Forecasting MISO Electricity Prices: A Threshold Autoregressive Approach with Load Data

Electricity price dynamics for the Illinois market are examined by estimating eleven dierent threshold autoregressive models and comparing according to t and forecasting performance. The threshold is endogenous and depends on load data in three of the cases. A theoretical model demonstrates that supply constraints could explain price spikes and that prices would display less persistence in those cases. Estimation results conrm that presence of non-linearity in the evolution of prices. However, inclusion of the load data does not improve performance, which provides evidence against this hypothesis. The model where the threshold depends on the change in the past price is best.