Different angle
Banks, underwriters, rating agencies and all other analysts
focusing on the prospects for merchant power investments are
revising the ways in which these risks are assessed. A merchant
power plant is not simply a moment to moment seller of power into
the spot market. Treating power plants like any other ?widget
maker? ? and focusing on predicting the price of widgets ? does not
apply here. In any event, predictions about the spot price of power
over the past three years have been too far off to be reliable.
Instead an analysis of the prospects for merchant power has to be more economic, focusing on the factors affecting supply/demand balance in particular geographic regions. The point is to shift the attention from price forecasts in market consultants' reports to the information on marginal fuels, capacity makeup, and average regional heat rates.
Now that three summers of market experience is available from most US regional markets, fairly reliable correlations have become observable between power prices and fuel prices, weather, forward markets, market heat rate averages, and the like which are more useful in analyzing the prospects for merchant projects, than mere price forecasts. A backward-looking tour of four US power markets reveals several correlations and trends that have positive as well as negative implications for financings and investments in merchant power assets. A few of the following economic considerations emerge:
1. Forecasting the price of power is less important than forecasting the relationship between power prices and fuel prices (i.e., the ?spark spread?).
2. In assessing regional Supply/Demand balance, comparing ?Reserve Margins? is less useful than noting the seasonality of weather or the presence of hydro power from region to region.
3. The ?volatility value? of a merchant power asset, as distinguished from its mere commodity value, is clearly observable in the relationship of spot and forward markets.
Critical Economic Factors
In reviewing how the price of power has behaved, several factors appear to be critical, but the most important is the supply/demand relationship in each market. Since electricity cannot be effectively stored, the balance between supply and demand is very volatile depending upon the region. High reliance on hydropower will make the effective supply side of the equation relatively volatile; a climate with wide swings in temperature (and consequently power use) will increase the volatility of the demand side of the equation.
Hot weather, the absence or presence of hydro capacity and the like, are short-term conditions that will induce temporary power scarcity and price spikes. The most remarkable feature of spot power markets in the US over the past three years have been the unexpectedly high price spikes which have appeared when power was scarce ? mostly in the summer. However, this has been a relatively infrequent condition; most of the time the price of power has been driven by cost rather than scarcity.
For this reason, the most important factor to note over the past year was the run-up of gas prices and its effect on power markets. Gas prices during the summer of 2000 were twice as high as the previous summer. There was an extraordinary run-up through the winter of 2000-01 to as high as $10/MMbtu, but a steady moderation of prices since then. The price prevails now (September 2001) in the $2.5-3.1/MMbtu range or even lower. This is well below the summer of 2000 ($4-4.5/MMbtu), but still above the $2-2.4/MMbtu price, which had prevailed for years before 2000. Gas is a winter peak fuel in the US, so more volatility in the coming months is likely.
In Tables 1, 2 and 3, below, are comparisons of spot market power prices for three different US regions: NEPOOL, PJM, and ERCOT. These markets have all demonstrated how scarcity induced by sudden or sustained increases in power demand (or ?load?) will cause price spikes ? when the power price is detached from the cost of production. Natural gas is the principal marginal cost of production in NEPOOL and ERCOT, and coal is the marginal fuel for PJM. The graphs also demonstrate how the cost-based power prices in these markets have been lifted over the past two years by the run-up in fuel prices.
The NEPOOL market is not materially dependent on hydropower, which minimizes volatility on the supply side, but the region experiences extremes in temperature, which raise the demand-side volatility. For this reason, a sudden temperature increase or a weeklong period of +90? weather will lead to load spikes, scarcity and consequently price spikes. Yet, most of the time, power is not scarce, and its price is driven by the marginal cost of the next-most expensive generator being dispatched by the market ? which is largely reliant on gas-fired generators. Consequently, NEPOOL experiences a high correlation between power prices and natural-gas prices. The effect of the run-up in gas prices over the summer to winter 2000 is quite apparent as power prices elevated in the region, and significant price spikes were induced during the winter of 2000-2001 in addition to the normal summer pattern.
The PJM sub-region includes Pennsylvania New Jersey and Maryland and is one of the country's largest in terms of generating capacity with over 58,000MW. Geographically, PJM is near NEPOOL and has similar seasonal weather patterns. Although the climate of PJM may be comparable to that of NEPOOL, the region's capacity makeup tends to differentiate its price trends. PJM's has a substantial base load generating capacity comprised of approximately 22% nuclear, 31% coal, and 5% hydro. Of the remaining 17%, only 5% is natural gas fired. It is coal that drives the market price for the majority of the time in PJM.
The ERCOT market, in contrast to the previous two, is a ?bilateral market?, functioning without a power pool or transparent exchange. Participants in this Texas market independently schedule and coordinate dispatch as well as set the market price through bilateral trades. Similar to PJM, ERCOT is a significantly large electric region consisting of a capacity base of over 65,000MW. Unlike PJM, ERCOT's capacity has a significant share of gas-fired generators, which comprise about 40% of the region's generating capacity; additionally, the region lacks hydropower and has minimal nuclear capacity making it even more dependent on natural gas. ERCOT diverges from NEPOOL and PJM with regards to its climate; ERCOT has more homogenous weather patterns with consistently higher temperatures throughout the year. In other words, temperatures above 85? are so common their occurrence does not necessarily induce power scarcity. Table 3 shows ERCOT's peak and off-peak prices since 1999. The data was compiled through Platts which obtains price data via daily surveys of energy traders in the region. Since the prices are daily (vs. hourly), volatility and price spikes are less extreme than the hourly spot prices in NEPOOL and PJM. Nevertheless, the data does however support the notion of evolving market efficiency as one can notice far fewer daily price spikes as one moves along the timeline.
Duration Curves
An alternative presentation of the price data is provided by duration curves, which portray prices as a function of time ? the duration in order of magnitude for which a certain price prevails in the market. This approach makes it easier to identify trends. To control for the weather seasonality, which is a variable for power demand, Tables 4, 5 and 6 below are focused on the summer months (through July 31) during which power markets experience the greatest and most numerous price spikes.
The left side of the graph in Table 4 (less than 16% on the x-axis) collects the price spikes and compares how the price of power was driven by scarcity during each of the summers. The balance of the graph compares how the cost-driven power price behaves each summer. In this latter regard, the higher natural-gas prices plainly explain the elevation of the 2000/01 prices for the cost-based portion of the duration curve over those prevailing in the summer of 1999.
The left side of the graph needs to be analyzed in conjunction with weather, which are a far more significant variable affecting periods when price is determined by scarcity. The summer of 2000 was one of the mildest on record, which probably explains why price spikes were least frequent during that summer. On the other hand, the summer of 2001 has been the most extreme of the three (i.e., highest number of +85? days), and yet the occurrence of price spikes this summer has been less frequent than in 1999. This may be evidence of several variables that are reducing power scarcity in NEPOOL: capacity additions, increased use of hedging, improved plant scheduling and other examples of increased market efficiency.
Table 5 below shows that prices in PJM were far less responsive to the gas run-up. Most notable in comparing the summer duration curves above is the fact that the doubling of gas prices between 1999 and 2000 did not raise power prices during the cost-based portion between the two summers in PJM ? summer 2000 power prices were generally lower than 1999. Prices have experienced a general elevation in 2001, but this may have more to do with the elevated coal prices. The left side of the duration curve comparison shows that power scarcity is becoming less frequent in this market. Price spikes were relatively infrequent in PJM this summer, even though PJM experienced higher temperatures and peak demand, which surpassed those experienced in either of the previous two years. The trend toward less power scarcity, and greater market efficiency, is apparent.
Table 6 above shows the striking summer duration curve comparison for ERCOT. The dissipation of price spikes each successive summer has truly marginalized the slope of the duration curve by 2001. This may be preliminary evidence of how capacity additions in a semi-tropical climate can virtually eliminate for a period the occasions when spot power prices are determined by scarcity rather than cost. The modest slope of the duration curve for summer 2001 resembles the power price duration curve for Spain and the U.K., other markets without extremes in temperature where the supply/demand balance has been weighted toward the supply side through recent capacity additions. Meanwhile the cost-driven portions of the three curves in Table 6 appear to rank themselves consistently with the gas price experience: highest during summer 2000, followed by 2001, then 1999. This provides further confirmation of the responsiveness of power prices to gas price fluctuations in the ERCOT region.
Implications
The implication of all this is that analyzing power markets is a lot more complex than thinking about the average market price of power. In PJM for example, the average market price of power for the twelve months ending 3/31/00 and 3/31/01 was $29.54/mwh and $30.31/mwh respectively. These prices are practically the same, and only a bit higher than consultants forecast in 1999. However, the distribution of prices that make up these averages is quite different (more price spikes in period ending 3/31/00, but cost-based prices at a higher level in the later period ending 3/31/01). So, even though the average market price of power is almost the same for both periods, peaking plants are clearly better off in the earlier period, base load coal plants clearly better off in the later period, however base load gas plants are worse off in the later period ? because fuel costs were higher.
The critical issue is the margin between power prices and the marginal operating costs, the ?spark spread?. The prospects for spark spreads have to be analyzed for two periods: the infrequent times when they are driven by scarcity, and the much more common times when they are driven by costs.
The market experience above shows that the general trend is for power markets to become more efficient, and for scarcity to become less frequent. Regional differences will affect this trend. The regions with seasonal weather are more likely to have some amount of scarcity in their markets ? this will be particularly true of those regions that are dependant on hydro power or nuclear plants that may become suddenly unavailable. For this reason, comparing reserve margins is too simplistic as an indication of the supply/demand balance in a power market.
Analyzing the prospects for spark spreads during cost-driven periods is more critical for investments intended to be base or intermediate load. A comparison of the cost-based portions of the duration curves in the regions above does confirm how responsive power prices are to gas costs in markets like NEPOOL and ERCOT where gas is the marginal fuel. This means that all things being equal, fuel cost increases will increase the spark spread, and consequently gas plant profitability in absolute terms; fuel cost decreases would have the opposite effect. But all things are not equal. The greatest concern for investments in generating assets is that too many capacity additions (?overbuild?) will reduce spark spreads regardless of fuel price fluctuations.
The effect of efficient capacity additions on the cost-driven side of the duration curve needs to be analyzed. In theory, the cost-based power price in a gas-driven market will for the most part be determined by the heat rates of the various plants being dispatched into the market and their relative shares of the market. For any given period, a market will have an average heat rate based on the actual weighted average heat rates being dispatched ? something hard to know ? but the databases in market consultants' models use these assumptions to forecast power prices. An implicit market heat rate can be observed by comparing the price of power and the price of fuel prevailing in markets for a given period. This provides a good indication of what that average heat rate might be. In any event, the average rate that is converting fuel costs to power prices is being lowered by capacity additions with 7,000 heat rates. This stands to reason, since so much of the existing capacity being displaced consists of older dual fueled plants with higher (less efficient) heat rates. The reduction of the dispatch factors for these plants will lower the average heat rate of a given market. However, since virtually no one is building or planning capacity additions with heat rates materially better than 7,000 heat rates, a market's average or implicit heat rate could only be lowered so far.
The California Phenomenon
California behaved much like a gas-driven, seasonal weather market before the summer of 2000. After that time, this market had several unique circumstances, and certainly unique spot power price behavior, which caution against projecting forward trends based on a backward-looking analysis here. Looking back, the exceedingly high spot prices which prevailed through the summer, fall and winter of 2000, are linked to five factors which are not likely to recur jointly in the medium term:
1. A fundamental imbalance between demand for power and capacity in the California market;
2. A draught which aggravated the supply/demand imbalance by removing much of the hydro capacity (normally 25%) from the market,
3. An exceedingly hot summer which pushed-up load;
4. The run-up of gas prices (noted already), multiplying fuel expenses five times between June and December 2000;
5. Market regulations, which required the three major utilities to purchase all of their requirements on the spot market which was the California PX.
The fifth factor ? perhaps the most critical to the financial side of the California energy crisis ? has been removed from the market. Of the other four, the first and second are still true, but the others have not been applicable during the past four months, and the spot power prices in California have moderated considerably. It would be premature to examine much further the trends for power prices in the California market; the variables noted above make the recent data too unreliable.
Lessons from Spot Markets
Despite the anomalies presented by the California experience, a backward-looking analysis of other spot power markets reveals predictable correlations: between power prices and the factors affecting scarcity, and between power prices and fuel costs. On a practical level lenders analyzing the prospects for ?over build? in a particular market need to forecast a duration curve with diminished price spikes, and contracted spark spreads. This is a much more useful approach than giving arbitrary ?hair cuts? to revenue forecasts.
As the two regions most mentioned as at risk for overbuild are gas-driven ERCOT and NEPOOL, a forecast of the duration curves in those regions needs to be informed by a forecast about gas prices. A fair ?downside? forecast would assume reasonably low gas costs and a conservative implicit heat rate for the two markets. With the help of consultants' databases, and another summer of experience, a fairly conservative downside margin in the range of $60-50/kW-year could be calculated for these regions for 7,000-heat rate plants, lower for the less efficient plants. This scenario will not be fatal to the low heat rate plants of these regions.
Forward Markets:
The discussion above has focused on the information provided by the spot markets, however bulk electricity is also purchased through forward markets ? both bilaterally and through exchanges. Because power prices have been so volatile in the summer months, for reasons explained above, most of the forward market trading is focused on hedging the summer price of power. Table 7 above is one way of presenting how forward markets have behaved in PJM.
Layered on top of one another in the graph above are the forward price-lines prevailing for various times over the past three years. Each distinct line depicts the settlement prices for contracts entered into during a certain bimonthly date (the left-most point on the line); each point along the line corresponds to the settlement date and price of the contracts. For example, the cross-hatched line corresponds to all contracts that were entered on May 23, 2000 for settlement during the following 18 Months. The first point along the line corresponds to a settlement price of $98/MWh to be executed during the month of June 2000; the last point along the line corresponds to a contract entered into on May 23, 2000 with a settlement price of $29.50/MWh and a settlement date of November 2001. The chart shows the high concentration of futures market activity that exists during the May-Aug summer months. Since the summer months are when demand is at its peak and power supply is scarce, it is during these months that most activity and hence volatility takes place.
As a general observation, it is evident that the market anticipated high prices leading up to the summer of 2000, but forward prices were contracted at lower levels leading up to the summer of 2001, with prices being contracted at still lower prices for next summer. This all follows logically from what actually occurred in the spot markets for PJM. The power scarcity and price spikes were highest in the summer of 1999 ? which (along with a uniquely high-priced day in early May 2000) contributed perhaps to the over-hedging in forward markets leading up to the summer of 2000. The summer of 2000 turned out to be such a mild spot market in PJM that forward prices anticipating summer 2001 moderated a bit ? although the run-up in fuel costs would have been a price-lifting factor here.
Table 8 below concentrates on the activity of the PJM futures market leading to the summer months. Each line depicts the settlement prices for contracts to be executed in either June, July or August, which were transacted during the months leading up to the summer corresponding to the date on the X-axis coordinate. For example, the dotted line illustrates the prices for transactions to be settled in August 2000. The leftmost point along the line corresponds to a transaction entered into on December 7, 1999, which will be settled in August 2000 at a contract price of $87/MWh. It's noteworthy that by 2001, the forward markets are not making a distinction between power delivered in July or August, so those prices appear to have merged into a single line.
A few generalizations can be made about the above comparison. Price swings are common leading up to each summer, but there is a variation in the amount of volatility in each of the years being observed. The most volatile upward trending forward market (spring 2000) was followed by the least volatile spot market (summer 2000). Power market consultants have predicted that price volatility would migrate between spot and forward markets. This may be evidence that over-hedging will moderate spot markets. In California during the spring of 2001, the California Department of Water Resources (DWR) pursued medium and long-term contracts for most of its power requirements (at prices raging from $180-288/kW-yr). By the summer of 2001 (which enjoyed mild weather), DWR was reselling this power into the spot market for on-peak prices in the range of $50 ? $70/MWh. Pacific Corp has also reported losses from over-hedging in the spring of 2001.
Volatility can be induced by a combination of hot weather and under-hedging. The forward markets leading up through the summer of 2001 for PJM (NEPOOL for that matter) were noticeably flatter than the prior year. Consistent with the theory that volatility migrates between spot and forward markets, late summer heat (July-August) did produce a revival of price spikes in the PJM and NEPOOL spot markets.
Table 9 below tracks the forward market price of power for Entergy-SERC prior to the past three summers. The trends are generally the same as described for PJM above with more volatility and generally higher price levels. The absence of a transparent spot market in SERC could be one explanation for this. Natural gas is much more of a marginal fuel for the Entergy portion of SERC (than in PJM), and this would also elevate forward prices and their volatility in the 2000-01 period
Focus on Volatility Value:
As mentioned already, a merchant power plant generates more value for investors than the simple moment-to-moment sale of energy into a power exchange. Power plants, and particularly portfolios of power plants, give their investors a natural long position in a highly volatile energy market, which represents other sources of value apart from mere physical commodity sales. The table below lists some of the sources of value to be realized from typical gas-fired plants; these sources are referred to by many investors as ?Volatility Value? or ?Optionality? because the common merchant power plant operator has more options besides commodity sales into the spot power market.
A look back at how regional forward markets for power behave illustrates the profit potential of trading in forward markets from a physical position. The routine price swings of summer power noted above would provide ample opportunity for profitable power trading from a physical position. If the forward sale cannot be covered profitably with purchased power, the generator can ultimately cover with a physical delivery (provided the fuel cost has been properly hedged).
Most structured financings of merchant power assets have not been based upon revenue streams that included significant volatility value or optionality value in the cash flow forecasts. In many financing structures, the SPV established to hold the merchant assets is selling all of its power and capacity to the power marketing affiliate of the sponsoring company. This is regarded as a ?safer? structure by many lenders because it isolates the risk of trading outside of the borrower. However, these types of structures exclude the value of optionality along with its possible risks.
Trading from the natural long position that generation ownership confers can be a relatively safe activity when the proper contracts and procedures are in place. The generator should be able to profit from its natural option on the spread between gas and electricity. Backward-looking evidence of volatility value can be found in the markets described above. Broader acceptance of optionality or volatility value by lenders or investors in the merchant power industry is also likely to be driven by earnings records as this industry matures.
Instead an analysis of the prospects for merchant power has to be more economic, focusing on the factors affecting supply/demand balance in particular geographic regions. The point is to shift the attention from price forecasts in market consultants' reports to the information on marginal fuels, capacity makeup, and average regional heat rates.
Now that three summers of market experience is available from most US regional markets, fairly reliable correlations have become observable between power prices and fuel prices, weather, forward markets, market heat rate averages, and the like which are more useful in analyzing the prospects for merchant projects, than mere price forecasts. A backward-looking tour of four US power markets reveals several correlations and trends that have positive as well as negative implications for financings and investments in merchant power assets. A few of the following economic considerations emerge:
1. Forecasting the price of power is less important than forecasting the relationship between power prices and fuel prices (i.e., the ?spark spread?).
2. In assessing regional Supply/Demand balance, comparing ?Reserve Margins? is less useful than noting the seasonality of weather or the presence of hydro power from region to region.
3. The ?volatility value? of a merchant power asset, as distinguished from its mere commodity value, is clearly observable in the relationship of spot and forward markets.
Critical Economic Factors
In reviewing how the price of power has behaved, several factors appear to be critical, but the most important is the supply/demand relationship in each market. Since electricity cannot be effectively stored, the balance between supply and demand is very volatile depending upon the region. High reliance on hydropower will make the effective supply side of the equation relatively volatile; a climate with wide swings in temperature (and consequently power use) will increase the volatility of the demand side of the equation.
Hot weather, the absence or presence of hydro capacity and the like, are short-term conditions that will induce temporary power scarcity and price spikes. The most remarkable feature of spot power markets in the US over the past three years have been the unexpectedly high price spikes which have appeared when power was scarce ? mostly in the summer. However, this has been a relatively infrequent condition; most of the time the price of power has been driven by cost rather than scarcity.
For this reason, the most important factor to note over the past year was the run-up of gas prices and its effect on power markets. Gas prices during the summer of 2000 were twice as high as the previous summer. There was an extraordinary run-up through the winter of 2000-01 to as high as $10/MMbtu, but a steady moderation of prices since then. The price prevails now (September 2001) in the $2.5-3.1/MMbtu range or even lower. This is well below the summer of 2000 ($4-4.5/MMbtu), but still above the $2-2.4/MMbtu price, which had prevailed for years before 2000. Gas is a winter peak fuel in the US, so more volatility in the coming months is likely.
In Tables 1, 2 and 3, below, are comparisons of spot market power prices for three different US regions: NEPOOL, PJM, and ERCOT. These markets have all demonstrated how scarcity induced by sudden or sustained increases in power demand (or ?load?) will cause price spikes ? when the power price is detached from the cost of production. Natural gas is the principal marginal cost of production in NEPOOL and ERCOT, and coal is the marginal fuel for PJM. The graphs also demonstrate how the cost-based power prices in these markets have been lifted over the past two years by the run-up in fuel prices.
The NEPOOL market is not materially dependent on hydropower, which minimizes volatility on the supply side, but the region experiences extremes in temperature, which raise the demand-side volatility. For this reason, a sudden temperature increase or a weeklong period of +90? weather will lead to load spikes, scarcity and consequently price spikes. Yet, most of the time, power is not scarce, and its price is driven by the marginal cost of the next-most expensive generator being dispatched by the market ? which is largely reliant on gas-fired generators. Consequently, NEPOOL experiences a high correlation between power prices and natural-gas prices. The effect of the run-up in gas prices over the summer to winter 2000 is quite apparent as power prices elevated in the region, and significant price spikes were induced during the winter of 2000-2001 in addition to the normal summer pattern.
The PJM sub-region includes Pennsylvania New Jersey and Maryland and is one of the country's largest in terms of generating capacity with over 58,000MW. Geographically, PJM is near NEPOOL and has similar seasonal weather patterns. Although the climate of PJM may be comparable to that of NEPOOL, the region's capacity makeup tends to differentiate its price trends. PJM's has a substantial base load generating capacity comprised of approximately 22% nuclear, 31% coal, and 5% hydro. Of the remaining 17%, only 5% is natural gas fired. It is coal that drives the market price for the majority of the time in PJM.
The ERCOT market, in contrast to the previous two, is a ?bilateral market?, functioning without a power pool or transparent exchange. Participants in this Texas market independently schedule and coordinate dispatch as well as set the market price through bilateral trades. Similar to PJM, ERCOT is a significantly large electric region consisting of a capacity base of over 65,000MW. Unlike PJM, ERCOT's capacity has a significant share of gas-fired generators, which comprise about 40% of the region's generating capacity; additionally, the region lacks hydropower and has minimal nuclear capacity making it even more dependent on natural gas. ERCOT diverges from NEPOOL and PJM with regards to its climate; ERCOT has more homogenous weather patterns with consistently higher temperatures throughout the year. In other words, temperatures above 85? are so common their occurrence does not necessarily induce power scarcity. Table 3 shows ERCOT's peak and off-peak prices since 1999. The data was compiled through Platts which obtains price data via daily surveys of energy traders in the region. Since the prices are daily (vs. hourly), volatility and price spikes are less extreme than the hourly spot prices in NEPOOL and PJM. Nevertheless, the data does however support the notion of evolving market efficiency as one can notice far fewer daily price spikes as one moves along the timeline.
Duration Curves
An alternative presentation of the price data is provided by duration curves, which portray prices as a function of time ? the duration in order of magnitude for which a certain price prevails in the market. This approach makes it easier to identify trends. To control for the weather seasonality, which is a variable for power demand, Tables 4, 5 and 6 below are focused on the summer months (through July 31) during which power markets experience the greatest and most numerous price spikes.
The left side of the graph in Table 4 (less than 16% on the x-axis) collects the price spikes and compares how the price of power was driven by scarcity during each of the summers. The balance of the graph compares how the cost-driven power price behaves each summer. In this latter regard, the higher natural-gas prices plainly explain the elevation of the 2000/01 prices for the cost-based portion of the duration curve over those prevailing in the summer of 1999.
The left side of the graph needs to be analyzed in conjunction with weather, which are a far more significant variable affecting periods when price is determined by scarcity. The summer of 2000 was one of the mildest on record, which probably explains why price spikes were least frequent during that summer. On the other hand, the summer of 2001 has been the most extreme of the three (i.e., highest number of +85? days), and yet the occurrence of price spikes this summer has been less frequent than in 1999. This may be evidence of several variables that are reducing power scarcity in NEPOOL: capacity additions, increased use of hedging, improved plant scheduling and other examples of increased market efficiency.
Table 5 below shows that prices in PJM were far less responsive to the gas run-up. Most notable in comparing the summer duration curves above is the fact that the doubling of gas prices between 1999 and 2000 did not raise power prices during the cost-based portion between the two summers in PJM ? summer 2000 power prices were generally lower than 1999. Prices have experienced a general elevation in 2001, but this may have more to do with the elevated coal prices. The left side of the duration curve comparison shows that power scarcity is becoming less frequent in this market. Price spikes were relatively infrequent in PJM this summer, even though PJM experienced higher temperatures and peak demand, which surpassed those experienced in either of the previous two years. The trend toward less power scarcity, and greater market efficiency, is apparent.
Table 6 above shows the striking summer duration curve comparison for ERCOT. The dissipation of price spikes each successive summer has truly marginalized the slope of the duration curve by 2001. This may be preliminary evidence of how capacity additions in a semi-tropical climate can virtually eliminate for a period the occasions when spot power prices are determined by scarcity rather than cost. The modest slope of the duration curve for summer 2001 resembles the power price duration curve for Spain and the U.K., other markets without extremes in temperature where the supply/demand balance has been weighted toward the supply side through recent capacity additions. Meanwhile the cost-driven portions of the three curves in Table 6 appear to rank themselves consistently with the gas price experience: highest during summer 2000, followed by 2001, then 1999. This provides further confirmation of the responsiveness of power prices to gas price fluctuations in the ERCOT region.
Implications
The implication of all this is that analyzing power markets is a lot more complex than thinking about the average market price of power. In PJM for example, the average market price of power for the twelve months ending 3/31/00 and 3/31/01 was $29.54/mwh and $30.31/mwh respectively. These prices are practically the same, and only a bit higher than consultants forecast in 1999. However, the distribution of prices that make up these averages is quite different (more price spikes in period ending 3/31/00, but cost-based prices at a higher level in the later period ending 3/31/01). So, even though the average market price of power is almost the same for both periods, peaking plants are clearly better off in the earlier period, base load coal plants clearly better off in the later period, however base load gas plants are worse off in the later period ? because fuel costs were higher.
The critical issue is the margin between power prices and the marginal operating costs, the ?spark spread?. The prospects for spark spreads have to be analyzed for two periods: the infrequent times when they are driven by scarcity, and the much more common times when they are driven by costs.
The market experience above shows that the general trend is for power markets to become more efficient, and for scarcity to become less frequent. Regional differences will affect this trend. The regions with seasonal weather are more likely to have some amount of scarcity in their markets ? this will be particularly true of those regions that are dependant on hydro power or nuclear plants that may become suddenly unavailable. For this reason, comparing reserve margins is too simplistic as an indication of the supply/demand balance in a power market.
Analyzing the prospects for spark spreads during cost-driven periods is more critical for investments intended to be base or intermediate load. A comparison of the cost-based portions of the duration curves in the regions above does confirm how responsive power prices are to gas costs in markets like NEPOOL and ERCOT where gas is the marginal fuel. This means that all things being equal, fuel cost increases will increase the spark spread, and consequently gas plant profitability in absolute terms; fuel cost decreases would have the opposite effect. But all things are not equal. The greatest concern for investments in generating assets is that too many capacity additions (?overbuild?) will reduce spark spreads regardless of fuel price fluctuations.
The effect of efficient capacity additions on the cost-driven side of the duration curve needs to be analyzed. In theory, the cost-based power price in a gas-driven market will for the most part be determined by the heat rates of the various plants being dispatched into the market and their relative shares of the market. For any given period, a market will have an average heat rate based on the actual weighted average heat rates being dispatched ? something hard to know ? but the databases in market consultants' models use these assumptions to forecast power prices. An implicit market heat rate can be observed by comparing the price of power and the price of fuel prevailing in markets for a given period. This provides a good indication of what that average heat rate might be. In any event, the average rate that is converting fuel costs to power prices is being lowered by capacity additions with 7,000 heat rates. This stands to reason, since so much of the existing capacity being displaced consists of older dual fueled plants with higher (less efficient) heat rates. The reduction of the dispatch factors for these plants will lower the average heat rate of a given market. However, since virtually no one is building or planning capacity additions with heat rates materially better than 7,000 heat rates, a market's average or implicit heat rate could only be lowered so far.
The California Phenomenon
California behaved much like a gas-driven, seasonal weather market before the summer of 2000. After that time, this market had several unique circumstances, and certainly unique spot power price behavior, which caution against projecting forward trends based on a backward-looking analysis here. Looking back, the exceedingly high spot prices which prevailed through the summer, fall and winter of 2000, are linked to five factors which are not likely to recur jointly in the medium term:
1. A fundamental imbalance between demand for power and capacity in the California market;
2. A draught which aggravated the supply/demand imbalance by removing much of the hydro capacity (normally 25%) from the market,
3. An exceedingly hot summer which pushed-up load;
4. The run-up of gas prices (noted already), multiplying fuel expenses five times between June and December 2000;
5. Market regulations, which required the three major utilities to purchase all of their requirements on the spot market which was the California PX.
The fifth factor ? perhaps the most critical to the financial side of the California energy crisis ? has been removed from the market. Of the other four, the first and second are still true, but the others have not been applicable during the past four months, and the spot power prices in California have moderated considerably. It would be premature to examine much further the trends for power prices in the California market; the variables noted above make the recent data too unreliable.
Lessons from Spot Markets
Despite the anomalies presented by the California experience, a backward-looking analysis of other spot power markets reveals predictable correlations: between power prices and the factors affecting scarcity, and between power prices and fuel costs. On a practical level lenders analyzing the prospects for ?over build? in a particular market need to forecast a duration curve with diminished price spikes, and contracted spark spreads. This is a much more useful approach than giving arbitrary ?hair cuts? to revenue forecasts.
As the two regions most mentioned as at risk for overbuild are gas-driven ERCOT and NEPOOL, a forecast of the duration curves in those regions needs to be informed by a forecast about gas prices. A fair ?downside? forecast would assume reasonably low gas costs and a conservative implicit heat rate for the two markets. With the help of consultants' databases, and another summer of experience, a fairly conservative downside margin in the range of $60-50/kW-year could be calculated for these regions for 7,000-heat rate plants, lower for the less efficient plants. This scenario will not be fatal to the low heat rate plants of these regions.
Forward Markets:
The discussion above has focused on the information provided by the spot markets, however bulk electricity is also purchased through forward markets ? both bilaterally and through exchanges. Because power prices have been so volatile in the summer months, for reasons explained above, most of the forward market trading is focused on hedging the summer price of power. Table 7 above is one way of presenting how forward markets have behaved in PJM.
Layered on top of one another in the graph above are the forward price-lines prevailing for various times over the past three years. Each distinct line depicts the settlement prices for contracts entered into during a certain bimonthly date (the left-most point on the line); each point along the line corresponds to the settlement date and price of the contracts. For example, the cross-hatched line corresponds to all contracts that were entered on May 23, 2000 for settlement during the following 18 Months. The first point along the line corresponds to a settlement price of $98/MWh to be executed during the month of June 2000; the last point along the line corresponds to a contract entered into on May 23, 2000 with a settlement price of $29.50/MWh and a settlement date of November 2001. The chart shows the high concentration of futures market activity that exists during the May-Aug summer months. Since the summer months are when demand is at its peak and power supply is scarce, it is during these months that most activity and hence volatility takes place.
As a general observation, it is evident that the market anticipated high prices leading up to the summer of 2000, but forward prices were contracted at lower levels leading up to the summer of 2001, with prices being contracted at still lower prices for next summer. This all follows logically from what actually occurred in the spot markets for PJM. The power scarcity and price spikes were highest in the summer of 1999 ? which (along with a uniquely high-priced day in early May 2000) contributed perhaps to the over-hedging in forward markets leading up to the summer of 2000. The summer of 2000 turned out to be such a mild spot market in PJM that forward prices anticipating summer 2001 moderated a bit ? although the run-up in fuel costs would have been a price-lifting factor here.
Table 8 below concentrates on the activity of the PJM futures market leading to the summer months. Each line depicts the settlement prices for contracts to be executed in either June, July or August, which were transacted during the months leading up to the summer corresponding to the date on the X-axis coordinate. For example, the dotted line illustrates the prices for transactions to be settled in August 2000. The leftmost point along the line corresponds to a transaction entered into on December 7, 1999, which will be settled in August 2000 at a contract price of $87/MWh. It's noteworthy that by 2001, the forward markets are not making a distinction between power delivered in July or August, so those prices appear to have merged into a single line.
A few generalizations can be made about the above comparison. Price swings are common leading up to each summer, but there is a variation in the amount of volatility in each of the years being observed. The most volatile upward trending forward market (spring 2000) was followed by the least volatile spot market (summer 2000). Power market consultants have predicted that price volatility would migrate between spot and forward markets. This may be evidence that over-hedging will moderate spot markets. In California during the spring of 2001, the California Department of Water Resources (DWR) pursued medium and long-term contracts for most of its power requirements (at prices raging from $180-288/kW-yr). By the summer of 2001 (which enjoyed mild weather), DWR was reselling this power into the spot market for on-peak prices in the range of $50 ? $70/MWh. Pacific Corp has also reported losses from over-hedging in the spring of 2001.
Volatility can be induced by a combination of hot weather and under-hedging. The forward markets leading up through the summer of 2001 for PJM (NEPOOL for that matter) were noticeably flatter than the prior year. Consistent with the theory that volatility migrates between spot and forward markets, late summer heat (July-August) did produce a revival of price spikes in the PJM and NEPOOL spot markets.
Table 9 below tracks the forward market price of power for Entergy-SERC prior to the past three summers. The trends are generally the same as described for PJM above with more volatility and generally higher price levels. The absence of a transparent spot market in SERC could be one explanation for this. Natural gas is much more of a marginal fuel for the Entergy portion of SERC (than in PJM), and this would also elevate forward prices and their volatility in the 2000-01 period
Focus on Volatility Value:
As mentioned already, a merchant power plant generates more value for investors than the simple moment-to-moment sale of energy into a power exchange. Power plants, and particularly portfolios of power plants, give their investors a natural long position in a highly volatile energy market, which represents other sources of value apart from mere physical commodity sales. The table below lists some of the sources of value to be realized from typical gas-fired plants; these sources are referred to by many investors as ?Volatility Value? or ?Optionality? because the common merchant power plant operator has more options besides commodity sales into the spot power market.
A look back at how regional forward markets for power behave illustrates the profit potential of trading in forward markets from a physical position. The routine price swings of summer power noted above would provide ample opportunity for profitable power trading from a physical position. If the forward sale cannot be covered profitably with purchased power, the generator can ultimately cover with a physical delivery (provided the fuel cost has been properly hedged).
Most structured financings of merchant power assets have not been based upon revenue streams that included significant volatility value or optionality value in the cash flow forecasts. In many financing structures, the SPV established to hold the merchant assets is selling all of its power and capacity to the power marketing affiliate of the sponsoring company. This is regarded as a ?safer? structure by many lenders because it isolates the risk of trading outside of the borrower. However, these types of structures exclude the value of optionality along with its possible risks.
Trading from the natural long position that generation ownership confers can be a relatively safe activity when the proper contracts and procedures are in place. The generator should be able to profit from its natural option on the spread between gas and electricity. Backward-looking evidence of volatility value can be found in the markets described above. Broader acceptance of optionality or volatility value by lenders or investors in the merchant power industry is also likely to be driven by earnings records as this industry matures.
Request a Demo
Interested in IJGlobal? Request a demo to discuss a trial with a member of our team. Talk to the team to explore the value of our asset and transaction databases, our market-leading news, league tables and much more.