WTI crude oil option implied VaR and CVaR: An empirical application

• Author: Marinela Adriana Finta,Chiara Legnazzi,Giovanni Barone‐Adesi,Carlo Sala
• DOI: http://doi.org/10.1002/for.2580
• Date: 01 September 2019
• Published date: 01 September 2019

Received: 20 September 2018 Accepted: 12 February 2019

DOI: 10.1002/for.2580

RESEARCH ARTICLE

WTI crude oil option implied VaR and CVaR: An empirical

application

Giovanni Barone-Adesi1Marinela Adriana Finta2Chiara Legnazzi1Carlo Sala3

1Swiss Finance Institute at Università

della Svizzera Italiana (USI), Institute of

Finance, Lugano, Switzerland

2Sim Kee Boon Institute for Financial

Economics, Lee Kong Chian School of

Business, Singapore Management

University, 188065 Singapore

3Department of Financial Management

and Control, ESADE Business School,

Ramon Llull University, Barcelona, Spain

Correspondence

Carlo Sala, Department of Financial

Management and Control, ESADE

Business School, Ramon Llull University,

Avenida de Torreblanca59, 08172 Sant

Cugat del Valles, Barcelona,Spain.

Email: carlo.sala@esade.edu

Funding information

Schweizerischer Nationalfonds zur

Förderung der Wissenschaftlichen

Forschung, Grant/AwardNumber: 153135

Abstract

Using option market data we derive naturally forward-looking, nonparametric

and model-free risk estimates, three desired characteristics hardly obtainable

using historical returns. The option-implied measures are only based on the

first derivative of the option price with respect to the strike price, bypassing the

difficult task of estimating the tail of the return distribution. We estimate and

backtest the 1%, 2.5%, and 5% WTI crude oil futures option-implied value at risk

and conditional value at risk for the turbulent years 2011–2016 and for both tails

of the distribution. Compared with risk estimations based on the filtered his-

torical simulation methodology, our results show that the option-implied risk

metrics are valid alternatives to the statistically based historical models.

KEYWORDS

backtest, elicitability, option prices, VaR and CVaR

1INTRODUCTION

The role of risk management is to predict future risks

in an effective and efficient way. Future risks are usu-

ally summarized by computing point risk forecasts, such

as the value at risk (VaR) and conditional value at risk

(CVaR). To date, the risk estimation process is performed

either parametrically,via more or less sophisticated Monte

Carlo simulation techniques, or by imposing a para-

metric structure to the probability density function of

future returns (e.g., the Normal approach or modeling

the tails using different extreme value theory specifica-

tions), or nonparametrically by means of historical simula-

tions. Detaching from the normality assumption and using

data-driven empirical innovations, the filtered historical

simulation (FHS) approach of Barone-Adesi, Giannopou-

los, and Vosper (1999) combines the aforementioned

simulation techniques. As a common point, all these mod-

els only rely on historical data to infer a risk measure

through more or less sophisticated statistical approaches.

Throughout the article, these models are labeled as statis-

tically based historical risk models and are compared to the

option-implied risk models presented in Section 2.

Statistically based historical models infer the future

profit and loss (henceforth P&L) distribution from the

time series of observed past returns. From an economet-

ric viewpoint, the estimation of such a density is far from

being trivial. First, since the past is rarely a good proxy for

the future, historical market data might be not informa-

tive, especially during highly volatile periods (Danielsson,

2002), and references therein) when investorsact in a more

heterogeneous way. Second, being valid under the physi-

cal measure, statistically based historical models require

the estimation of complex economic quantities, such as the

552 © 2019 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/for Journalof Forecasting. 2019;38:552–563.