2008, and then further elaborated on by Carr and Nadtochiy, 2017 (CN2017), and Carr and Itkin, 2018 (CI2018). As compared with the latest version of the model developed in CI2018 and called the ELVG (the Expanded Local Variance Gamma model), here we provide two innovations. First, in all previous papers the model was constructed based on a Gamma time-changed {\it arithmetic} Brownian motion: with no drift in CI2017, and with drift in CI2018, and the local variance to be a function of the spot level only. In contrast, here we develop a {\it geometric} version of this model with drift. Second, in CN2017 the model was calibrated to option smiles assuming the local variance is a piecewise constant function of strike, while in CI2018 the local variance is a piecewise linear} function of strike. In this paper we consider 3 piecewise linear models: the local variance as a function of strike, the local variance as function of log-strike, and the local volatility as a function of strike (so, the local variance is a piecewise quadratic function of strike). We show that for all these new constructions it is still possible to derive an ordinary differential equation for the option price, which plays a role of Dupire's equation for the standard local volatility model, and, moreover, it can be solved in closed form. Finally, similar to CI2018, we show that given multiple smiles the whole local variance/volatility surface can be recovered which does not require solving any optimization problem. Instead, it can be done term-by-term by solving a system of non-linear algebraic equations for each maturity which is fast.