Volatility Timing using Machine Learning - An Application to a Signal Based Portfolio
(2022) DABN01 20221Department of Economics
- Abstract
- Recent events such as the covid-19 pandemic and the Russian-Ukrainian war have led to a tremendous increase in volatility, making financial markets riskier for investors. To see whether investors can counteract or profit from such risk, we develop a volatility timed trading strategy. To do so, a volatility forecast is used, in our case, the direction of the VIX. We predict the one week ahead movement of the VIX by deploying several machine learning algorithms, resulting in the support vector machine with a linear kernel being the best model. The volatility timed portfolio is based on the forecast acting as a signal, investing in equities if the volatility decreases and in fixed income if it increases. To see whether an investor is better... (More)
- Recent events such as the covid-19 pandemic and the Russian-Ukrainian war have led to a tremendous increase in volatility, making financial markets riskier for investors. To see whether investors can counteract or profit from such risk, we develop a volatility timed trading strategy. To do so, a volatility forecast is used, in our case, the direction of the VIX. We predict the one week ahead movement of the VIX by deploying several machine learning algorithms, resulting in the support vector machine with a linear kernel being the best model. The volatility timed portfolio is based on the forecast acting as a signal, investing in equities if the volatility decreases and in fixed income if it increases. To see whether an investor is better off with that portfolio than a 60% equity and 40% fixed income portfolio, we compare the two using the Sharpe ratio. Furthermore, a split on market regimes is done to see whether there are performance differences between the portfolios for different levels of the VIX. We find that the signal-based portfolio is unable to outperform the 60/40 portfolio over the whole period, while for certain market regimes, it has a substantial advantage. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9084440
- author
- Lövgren, Filippa LU and Ulmer, Julian Marvin LU
- supervisor
- organization
- course
- DABN01 20221
- year
- 2022
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Machine Learning, Support Vector Machines, VIX, Volatility Timing, Portfolio Construction
- language
- English
- id
- 9084440
- date added to LUP
- 2022-06-08 12:51:23
- date last changed
- 2022-06-08 12:51:23
@misc{9084440,
abstract = {{Recent events such as the covid-19 pandemic and the Russian-Ukrainian war have led to a tremendous increase in volatility, making financial markets riskier for investors. To see whether investors can counteract or profit from such risk, we develop a volatility timed trading strategy. To do so, a volatility forecast is used, in our case, the direction of the VIX. We predict the one week ahead movement of the VIX by deploying several machine learning algorithms, resulting in the support vector machine with a linear kernel being the best model. The volatility timed portfolio is based on the forecast acting as a signal, investing in equities if the volatility decreases and in fixed income if it increases. To see whether an investor is better off with that portfolio than a 60% equity and 40% fixed income portfolio, we compare the two using the Sharpe ratio. Furthermore, a split on market regimes is done to see whether there are performance differences between the portfolios for different levels of the VIX. We find that the signal-based portfolio is unable to outperform the 60/40 portfolio over the whole period, while for certain market regimes, it has a substantial advantage.}},
author = {{Lövgren, Filippa and Ulmer, Julian Marvin}},
language = {{eng}},
note = {{Student Paper}},
title = {{Volatility Timing using Machine Learning - An Application to a Signal Based Portfolio}},
year = {{2022}},
}