بازده بازار و اجرت های در معرض خطر در قیمت های رو به جلو کوتاه مدت

Using recursive estimation and rolling windows over extended sample periods we examine the time-varying relationship between spot and short-term forward prices in the Pennsylvania–New Jersey–Maryland (PJM) wholesale electricity market. We examine theoretical models of forward risk premia in electricity markets and show that recent data do not provide support for existing models. The results indicate that short-term forward prices have converged towards unbiased predictors of the subsequent spot prices.

In forward markets the risk premium, defined to be the difference between the forward price and the expected spot price, plays a central role in understanding market dynamics. Since the introduction of financial electricity markets, the relationship between electricity spot and forward prices has been subject to growing interest among researchers and practitioners. The well-documented properties of electricity prices, such as strong seasonality and price spikes,2 have led researchers to develop equilibrium models to account for the stylized features of the underlying price process and to determine specific conditions that would induce a risk premium in forward prices. The seminal example of such a model is presented in Bessembinder and Lemmon (2002). For more than a decade, agents in the liberalized Pennsylvania–New Jersey–Maryland (PJM) market have had the opportunity to hedge against real time price fluctuations by taking positions in the short-term, day-ahead forward market. Longstaff and Wang (2004) use these short-term forward prices and the subsequent realizations of spot prices to test the model of Bessembinder and Lemmon (2002). Their results indicate that there exist significant risk premia in forward prices at PJM. As liberalized electricity markets have developed only recently, it is possible that risk premia have changed. We repeat the tests of Longstaff and Wang (2004) using the most recent data from PJM. We present results based on rolling windows and recursively extended samples. To our knowledge, this is the first such attempt in the literature. These analyses enable us to test for time variation in the relationship between short-term forward prices and realized spot prices, and thus indirectly allow us to examine potential market learning as well as time variation in market efficiency and/or risk premia. We make three contributions. First, our analyses show that the simple reduced form models of higher order moments of the spot price or by demand characteristics as in Bessembinder and Lemmon (2002) are not supported by the recent data. Second, in contradiction to previous studies, we find a striking lack of evidence for significantly biased predictions in recent short-term electricity forward prices at PJM. Third, we provide evidence that including other information known to market participants does not significantly improve forecasts of future spot price compared to forecasts based on the forward price alone. We conclude that either (1) market efficiency has increased, (2) risk premia are reduced, or (3) both, as agents have gained experience. The rest of this paper is organized as follows. Section 2 reviews risk premia in forward markets with a focus on electricity markets. Section 3 describes the data and presents preliminary analyses. Section 4 updates tests of the Bessembinder and Lemmon (2002) model using recent data. Section 5 provides empirical evidence on the relationship between spot and forward prices by using static, rolling, and recursive estimations. Section 6 summarizes and concludes.

In this paper we examine the time-varying relationship between spot and short-term forward prices at the eastern hub of the liberalized PJM electricity market. The main findings of this study can be summarized in the following. First, we use recent data and are unable to find support for the equilibrium forward pricing model of Bessembinder and Lemmon (2002). Rolling and recursive estimations reveal highly unstable parameter values and diminishing explained variance. Second, by extensive use of rolling windows and recursively increasing sample periods, we are able to document the dynamic properties of the data. The overall conclusion is that as the market has matured, the short-term forward prices has converged towards unbiased predictors of the subsequent spot prices. Tests of the unbiased forward rate hypothesis find no consistent evidence that constant and slope terms differ from zero and one, respectively. Third, we re-estimate the UFH including other information known to market participants. The results show very small improvements relative to forecasts using only forward prices. Our results imply that it is unlikely that agents can gain consistent economic profits by trading on recent past spot price or demand characteristics alone.