Short term forecasting of electricity prices for MISO hubs: Evidence from ARIMA-EGARCH models

This study estimates three time series models (ARIMA, ARIMA-EGARCH, and ARIMA-EGARCH-M) for hourly real time electricity prices for each of the five hubs of the Midwest Independent System Operator (MISO) and examines the in- and out-of-sample forecasting performance of the respective models. The results from the ARIMA models reveal the presence of autoregressive conditional heteroskedasticity. Recognizing the possibility of asymmetric time-varying volatility, the EGARCH specification for the variance equation demonstrates the presence of an inverse leverage effect in electricity prices for each hub. With respect to forecasts, no one model clearly dominates the others in terms of in-sample forecasting performance based on four forecast evaluation statistics. However, the ARIMA-EGARCH-M model outperforms the other models (Michigan hub is the exception) in terms of the out-of-sample forecasting performance.