/WigGARCH

Bachelor's diploma work focusing on predicting volatility of the Polish WIG20 index with GARCH models.

Primary LanguageR

©️ Tags

  • Models: ARCH, GARCH, EGARCH, GLR-GARCH, APARCH
  • Areas: Time Series, Econometrics, regression, prediction, simulation, anomaly detection

💡 About

A project focusing on predicting the volatility of the Polish WIG20 stock index and assessing the accuracy of the value at risk (VaR) and expected shortfall (ES) forecasts using models from the GARCH family.

📂 Content

  • research - main file containing the comprehensive research ✍️
  • analysis – analysis, modeling, and testing procedure stored as R script ⚗️
  • dataset – raw dataset 📀

🧪 Methodology

  • Returns distribution was selected based on the AIC, BIC, HQIC information criteria 🔍
  • The best model was selected based on the above criteria and comparison with the historical simulation 👾

🏆 Findings

  • t-Student distribution was the best fit for the returns of the WIG20 index.
  • Nelson's exponential GARCH model EGARCH(1,1) had fewer exceedances of the Estimated Shortfall at the 5% and 1% confidence levels compared to the historical simulation.
  • EGARCH's accuracy proved to be time-invariant, which means that the model can effectively predict future losses of WIG20.