Weather Risk Management for Regulated Utilities

Why hedging can make sense, even for companies covered by weather-normalized rates.

Weather risk management is growing, but utilities may be losing out.

A recent survey suggests that the number of transactions involving financial derivatives to hedge weather-related risks grew by 43 percent against the prior year for the twelve months ended March 31.¹ Yet regulated utilities continue to show reluctance to embrace weather derivatives.

For regulated utilities, the risk of volatility in commodity costs may be passed directly through to retail customers, using rate-making tools such as a purchased fuel adjustment clause.² That leaves sales volume as the most important weather-related risk factor for utilities. Nevertheless, some utilities believe they have a good alternative to weather derivatives. That alternative comes in the form of weather-normalized (WN) rates approved by the state public utility commission.³

Are the utilities right?

In this article we compare WN rates with various types of financial derivatives for hedging the weather, taking as an example a typical natural gas local distribution company (LDC). We show why WN rates fail to insure against some specific weather-related types of earnings risk. We also show why weather-related derivatives might in fact make sense in some cases for a regulated utility covered by WN rates.

Of course, recent market failures in the energy industry have spawned a tightening of credit that has led a number of energy trading and marketing companies, such as Reliant and Aquila, either to exit or sharply curtail their activities. Their place is being taken by financial institutions, primarily (re)insurers such Element Re and Swiss Re, expanding from insurance into the derivative market, and by banks like Société Générale.

But these events should not color our analysis of WN rates and weather hedging. That analysis begins with some general principles to help energy companies quantify weather risk and calculate the value of a weather-hedging derivative. After that, we examine how WN rates protect utilities from weather risks in some cases, but not in others.

Quantifying Weather Risk

The Heating Degree Day ("HDD")⁴ is commonly used in the utility industry to quantify weather and measure how earnings are related to weather.⁵

Correctly estimated, HDDs often show a remarkably linear and tight relationship to volumes and earnings over a significant range of HDDs. Although linearity is not crucial, the tightness of the fit (i.e., the standard deviation of volumes consumed for a given HDD) is important. Since weather varies widely by location, it is important to determine the company's volume to HDD relationship by aggregating on a location- and business segment-weighted basis. Figure 1(a) shows the volume to HDD relationship for residential customers at a variety of locations. Figure 1(b) shows the load-weighted volume to HDD relationship. The load-adjusted $/HDD relationship is what should be used when considering weather protection.

An energy company wishing to protect itself from this weather variability can buy a weather contract that provides a pay-off in the event of unfavorable weather. This type of weather derivative is commonly called a "floor," and requires a premium. In the alternative, a buyer could "pay" for the protection by giving away some of its potential upside with a cap. Together a floor and a cap provide a "collar," which limits a company's earnings to within the band of the collar.

Figure 2(a) shows a collar, but where the floor offers protection only over a certain range. This, in turn, imposes a limit on how much the seller has to pay in the event of extremely warm winter conditions, and is common in practice. Figure 2(b) shows the pay-off structure of a weather derivative contract that provides such protection. The other common type of weather derivative is known as a "swap," where the company exchanges volatile earnings for constant earnings. This is equivalent to the floor and cap being set at the same point in Figure 2(a).⁶

For a buyer to implement a particular risk profile, the company also needs to understand its portfolio of risk, and recognize the impact of naturally offsetting hedges. For example, an LDC that operates around the country may be partially naturally hedged.⁷

What Price for a Hedge?

Some buyers may worry that the seller will sell them too much of the wrong product at too high a price-a result of the seller's and the buyer's economic incentives not being fully aligned. These concerns have at least some foundation in fact.

A review of quotes obtained from the Internet for an energy trading and marketing company showed that in the beginning of this year, unsophisticated consumers often were offered annual premiums of 100 percent over the contract's expected payout.⁸ Premiums, of course, need to be measured relative to the baseline payout for "normal" weather. The derivative industry tends to use the last 10 years-which on the whole have been warmer than the last 30 years; the latter measure commonly is used by energy companies for regulatory proceedings. Interviews with a number of market makers suggest that by shopping around, these premiums often can be reduced to 10 to 30 percent above the expected payout. The premium should reflect the likelihood of a particular payment. An alternative pricing formula puts the premium at a similar percentage of the standard deviation of the expected payout.

The issue of counter-party credit worthiness also has grown in importance for potential buyers following the collapse of Enron and the ensuing credit issues faced by many energy trading and marketing companies. As mentioned earlier, some of their activities are being taken up by financial institutions. Beyond the issue of premium charges and credit worthiness, buying a derivative is not a particularly risky enterprise for the buyer.⁹

How to Value a Contract?

When a seller sets the price of the weather derivative, the buyer needs to consider the implicit premium being paid. A quick way to do this is to estimate the intrinsic (i.e., pre-premium) value of the weather derivative. This back-of-the-envelope approach simply estimates what the average pay-off of such a contract would have been over the last 10 years (the "burn rate"), and then compares the expected payment with the price being charged by the seller.

Figure 3 shows how to estimate properly the intrinsic value of various types of weather derivatives. The figure shows that valuing a weather derivative is relatively straightforward in principle. The payoff-function ($/HDD) is simply multiplied by the probability that the HDD occurs, summed over the relevant HDD range. This is not rocket science. The main difficulty, and it can be a real problem, is in estimating the mean and the shape of the true HDD distribution on a going-forward basis.

Sellers of weather derivatives do so expecting to make at least a fair return.¹⁰ The use of multiple bids for the same payout structure is probably the best way for a potential buyer to ensure receiving a fair premium, although at a minimum the buyer should also do a burn rate analysis to check for reasonableness. Ideally, if companies want to be protected, they need to buy early, rather than try to bet on the weather. Multi-year contracts should also be considered, since they ought to result in a lower premium (because for a given expected annual payout, the distribution will be narrower for a multi-year period). However, the recent warm weather may make multi-year deals expensive in the short-term.

Three bids is a useful rule of thumb. Too few bids obviously can lead to a non-competitive price, while too many bids can reduce the interest of potential sellers and move the price against the buyer in the secondary markets. In some cases potential buyers work directly with brokers, such as Aon, Marsh, and Willis, that handle weather contracts. In other cases potential buyers may work directly with one of the potential sellers.

In the end, pricing is a combination of art, science, and what the market will bear. A fair price depends on your view of what the "normal" weather is, since this determines the expected payout.¹¹ Stung by a succession of recent warm summers, there has been a move by some sellers to try to convince buyers that the last five-year time horizon is more representative of "normal" than the last 10 years. The statistical validity of such a claim is doubtful, and leads to the risk that some sellers may-at least in the near-term-price themselves out of the market.

Understanding Weather-Normalized Rates

For regulated energy companies that are interested in mitigating the impact of weather, the decision on using weather derivatives is more complex than for their unregulated counterparts. This is because, in many areas of the United States, regulated companies also have the option of implementing WN rates.

A recent survey showed that 40 gas companies in 19 states have a weather normalization adjustment (WNA) clause in their rates. This also suggests, however, that a large number of states do not have WNAs. Moreover, some LDCs' applications for WN rates have been rejected, and so for these firms, weather insurance may be the only available alternative.

When a choice is available, the relative merits of using weather derivatives versus weather-normalized rates are not universally understood or agreed upon in the industry. Some companies use weather derivatives, (e.g., Washington Gas), some use WN rates (e.g., Piedmont Natural Gas), some use neither (e.g., Cascade Natural Gas), while some do both (e.g., Atmos Energy). To understand the relative merits of WN rates and weather derivatives, it is necessary to briefly review how LDCs recover rates and how WN rates work.

While most of a utility's operating and capital costs are typically fixed in nature, less than half of these fixed costs of providing service are recovered through fixed charges. A significant amount of these costs are recovered volumetrically in the utility's commodity rates. These volumetric unit rates ("commodity rates"), used to recover the balance of fixed costs, are usually designed based on "normal weather," using a 30-year average number of HDDs and calculating a heating-load use per HDD. As a result, when the weather is warmer (colder) than normal, the utility realizes less (more) revenues than it was designed to recover in its fixed costs, regardless of whether it utilizes flat commodity rates, or declining block commodity rates.¹²

A WNA is one type of mechanism that may be employed by the utility to recover revenue under-collection or refund revenue over-collection. If the mechanism works perfectly, it is effectively a costless swap; the utility's variable weather-dependent earnings are replaced with fixed earnings. ()

WN rates have many of the characteristics that we have already discussed for weather derivatives, including more stable earnings, better debt ratings, and less need for short-term cash. The main advantage of WN rates compared to weather derivatives is that there is no premium required from shareholders. However, a potentially significant disadvantage is that WN rates do not provide a perfect earnings hedge, particularly in extreme weather conditions. Additionally, as swaps, WN rates do not allow the utility to choose to take upside risk. Table 1 outlines the pros and cons of using weather derivatives vs. weather normalized rates.

Discussions with clients suggest there are two distinct ways that WN rates can fail:

Simple linear WN rates formulas do not capture extreme conditions; and

WN rates formulas are not forward looking, so they do not reflect changing conditions over time.

Where WN Rates Breaks Down

The first way WN rates can fail reflects the tendency for regulators to seek simple rules. As a result, many LDCs have simple rate adjustment formulas that assume that the consumption and $/HDD relationship is linear (). This may work well for weather conditions that are 10 percent warmer or 10 percent colder than normal, but might not be true if the weather conditions are more extreme (e.g., 30 percent warmer or colder than normal). In extreme temperature conditions the consumption and hence $/HDD can become non-linear. This can result in LDCs under-recovering in very warm winters and over-refunding in very cold winters.

In the case of a really cold winter, the actual change in consumption per degree-day can fall. One reason for this is that when it is extremely cold, a consumer's heater may be operating at its maximum output. If it then gets colder, gas usage will remain the same. Since the WNA formula assumes that usage and hence earnings will increase, this results in over-refunding by the LDC. This leads to the rather ironic result that an LDC experiencing very cold weather has lower earnings than if the weather had been normal.

At the other end of the spectrum, an LDC may also suffer in very warm winters. If it gets warm enough, consumers may just switch off their heating equipment over a range of HDDs. This can also result in a downward discontinuity in the true consumption per HDD for very low HDDs. In this case a utility's volumes and hence earnings fall faster than the WNA formula anticipates, and so the LDC under-recovers ().¹³

Even for an LDC whose WN rates correctly estimate the $/HDD relationship for a given range, the subsequent rate adjustments will not always exactly match lost (gained) revenues when the weather is warmer (colder) than normal. This is because of the phenomenon of declining usage on an absolute basis, and changes in consumption per HDD relationship over time.

Figure 4 shows revenue from residential consumption per HDD over a range of HDDs spanning three years. In this example, the average consumption per customer for a given HDD falls each year. The slope of the line, the heat load factor, represents the consumption per HDD. In our example the slope has flattened over time, corresponding to a reduced heat load factor.

The declining usage for a given HDD will lead the utility to under-recover during normal weather conditions. The utility would need to seek a rate case to recover this loss, something it may not want to do on an annual basis. The change in slope also leads to a change in the consumption per HDD. Figure 4 shows how this could impact earnings for an LDC. The difference between the dotted line and the solid line shows the earnings that would be unprotected if the consumption per HDD changed over time while the WNR structure remained unchanged. In this example, a cold winter results in the LDC refunding too much.¹⁴

Weather derivatives therefore offer a potential complement as well as an alternative to the use of weather normalized rates. For example, an LDC with WN rates that failed at extreme weather conditions could buy a collar that only kicked in during extreme weather conditions, and gave no payout over the range of HDDs where the WN adjustment was effective.

In the current climate, managing earnings uncertainty increasingly is likely to be rewarded by investors. Utilities need to be aware of the different options available for managing weather risk. Weather derivatives could play an important role in managing this risk, provided they are priced reasonably.

Current survey from April 1, 2001 to March 31, 2002. Joint survey by Weather Risk Management Association and PwC.

The ability to pass through commodity costs is not without risks to earnings. When the price of gas rose abnormally in the winter 2000-01, LDCs' pass through of these costs led to a significant rise in unpaid bills. As result, many LDC companies have subsequently considered hedging the price for some part of their supply portfolio.

Weather Normalization Adjustments ("WNAs") provide the utility recovery of its fixed costs, regardless of volumetric throughput, through a surcharge mechanism. If the weather is warmer than normal, the utility assesses a surcharge for revenues that would otherwise be lost. If the weather is colder than normal, the utility refunds its excess collections.

HDDs are particularly relevant for measuring how warm or cold the winter season is, while cooling degree days (CDDs) perform a similar function for energy companies whose earnings are susceptible to mild or hot summers.

Estimating the correlation between volumes and/or earnings and some measure of the weather is the most common approach.

The example shown in Figure 2 assumed that there was a perfect relationship between earnings and HDDs (corresponding 100 percent correlation). This does not need to be the case. Another company may have earnings that are only 85 percent correlated to HDDs. The company could still significantly benefit from hedging using the pay-off structure in Figure 2(b). In this case, however, it will not be fully hedged.

HDDs are not the only measure of weather relevant to energy companies. For example, low rainfall or snowfall could limit the ability of a hydro plant to deliver energy for a sustained period, due to lack of water. If such a plant also had pre-existing contracts to supply energy it may be forced to buy power in the open market to meet its shortfall. To hedge against such an event, the firm could buy a weather instrument that paid out if the rainfall at a given location fell below a certain amount over a three-month period. While temperature-based contracts remain by far the most popular, the recent WRMA/PwC survey showed that the number of rain-based contracts increased roughly four-fold for 1.6 percent to 6.9 percent between 2000-01 and 2001-02.

An easy way to get a feel for the premium is to estimate what the instrument would have paid out on average over the last 10 years; this is known as the "burn rate." This can be compared with the annual cost of buying the instrument. We have observed quotes for $10,000 to sell an instrument that has a historically expected payout of $5,000 or less. While the last 10 years is not necessarily fully representative of future weather, it is considered important information by the seller when setting the rate.

The buyer is explicitly purchasing the derivative to eliminate or reduce risk. Sellers of such derivatives, on the other hand, do carry significant risk, unless they are properly diversified (e.g. through matching buyers with sellers).

A fair return for a weather derivative is made up of three components: the risk free rate, a premium that reflects width of the HDD distribution, and an additional term that reflects uncertainty over the expected mean of future HDD distribution.

Comment: This refers to six months prior to the winter season, when no meaningful predictions are available. Obviously closer to the winter season, weather predictions can impact the price significantly.

Generally, when the weather is warmer than normal, declining block rates provide some protection to the revenue stream to the extent that the load falls off in the lower-priced tailblocks. Conversely, when the weather is colder than normal, flat commodity rates cause the revenue stream to increase at a greater rate than declining block rates.

Another factor that could impact recovery is the asymmetric way a HDD is defined: positive if below 65 degrees, zero if above 65 degrees. Above 65 degrees it does not matter if it is 67 degrees or 71 degrees; both give rise to zero HDDs. Thus abnormally warm weather can be underestimated using a HDD approach.

Conversely, in mild winters the utility may over-collect from customers.

Articles found on this page are available to Internet subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.