Natural-Gas Procurement: A Hard Look at Incentive Mechanisms


Better designs are needed to realize the goal of lower-cost gas.

Fortnightly Magazine - February 2006

As customers receive substantially higher bills than they received last winter from local natural-gas distribution companies (LDCs), questions arise as to what more the utilities could be doing to contain their gas-supply costs, and what regulatory changes could contribute to lower gas costs for the future.

The gas-procurement choices LDCs make, how they manage assets that are held for this function, and the extent to which costs are hedged far in advance determine the cost of gas their customers ultimately pay. For most LDCs, these costs have been fully recovered from customers, subject to prudence review, under a purchased gas adjustment (PGA) tariff clause. Consequently, utilities have had no direct incentive to purchase supply and manage the associated assets most economically. Furthermore, the threat of a prudence review and possible disallowance can discourage utilities from actions (such as hedging) that might be in the customers’ long-term interest, but lead to higher gas costs in some years.

With the unbundling and increasing competitiveness of the natural-gas industry during the past two decades, LDCs have faced a much broader array of choices for acquiring gas supply for customers. This has made gas procurement more complex, and effective regulatory oversight less feasible and more costly.

Recognizing that stronger incentives for the gas procurement function could lead to improved utility performance while reducing the need for detailed prudence reviews, state regulators in more than a dozen states, including California, Oregon, Wisconsin, Indiana, Missouri, and Rhode Island, have approved gas procurement incentive mechanisms (GPIMs) for one or more jurisdictional utilities (see Figure 1).

Gas cost to consumers should decline as a result of an incentive mechanism if the incentive leads to a reduction in the utility’s actual cost of purchased gas (compared with what it would have been without the incentives) that exceeds the incentive payment earned by the utility under the mechanism. With stronger incentives, LDCs could lower gas cost by applying more resources to the function (employing more highly qualified staff, acquiring the best market intelligence, etc.); taking greater advantage of the considerable buying power that LDCs have in the marketplace by virtue of the large, stable, firm loads they represent; managing the substantial flexibility LDCs typically have within their portfolios to maximum advantage; harvesting the full value of the transportation and storage assets they control; and taking calculated risks that, lacking incentives, they might be unwilling to take.

While the objectives and broad structures of the various GPIMs implemented around the country have many similarities, there are substantial differences in the detailed structures that are not readily apparent from a general description, and these details can have a large impact on their incentive properties and the regulatory outcomes for the utility and customers. Most utilities across the country operating under GPIMs have regularly beat their benchmarks and earned incentive rewards. However, many GPIMs have been criticized and frequently modified, and some have led to disputes or allegations of utility misconduct.1 Some GPIMs have been terminated, either by the initiative of the regulatory commission or the utility itself.2

GPIM Objectives, Scope, and Structure

The objective of most GPIMs is to provide incentives for achievement of lower short-term procurement costs. As such, they are generally not designed to incent or reward provision of reliability or price stability, which are different and competing objectives for which it would be difficult or impossible to provide effective incentives within a single incentive mechanism. Nor do GPIMs attempt to provide incentives with respect to decisions concerning long-term firm pipeline and storage reservations, which usually are reviewed and approved separate from a GPIM. However, management of such assets in the short term, including revenues from gas sales or capacity release, generally is within the scope of a GPIM’s incentives.

Many current GPIMs were designed in the mid-1990s when gas prices were relatively low and stable, and few if any utility commissions were encouraging utilities to hedge gas costs. So it is not surprising that many GPIM benchmarks assume only short-term gas purchases (with monthly and daily pricing), and little or no hedging other than through storage.3

However, as natural-gas prices have increased and become more volatile in recent years, interest has grown in providing customers with greater price stability, which can be achieved through physical forward purchases or financial hedging. When a GPIM benchmark includes only short-term purchases, it places the utility at risk through the GPIM for the costs and outcomes of any substantial amount of forward physical purchases or financial hedges. This discourages hedging.

To accommodate an approved level of hedging, a GPIM can include a target schedule and quantity of hedging in the benchmark.4 The utility is permitted to hedge more or less, or sooner or later, than the benchmark schedule (perhaps limited to some range), and is at risk for whether its adjustments to the schedule ultimately result in higher or lower gas cost. Another approach to accommodating hedging is to exclude both the costs and impacts of hedges from a GPIM.

With respect to short-term procurement-related costs, many GPIMs are designed to provide incentives for all or nearly all such costs and associated revenues, including supply, transportation, and storage.5 (Note that a GPIM provides an incentive for a particular cost or revenue category if it establishes a separate benchmark value for it; if instead the cost category is excluded from the calculation of benchmark and actual costs, or if the actual cost value is included in the benchmark, no incentive is created.) GPIMs that encompass as broad a scope of interdependent procurement-related costs and revenues as is feasible, within a single incentive calculation applicable to all costs, provide the best incentives.

The actual utility reward or penalty under a GPIM is determined in a periodic (usually 12-month) GPIM accounting, in which the GPIM’s sharing rules are applied to the difference between benchmark and actual costs. Typically, 50, 25, or 10 percent of the difference may be assigned to the utility. A higher percentage provides a stronger incentive, but also imposes more risk and could lead to more risk-averse, cautious actions; the utility might tend to forgo opportunities that are attractive in expectation but somewhat risky, raising gas costs. A larger percentage assigned to the utility may also lead to very large rewards because of external conditions that cause a greater difference between benchmark and actual gas costs. In general, a GPIM’s sharing parameters are set to provide strong incentives while also balancing these other concerns and objectives under particular utility circumstances.

Tolerance bands, within which there is no reward, are common, as are caps on rewards or penalties, and sharing percentages that vary by the size of the reward. These features tend to reduce incentives, and they also can lead to some distortion of incentives.

Design of the Benchmark

The central element of GPIM design is the benchmark formula, which affects the relative likelihood of reward or penalty, the strength and nature of incentives, and the extent to which the utility is exposed to external risks as a result of the GPIM.

The simplest possible GPIM would entail a benchmark gas cost that is a fixed dollar figure set in advance, representing a forecast of gas cost. However, this simple approach would expose the utility to substantial risk due to factors that it cannot control, such as the weather-induced variability of gas loads or large movements in natural gas market prices. To remove these risks from the mechanism and focus incentives on factors under utility control, benchmark formulas typically take into account load levels and market-price indexes.

The overriding objective of a GPIM is to provide incentives for performance surpassing that which would be reasonable and expected under traditional regulation. Under a GPIM, there is no incentive reward or penalty if actual costs equal (or are within a tolerance band around) the benchmark, and the utility receives an incentive reward if it beats the benchmark. This suggests that, in principle, the benchmark should be designed to reflect the gas cost that would result from a reasonable procurement strategy reflecting acceptable, but not superior, performance deserving of no reward or penalty. However, in practice, benchmark formulas are kept simple, and, as a result, benchmarks often reflect implied purchasing rules that can be easy to beat.

For example, under some GPIMs, the benchmark assumes purchases from available supply basins in fixed proportions, based on historical averages, a gas supply forecast, or firm pipeline reservation quantities.6 The resulting benchmark is easy to beat if relative prices change and the utility is able to optimize purchase locations accordingly. Other GPIMs determine the purchase locations and timing to be assumed in the benchmark through a formula that adapts to relative prices to some extent.7 However, designing a benchmark formula to better reflect how procurement should adapt to external conditions will tend to increase its complexity, which can render it more costly to audit and increases the potential for misunderstandings or disputes.

While it is important for a GPIM benchmark to reflect external conditions that are outside of utility control, such as load and prices, it is also important that a GPIM benchmark not use parameters that are under utility control or reflect utility choices. It is a fundamental principle of the design of incentive mechanisms that a benchmark should provide an external and independent basis for evaluating company performance. This means that the benchmark must be calculated using only parameters and assumptions that are independent of the utility’s actual purchasing decisions (we call this an exogenous benchmark). If the benchmark is exogenous, the utility can only increase its reward by lowering actual gas cost, not by raising the benchmark, with utility and customer interests aligned.

Any assumptions in the benchmark calculation that are affected by utility choices result in a benchmark that is not exogenous, and some distorted incentives. Such a GPIM sometimes will reward actions that are not in the customers’ interest, or fail to reward actions that are in the customers’ interest.

Common Pitfalls in GPIM Design

The preceding discussion suggests that GPIMs focus on providing incentives for short-term procurement. Their designs should apply a strong incentive—not blunted by caps on rewards or large tolerance bands—equally to all costs and revenues related to short-term procurement. They should use a benchmark that approximates a reasonable procurement strategy, is independent of utility choices (exogenous), and is designed to adjust to external conditions such as market prices and load levels. GPIMs with such designs provide strong incentives to procure gas supplies at least cost.

Our review has found that existing GPIMs generally do not achieve all of these design objectives, and, consequently, they provide weak or distorted incentives for some types of procurement actions, or they expose the utility and customers to some risk of rewards or penalties that may at times be excessive and undeserved.

Of course, utilities will not necessarily act according to the incentives of their GPIM, especially when there is a known conflict with customers’ interests. This does not mean such conflicted incentives are benign. To the extent such conflicts exist, the incentives serve no useful purpose while increasing the risk of actions contrary to customers’ interests and the potential need for more detailed review of purchasing decisions.

One common GPIM design characteristic that can lead to incentive problems is the application of GPIM incentives only to certain cost components.8 This may allow the utility to increase its GPIM reward by spending relatively more in areas where incentives do not apply, to reduce other costs to which stronger incentives apply. For example, some GPIMs exclude revenues from release of interstate transportation capacity, passing all such revenue through to customers. When such capacity is released, the utility may incur additional cost for short-term transportation or downstream purchases, and it also may lose opportunities for profitable off-system gas sales. The potential revenue from a capacity release may be much greater than the incremental supply and transportation cost, lowering total cost, but the capacity release would be discouraged by a GPIM that includes the supply and transportation costs but ignores the capacity release revenue. In this example, the LDC’s incentive under the GPIM can be contrary to the interests of its customers.

Another very common design compromise is the use of actual utility purchase volumes in the benchmark calculation.9 The motivation for doing this is clear—it is a simple approach to creating a benchmark that adapts to a broad range of external circumstances, such as changing relative prices and overall load levels. However, when a GPIM benchmark uses actual utility purchase or net purchase volumes (or weights reflecting them), it means that the utility’s actions affect the benchmark, and it is no longer exogenous. The result is that incentives are distorted to a surprising extent.

As a simple example, consider a utility that can purchase supply from either of two basins, A or B. Figure 2 illustrates that when actual purchases are used in the benchmark, each purchase is essentially benchmarked to a price index at its location. As a result, the GPIM incentive is not to purchase the least-cost supply, but to make the deal that allows beating the respective locational price index by the largest amount, earning the largest incremental reward. By contrast, if the volumes used in the benchmark are independent of utility choices (for instance, set by a rule that is a function of actual loads and market prices, which the utility does not control), the utility always has the proper incentive to purchase the lower cost supply.

Use of actual net-purchase volumes in the benchmark creates similar incentive problems with respect to use of storage and the timing of purchases, as it reflects actual storage choices in the benchmark. As an illustration, suppose a utility chooses to delay storage injections during spring and early summer, planning to catch up later in the summer. An exogenous benchmark would include a specific storage injection schedule, or one based on parameters independent of the utility’s actual choices, placing the utility at risk for a share of the cost impact of deviations from the benchmark schedule. With a benchmark that uses actual net purchases reflecting storage, the utility would be at no risk for whether its choice turned out well or poorly, as the benchmark would reflect the choice. The consequences, however, could be costly for customers.10

Use of net-purchase quantities (reflecting storage) in the benchmark can create more serious and frequent incentive problems depending upon other benchmark characteristics, such as use of first-of-month prices in the benchmark that apply to incremental sales or purchases on the daily market.11

Use of actual purchase volumes in the benchmark also distorts incentives with respect to additional contract provisions that change the value of a contract relative to the price indexes used in the benchmark. For example, a utility buyer can receive a discounted price on a deal by offering a seller valuable flexibility, such as rights to recall the supply or terminate the contract at any time during the month. However, such recall rights could lead to costly replacement purchases when the supply is recalled under tight market conditions. If the utility is able to arrange the replacement purchases so that they are priced at benchmark (perhaps by initially relying on storage, to defer replacement to a later month when the gas can be purchased at index), it will have an incentive to enter into such contracts to increase its GPIM reward, even if total customer gas cost is raised as a result.12

These examples suggest just a few of the ways compromises to the GPIM design principles can lead to distorted incentives and opportunities for the utility to increase its GPIM reward through actions that are contrary to the interests of customers. How frequently these opportunities arise, and the strength of the distorted incentive when they do arise, will depend upon other GPIM characteristics and the particular utility circumstances. A more detailed treatment of GPIM incentives and incentive problems is included in the authors’ longer paper on this topic.

GPIM Review and Assessment Procedures

In principle, strong, well-aligned incentives provided by a GPIM obviate the need for detailed regulatory review of the reasonableness of a utility’s performance of the gas procurement function. To the extent a GPIM provides distorted incentives, the need for regulatory review of some aspects of utility decision-making actually is greater than in the absence of direct incentives, such as when there is no GPIM.

Our review of multiple GPIMs, of documents representing reviews of GPIM results, and of regulatory commission orders approving the associated rewards, suggests that in a number of instances, the regulatory staff reviewing the GPIM filing may be unaware of the extent of a GPIM’s shortcomings and the areas where it can provide weak or distorted incentives. In these instances, staff and the commissions may be unaware of the full scope of opportunities to increase the utility reward at the cost of customers, and, consequently, may not be adequately reviewing utility actions to ensure that such gaming has not taken place to a significant extent. In some instances, staff reviews and commission orders on GPIMs apparently assume that if actual costs are below the benchmark, customers have benefited, and there is little need for a detailed review.

This brings us to another common problem with the GPIM review process—misinterpretation of the difference between the benchmark gas cost and actual gas cost. A utility operating under a GPIM often will interpret this difference as “savings” for customers, suggesting that it reflects superior utility performance under the GPIM. Such claims are echoed by the regulatory commission staff or consumer advocate staff responsible for reviewing the GPIM filing,13 and reflected in the commission order approving the incentive reward.14 However, this claim is true only if the benchmark formula accurately estimates the costs that would result from the utility pursuing a reasonable procurement strategy absent the incentives provided by the GPIM. It is unlikely that this is ever the case. Instead, GPIM benchmarks are based on simplified formulas that in some ways reflect relatively mechanical procurement strategies that are often easy to beat. In addition, for GPIMs that provide distorted incentives, the magnitude of the difference between benchmark and actual costs partially may reflect utility actions that raised the benchmark. Consequently, while the utility may have achieved “savings” relative to its benchmark formula, it may not have achieved savings relative to the procurement strategy it would have pursued absent the GPIM incentives, or relative to a merely reasonable procurement strategy that another, similarly situated utility might have pursued.

Final Observations

A GPIM that provides strong incentives for a broad range of procurement-related costs and revenues, using a benchmark that is both exogenous and adaptive to external circumstances, can benefit consumers through lower gas costs and reduced need for regulatory oversight of the procurement function. A GPIM can incent more active use of utility storage and other assets, which can contribute to market efficiency and mitigate market volatility, benefiting both utility customers and the broader market. Incentive mechanisms also can encourage hedging to some extent. If a target level and schedule of hedging is reflected in the benchmark, the utility minimizes its risk by staying close to the targets.

However, GPIM designs often reflect tradeoffs between competing principles, and the best balance between the various principles will depend upon each utility’s particular circumstances. Under some circumstances it may be difficult to design a GPIM that provides sound incentives while not exposing the utility to substantial risks, or customers to the potential cost of windfall rewards, without significant complexity in the benchmark formula. Accordingly, GPIMs may not be appropriate for some utility circumstances.



1. See, for instance, “Madigan Calls For $160 Million Refund For Nicor Customers,” press release dated Nov. 21, 2003, by Illinois Attorney General Lisa Madigan, alleging that Nicor Gas improperly sold low-cost gas reserves in order to profit under its gas procurement incentive mechanism. The incentive mechanism has been cancelled and the refunds are a subject of Illinois Commerce Commission Docket No. 01-0705.

2. As examples, GPIMs were once in place but have since been terminated for Nicor Gas (IL), Minnegasco (MN), Columbia Gas of Pennsylvania (PA), and Avista Utilities (WA). Laclede Gas (MO) had a GPIM that was terminated but later redesigned and reinstated.

3. Examples of GPIMs with benchmarks that assume little or no hedging are PG&E (CA), SoCalGas (CA), Laclede Gas (MO), Superior WL&P (WI), LG&E (KY), MidAmerican (IA), among others.

4. An example of a GPIM benchmark that includes a schedule for hedging is New England Gas (RI).

5. Examples of fairly comprehensive GPIMs are PG&E (CA), LG&E (KY), Atmos (TN), and Nashville (TN).

6. Examples are Laclede Gas (MO), whose benchmark uses fixed percentages by location; Alliant/Wisc. P&L (WI), whose benchmark uses pipeline reservation quantities; and Superior WP&L (WI), whose benchmark uses volumes from its Gas Supply Plan.

7. An example of a benchmark that adapts to changes in relative prices is PG&E (CA).

8. Examples of GPIMs with more limited scope are MidAmerican (IA), which provides incentives for commodity and pipeline reservation costs, but excludes gas sales and capacity release, and New England Gas (RI), which applies only to the upstream commodity portion of certain “discretionary purchases,” with a separate incentive plan for fixed costs associated with gas supply, storage, and transportation, including capacity release.

9. Examples of GPIMs that use actual volumes in the benchmark calculation are SoCalGas (CA), Southwest Gas (CA), NIPSCO (IN), Louisville G&E (KY), Atmos (TN), and MidAmerican (IA).

10. An example is the SoCalGas incentive mechanism during the 2000-2001 period. In 2000, SoCalGas injected relatively little gas for its customers during the spring and early summer, when prices into the SoCal system (SoCal Topock) were in the $2 to $4 range, and had to catch up during July to October when prices rose to the $4 to $6 range. In 2001, SoCalGas injection had the opposite pattern, with a very large amount injected before July 1 and very little after July 1; but prices were above $10 until June-July, after which they fell back under $4. The SoCalGas incentive mechanism reflects the actual pattern of storage injections in the benchmark, so the cost impact of the actual injection pattern relative to, for instance, a ratable injection pattern is not reflected in the calculations of ratepayer benefits or utility reward. In its reviews of the SoCalGas incentive mechanism during this period, the Office of Ratepayer Advocates stated that SoCalGas did an effective job of managing gas procurement, saving ratepayers $192.7 million and $192.4 million in 2000-2001 and 2001-2002, respectively. SoCalGas received large incentive rewards for each year that are subject to modification as a result of an ongoing investigation in CPUC Docket No. I.02-11-040. More recently, ORA has recommended that SoCalGas be required to inject gas ratably.

11. SoCalGas’ incentive mechanism has this characteristic. See Comparison of Incentives Under Gas Procurement Incentive Mechanisms, prepared testimony of James F. Wilson on behalf of Pacific Gas and Electric Co., filed Dec. 10, 2003, in California Public Utilities Commission Docket No. I-02-11-040, p. 21.

12. NIPSCO’s incentive mechanism explicitly states that contracts with such supplier recall rights are benchmarked to first-of-month prices, with “any replacement gas” also benchmarked to first-of-month prices. However, virtual storage deals, and parks and loans, are not benchmarked, and these or other approaches potentially could be used to replace any recalled gas with little or no impact on incentive reward.

13. See, for example, Monitoring and Evaluation Report of Southern California Gas Company’s Gas Cost Incentive Mechanism for April 1, 2004 through March 31, 2005, Office of Ratepayer Advocates, California Public Utilities Commission, Docket No. A.05-06-030, Nov. 30, 2005 (p. 1-1, “ORA’s review also confirmed that application of the sharing mechanism approved in D.02-06-023 results in a ratepayer benefit of $28.9 million and a shareholder reward of $2.5 million” and p. 1-6 Table 1-2, summarizing “Total Gas Cost Savings For Ratepayers”).

14. See, for example, Order Allowing Incentive Gas Supply Procurement Plan Award and Granting Extension of Plan, Iowa Utilities Board, Docket No. RPU-94-3, Nov. 29, 2004 (p. 3, “Through the IGSPP, MidAmerican customers have realized a savings, relative to the reference price, of $58.4 million …”).