Not So Fast


Proving market performance requires detailed analysis.

Fortnightly Magazine - January 2010

Now that fuel prices have fallen recently from the highs seen in 2008 and wholesale electricity prices also have decreased, it might be tempting to attribute the lower prices to the restructuring of the wholesale electricity markets. Unfortunately, it’s a little more complicated than that.

To say fuel prices have been volatile in recent years would be a gross understatement. Natural gas prices in this decade have seen average monthly wellhead prices as low as $2.2 per thousand cubic feet (Mcf) in early 2002 to a high of $10.8/Mcf in June 2008. Then prices dropped again last fall to just below $3/Mcf in September 2009.1 At this writing in early December 2009, futures prices again are well north of $4/Mcf. Coal prices too have been volatile, particularly northern and central Appala-chia coal spot prices that more than tripled to above $140 a short ton, before dropping back to the pre-2008 levels in late 2009. Other coal prices also saw significant price increases on a percentage basis in 2008, followed by declines in 2009.

Power prices too have followed a similar pattern. Looking only at PJM’s annual average real-time market prices, they increased from $31.6/MWh in 2002 to just over $71/MWh in 2008.2 Monthly averages peaked in June 2008 above $100/MWh. In 2009, the monthly average prices have decreasedsteadily and were below $40/MWh for each month of April through September in 2009. These monthly averages were lower than any of the previous four years for the same months.

So what is the connection between fuel costs and electricity prices? Obviously there is a connection, but a few points should be kept in mind. First, natural gas has had a disproportionate impact on the price of electricity in RTO energy markets in some regions—that is, disproportionate to the amount of natural gas actually used to generate electricity. U.S. electric utilities saw natural gas costs increase from $2.6 per million Btu (MMBtu) in 1999 to about $10/MMBtu in the middle of 2008, a 285-percent increase. Coal costs for utilities also have increased, but at a relatively more modest pace, increasing from $1.2/MMBtu in 1999 to about $2/MMBtu for the same time period, a 67-percent increase. For PJM, average LMPs increased almost 127 percent from 1999 to the September 2008 price. During this time period, natural gas costs and the LMP seem to move up and down in near (but not complete) unison.

While natural gas cost appears correlated to the energy prices, gas accounted for only 7.3 percent of the generation in PJM during 2008; coal and nuclear sources accounted for almost 90 percent of PJM generation. This proportion of the fuels used to generate electricity has been about the same for several years.

The typical explanation for this disproportionate impact of natural gas on wholesale power prices is that natural gas often is the marginal fuel. In PJM, as in several other RTOs, the price for the units selected for dispatch is set by the highest offer price from a dispatched unit, or the marginal unit. During peak hours, relatively more expensive units are used to meet demand and often these units use natural gas. As a result, the wholesale price can climb quickly—and to hundreds of dollars per megawatt hour—when these units are dispatched.

However, while natural gas frequently might be on the margin, especially during peak hours, it isn’t the fuel that’s most often on the margin during the year in PJM; coal is on the margin for more hours. For total hours during the year in 2008, coal was the marginal fuel 78 percent of the hours, with natural gas 17 percent of the hours, and a mix of several different energy sources used for the remaining hours. Again, as with percent of generation, natural gas appears to have a disproportionate impact on the price of electricity.

A more detailed examination shows that fuel cost price changes don’t always precisely match the movement of electricity prices on a shorter time scale, for example, day-to-day or month-to-month.3 The two might even move in opposite directions, as they did in 2005 in the aftermath of Hurricane Katrina. Power prices were rising through the summer of 2005, but after Katrina hit at the end of August and natural gas prices soared the next couple months, power prices actually fell from September through November.

The explanation is that this is typically when load drops as the region moves into the cooler fall months. The dramatic run-up in natural gas prices was offset by the drop in electricity demand. Clearly, customer load changes are a significant factor in influencing electricity prices—and even can be more important than a fuel-cost change in the opposite direction, depending on the relative size of the changes. Overall, both demand and fuel costs are significant determinants of electricity prices.

Currently, electricity demand has been falling due to the economic slowdown. Overall, electricity sales have decreased 4.3 percent for the 12-month period ending in August 2009; industrial sales are down more than 11 percent for that same period. At the moment, fuel prices and demand have been moving in the same general direction, unlike the fall of 2005.

But what do these correlations say about the market’s performance? Can we draw any conclusions from this relationship and assess the market’s performance? In a word: No. Conclusions can’t be drawn about the competitiveness or performance of the market based solely on the assertion that the price variation is due to fuel costs or demand. Wholesale electricity price increases don’t mean generators have market power any more than price decreases mean they don’t have market power and the prices are the result of competitive forces. Observing electricity price changes and their correlation with fuel costs and demand aren’t a substitute for careful analysis of market performance.

Of course for any competitive industry, lower prices for an important input would be expected to decrease the price of the final product, assuming other factors are the same. But this is true also for a monopoly or oligopoly market structure. A monopolist also would be expected to lower prices for its final product when input prices decrease. Even if fuel costs and wholesale prices were perfectly correlated, it would say very little about the performance of the market.

Likewise, a decrease in demand also will lower the price of the final product—again whether the product is provided by a competitive firm or a mono- polist. Put the two factors together, fuel costs and demand, and it would be a big surprise if we didn’t see wholesale electricity prices falling; even a pure monopolist would behave that way.

Real Market Analysis

All RTOs that have been approved by the Federal Energy Regulatory Commission (FERC) have a market monitoring function. Most of these monitors use concentration measures such as the Herfindahl-Hirschman Index (HHI) and pivotal supplier indices. These measures are useful tools to characterize market structure. However, they are screening tools to decide whether further investigation is necessary. They don’t provide a definitive answer on the exercise of market power or of market performance any more than does correlating fuel costs and demand with market prices. Producing large volumes of concentration ratios and indices doesn’t compensate for the inability of these measures to determine the performance of a market being examined. Even worse, overreliance on such measures can lead to incorrect or misleading conclusions on market performance.

The PJM market monitor repeatedly has concluded that energy markets, capacity markets and ancillary services markets are competitive and that prices aren’t the result of the exercise of market power.4 This claim isn’t supported by the market monitor’s analysis, or any other publicly-available analysis.

To be sure, PJM’s market monitor goes further than do most RTO market monitors.5 In addition to the standard concentration measures, PJM’s market monitor also estimates the mark-up of price over marginal cost, as a percentage of price (similar to the Lerner index). However, the usefulness of this measure depends critically on the scope of the analysis—for example, the products and the geographic area that are selected—as well as the time period used, and how the marginal cost is estimated. Simply put, if the measure is too broad or aggregated, the results will be of little or no value. Also, the quality of the data obviously will affect the results, and the estimation of the marginal cost presents perhaps the biggest challenge for the market analyst.

There are two basic problems with this approach and relying on it to make pronouncements of market competitiveness. First, even if a supplier were exercising unilateral market power, this approach wouldn’t necessarily detect it since the focus is on determining whether the marginal unit is operating at or near its marginal cost. Ignoring problems of estimating the marginal cost and assuming an approximate value can be determined, a firm exercising market power that’s able to raise the price above a competitive level wouldn’t need to raise the offer price above that unit’s marginal cost to enjoy higher prices than a competitive outcome would produce. This is because a strategy can be used of withholding capacity to cause a higher marginal cost unit to be dispatched and set the price, even if all units were offered at their respective marginal costs. The mark-up estimation value would look insignificant, but unilateral market power was exercised nonetheless.

This leads to the second problem with relying on this approach to decide whether markets are competitive; it can’t detect strategic behavior suppliers might use to raise the price above a competitive level. Academic studies using simulations or experiments with students has suggested that even with a relatively large number of suppliers, suppliers can, over time with repeated iterations, learn what others are doing and adjust in ways to raise the price above a competitive level. This doesn’t require direct—and illegal—explicit collusion, but because of the nature of the market conditions (i.e., repeated bidding, some knowledge of others’ actions in certain circumstances, etc.), suppliers can lean and tacitly collude—that is, without coordin- ated activity—even when they have no unilateral market power themselves. Market concentration measures and the way markup estimates are calculated simply can’t detect strategies that raise prices above competitive levels.

While PJM has made a considerable amount of price and demand data available to the public, PJM doesn’t release the data that would be required to conduct an analysis of strategic bidding behavior or whether that bidding has been successful and, if so, how much it has affected prices. FERC also has access to this data, but hasn’t and doesn’t conduct analysis in sufficient detail. These data also should be made available to the states to conduct their own analysis. While state commissions and others have tried to obtain this information, with the promise of not releasing the results in a form that may reveal the identity of individual suppliers, to the author’s knowledge these data never have been provided.

FERC—not PJM or any other RTO—bears the responsibility to require this information be released to the states. Once the data are released, it would be up to FERC and the states to ensure that independent analysis is conducted.

Better Late Than Never

Care needs to be given when considering the impact of fuel costs on electricity prices. Fuel costs and demand are correlated with electricity prices, but this correlation says nothing about the perfor- mance of wholesale electricity markets. The way to determine the market’s performance is through more detailed and independent analysis—something that isn’t available at this time. FERC can and should require the necessary data to be made available to the states and allow them to conduct their own independent analysis.

In many respects we need serious analysis now more than ever. Demand is down, fuel prices are lower, and electricity prices are lower too. It’s a good time to take a careful look and make a rational assessment, rather than waiting for the next run-up of fuel prices or whatever predicament the future holds in store.

Also, with the industry perhaps being required to make a serious attempt at reducing its carbon-dioxide emissions soon and quickly, a thorough examination would be especially useful at this time, so that customers don’t pay more than necessary to accomplish goals set by policymakers.

This analysis is late and overdue, but it isn’t too late.



1. The national price and sales figures used here are derived from U.S. Department of Energy’s Energy Information Administration data.

2. PJM price figures are based on data from PJM’s market monitor, Monitoring Analytics.

3. This is discussed in more detail in “The Impact of Fuel Costs on Electric Power Prices,” June 2007, prepared for the American Public Power Association (APPA).

4. Since 2008, PJM’s market monitoring has been conducted by Monitoring Analytics. Their analysis is available here.

5. A notable exception to this is the analysis of the California power markets, particularly after the Western power crisis that produced some of the best analysis of any market to date.