Option Models Calibration: A Hedging Criterion

Details

Supervisors

Vrins, Frédéric

Faculty

Louvain School of Management

Degree label

Master [120] en sciences de gestion (horaire décalé à Mons)

Abstract

This thesis aims to study the impact of the parameter calibration step on the hedging performances of the Black Scholes model. Generally, option models are actually option pricing models and, as such, are considered properly calibrated when they can best replicate market quotes. However, this calibration method might not be the one that leads to the best hedging performances. In this study, we calculate different volatility values than the one obtained through a pricing-based calibration and we analyse the impact of such values on payoff replication performances. Globally, we find that the pricing-based calibration leads to satisfying results when compared to other methods. Because we observe that option models have a tendency to over-hedge, we also introduce the concept of transaction costs through the payoff replication procedure. We find that doing so improve hedging performances noticeably. Finally, this thesis presents two ways in which machine learning techniques can be used to analyse hedging performances. We perform a classification with support vector machines in order to classify between overpriced and underpriced options. This leads to promising results that need to be generalised in a further study. We also perform regression in order to estimate the optimal hedging volatility of an option. While this last approach failed to improve hedging performances, we point out future research perspectives that might improve the results of this method in a significant way.