Evaluation of Machine Learning Methods for Portfolio Replication of VIX Futures

During the past two decades, the CBOE Volatility Index (VIX® Index), a key measure of investor sentiment and 30-day future volatility expectations, has generated much investor attention because of its unique and powerful features. The introduction of VIX futures in 2004, VIX options in 2006, and other volatility-related trading instruments provided traders and investors access to exchange-traded vehicles for taking long and short exposures to expected S&P 500 Index volatility for a particular time frame. Certain VIX-related tradable products may provide benefits when used as tools for tail-risk hedging, diversification, risk management, or alpha generation. However, all such VIX relevant derivatives simply do not react quickly enough to movements in the spot VIX, which leads to a non-effective hedging performance. On the other hand, there are many other derivatives that can help us to reduce the volatility risk. For example, Peter Carr shows us a promising result using the SPX option only to forecast VIX index by some modern approaches, such as Machine Learning. In this project, we will look at different methods to extend Peter Carr’s work, and their comparative effectiveness in using a combination of SPX options to predict VIX futures.

Faculty Supervisor:

Matheus Grasselli

Student:

Jieyi Zhu

Partner:

QTS Capital Management LLC.

Discipline:

Statistics / Actuarial sciences

Sector:

Finance, insurance and business

University:

McMaster University