The browser you are using is not supported by this website. All versions of Internet Explorer are no longer supported, either by us or Microsoft (read more here: https://www.microsoft.com/en-us/microsoft-365/windows/end-of-ie-support).

Please use a modern browser to fully experience our website, such as the newest versions of Edge, Chrome, Firefox or Safari etc.

Performance of Stochastic Volatility and GARCH Models in Different Market Regimes

Author

  • Felix Viitanen
  • Erik Lundgren

Summary, in English

Reliable methods for estimating financial return volatility are crucial in many areas of trading and investing. Two such frameworks, the GARCH and SV, have been of particular interest to academics and practitioners alike. The GARCH model describes the variance of the current innovation as a function of the actual sizes of the previous innovations. In contrast, the stochastic volatility model describes volatility as a latent variable following
a stochastic process. This thesis attempts to extend the research conducted by Lopes and Polson (2010) by analyzing the performance of the Gaussian GARCH(1,1) and basic SV model on the SP500 and OMXS30 before and during the endogenous credit crisis, as well as before and during the exogenous COVID-19 pandemic. The results indicate that the SV model consistently fits the data better than the GARCH model on all data sets, while the fit for both models became worse during the periods of market stress, and even more so for the pandemic. In regards to the volatility estimation performance, the GARCH model tends to be better for periods with low volatility, while the performance is similar
in highly volatile climates. Finally, the pandemic appeared to be the stress event that had the largest negative impact on the model validation.

Summary, in English

Reliable methods for estimating financial return volatility are crucial in many areas of trading and investing. Two such frameworks, the GARCH and SV, have been of particular interest to academics and practitioners alike. The GARCH model describes the variance of the current innovation as a function of the actual sizes of the previous innovations. In contrast, the stochastic volatility model describes volatility as a latent variable following
a stochastic process. This thesis attempts to extend the research conducted by Lopes and Polson (2010) by analyzing the performance of the Gaussian GARCH(1,1) and basic SV model on the SP500 and OMXS30 before and during the endogenous credit crisis, as well as before and during the exogenous COVID-19 pandemic. The results indicate that the SV model consistently fits the data better than the GARCH model on all data sets, while the fit for both models became worse during the periods of market stress, and even more so for the pandemic. In regards to the volatility estimation performance, the GARCH model tends to be better for periods with low volatility, while the performance is similar
in highly volatile climates. Finally, the pandemic appeared to be the stress event that had the largest negative impact on the model validation.

Publishing year

2022

Language

English

Document type

Student publication for Bachelor's degree

Topic

  • Mathematics and Statistics

Supervisor

  • Krzysztof Podgorski (Professor)
  • Alessandro Mercuri