Applying Hybrid ARIMA-SGARCH in Algorithmic Investment Strategies on S&P500 Index

Area of Science:

• Quantitative Finance
• Econometrics
• Time Series Analysis

Background:

• Algorithmic trading strategies require accurate financial market forecasting.
• Autoregressive Integrated Moving Average (ARIMA) models are standard linear tools.
• GARCH family models capture volatility clustering in financial returns.

Purpose of the Study:

• To compare the forecasting performance of ARIMA against hybrid ARIMA-GARCH models for S&P500 log returns.
• To evaluate the effectiveness of hybrid models in constructing superior algorithmic investment strategies.
• To determine if hybrid models better capture time-series characteristics than ARIMA alone.

Main Methods:

• Utilized daily S&P500 log return data from 2000-2019.
• Employed a rolling window approach for model comparison.
• Assessed forecasting accuracy using error metrics (MAE, MAPE, RMSE) and performance metrics (returns, drawdown, information ratio).

Main Results:

• Hybrid ARIMA-SGARCH and ARIMA-EGARCH models demonstrated superior performance over the simple ARIMA model.
• The hybrid models outperformed the benchmark Buy & Hold strategy.
• Results remained robust across different window sizes and GARCH model specifications.

Conclusions:

• Hybrid ARIMA-GARCH models offer enhanced predictive power for S&P500 log returns.
• These models are more effective for developing long-term algorithmic investment strategies.
• The findings highlight the importance of incorporating volatility modeling in financial forecasting.