Market Impacts of a Nuclear Power Plant Closure in California

Nuclear Power Industry

Lucas Davis and Catherine Hausman (2015) reported that falling revenues and rising costs have put U.S. nuclear plants in financial trouble, and some threaten to close. To understand the potential private and social consequences, they examined the abrupt closure of the San Onofre Nuclear Generating Station (SONGS) in 2012. Using a novel econometric approach, they showed that the lost generation from SONGS was met largely by increased in-state natural gas generation. In the twelve months following the closure, natural gas generation costs increased by $350 million. The closure also created binding transmission constraints, causing short-run inefficiencies and potentially making it more profitable for certain plants to act non-competitively.

Nuclear power has historically supplied a substantial portion of electricity – 20 percent in the United States and 14 percent worldwide for 2000 to 2012. As recently as 2008, the outlook for the industry was robust, with nuclear plants earning large profits. Since 2009, however, prospects for nuclear power – even at existing facilities – have substantially waned, with the closure of several large facilities and predictions of more closures to come (EIA 2014). As they described in detail, multiple factors have contributed to the recent closures of nuclear plants.

Peak wholesale electricity prices fell around 50 percent in real terms from 2007 to 2012, a result of both falling natural gas prices and stagnant electricity demand. At the same time, costs for nuclear plants have been rising, a combination of rising wages and fuel prices, stricter safety regulations, and the aging of decades-old equipment.
To many observers, low profitability at existing nuclear plants is surprising, since the marginal cost of generation is very low at nuclear plants. However, while marginal costs hour-to-hour are low, fixed operating costs (e.g., keeping employees on staff) are high. Total operations and maintenance (O&M) costs at U.S. nuclear plants have increased by about 20 percent in real terms since 2002 and today are more than twice as high as O&M costs at natural gas plants. These higher costs reflect the fact that nuclear plants have substantially higher requirements for safety, security, and testing.

In this paper, they used evidence from a nuclear power plant closure to examine the rapidly evolving economics of nuclear power and to assess the potential private and social consequences of plant closures. While in operation, the San Onofre Nuclear Generating Station (SONGS) generated an average of 16 million megawatt hours of electricity annually, making it the second largest electric generating facility in California. During this period, SONGS generated enough electricity to meet the needs of 2.3 million California households – about 8 percent of all electricity generated in the state. SONGS was closed abruptly in February 2012, when workers discovered problems with the plant’s steam generators. Although it was not known at the time, SONGS would never operate again.

The first-order effect of the plant’s exit was a large inward shift of the electricity supply curve. Like other nuclear power plants, SONGS produced electricity at very low marginal cost. Consequently, the plant was always near the bottom of the supply curve, operating around the clock and providing a consistent source of electricity. When SONGS was closed, this generation had to be made up for by operating other generating resources with higher marginal cost. We use rich micro-data from a variety of sources and a novel econometric Peak wholesale prices at various hubs for ICE contracts; source: EIA. Prices throughout are deflated tomethod to identify those marginal resources that would be expected to increase production.

We find that the lost generation from SONGS was met largely by in-state natural gas plants. Bringing these additional plants online cost an average of $63,000 per hour in the twelve months following the closure. The SONGS closure also had important implications for the environment, increasing carbon dioxide emissions by 9 million tons in the first twelve months. To put this in some perspective, this is the equivalent of putting 2 million additional cars on the road.

There was also a second-order, but not insignificant, additional impact on the market. SONGS was even more valuable than these numbers suggest because of its location between Los Angeles and San Diego, two enormous demand centers. Although there is transmission that connects Southern California to the rest of the state, the capacity is limited. Prior to the closure, transmission capacity between Northern and Southern California was almost always sufficient, so that wholesale prices equalized in the two regions during the vast majority of hours. However, beginning with the closure in 2012, we document a substantial divergence in prices between Northern and Southern California. This binding transmission constraint and other physical constraints of the grid meant that it was not possible to meet all of the lost output from SONGS using the lowest cost available generating resources.

These second-order effects are reflected in our model as “residuals,” measured as deviations from predicted plant behavior. We find that during low demand hours, the change in generation closely follows predictions based on pre-closure behavior, with about half of the increased generation coming from Southern California and the other half coming from Northern California. During high demand hours, however, we find significant residual effects: higher cost generating units coming online more than predicted. In high demand hours in 2012, we find that as much as 75 percent of the lost generation was met by plants located in Southern California. On average, these constraints increased generation costs by an average of $4,500 per hour, implying that the total cost of additional natural gas generation was almost $68,000 per hour in the twelve months following the closure.

These residuals also potentially reflect non-competitive behavior. Tight market conditions make it more profitable for certain firms to exercise market power, and using our model they are able to determine which individual plants changed their behavior the most after the SONGS closure. Because of the transmission constraints, the largest positive residuals are at Southern plants, and the largest negative residuals are at Northern plants. Surprisingly, they also found large negative residuals during high demand hours at two Southern plants: Alamitos and Redondo, both owned by the same company. This was unexpected but, as it turns out, not coincidental. The Federal Energy Regulatory Commission recently alleged market manipulation at these plants over the period 2010 to 2012, for which JP Morgan paid fines of over $400 million. The fact that the results clearly identified these two plants suggests that our approach may serve as a useful diagnostic tool. Although a large residual effect does not prove that a plant is exercising market power, it is a good indicator of unusual behavior.

Overall, they found that the SONGS closure increased generation costs at other plants by $350 million during the first twelve months. This is a large change, equivalent to a 13 percent increase in total in-state generation costs. Annual O&M costs at SONGS were about the same amount, so the decision to close probably made sense from a private perspective. Incorporating externalities makes it less clear. Their estimates of the increase in carbon dioxide emissions imply external costs of almost $320 million during the first twelve months. If plant closure decisions are to be made efficiently, it is important that these environmental impacts be taken into account. Historically, state and federal policies aimed at decreasing carbon emissions have not been designed to incentivize nuclear plants.

Their paper contributes to several strands of the literature. Several recent papers have focused on the prospects for nuclear power, particularly as concern about climate change hasincreased (Joskow and Parsons, 2009; MIT, 2009; Davis, 2012; Joskow and Parsons, 2012; Linares and Conchado, 2013). However, the literature has almost exclusively focused on the outlook for new nuclear plants. The decision to enter the market is quite different from the decision to exit. Entry decisions are driven in large part by the large capital (and financing) costs of nuclear plants. In contrast, because these costs are sunk for existing nuclear plants, exit decisions are driven by wholesale electricity prices and operating expenses. With construction of new nuclear almost completely halted, we argue that exit will be the more
policy-relevant margin for the foreseeable future.

Their paper also adds to a small literature on the value of geographic integration in electricity markets (Mansur and White, 2012; Birge et al., 2013; Wolak, 2014a; Ryan, 2014). Economists have long written about the importance of transmission constraints, but previous studies have either used stylized theoretical models (Cardell, Hitt and Hogan, 1997; Joskow and Tirole, 2000), or Cournot simulations (Borenstein, Bushnell and Stoft, 2000; Ryan, 2014), rather than econometric analysis. Their methodology is novel, because it quantifies the impact of transmission constraints without requiring strong assumptions about the firms’ objective function or an explicit representation of the physical constraints of the electric grid. While their estimates are not directly applicable to other markets, theysee broad potential for applying this general method elsewhere. Their approach relies entirely on publicly-available data, so it would be relatively straightforward to perform similar analyses in other markets, both for
quantifying the impacts of large changes in generation and transmission infrastructure, and for detecting unusual changes in firm behavior.