Inefficiency in deregulated wholesale electricity markets: the case of the New England ISO

Abstract

The efficiency of the New England Independent System Operator electricity market is examined. Using peak hour prices from March 2003 to February 2007, significant differences between Day Ahead (DA) and spot prices are found, indicating that profitable arbitrage opportunities existed in the DA market. It is also found that persistence of price volatility is higher in the DA market, indicating that information is not incorporated into DA prices as quickly as it is in the spot market.

Notes

¹In a similar vein, Von Hirschhausen et al. (2006) use data envelopment analysis to examine the efficiency of German electricity distribution utilities.

²Wholesale spot markets on ISO-NE began operation in 1999. A DA market was added in 2003.

³ Annualized volatility = SD of daily returns (from Table 2) × 250^1/2.

⁴ In the empirical work below, lagged values of R_t were added to the mean equation to correct for autocorrelation as identified by inspection of the Ljung–Box Q-statistics. This correction had, generally, a minor effect on the ARCH and GARCH terms.

⁵The monthly dummies generally were not statistically significant, so presentation of only the coefficients on the other variables in the variance equation is shown. Spot monthly effects were statistically significant for just the two northern-most zones, Maine and New Hampshire, for February, April, May and September to December. The SEs are robust heteroskedastic consistent estimators using the methods described by Bollerslev and Wooldridge (1992).

⁶The alternative model of Bessembinder and Lemmon (2002) considers an electricity market that has no outside participants. That is, one in which prices are determined solely by generators and distributors of electricity. The premium in their model can be positive, negative or zero, depending on the properties of spot prices in the delivery period. Their model is summarized

where F _t,T is the futures price at time t, E(S_T ) is the expected average wholesale spot price during the delivery period, Var(S_T ) and Skew(S_T ) are the variance of the skewness of wholesale spot prices during the delivery period, and α and γ are functions of the number of generators and retailers, marginal cost, wholesale price and retail price. The premium, E(S_T ) − F_t,T , is [−αVar(S_T ) − γSkew(S_T )]. Bessembinder and Lemmon (2002, p. 1359) explain that in this model, α is strictly negative. Thus, the forward price is downward biased relative to the expected spot price and the premium is positive as the variance increases. The downward bias in the forward price in this case reflects retailers’ net hedging demand. The parameter γ is positive, implying that, ceteris paribus, the forward price increases with the skewness of spot prices, reducing the premium. Thus, if the participants in the electricity futures markets are generators and retailers, it is unclear whether the premium will be zero, positive, or negative. In a market such as the ISO-NE, with more than 260 participants and virtual trading comprising roughly 20% of all bids, one would expect the market to reflect a great deal of openness. Thus, Bessembinder and Lemmon's model seems inappropriate.

⁷Zalcman (2007) reports that in New York State particulate matter from power plants causes 1200 premature deaths and 2500 heart attacks per year. He also reports that the electricity industry nationally, in total (not just the inefficient peak generators), is responsible for 67% of sulphur dioxide emissions (acid rain), 25% of nitrous oxide emissions (smog), 40% of carbon emissions (global warming) and 33% of mercury emissions.