The contingent claims valuation of physical assets and financial derivatives depends critically on the specification and estimation of the stochastic process that describes the price path. Accurate valuation of claims based on competitive electricity prices has proved problematic, as electricity price data are not well represented by traditional commodity price models of Brownian motion. Observed on-peak (high demand period) electricity spot prices are highly volatile and strongly mean reverting, infrequently punctuated by large upward jumps which quickly drop toward the mean price level.1 Existing commodity price characterizations do not capture this dynamic, though they are often used as there is no established alternative.2 Based on these stylized facts, continuous time models for real options and financial derivatives where the underlying state variable is the spot price of electricity have been proposed (Ethier 1997, 1999, Barz and Johnson 1998, Deng 1998). To date, these models have not been fit to market data, nor has econometric testing been undertaken.
This paper tests the stylized facts which have evolved with the accumulation of data from competitive electricity prices, estimating a mean reverting price process with stochastic regime switching which allows discontinuous jumps in electricity prices. The non-linear econometric model allows complex state dynamics and the standard models of commodity prices are special cases of the proposed model. Although the model allows complex transition dynamics, it remains tractable for financial applications and requires only observed price data for estimation and forecasting. Thus the problem is a fundamental one: to characterize the marginal distribution of electricity prices. This task is also a logical precursor to the estimation of joint distributions potentially of interest to electricity market participants (e.g. electricity prices and natural gas spot prices).