Shareholder Incentives for Utility-Delivered

Energy Efficiency Programs in California

(Working Paper )

Jiyong Eom∗,a,1, James L. Sweeneya,1

aPrecourt Energy Efficiency Center, Stanford University, Y&Y Energy $ EnvironmentBuilding, 473 Via Ortega, Room 172, Stanford, CA 94305

Abstract

The State of California has committed an unprecedented sum of $2.2 billion inratepayer funds to utility-delivered energy efficiency programs from 2006 through2008; the State finalized in 2007 the determination of the shared-savings incentivemechanism for the 2006-2008 programs and beyond. This study seeks to examinewhether the adopted incentive mechanism would ensure an efficient delivery ofthe programs, and what reforms, if any, could be proposed to meet this end. Wedevelop an economic model for the implementation of the programs, in which aregulator adopts an energy savings target and a shared-savings incentive mecha-nism before a utility firm proposes program funding, gets it authorized, and beginsto manage it. The study reveals that each utility firm requires a certain minimumincentive rate to ensure that the firm will be encouraged to achieve the energysavings target, eventually bringing non-negative bill savings to its customers. Italso reveals that depending on market and regulatory circumstances, a higher-than-minimum incentive rate can be warranted to achieve not only a greater netsocial benefit but also greater bill savings for customers. Model-based analysis ofCalifornia energy efficiency programs suggests that a higher-than-adopted incen-tive rate is warranted and that social efficiency would be improved by customizing

∗Corresponding author: E-mail: jyeom@stanford.edu1The authors gratefully acknowledge the helpful suggestions by Lawrence H. Goul-

der and John P. Weyant. This paper is partly based on the first author’s unpublishedPh.D. dissertation, “Incentives and Politics of Utility-Based Energy Efficiency Programsin California.” This article copyrighted and reprinted by permission from the InternationalAssociation for Energy Economics/United States Association for Energy Economics. Thematerial first appeared in the online proceedings of the 28th USAEE/IAEE North Amer-ican Conference (New Orleans, December, 2008).

Preprint submitted to Elsevier April 30, 2009

incentive mechanisms for individual utilities and updating them on a regular basis.

Key words: Energy Efficiency, Utility Regulation, Shareholder Incentives,Shared Savings Incentive Mechanism, Ratepayer Funds

1. INTRODUCTION

Energy efficiency is increasingly being recognized as a resource warrant-ing aggressive public investment. For policy makers, escalating energy costsreflected in the economy, mounting concerns over global warming and as-sociated regulatory demand for mitigation measures, and increasing energydependency on politically unstable regions all contribute to a heightenedawareness of the need for increased energy efficiency. For utilities, particu-lar concerns are rising fuel prices; increasing uncertainty about cost-recoveryfor “steel-in-the-ground” investments; and intimidating environmental costs,in particular, potential greenhouse gas reduction costs. Improving energyefficiency is one of the most cost-effective ways to positively address thesechallenges.

Under these favorable circumstances, policy makers, regulators, and util-ities are taking a serious look at what policy instruments might be requiredto justify a substantial expansion of utility-delivered energy efficiency pro-grams. Among types of utility-delivered energy efficiency programs, thoseadministered by shareholder-owned utilities or, equivalently, investor-ownedutilities (IOUs), are attracting the most attention of policy makers becausethe alignment of the interests of ratepayers and IOUs is considerably moredifficult than that of ratepayers and municipal or cooperative utilities, whichare owned by the ratepayers themselves. Policy instruments that align theinterests of ratepayers and IOUs with regard to energy efficiency are also im-portant because IOUs are serving a majority of U.S. energy customers, whichmeans that if the instruments are not properly designed, a vast amount ofratepayer funds in energy efficiency will not produce the greatest possiblebenefits for the ratepayers.2

2Among types of electric utilities, IOUs have the largest service territories and sales.In 2006, IOUs in the U.S. served 68% of the total number of electricity customers andsold 66% of total electricity in MWh; IOUs in California served 78% of its total numberof electricity customers and sold 73% of total electricity in MWh (calculated from EnergyInformation Administration (2006))

2

As a necessary first step, many utility regulators are currently taking aserious look at a variety of utility shareholder incentives for energy efficiencyprograms that are being considered or are already in place in many statesin the U.S. Along with the interest in shareholder incentives, the State ofCalifornia has committed the unprecedented sum of $2.2 billion in ratepayerfunds to utility-delivered energy efficiency programs from 2006 through 2008and finalized in 2007 the determination of its shared-savings incentive mech-anism for the 2006-2008 programs and beyond (CPUC, 2007). A rigorousscrutiny of California energy efficiency programs and the associated incentivemechanism could assist the state, as well as other states contemplating ex-panding the programs, in delivering energy efficiency most effectively, whileensuring political support for the program implementation.

This research focuses on California energy efficiency programs, which em-ploy the shared-savings incentive mechanism. The incentive mechanism, bydefinition, specifies how large a share of the net benefits created by programimplementation the utility shareholders would be allowed to acquire once theutility achieves a certain preset performance target. This paper aims to ex-amine whether the adopted shared-savings incentive mechanism can ensurean efficient delivery of the energy efficiency programs, and what reforms, ifany, could be proposed to meet this end. The crucial question is whether theadopted incentive rate for the mechanism is appropriate not only to promptutility managers to commit their resources to the achievement of the en-ergy savings targets adopted by the California Public Utilities Commission(CPUC) but also to maximize net social benefits.

Versions of shared savings incentive mechanisms adopted by several statesvary in their specification of an incentive rate. Eto et al. (1998) reportedthat, before the market restructuring, shared-savings incentive payments toseven electric utilities in states such as California, New York, and New Jerseyaccounted for from 8% to 27% of the net benefits derivable from demand-side management programs. Most of these incentive rates provide a largerfraction of the net benefits to utilities than does the one adopted by theCPUC, which ranges only from 9% to 12%. Despite the variation in theseincentive rates, there is no evidence that serious scrutiny has been given totheir establishment. The CPUC has stated that “Establishing the level ofearnings opportunity for a shareholder risk/reward incentive mechanism isultimately a judgment call that the Commission must make, and not a precisescience.”(CPUC, 2007)

There is no consensus in the literature either. On the basis of a simplified

3

shared-savings incentive model, Stoft et al. (1995) proposed that the incen-tive rate of 100% is efficient because then the utilities’ profit would equal thenet social benefits.3 However, their suggestion is not appropriate for Califor-nia energy efficiency programs because their study presumes that (i) unlikein the California case, regulators do not have any preset performance targetsand only care about the net social benefits and (ii) there is no strategic inter-action between utilities and regulators, which is likely to be prevalent duringthe process of funding proposals and authorization. More importantly, theproposition made by Stoft et al. (1995) is not politically feasible because theincentive mechanism would then guarantee a monopoly position for utilitiesin the delivery of energy efficiency. While they advise the imposition of alump-sum charge on utilities, none of the reported mechanisms has offeredan incentive rate close to 100% along with such a lump-sum charge (Eto etal., 1998).

Considering a more general setting, called the “sharecropping” model,several authors provide some insights into the landlord-worker relationship,which appears to be similar to the CPUC-utilities relationship in Californiaenergy efficiency programs. Hurwicz and Shapiro (1978) found that a 50%split is efficient for a broad class of reward schemes to maximize the landlord’sresidual gain, namely, the part of output net of reward payment to the worker.Based on the model in which a random state of nature as well as actionstaken by the worker determines an overall gain, Holmstrom (1979) showedthat, without removing an efficiency loss due to the risk-averse nature of theworker, the performance-based sharing policy adopted by the risk-neutrallandlord could encourage the worker to take costly action that produces thegreatest possible ex-ante residual gain for the landlord. These sharecroppingmodels, however, do not apply to the California case. In addition to thesame problems that arise with the proposal of Stoft et al. (1995), the modelssuffer from the fact that what is to be maximized in the California case canbe the net social benefit, not necessarily the ratepayers’ bills savings as inthe models, and that the utility expenditures are funded by the ratepayersthemselves, not by the utility shareholders. To our knowledge, no study hasinvestigated the efficient design of a shared-savings incentive mechanism in

3This proposal is consistent with the one made earlier by Loeb and Magat (1979) that,in utility regulation, it is efficient to allow the utility to name its own service price and tosubsidize the utility on a per unit basis equal to the consumer surplus at the price.

4

connection with landlord-funded programs that are in place to accomplishthe maximum net social benefit while ensuring the achievement of presetperformance targets.

It is also important to note that the California case allows us to relateallocative efficiency of performance-based incentives, which has been inten-sively studied in the literature, to its distributional issues, which have mostlybeen underemphasized therein. A possible reason for the lack of this con-nection in the literature is the researchers’ practice of constructing a genericmodel based on simplifying assumptions, which may enable them to derivequalitative requirements for social efficiency but often keep them from ex-tracting useful distributional insights from the model. Other reasons may bethat there is no objective basis for assigning particular weights for the welfareof different stakeholders, or that welfare distribution has not, in fact, beenperceived as a critical issue. In the California case, however, the welfare dis-tribution must be seriously taken into account. In fact, how much earningsthe utilities should be allowed to garner in comparison with their customers’bill savings has been one of the most debated issues throughout the rule-making process of designing the shared-savings incentive mechanism. Thispaper answers the questions, (i) how would the welfare distribution acrossthe stakeholders vary with the specifications of the mechanism? and (ii) howwould the program efficiency be affected accordingly?

The rest of this paper is organized as follows: Section 2 presents back-ground of the research by reviewing the proceedings on California energyefficiency programs. Section 3 builds an economic model and analyzes theimplementation of the programs based on the shared-savings incentive mech-anism. Section 4 then presents numerical analysis results drawn from theprevious economic model and discusses policy implications. Finally, Section5 reviews the important findings presented in this paper.

2. BACKGROUND OF THE STUDY

Under traditional ratemaking practices, no utility has an incentive todeliver energy efficiency even if the utility’s direct program costs associatedwith program administration, implementation, and incentives to programparticipants are guaranteed to be reimbursed. This is primarily because oncerates are set, the utility earns extra return if ex-post sales exceed projected

5

sales, but loses if ex-post sales fall below the projection.4 It follows that theutility would have incentives to be in favor of supply-side investment evenif cost-effective energy efficiency measures were available. To reconcile theconflict between utilities’ interest in higher sales and the increasing publicinterest in energy efficiency, many Public Utilities Commissions (PUCs) haveauthorized or are considering revenue decoupling mechanisms for electric ornatural gas utilities. The term revenue decoupling refers to a rate adjustmentmechanism that simply ensures that utilities’ approved costs, including anallowed rate of return, are recovered regardless of the ex-post fluctuation ofretail sales.5

It should be noted, however, that revenue decoupling may not be sufficientfor encouraging utilities to deliver energy efficiency (Bachrach et al., 2004).In effect, under most ratemaking structures with revenue decoupling, utilityshareholders’ earnings opportunities are restricted only to a return on supply-side investments that can be added to the rate base. Thus utility managers,who in principle aim to first and foremost consider shareholder value in theirbusiness activities, have a clear incentive to promote supply-side resources,rather than energy efficiency. That is, more than revenue decoupling can berequired to gain utilities’ acceptance of energy efficiency programs.

Shareholder incentives are a commonly used approach for states that havea regulatory commitment to utility-delivered energy efficiency which goes be-yond addressing problems of lost revenue. As of 2007, eighteen states haveprovided various types of energy efficiency program incentives to investor-owned utilities (Kushler et al., 2006; National Action Plan for Energy Effi-ciency, 2007). Among these shareholder incentives, shared-savings incentivesare receiving the greatest attention because they explicitly account for boththe benefits and the costs of energy efficiency programs and therefore mayensure an economically efficient delivery of energy efficiency. By 1998, shared-

4While a utility’s revenue increases with electricity sales, its earnings may not increasedepending on the cost structure.

5As of 2006, at least seven states - California, Oregon, Maryland, North Carolina, Ohio,Utah, and New Jersey - have authorized revenue decoupling mechanisms for electric ornatural gas utilities, while more than five other states are considering such mechanisms(Kushler et al., 2006). The state of California, which eliminated a revenue decouplingmechanism called the Electric Rate Adjustment Mechanism (ERAM) in 1996, mandateda return to decoupling the utilities revenues and sales in 2001, stating: “The commissionshall ensure that errors in estimates of demand elasticity or sales do not result in materialover- or under-collections of the electrical corporation.”

6

Figure 1: California’s shared-savings incentive mechanism (Source: CPUC(2007))

savings incentives had been adopted by utilities in at least sixteen states (Etoet al., 1998) and, in the year of 2008, versions of the incentives are currentlyin place in six states including the state of California (National Action Planfor Energy Efficiency, 2007).

Figure 1 illustrates the shared-savings incentive mechanism adopted bythe CPUC in 2007. The figure indicates that the adopted mechanism is, ineffect, a “letter-graded” scheme: (i) utilities start to receive tiers of rewardover the threshold percentage savings of 85% relative to the CPUC-set energysavings targets; (ii) utilities are penalized under the threshold percentagesavings of 65%; and (iii) utilities receive neither reward nor penalty over thedead-band between the two thresholds. In particular, the reward structureis tiered such that shareholder earnings accrue at a marginal incentive rateof 9% if performance reaches 85% of the energy savings targets, and theincentive rate jumps to 12% if performance achieves 100% of the targets.That is, shareholder earnings are 9% of the net program benefit under thefirst earnings regime, and the earnings increase from 9% to 12% of the netprogram benefit under the second earnings regime. The CPUC estimatesthat, once the targets are precisely met, the 2006-2008 programs based onthe incentive mechanism will create a net program benefit of $2.7 billion ($2.4billion ratepayer savings and $0.3 billion shareholder earnings) from a $2.2billion investment in energy efficiency (CPUC, 2007). It is also important to

7

note that the CPUC has made it clear that the adopted incentive mechanismwill be in effect over the 2006-2008 programs and subsequent program cyclesuntil further notice (CPUC, 2007).

The letter-graded mechanism makes utilities’ earnings dependent not onlyon the net program benefits produced by the management of energy efficiencyprograms but also on their ex-post energy savings. With the incentive mech-anism, utilities potentially earn nothing and may even lose money if theyfail to accomplish the targets, despite having created substantial net benefitsfrom the management of energy efficiency programs. This characteristic ofthe incentive mechanism is consistent with the CPUC’s goals for energy effi-ciency programs, namely, maximizing net social benefits and meeting presetenergy savings targets, as pointed out in the Energy Efficiency Policy Manual:“The Commission’s overriding goal guiding its energy efficiency efforts is topursue all cost-effective energy efficiency opportunities over both the short-and long-term...The Commission translated this policy into specific annualand cumulative numerical goals for electricity and natural gas savings byutility service territory” (CPUC, 2005b).

In theory, the purpose of any regulatory incentive is to encourage itsagents to achieve certain regulatory objectives in the most efficient manner.It is expected that these objectives will ultimately maximize social welfare.When it comes to energy efficiency programs, such regulatory objectives haveoften taken the form of “dual goals”: to attain the maximum net socialbenefits while ensuring the achievement of preset targets.6 With this dual-

6There are two possible reasons why the incentive mechanisms for utility-deliveredenergy efficiency programs pursue the dual-goals approach rather than that of single net-benefit maximization. First, policy makers can elicit a greater net social benefit by de-ploying a series of energy efficiency programs, guided by particular energy savings targets,which can be updated when necessary. In doing so, the policy makers can advance energyefficiency in accordance with the resolution of a variety of market uncertainties that mayaffect the cost-effectiveness of pursuing energy efficiency. These uncertainties include notonly unpredictable energy demands but also technological changes and utilities’ effective-ness in delivering energy efficiency. This rationale is consistent with the CPUC’s Decision04-09-060, which established its energy savings targets for 2006 and beyond, permittingpossible updates of these targets. The second possible reason for policy makers to takethe dual-goals approach concerns the funding for energy efficiency programs. Even in theabsence of the abovementioned uncertainties, ratepayers may not support a substantialamount of energy efficiency funding, which increases their rates in the short-run, until theyperceive significant bill savings over time. This, in turn, leaves utilities open to the risk ofnot recovering all of their program expenditures, which might lead to “window-dressing”

8

goals approach, policy makers authorize a series of ex-ante efficient energysavings targets, while restricting ratepayer funding for individual programcycles. Indeed, this strategy is being employed by the CPUC, as stated in itsDecision 04-09-060: “In submitting proposed energy efficiency program plansand funding levels to meet the savings goals adopted by the Commission,the programs administrator(s) shall demonstrate that their proposed level ofelectric and natural gas energy efficiency program activities and funding isconsistent with the Commission’s-adopted electric and natural gas savingsgoals...Today’s adopted goals take into consideration the practical limits toeffectively increasing program funding and ramping up programs to capturethe full economic potential of energy efficiency in the near term”(CPUC,2004).

Note, however, that while the letter-graded nature of the shared-savingsincentive mechanism is necessary to accomplish the dual regulatory goals,the mechanism itself may not be sufficient to align shareholder interests andratepayer interests. To achieve an alignment of these interests, the level offorecasted earnings under the incentive mechanism must represent a sizeablefinancial opportunity for utility shareholders. In this regard, it is crucial torecognize that forecasted shareholder earnings from the programs must beno less than utilities’ opportunity costs associated with the management ofthe programs, which are not apparent to regulators and thus not reimbursed.Such opportunity costs include (i) hidden costs associated with paying man-agerial attention to energy efficiency programs or adjusting limited orga-nizational resources for the efficient management of the programs and (ii)shareholder earnings from supply-side projects that are displaced by energyefficiency programs (Stoft et al., 1995).

Furthermore, utilities’ earnings opportunities suggested by a particularshared-savings incentive mechanism must take into account a dynamic pro-cess of program implementation, in which different entities strategically in-teract with each other. As discussed, California energy efficiency programsare implemented in such a way that, given the CPUC-set energy savings tar-gets, individual utilities propose their own program plans and funding levelsto meet the targets, get the proposals authorized by the CPCU, and begin

efforts by the utilities in energy efficiency. Huge program funding is not supported alsofrom the social efficiency point of view because ratepayer funding typically involves non-trivial economic distortions, which define the opportunity costs of the program funding.

9

to manage the programs. In this process, strategic interaction is likely tooccur between utilities seeking to receive the greatest possible funding andthe CPUC seeking to authorize the least possible funding that ensures theachievement of its dual goals.7

Based on these considerations, we construct an economic model, in whichboth a regulator who establishes a shared-savings incentive mechanism anda utility firm facing the opportunity costs of program management behaveoptimally at any stage of program implementation. Because a higher levelof abstraction will not offer richer insights into the model, we instead choosenon-generic functions that allow for a close examination of the Californiacase while permitting moderate loss of generality.

This research by its nature applied. The main focus is to investigate theshared-savings incentive mechanism that has been adopted by the CPUCunder its dual regulatory goals for the energy efficiency programs, not toobtain the first-best optimal regulatory process or mechanism for the deliv-ery of public programs in general. Many of the normative suggestions fromthe economic analyses that will follow are provided in the context of Cali-fornia’s programs and its incentive mechanism. This more realistic approachis, nonetheless, inherently different from what is used for the ex-post assess-ment of program delivery and impacts. With the approach taken by thisresearch, we can describe and predict how the energy efficiency programswill be delivered under the incentive mechanism and how the changes in avariety of market circumstances will affect the effectiveness of the programs.This paper starts with this perspective and direction.

3. THE ECONOMIC MODEL

The economic model presented here analyzes California’s shared-savingsincentive mechanism, which aims to achieve the dual goals of utility-deliveredenergy efficiency programs, that is, to maximize the net social benefit while

7Funding for energy efficiency programs will cover the so-called Program Administra-tion Costs (PAC), which consists of customer incentives and rebates, installation costs,operating costs, and marketing/outreach costs. The PAC differs from the Total ResourceCosts (TRC), which also includes the costs paid by participants. Note that the incentivemechanism adopted by the CPUC employs the Performance Earning Basis (PEB) as itsearnings basis, which is defined as the sum of 2/3 of the TRC-based net benefit and 1/3 ofthe PAC-based net benefit. The PAC-based net benefit, which is employed as the earningsbasis in this study, is an overestimation of the PEB.

10

ensuring the achievement of preset energy savings targets. Note that success-ful implementation of the programs can be impeded by potential informationasymmetry between utilities and their regulators. In particular, the regula-tors may not have complete information about how effective the utilities willbe in the management of the programs, and the utilities may want to exploitthe regulators’ incomplete information (Baron & Myerson, 1982; Joskow &Schmalensee, 1986). The accommodation of information asymmetry couldallow the analysts of the regulatory process to achieve more realistic insightsinto the implementation of energy efficiency programs.8

However, a benchmark analysis performed in the absence of informationasymmetry is valuable because ignoring utilities’ strategic behavior based onasymmetric information allows for a more focused look at the implicationsthat the shared-savings incentive mechanism itself has for the implementationof energy efficiency programs. In this paper, we present an economic model bypresupposing information symmetry according to which utilities, customers,and regulators all have the same information. Consistent with the Californiacase, the entire process of program implementation is structured as follows:

• Stage 1: The regulator, given an energy savings target of t ∈ (0,∞),chooses a marginal incentive rate of r ∈ (0, 1] for the incentive mecha-nism.

• Stage 2: The utility firm with program management efficiency, θ ∈(0,∞), proposes funding K ∈ (0,∞), which will be used for energyefficiency activities directed toward a customer base of size β. Thenthe regulator authorizes the proposal as long as K is no greater thanthe funding level that the fully-informed regulator estimates in orderfor the firm to achieve t.

• Stage 3: Once K is authorized, the firm manages the programs in sucha way that it produces an energy savings level of x ∈ [0,∞) with anenergy savings productivity of e ∈ [0,∞).

In this model, e is defined as energy savings achieved per dollar of pro-gram expenditure, and θ is a measure representing how much less a utility

8For a detailed analysis of utility-delivered energy efficiency programs under asymmetricinformation, see Eom (2009).

11

firm would lose to achieve a given level of e, all other factors held constant.We will analyze the equilibrium of the model, in which the regulator and theutility firm behave optimally at any stage. Below are the definitions of keyfunctions that will be used for the analysis.

The Incentive Mechanism: There is a shared-savings incentive mech-anism by which earnings of the utility firm begin to accrue at a uniformmarginal incentive rate of r ∈ (0, 1] if the firm’s ex-post energy savings xmeets or surpasses regulator-set target t.9 That is, if x is greater than orequal to t, for every dollar of net benefits, the firm appropriates r and theconsumers take the rest, 1− r, whereas if x is lower than t, the firm acquiresno earnings. Incentive mechanism R is then represented as follows:10

R =

{r, if x ≥ t0, if x < t.

(1)

This representation indicates that the marginal incentive rate is zero over0 ≤ x < t, which is called the dead-band region, but surges to infinity aroundx = t. Therefore, the firm with this mechanism will have no incentives toincrease ex-post energy savings x within the dead-band region but will havesubstantial incentives to move at least to the upper-end of this region.

The Gross Benefit of Energy Savings: The gross benefit of energy sav-ings produced by the energy efficiency programs is equivalent to the totalavoided costs to the utility firm—supply-side generation, transmission, dis-tribution, and environmental costs avoided by the energy savings. The grossbenefit, denoted as B [$], is a function of produced energy savings x [MWhannual savings] and marginal energy savings benefit η [$/MWh annual sav-

9For the welfare impacts of introducing another tier into the incentive rate, see Eom(2009).

10In fact, the CPUC-adopted shared-savings incentive mechanism shown in Figure 1 ismore sophisticated than the simplified mechanism (1). However, the simplified mechanismcan be valid for the analysis of the California case (i) if r is set high enough for the firmto always be encouraged to accomplish t and (ii) if there is no significant uncertaintyassociated with forecasting customers’ participation and associated load impacts. If theseassumptions hold, the firm will achieve at least t and thus appropriate r for every dollarof net program benefit because regulatory provision ensures that program funding willbe developed to achieve t (CPUC, 2004). I presuppose that the assumptions are fulfilledthroughout this paper.

12

ings]. Marginal energy savings benefit η refers to the change in total avoidedcosts arising when achieved energy savings change by one unit. η allows forthe variation across different program cycles or different utility service ter-ritories.11 We assume that η is constant over x, that is, that the size of thesupply-side investments avoided by the firm increases proportionately withits energy savings achievement.12 We get, then

B = η x. (2)

Energy Savings: The utility firm produces energy savings x from programfunding K with an energy savings productivity of e. Energy savings produc-tivity e increases with customer-oriented activities associated with pursuingpotential avenues for cost-effective energy savings opportunities throughouta program cycle in such a way that more energy savings can be elicited fromauthorized funding. For instance, having learned how to facilitate the pen-etration of high-potential measures into the customer base at lower costs,the firm may implement such measures on a larger scale by replacing somehigh-cost and low-potential measures, resulting in a decrease in the ex-postaverage cost of energy savings.13

To keep the analysis mathematically tractable, we assume that ex-postenergy savings x [MWh annual savings] are proportional to public funding K[$] and that the marginal return of the fund is equivalent to the firm’s energy

11For example, the avoided costs of transmission and distribution (T&D) capacity arelocation-specific. The T&D cost avoidance is achieved only if it occurs in part of theelectric system where the T&D investments would have been required.

12To be precise, the proportionality assumption may not hold if the avoided costs dependon the time when any energy savings occur. In fact, in the 2006 Avoided Cost Update(CPUC, 2006), the CPUC adopted correction factors to adjust the avoided costs duringpeak hours.

13The implementation process of energy efficiency programs determined by the CPUCallows individual utilities to enhance their energy savings productivity regarding the autho-rized programs. In fact, the funding for 2006-2008 includes funding for market penetrationstudies and process evaluations to assess which measures are working and which measuresare not. On the basis of such information, the utilities can shift resources among bud-get categories within programs as well as across programs under the fund shifting rules,providing the utilities with the flexibility to use their authorized funding to achieve evengreater energy savings (CPUC, 2005c).

13

savings productivity e [MWh annual savings/$], represented as follows:

x = e K. (3)

The Opportunity Costs of Managing Energy Efficiency Programs:Utilities’ energy efficiency activities incur sizeable opportunity costs, whichmay not be apparent to utility regulators and thus are not reimbursed throughprogram funding. The opportunity costs include hidden managerial costs as-sociated with the management of energy efficiency programs and potentialforegone shareholder earnings from supply-side projects that are displacedby energy efficiency (Stoft et al., 1995).14

It is assumed that the opportunity costs are increasing and convex withthe firm’s energy savings productivity e. The convexity assumption is sensiblebecause the firm can only enhance its energy savings productivity with theadjustment of its organizational and financial resources, which are generallyconstrained, so that greater productivity incurs disproportionately highercosts to the firm. To keep the problem tractable without losing much gener-ality, we chose a quadratic opportunity cost function, which is as follows:

ψ(e; β, θ) =β

2θe2, (4)

Here, θ [MWh annual savings squared/$cubed] is the firm’s program man-agement efficiency, which contributes to lowering ψ [$], and β is the size ofthe firm’s customer base, which raises ψ proportionately. That is, to achievethe same level of e, a firm will lose less when it has higher θ and a smaller β.

The variation in program management efficiency θ across utility firms isattributable to their differences in codified or even tacit knowledge of pro-gram management, which has been acquired from prior implementation andevaluation of the programs. Such a knowledge asset can be embodied in well-established communication channels with the programs’ participants, such ascustomers and subcontractors, and in dexterity at incorporating informationabout energy savings opportunities into program designs. It would be fair

14Note that an economic interpretation of supply-side foregone earnings should accountfor the possibility that utility shareholders can put their investment money elsewhere,earning a return similar to the one they would have earned on the displaced supply-sideinvestments.

14

to say that the individual utilities’ program management efficiency is an id-iosyncratic resource and is, to some extent, private information.15 In thispaper, however, we assume that θ is perfectly known to the customers andthe regulator.

Having stipulated the key functions, we now analyze the equilibrium ofthe model by employing the usual backward induction.

Stage 3: The Firm’s Program ManagementIn the pursuit of energy efficiency, utility managers, who first consider share-holder value in their business, balance gross shareholder earnings—a share ofa net program benefit—with the associated opportunity costs borne by theshareholders. In other words, by choosing an optimal level of energy savingsproductivity e, the utility firm maximizes its cost-adjusted earnings from en-ergy efficiency programs or, equivalently, net earnings on energy efficiency.The firm’s maximum net earnings on energy efficiency are then representedas follows:

U = maxe

R(B −K)+ − ψ

= max

{maxe≥ t

K

r(ηeK −K)− β

2θe2, max

e< tK

− β

2θe2

}. (5)

As shown, a higher marginal energy savings benefit (i.e., higher η), a higherincentive rate (i.e., higher r), or greater program funding (i.e., higher K)enhances the firm’s gross earnings, whereas higher program managementefficiency (i.e., higher θ) or a smaller size of the customer base (i.e., lower β)reduces the firm’s opportunity costs. For convenience, we call the first groupof changes (i.e., higher η, higher r, and higher K) the earnings-enhancingfactors, and the latter change (i.e., higher θ and lower β) the cost-reducingfactors.

The right term in the brackets of (5) indicate the case in which the utilityfirm expects that no matter how productive it is in the management of theenergy efficiency programs, the programs will result in a net cost (i.e., there

15To lessen the extent to which utilities’ informational advantage could impede effi-cient regulatory oversight, the CPUC ordered non-financially interested “Peer ReviewGroups”(PRGs) to be formed and review the utilities’ proposed plans and funding pro-posals before they are submitted (CPUC, 2005c).

15

will be no earnings basis that can be shared between the firm and its cus-tomers); the firm therefore decides not to produce any energy savings, whichresults in the net earnings of zero. The zero earnings can be thought of asreservation earnings for the firm, which are given regardless of its programperformance.

However, if a sufficient level of net program benefit is expected, the firmwill be encouraged to manage the energy efficiency programs with the fol-lowing level of energy savings productivity:

e∗ =ηrθK

β. (6)

We call this the profitable-and-optimal level of energy savings productiv-ity, by which the firm can acquire non-negative and maximum net earnings—the left term in the brackets in (5). Equation (6) establishes that either theearnings-enhancing factors (i.e., higher η, higher r, and higher K) and thecost-reducing factor (i.e., higher θ) encourage the firm to enhance its energysavings productivity.

Substituting (6) for (5) results in the firm’s maximum net earnings withthe profitable-and-optimal energy savings productivity e∗, given as

U = max

{η2r2θ

2βK

(K − 2β

η2rθ

), 0

}. (7)

It follows that e∗ is chosen by the firm only when the energy efficiency pro-grams and the associated funding fulfill the following constraint:

K ≥ 2β

η2rθ. (8)

This is referred to as the firm’s participation constraint, which ensures thatthe firm will produce energy savings by choosing e∗, rather than zero pro-ductivity for seeking the reservation earnings. Constraint (8) suggests thatthe earnings-enhancing factors and the cost-reducing factor can be compen-sated for by decreased program funding to ensure the firm’s participation inprogram management. Note that the lower bound of the constraint is pro-portional to the size of the customer base.

Stage 2: The Firm’s Funding ProposalFollowing the provision that requires any funding proposal to be developed

16

to meet t, the regulator expects K to be no greater than what the utility firmneeds to just meet t with its profitable-and-optimal level of energy savingsproductivity e∗, so that

x∗ = e∗K ≤ t. (9)

Then the firm will propose the greatest funding K̃ fulfilling constraint (9).Plugging (6) into the equality form of (9) yields the firm’s funding proposal

K̃ = β

(1

ηrθ

t

β

)1/2

= β

(t̄

ηrθ

)1/2

, (10)

which we term the justifiable funding level. For simplicity of exposition, weuse the shorthand notation t̄ for t/β and call it per-customer energy savingstarget. Equation (10) indicates that higher t̄ allows the firm to propose higherK̃ and that if t̄ is held constant, the firm’s proposal K̃ becomes proportionalto β. That is, an increase in the size of the customer base will have a positiveand proportional effect on the justifiable funding level as long as the increaseis accompanied by a proportional increase in energy savings target.

Moreover, (10) indicates that the firm must propose lower K̃ either withthe earnings-enhancing factors (i.e., higher η and higher r) or with the cost-reducing factor (i.e., higher θ). The intuitive explanation is as follows: Withthe earnings-enhancing factors or the cost-reducing factor, the firm is encour-aged to achieve higher productivity e∗, as indicated by (6). By utilizing thefirm’s motivation, the regulator can authorize lower program funding whilestill ensuring the achievement of the target t.

Consequently, the equilibrium of program implementation is that the util-ity firm proposes K̃, gets the proposal authorized by the regulator, andchooses e∗ in program management to achieve t. Substituting (10) for (6) re-sults in the firm’s profitable-and-optimal level of energy savings productivityat equilibrium, given by

e∗ = (ηrθt̄)1/2 . (11)

This equation, in comparison with (6), suggests that the productivity aug-menting effects of the earnings-enhancing factors (i.e., higher η and higherr) and the cost-reducing factor (i.e., higher θ) are, to some extent, mitigatedby the regulator’s incentive to exploit the augmented productivity in termsof authorizing lower program funding.

The utility firm’s opportunity costs at equilibrium are represented as

ψ(e∗, β, θ)|K=K̃ =ηrt̄

2β, (12)

17

establishing that the earnings-enhancing factors (i.e., higher η and higherr), not the cost-reducing factor (i.e., higher θ), raise the firm’s opportunitycosts. Note that the firm’s cost-reducing factor, in effect, has nothing todo with its opportunity costs. The reason is that the direct impacts of thecost-reducing factor on the opportunity costs is, at equilibrium, offset by thefactor’s indirect impacts on the opportunity costs through an increase in e∗.Yet, the earnings-enhancing factors contribute to an increase in the firm’sopportunity costs by raising e∗.

Equation (12) also indicates that the opportunity costs increase with per-customer energy savings target and that if the per-customer target is heldconstant, the costs become proportional to the size of the customer base.This result is sensible for most of the energy efficiency programs because, asthe programs expand on a customer size basis, utilities’ hidden managerialcosts and foregone supply-side earnings are likely to increase proportionately.

Stage 1: The Regulator’s Incentive DecisionGiven the motivation of the regulator and the firm, we derive a set of in-centive rates to ensure that the firm choosing energy savings productivity e∗

will achieve target t by utilizing justifiable funding K̃. Such a set is obtainedby plugging the justifiable funding level (10) into the firm’s participationconstraint (8) and is represented as follows:

4

η3θt̄≤ r ≤ 1. (13)

This constraint implies that the regulator can establish a lower incentiverate (i.e., lower r) with the earnings-enhancing factors (i.e., higher η andθ) or with higher per-customer energy savings target (i.e., higher t̄). Theregulator can do so because all of these factors offer the firm greater earningsopportunities, which can be offset by lowering the incentive rate, while stillensuring the firm’s participation in the programs.

We call constraint (13) the jointly sufficing constraint, in the sense thatany r fulfilling the constraint guarantees that the firm’s program managementwill achieve the regulator-set target, while bringing non-negative net earningsto the firm and non-negative bill savings to the customers. In other words, anincentive rate fulfilling the jointly sufficing constraint can lead to a Pareto-superior allocation relative to the status quo, by aligning shareholder interestswith customer interests regarding energy efficiency programs. The alignmentof interests is achieved by the firm viewing its management of the programs

18

as one that can generate meaningful shareholder earnings—earnings thatare no less than the opportunity costs of program management—and bythe customers expecting that they will receive bill savings and be protectedagainst financial losses on their investment.

As long as constraint (13) is fulfilled, the programs pass not only thenet program benefit criterion, namely, ηx − K ≥ 0, but also the net so-cial benefit criterion that accounts for the firm’s opportunity costs, namely,ηx − K − ψ ≥ 0; the satisfaction of both of these criteria justifies the im-plementation of the programs. Put differently, the regulator must explicitlyconsider the jointly sufficing constraint in the design of an incentive mecha-nism because, otherwise, the mechanism might preempt the firm’s incentiveto deliver energy savings, resulting in net costs to society, which will com-pletely eliminate political support for the implementation of the programs.This requirement is summarized as follows:

Proposition 1. The regulator’s choice of r must fulfill the jointly sufficingconstraint, so that the firm’s program management will make both the firmand its customers weakly better off.

For ease of terminology, we call any incentive rate fulfilling the jointlysufficing constraint (13) a jointly sufficing rate, and we call the lower boundof the constraint, 4β/η3θt, the minimum sufficing rate, which is denoted asrmin. Here we introduce a useful synthetic parameter, ξ, defined as

ξ (t, η, β, θ) ≡ η3θt̄

4=

1

rmin

. (14)

Then the jointly sufficing constraint (13) is rewritten as

1/ξ ≤ r ≤ 1. (15)

We call the synthetic parameter, ξ, the design flexibility because it indicateshow much earnings opportunities the programs can provide to the firm andthus represents the extent to which the regulator is flexible in its choice of ajointly sufficing rate. As indicated by (15), higher ξ broadens the range ofjointly sufficing rates and thus offers a more complete set of design choicesto the regulator.

Design flexibility ξ has important characteristics. First, ξ is not depen-dent on the size of the customer base β, but only on per-customer energy

19

savings target t̄. Thus, as long as energy savings target increases propor-tionately with the size of the customer base, ξ will be the same regardlessof the size of the customer base to which the programs are delivered. Sec-ond, ξ is determined by per-customer energy savings target t/β, marginalenergy savings benefit η, and the firm’s program management efficiency θ ina substitutive manner. For instance, for the same level of ξ, lower t̄ can becompensated for by either higher η or higher θ. Thus the regulator may facethe same set of jointly sufficing rates even under dissimilar market and regu-latory circumstances exhibiting different t̄, η, or θ, as long as the parameterscollectively lead to the same ξ. Third, if any working incentive mechanism isto exist, ξ must be no less than one. In other words, if some of the param-eters were so low that the other parameters could not compensate in such away as to yield ξ ≥ 1, no incentive rate would be workable for the incentivemechanism.

Having analyzed the equilibrium of the entire process of the implemen-tation of energy efficiency programs, we go back to the question of how theprograms can accomplish the dual regulatory goals, namely, maximizing netsocial benefit while ensuring the achievement of preset targets. Under the sce-nario in which the regulator chooses any jointly sufficing rate, meeting targett is accomplished through the equilibrium implementation of the programs:given the jointly sufficing rate (at Stage 1), the firm proposes justifiablefunding K̃, gets the proposal authorized by the regulator (at Stage 2), andmanages the programs with the profitable-and-optimal level of productivitye∗, achieving target t, no more or less (at Stage 3).

What remains is the maximization of the net social benefit created bythe implementation of energy efficiency programs. We define the net socialbenefit in terms of joint welfare for the firm and its customers, recognizingthat the utility shareholders themselves can be better off with the imple-mentation of the programs. Therefore, what is to be maximized is the netsocial benefit, which accounts also for the firm’s opportunity costs, not thenet benefit metric that precludes such costs.16

To derive conditions ensuring the maximization of the net social benefit,we first identify both the firm’s net earnings and the customers’ bill savings

16The net benefit that precludes opportunity costs is a metric is called the net programbenefit and is used in this paper as an earnings basis for the shared-savings incentivemechanism.

20

at equilibrium. Given r, t, and η, the utility firm with β and θ acquires thefollowing net earnings on energy efficiency:

U(r; t̄, η, β, θ) = r(ηt̄β − K̃)− ψ(e∗, β, θ)|K=K̃

= r

{ηt̄

2−(

t̄

ηrθ

) 12

}β. (16)

As a result, the firm’s customers receive, in the aggregate, the following billsavings:

V (r; t̄, η, β, θ) = (1− r)(ηt̄β − K̃)

= (1− r)

{ηt̄−

(t̄

ηrθ

) 12

}β. (17)

Again, as pointed out by Proposition 1, both U(r; t̄, η, β, θ) and V (r; t̄, η, β, θ)are non-negative as long as adopted incentive rate r fulfills the jointly sufficingconstraint (15). The net social benefit are then given by

W (r; t̄, η, β, θ) = ηt̄β − K̃ − ψ(e∗, θ)|K=K̃

=

{ηt̄(

1− r

2

)−(

t̄

ηrθ

) 12

}β. (18)

Equations (16) through (18) indicate that, U , V , and W all increase withper-customer energy savings target t̄, and that if t̄ is held constant, they areproportional to the size of the customer base β.

Having specified the complete representation of the net social benefit,the incentive rate that induces the firm to produce the maximum net socialbenefit is obtained by solving the usual maximization problem for equation(18) subject to the jointly sufficing constraint (15). The socially efficientincentive rate is given by the following proposition:

Proposition 2. The socially efficient incentive rate rS∗ is given by

rS∗ =

{(1/4ξ)1/3, if ξ > 2

1/ξ, if 1 ≤ ξ ≤ 2.

This proposition indicates that the socially efficient incentive rate rS∗ ispiece-wise over the ranges of design flexibility ξ , where ξ ≥ 1 must hold

21

Figure 2: The socially efficient incentive rate with respect to design flexibility

for the existence of any jointly sufficing rate. If ξ > 2, it is socially efficientto adopt the interior solution, r = (1/4ξ)1/3, instead of the corner solution,r = 1/ξ ≡ rmin. This is because such a sizeable level of ξ offers a completeset of design choices, within which the regulator can establish the interiorsolution that balances the marginal social benefit of adopting higher r (i.e.,a decrease in justifiable funding K̃) and associated marginal social cost (i.e.,an increase in opportunity costs ψ∗). In this case, the firm acquires strictlypositive net earnings.

In contrast, if 1 ≤ ξ < 2, the corner solution, r = 1/ξ, must be chosenfor social efficiency. The reason is that such a limited level of ξ alreadyrequires an incentive rate r so high that no interior solution could producegreater joint welfare. In this case, the firm, in effect, gains nothing from themanagement of the programs because its gross earnings are exactly offset byits opportunity costs. Consequently, the level of ξ determines whether or notthe interior solution, which is a higher-than-minimum incentive rate, mustbe established and whether or not the firm will eventually become better offwith the management of the programs.

Figure 2 illustrates the socially efficient incentive rate rS∗ with respect toξ, where corresponds to the upper envelope of the two crossing curves. Thefigure indicates that, regardless of ξ , the higher the level of ξ, the lower thelevel of rS∗. This is represented as a simple corollary to Proposition 2.

22

Corollary 1. rS∗ decreases with ξ, that is, drS∗/dξ < 0.

Note that rS∗ decreases with ξ even under ξ > 2. This is because, withan increase in ξ, the marginal cost of adopting a higher incentive rate dueto an increase in ψ∗ rises faster than the associated marginal benefit due toa decrease in K̃. Likewise, with a decrease in ξ, a higher incentive rate canbe established because the associated marginal benefit rises faster than theassociated marginal cost.

Corollary 1 has significance for an efficient design of a shared-savings in-centive mechanism. It indicates that, ceteris paribus, a lower incentive rate isefficient for energy efficient programs with a higher per-customer energy sav-ings target (i.e., higher t̄), for those managed by a firm with higher programmanagement efficiency (i.e., higher θ), and for those implemented in utilityservice areas with a higher marginal energy savings benefit (i.e., higher η).

Emerging from the above analysis is the fact that the socially efficientincentive rate or any single incentive rate, in general, may not best representthe preferences of the utility firm and its customers. To examine the implica-tions of any adopted incentive rate for the firm and its customers, we obtaintheir preferred incentive rates by solving the maximization problems for (16)and (17), both subject to (15), which leads to the following propositions:

Proposition 3. The firm prefers to receive all of a net program benefit gen-erated by its management of the programs. That is, the firm’s preferredincentive rate rF∗ is one.

Proposition 4. The customers’ preferred incentive rate rC∗ (which is de-fined as the share provided to the firm) is given by

rC∗ =

{ρC(ξ), if ξ > 31/ξ, if 1 ≤ ξ ≤ 3,

where ρC(ξ) is a function that is given by the first-order optimality conditionfor (17).

While the firm’s solution rF∗ is straightforward, the customers’ preferredincentive rate rC∗ is piece-wise over the ranges of ξ, as is the socially efficientincentive rate rS∗. Proposition 4 indicates that if ξ > 3, the customers preferthe interior solution, r = ρC(ξ), rather than r = 1/ξ ≡ rmin. This is becausesuch a sizeable level of ξ offers a more complete set of design choices, within

23

which the customers can find the interior solution; this solution balances thecustomers’ marginal benefit of allowing higher r (i.e., an increase in a netprogram benefit due to a decrease in K̃) and the associated marginal cost(i.e., a decrease in their share of the net program benefit). In this case, thecustomers’ solution permits the firm to acquire strictly positive net earnings.

By contrast, if 1 ≤ ξ ≤ 3, the customers prefer the corner solution,r = 1/ξ. This preference is warranted because such a limited level of ξalready requires substantially high r, so that no interior solution could pro-duce greater bill savings to the customers. In this case, the firm gains nothingfrom the management of the programs because its net earnings are exactlyoffset by the associated opportunity. Consequently, the level of ξ determineswhether or not the customers will support the higher-than-minimum interiorsolution and whether or not the solution will make the firm better off withthe management of the programs.

Propositions 2 through 4 collectively suggest that the regulator’s choice ofan incentive rate can have very different impacts on the firm, its customers,and society as a whole. Figure 3 illustrates the firm’s net earnings U , thecustomers’ bill savings V , and net social benefit W as a function of adoptedincentive rate r, which is jointly sufficing. These relationships are also shownunder different market and regulatory circumstances represented by differentranges of design flexibility ξ, that is, 1 ≤ ξ ≤ 2, 2 < ξ ≤ 3, and ξ > 3;due to the difference in the ranges of ξ, the stakeholders’ earnings/savingsopportunities vary considerably, resulting in the differences in the levels ofminimum sufficing rates and in the sizes of the figures. Figure 3 indicatesthat while U is convex and that both V and W are concave in r regardlessof ξ, the first-order impacts of r on V and W differ by the ranges of ξ:(a) for 1 ≤ ξ ≤ 2, both V and W are monotonically decreasing in r; (b)for 2 < ξ ≤ 3, V is monotonically decreasing but W is non-monotone inr; and (c) for ξ > 3, both V and W are non-monotone in r. These first-and second-order impacts, along with Propositions 2 through 4, provide thefollowing proposition:

Proposition 5. When there exists any jointly sufficing incentive rate, thatis, ξ ≥ 1, the socially efficient incentive rate rS∗ lies between the customers’preferred incentive rate rC∗ and the firm’s preferred incentive rate rF∗, whichis represented by

1

ξ≤ rC∗ ≤ rS∗ ≤ rF∗ = 1,

24

Figure 3: Impacts of jointly sufficing rate r on U , V , and W under differentranges of design flexibility ξ : (a) 1 ≤ ξ ≤ 2, (b) 2 < ξ ≤ 3, and (c) ξ > 3

25

where the strict inequality rS∗ < rF∗ holds for ξ > 1, rC∗ < rS∗ holds forξ > 2, and 1/ξ < rC∗ holds for ξ > 3.

This proposition answers the overarching question of whether or not analignment of the interests of a utility firm and its customers regarding theimplementation of energy efficiency programs would be sufficient to accom-plish the maximization of their joint welfare. Proposition 5 indicates that themere alignment of the interests may not be sufficient. Suppose, for instance,the regulator adopts the minimum sufficing rate in an attempt not only toalign the stakeholders’ interests but also to avoid potential backlash from thecustomers. Such an incentive rate, however, encourages the firm to producethe maximum net social benefit only if ξ is limited such that 1 ≤ ξ ≤ 2. If ξ issizeable such that ξ > 2, a higher-than-minimum incentive rate is warrantedfor social efficiency.

More insights are derived from the extreme case when the regulator is con-cerned only about the customers’ welfare in the design of the incentive mech-anism. Proposition 5 suggests that if 1 ≤ ξ ≤ 3, the minimum sufficing ratenot only aligns the interests of the stakeholders but also generates maximumbill savings for the customers. If ξ > 3, however, a higher-than-minimumincentive rate must be established to maximize bill savings. Consequently,the level of design flexibility determines whether or not the alignment of thestakeholders’ interests merely with the minimum sufficing rate will suffice tomaximize the net social benefit and the customers’ bill savings and whetheror not a higher incentive rate is required.

Furthermore, Proposition 5 suggests that a conflict of interests regardingan incentive rate can exist between the firm and its customers and that,in some sense, the regulator’s choice r determines how the conflict will bereconciled. In the sensible case where the regulator adopts r between rC∗

and rF∗, the closer the adopted r is to rC∗, the greater the customers’ billsavings and the smaller the firm’s net earnings; the closer the adopted r isto rF∗, the greater the firm’s net earnings and the smaller the customers’ billsavings.

Note, however, that in the particular case where the regulator choosesrS∗ to maximize the joint welfare, the choice may appear to be preferentialtreatment to the customers from the perspective of the utility firm. As Figure3 demonstrates, while the choice rS∗ under ξ > 2 can be seen as one thatreconciles the interests of the firm and its customers, the choice rS∗ under1 ≤ ξ ≤ 2 may seem to favor the customers without any attention being paid

26

to the firm’s interests. This seemingly preferential treatment given to thecustomers under the social efficiency scenario is summarized in the followingcorollary to Proposition 5.

Corollary 2. The socially efficient incentive rate under 1 ≤ ξ ≤ 2, in effect,allows the customers to receive the greatest possible bill savings without per-mitting any net earnings to the firm, whereas, the socially efficient incentiverate under ξ > 2 allows the firm to acquire positive net earnings by forfeitingsome of the customers’ greatest possible bill savings.

4. ENERGY EFFICIENCY PROGRAMS IN CALIFORNIA

The economic model developed in the past sections can be used to illus-trate the performance of California energy efficiency programs over 2009-2011and beyond, in that the implementation processes for each of those programcycles will be consistent with the timing of the model: the CPUC adopts en-ergy savings targets and the share-savings incentive mechanism before util-ities propose their energy efficiency program plans and associated fundinglevels, get the proposals authorized by the CPUC, and begin to manage theprograms. Recognizing that each utility’s production of energy savings willbe optimal under the incentive mechanism, the CPUC authorizes the utility’sfunding proposal as long as it is no greater than the level the CPUC esti-mates for the optimizing utility to meet its energy savings target. Then theCPUC-set energy savings targets are achieved by the equilibrium of programimplementation: once the CPUC sets a jointly sufficing incentive rate, eachutility proposes the CPUC-projected level of program funding, gets the pro-posal authorized, and manages the programs with its optimal energy savingsproductivity. If the jointly sufficing rate is identical to the socially efficientincentive rate, the CPUC accomplishes the maximization of net social benefitas well.

The model-based illustration of program performance should be treatedwith caution because the economic model is built on five simplifying assump-tions, which may be too restrictive to reflect practical features of Californiaenergy efficiency programs.

First, the economic model does not consider the risk that, given ex-anteappropriate funding, utilities will fail to achieve preset energy savings tar-gets. The utilities’ risk of substandard performance may be attributable to

27

uncertainties associated with forecasting the penetration of individual mea-sures into the market and their associated energy savings impacts. However,this risk may not be significant in the California case not only because theutilities can manage the risk by diversifying their program portfolios, but alsobecause the CPUC allows the utilities to shift resources within programs aswell as across programs as they become more informed about which measuresare working and which are not (CPUC, 2005c).

Second, utilities’ supply of energy savings and their associated opportu-nity costs may not exhibit the nice properties assumed by the model. Forinstance, the linearity assumption made for program funding and resultantenergy savings does not take into account the fact that, particularly with theconsiderable amount of program funding available in the California case, theprogram portfolios can comprise a broad range of energy efficiency measuresof non-uniform cost-to-savings ratios.17 The simple quadratic opportunitycost assumption can constitute another source of inaccuracy by misrepre-senting the extent to which each utility’s energy efficiency activities affect itsopportunity costs.

Third, while our model defines an earnings basis for the incentive mecha-nism as the net program benefit based only on program administration costs(PACs: customer incentives and rebates, installation costs, operating costs,and marketing/outreach costs), the CPUC-adopted earnings basis metricreflects an additional two-thirds of participation costs paid by utility cus-tomers. Therefore, ceteris paribus, both the minimum sufficing rate and thesocially efficient incentive rate for the real incentive mechanism adopted bythe CPUC are likely to be higher than those suggested by the model.

Fourth, the economic model does not account for the CPUC’s require-ment that the utilities put a minimum of 20% of their portfolio plans to bidby third-party implementers in order to achieve improved portfolio perfor-mance. If the bid solicitation process is competitive and designed primarilyfor portfolios that specialized third-party implementers can accomplish ina more cost-effective manner, utility-based energy efficiency programs withsuch third-party implementers can exhibit higher energy savings productivitythan those without. This possibility may provide justification for adopting

17Compared to the linearity assumption, an increasing marginal cost function leads to agreater program funding but the same opportunity costs, requiring both a higher minimumsufficing rate and a higher socially efficient rate. This does not change the main insightsderived from the economic model.

28

a lower incentive rate for the shared-savings incentive mechanism than theone suggested by the economic model.

Last, the model cannot accommodate the feature that the CPUC evalu-ates the utilities’ performance as an average of the percentage achievementsof their cumulative targets for electricity savings [GWh annual savings],peak savings [MW savings], and end-use natural gas savings [Mtherm an-nual savings]–all of which have different benefit impacts.18 This is becausefunding for the energy efficiency programs cannot be separated by the threetypes of energy savings. Subsequent numerical analysis thus assumes thatelectricity savings performance will vary proportionately with peak savingsand natural gas savings, so that a utility’s electricity savings performance isrepresentative of the utility’s overall program performance.

4.1. Parameters for the Economic Model

Given these caveats, we calculate two important model parameters, whichwill be used to project the performance of the 2009-2011 energy efficiencyprograms. One is the marginal energy savings benefit [$/MWh annual sav-ings], which refers to the change in lifetime avoided costs to a utility firmarising when energy efficiency measures produce one additional MWh an-nual savings. The second model parameter is the utility’s program manage-ment efficiency [MWh annual savings squared/$cubed], which is a measurerepresenting how much less the utility would lose due to energy efficiencyactivities. The two model parameters are calculated by using informationabout the 2006-2008 programs, whose details are stipulated in the CPUCproceedings for energy savings targets and funding authorization.

According to the CPUC proceedings, (i) the three-year cumulative elec-tricity savings targets for the 2006-2008 programs are 2,826GWh for PG&E,3,135GWh for SCE, and 850GWh for SDG&E,19 (ii) the authorized pro-gram administration costs (PACs) over the program cycle are $959 millionfor PG&E, $661 million for SCE, and $266 million for SDG&E,20 and (iii)

18Under the CPUC-adopted incentive mechanism, the utilities’ energy savings perfor-mance of a three-year program cycle is assessed against their cumulative annual energysavings targets for that program cycle, so within the cycle, it does not make any differencewhen any particular measure is installed. However, total avoided costs (or, gross energysavings benefits), which are used to calculate an earnings basis, are based on the lifetimeenergy savings of any particular measure.

19See CPUC (2004).20See CPUC (2005c).

29

Table 1: Average impacts of 2006-2008 energy efficiencymeasures (per MWh annual savings) with the preciseachievement of the targets

PG&E SCE SDG&EEnergy Savings Benefits $760 $750 $680Program Administration Costs $340 $210 $310Net Program Benefitsa $420 $540 $370Shareholder Earningsb $50 $65 $44

aParticipation costs not consideredbWith the CPUC-adopted incentive rate of 12%

the gross benefit estimates or, equivalently, lifetime avoided cost estimatesfor the program cycle are $2,148 million for PG&E, $2,366 million for SCE,and $579 million for SDG&E.21

Dividing the gross benefit estimates by the electricity savings targets givesutility-specific estimates of marginal energy savings benefit regarding the2006-2008 programs: 760 [$/MWh annual savings] for PG&E’s portfolios,750 for SCE’s, and 680 for SDG&E’s. That is, an energy efficiency measureproducing one MWh annual savings for each of its lifetime years is esti-mated to be worth, on average, $760 for PG&E, $750 for SCE, and $680 forSDG&E. The reduction in electricity sales—and thus in electricity generatedand distributed—has similar values for PG&E and SCE, but slightly smallervalues for SDG&E.

Furthermore, the information from the CPUC proceedings suggests thatthe CPUC believed that PG&E would require, on average, $340 PACs (Pro-gram Administration Costs) to induce its customers to install an energy effi-ciency measure producing one MWh annual savings; as noted, such a measurewill create the benefit of $760 for PG&E customers over its lifetime.22 The

21The CPUC projected PAC benefit-cost-ratios (i.e., the ratios of lifetime avoided costsproducible by installed measures to the total PAC expenditure) of 2.24 for PG&E’s port-folios, 3.58 for SCE’s, and 2.18 for SDG&E’s (CPUC, 2005c). These figures, along withthe utility-specific PACs, give the gross benefit estimates.

22PG&E’s average PACs, 340 [$/MWh annual savings], are equivalent to paying annually$42 to avoid one MWh of electricity over average energy-efficiency measure life of 12 years(with the discount rate of 8%), which is below the current price of electricity.

30

counterpart numbers are $750 of benefits and $210 of PACs for SCE; $680 ofbenefits and $310 of PACs for SDG&E. Table 1 summarizes the impacts ofa 2006-2008 average energy efficiency measure producing one MWh annualsavings over its lifetime and the associated shareholder earnings under theCPUC-adopted incentive rate of 12%. As shown, SCE’s shareholders arelikely to gain the greatest earnings per one MWh annual savings largely dueto its lowest average PACs.

Consistent with the structure of the economic model, we can interpret theinverse of the individual utilities’ average PACs as their energy savings pro-ductivity over the 2006-2008 programs. The energy savings productivity is2,940 [MWh annual savings/$million] for PG&E, 4,740 for SCE, and 3,190 forSDG&E. Seen from the fact that SCE was authorized to use relatively limitedprogram funding ($661 million) for the accomplishment of its most ambitioustarget (3,135GWh), the CPUC could have believed that SCE would exhibitthe greatest energy savings productivity in program management.

While the implementation process of the 2006-2008 programs has notbeen consistent with the timing of the economic model developed in thisresearch, the utilities’ funding proposals and the above estimated levels ofenergy savings productivity could have reflected the utilities’ optimizing be-havior as suggested by the model, as long as the utilities and the CPUChad held the same belief of what an incentive rate would be like.23 We hereconjecture that the utilities and the CPUC alike believed that an incentiverate for the incentive mechanism would range from 10% to 25%. This isa reasonable conjecture, in that shared-savings incentive payments to sevenelectric utilities before the market restructuring accounted for from 8% to27% of the net benefits derivable from demand-side management programs(Eto et al., 1998).

Assuming that the above estimated levels of energy savings productivityare efficient for the individual utilities, we now develop ballpark estimatesfor their program management efficiency. With the presumed incentive rateof 10-25% and the individual utilities’ customer base,24 the structure of the

23The CPUC started, in April 2006, the rulemaking process of designing the incentivemechanism for the 2006-2008 and beyond. However, the funding authorization process forthe 2006-2008 programs was finalized in September 2005.

24It is assumed that the number of each utility’s customers is representative ofthe size of its customer base: 5.00 million for PG&E, 4.67 million for SCE, and1.32 million for SDG&E (Energy Sales and Prices in 2005, the Energy Commission,

31

model returns the program management efficiency of 0.08-0.20 for PG&E,0.18-0.45 for SCE, and 0.08-0.21 for SDG&E. SCE appears to be most effi-cient in program management, whereas the efficiency of PG&E and SDG&Eis about half that of SCE. These estimates of program management effi-ciency predict that, over the 2006-2008 programs, the utiltiies’ opportunitycosts with the precise achievement of the targets would range from $255 to$510 million in aggregate: $106-264 million for PG&E, $117-294 million forSCE, and $32-80 million for SDG&E. It is useful to have these estimatesbecause they establish the minimum earnings requirement that will inducethe utilities to commit their organizational resources to the achievement ofthe targets.

Part of the utilities’ opportunity costs associated with energy efficiencyactivities may be verified if one calculates the levels of reduction in share-holder earnings from supply-side projects that are displaced by the 100%achievement of the energy savings targets.25 Such foregone shareholder earn-ings are referred to as “supply-side comparable earnings.” The interpretationof supply-side comparable earnings as part of the utilities’ opportunity costscan be valid, in that the greater the energy savings that the utilities produce,the lower the returns that the utility shareholders will be authorized to earnfrom their supply-side investment. When the utilities’ hidden costs of pay-ing managerial attention to energy efficiency are also taken into account, theopportunity costs will surpass the supply-side comparable earnings.

It should be noted, however, that an economic interpretation of supply-side comparable earnings should account for the possibility that the share-holders can put their investment money elsewhere, earning a return similarto the one they would have earned on the displaced supply-side investments.Then the supply-side comparable earnings that account for such alternativeinvestment opportunities will be by far less than the returns that the utilitieswould be authorized to collect in rates if it were not for energy efficiency pro-

http://www.energy.ca.gov/electricity/)25Even under the provision that requires the decoupling of sales from revenue require-

ment, called “revenue decoupling,” utilities’ energy efficiency programs will incur a re-duction in shareholder earnings. This is because shareholders’ earning opportunities arerestricted to a return on supply-side investments that can be added to the rate base. De-coupling only eliminates a short-term financial penalty for energy efficiency activities anddoes not make utility managers prefer energy efficiency to supply-side investment.

32

Table 2: Model-based opportunity costs and projected supply-side comparable earnings for the 2006-2008 programs

PG&E SCE SDG&E TotalModel-based program 0.20-0.08 0.45-0.17 0.21-0.08 -management efficiencyModel-based 106-264 117-294 32-80 265-638opportunity costsa

TURN-estimated 0-36 0-41 0-10 0-87comparable earningsb

CPUC-estimated 177-275 201-312 50-77 428-664comparable earningsb

Utility-estimated 272 312 62 646comparable earningsb

aMillion dollarsbMillion dollars, Source: CPUC (2007)

grams.26 Nevertheless, the adjusted supply-side comparable earnings mightnot be zero. This is because the CPUC-adopted return on equity (ROE) ismeant to be reasonable and adequate for utilities to maintain and supporttheir financial credit ratings, so that the utilities can attract investors andthus fulfill their public service obligations.27

Table 2 shows (i) the model-based program management efficiency andthe associated opportunity costs estimates and (ii) the projections for supply-side comparable earnings regarding the 2006-2008 programs that were sub-mitted by the Utility Reform Network (TURN), the CPUC, and the threeutilities. Because the shareholders’ alternative investment opportunities wereaccounted for only by TURN, not by the utilities and the CPUC, the TURN-projected supply-side comparable earnings ($0-87 million) are by far lowerthan those of the utilities ($646 million) and the CPUC ($428-664 million).The comparison between the TURN-projected supply-side comparable earn-

26In fact, this perspective has been maintained by customer advocacy groups such asTURN and DRA throughout the incentive rulemaking process.

27See CPUC (2005d). In the rulemaking process, a TURN witness also testified thatTURN believes the CPUC establishes a return on equity that is greater than the cost ofequity (CPUC, 2007).

33

ings, which we believe to be most appropriate, and the model-based oppor-tunity costs ($256-638 million) reveals that, under the economic model, asignificant portion of the utilities’ opportunity costs could be attributed tohidden (and unobservable) managerial costs.

For ease of exposition, the subsequent numerical analyses employ themiddle points of the ranges of model-based program management efficiencyestimates: 0.14 for PG&E, 0.31 for SCE, and 0.15 for SDG&E. The nu-merical analyses also presume that the marginal energy savings benefit of750 [$/MWh annual savings] is representative of the 2009-2011 programs,regardless of the utility service territories.28 With uncertainties associatedwith the market for energy efficiency, however, these ballpark estimates maynot represent the utilities’ real levels of program management efficiency andmarginal energy savings benefit over the 2009-2011 programs. Nevertheless,the estimates are a reasonable case from which to begin sensitivity analyses.

4.2. Base Projections for the 2009-2011 programs

Given the ballpark estimates for the model parameters thus far, the eco-nomic model can be used to provide qualitative insights into the implica-tions of the CPUC-adopted incentive mechanism for the implementation ofthe 2009-2011 energy efficiency programs, whose energy savings targets havealready been adopted, although their funding authorization is as yet to begiven. The CPUC-adopted cumulative annual electricity savings targets forthe 2009-2011 programs are 3,169 GWh for PG&E, 3,528GWh for SCE, and818GWh for SDG&E.

Figure 4 depicts how the utilities’ net earnings U , their customers’ bill sav-ings V , and the associated net social benefit W with regard to the 2009-2011programs will vary with adopted incentive rate r for the incentive mechanism.Given the CPUC-adopted incentive rate, 12%, which accrues once the utili-ties precisely meet the targets, PG&E reaps the net earnings of $10 million(with shareholder earnings of $150 million) from its justifiable funding of $1.1billion, resulting in bill savings of $1.1 billion to its customers; SCE gains thenet earnings of $70 million (with shareholder earnings of $230 million) andits customers’ bill savings of $1.7 billion by utilizing the justifiable funding

28With the marginal energy savings benefit, 750 [$/MWh annual savings], customersreceive the annual bill savings of $92 from energy efficiency measure(s) that avoids annuallyone MWh of electricity over average measure lifetime of 12 years (with the discount rateof 8%).

34

Figure 4: Performance projections for PG&E’s 2009-2011 programs (top),SCE’s (middle), and SDG&E’s (bottom)

35

of $0.8 billion; and SDG&E gains almost nothing (with shareholder earn-ings of $40 million) and its customers’ bill savings of $0.3 billion by utilizingthe justifiable funding of $0.3 billion. In sum, all of the modeling assump-tions held, the CPUC-adopted incentive rate is projected to generate thetotal net social benefit of $3.1 billion out of the total ratepayer investment of$2.2 billion, creating $80 million net earnings (with $410 million shareholderearnings) and $3.0 billion ratepayer savings from the three utilities’ programscombined. Note that SCE produces the greatest net social benefit. This isbecause SCE’s highest program management efficiency, combined with itsrelatively large number of customers under roughly similar per-customer tar-gets across the utilities, encourages SCE to produce both the greatest netearnings and the greatest bill savings.

4.3. Implications of the Economic Model

4.3.1. Alignment of Shareholder and Ratepayer Interests

The alignment of shareholder and ratepayer interests has been the centralissue of California’s incentive rulemaking. In the rulemaking, the CPUCattempted to establish a so-called “win-win alignment of interests,” whichwill provide shareholders with meaningful earnings that compensate for theirforegone supply-side business, while ensuring that ratepayers will receive byfar the most of the programs’ net benefit (CPUC, 2007). Consequently,throughout the rulemaking process, the utilities and their customer coalitionsargued with each other about the appropriate determination of the utilities’supply-side comparable earnings. By reconciling their arguments, the CPUCeventually established the supply-side comparable earnings, without takingthe shareholders’ alternative investment opportunities into account. Thenthe supply-side comparable earnings were directly divided by a projectedearnings basis (i.e., a projected net program benefit) to elicit the adoptedincentive rate of 12%.

While it turned out that the CPUC-adopted incentive rate (12%) is aboutthe same as the greatest (11%) of the model-based minimum sufficing raterequirements for the three utilities (see Figure 4), the adopted incentive ratemay not suffice to return the maximum net social benefit, nor to return themaximum bill savings to their customers.29 Regarding the PG&E programs,

29The coincidence seems not to arise because the CPUC reasonably calculated the util-ities’ minimum earnings requirement (or, equivalently, their opportunity costs associated

36

for instance, the CPUC’s decision would induce the utility to create $1.1billion net social benefit, while incurring $200 million loss in the net socialbenefit relative to the social optimality scenario and $90 million loss in billsavings relative to the customer optimality scenario.

The CPUC’s incentive-rate decision fails to account for the possibilitythat a higher-than-minimum incentive rate can prompt the utilities to creategreater bill savings for the customers as well as a greater net social bene-fit. The economic model suggests that such a seemingly aggressive incen-tive rate can be warranted if program parameters—per-customer energy sav-ings targets, marginal energy savings benefits, and program managementefficiency—offer a sizeable flexibility in the incentive design, which is likelyto be the case in the 2009-2011 energy efficiency programs. The higher anyof these program parameters, the greater the design flexibility and the morea higher-than-minimum incentive rate is likely to be justified. With the ball-park estimates, PG&E, SCE, and SDG&E will be under the design flexibilityof 9, 25, and 10, respectively—all of which are sufficiently greater than thethreshold value of 3. With this range of design flexibility, a higher-than-minimum incentive rate, which provides strictly positive net earnings to theutilities, are warranted not only for a greater net social benefit but also forgreater ratepayer bill savings.

Sensitivity analyses confirm that a higher-than-adopted incentive rate canbe established to produce not only a greater net social benefit but also greaterbill savings to the customers. To provide more robust insights from the analy-ses, we examine PG&E’s 2009-2011 programs, whose parameters collectivelyprovide the least flexibility in the incentive design among the three utilities’programs. Figure 5 illustrates the impacts of the marginal energy savingsbenefit (Panel A) and the program management efficiency (Panel B) on thesocially efficient incentive rate rS∗, the customers’ preferred incentive raterC∗, and the minimum sufficing rate rmin for PG&E’s 2009-2011 programs.The results indicate that, given any not unreasonable misrepresentation ofthe programs, a higher-than-adopted incentive rate is warranted not only forthe maximum net social benefit but also for the maximum bill savings tothe customers; the lower the marginal energy savings benefit or the program

with energy efficiency). Rather, it may be an artifact of both substantially high supply-sidecomparable earnings, which does not account for the shareholders’ alternative investmentopportunities, and missing hidden managerial costs borne by the utilities.

37

Figure 5: Impacts of the marginal energy savings benefit(Panel A) and theprogram management efficiency (Panel B) on the socially efficient rate rS∗,the customers preferred raterC∗, and the minimum sufficing rate rmin forPG&Es 2009-2011energy efficiency programs

management efficiency, the greater the required upward adjustment in the in-centive rate. Therefore, as the marginal energy savings benefit declines overtime, which is likely, a series of similar-sized energy efficiency programs basedon the current incentive rate (12%) can suffer increasingly greater efficiencylosses.

4.3.2. Customization of the Incentive Mechanism

Aside from the proposition that a more aggressive incentive rate can beinstigated for all utilities’ programs, there is room for even greater improve-ment. The CPUC may consider deploying customized incentive mechanismsfor the individual utilities as long as the associated improvement in social ef-ficiency can more than compensate for the political and administrative costsof undertaking utility-specific incentive rulemaking processes.30 According to

30With asymmetric information, however, the customization of the incentive mechanismmay lead to more intense gaming by the utilities during the rulemaking processes, poten-tially resulting in lower social efficiency than the universal incentive mechanism wouldhave achieved.

38

the model, the customized deployment of socially efficient incentive rates forPG&E’s programs (30%), SCE’s (22%), and SDG&E’s (29%) would enhancesocial efficiency by about $320 million compared to the extensive deploymentof the CPUC-set incentive rate (12%), and it would also improve social effi-ciency by about $30 million even when compared to the extensive deploymentof the incentive rate of 22%, which is efficient for the SCE programs.

Furthermore, while the CPUC made it clear that the adopted mechanismwill be in effect also for subsequent program cycles until further notice,31

this study suggests that social efficiency can be improved by updating theincentive mechanism on a regular basis. This is so because all programparameters—per-customer energy savings targets, marginal energy savingsbenefits, and utilities’ program management efficiency—can change, to someextent, over a series of energy efficiency program cycles. Any such changeswill affect the stakeholders’ earnings/savings opportunities and the associatedlevels of the minimum sufficing rate rmin and the socially efficient incentiverate rS∗: the lower the level of any of those three program parameters, thehigher the levels of rmin and rS∗. In particular, a decline in marginal en-ergy savings benefits or in program management efficiency over subsequentprogram cycles without updates of the CPUC-set incentive rate (12%) be-ing permitted might completely eliminate the utilities’ incentive of pursuingenergy efficiency activities, which would impede an efficient delivery of theprograms. Figure 5 shows that, with a sizeable decline in the marginal energysavings benefit or the program management efficiency, the current incentiverate may fall below the minimum sufficing rate rmin.

5. CONCLUSIONS

In this paper, we have examined the implementation of utility-deliveredenergy efficiency programs that employ a shared-savings incentive mecha-nism. We developed an economic model that accounts for the conflictingincentives between a regulator and an investor-owned utility in the authoriza-tion of funding for the programs. Consistent with the structure of Californiaenergy efficiency programs, the model assumes that the regulator adopts anenergy savings target and a shared-savings incentive mechanism before theutility firm proposes program funding, gets it authorized, and begins to man-age it. In doing so, the regulator aims to accomplish dual goals: to produce

31See CPUC (2007).

39

the maximum net social benefit while ensuring the achievement of the energysavings target. Two important findings emerged from this work.

First, the utility firm will achieve the regulator-set energy savings target,eventually producing non-negative net earnings for itself and non-negativebill savings for its customers, if the adopted incentive rate is jointly sufficingand the implementation of the programs is at equilibrium: with the energysavings target and the jointly sufficing rate in place, the firm proposes thefunding level that the regulator estimates in order for the firm to meet thetarget, gets the proposal authorized, and manages the programs by choosingits optimal level of energy savings productivity. This jointly sufficing incen-tive rate is no less than the minimum sufficing rate. The minimum sufficingrate declines with increases in the value of a parameter, which we call designflexibility ξ, the level of which is collectively determined by, and increaseswith, per-customer energy savings target t/β, marginal energy savings ben-efit η, and the firm’s program management efficiency θ. Thus the minimumsufficing rate is lower for programs with higher t/β, for programs managed bya firm with higher θ, and for programs implemented in utility service areaswith higher η.

Second, a higher-than-minimum incentive rate can be required to producea greater net social benefit, as well as greater bill savings for the customers.While the minimum sufficing rate under the low range of design flexibility(1 ≤ ξ ≤ 2) produces both the maximum net social benefit and the maximumcustomer bill savings, the minimum sufficing rate under the mid-to-high rangeof design flexibility does not. Under the mid range of design flexibility (2 <ξ ≤ 3), a higher-than-minimum incentive rate is required to achieve a greaternet social benefit. Under the high range of design flexibility (ξ > 3), a higher-than-minimum incentive rate needs to be established to produce not only agreater net social benefit but also greater bill savings for the customers.This is because such an incentive rate raises net program benefit by loweringfunding requirements, which more than compensates for both an increasein the firm’s opportunity costs and a decrease in the net benefit share tothe customers. Specifically, if the adopted incentive rate is socially efficient,the energy efficiency programs will accomplish the dual regulatory goals. Thesocially efficient incentive rate depends again on design flexibility: the higherthe level of design flexibility, the lower the socially efficient incentive rate.

A closer examination of California energy efficiency programs in light ofthe above findings suggests that a higher-than-adopted incentive rate is war-ranted to achieve not only a greater net social benefit but also greater bill

40

savings for the customers. The main reason is that current market and regu-latory circumstances offer sizeable design flexibility, which renders untenablethe disputes on the determination of the minimum earnings requirementsoccured during the incentive rulemaking process. Furthermore, the exam-ination suggests that social efficiency would be improved by customizingincentive mechanisms for individual utilities and updating them on a regularbasis. This is because any of the crucial program parameters can vary acrossthe utilities and over a series of energy efficiency program cycles, requiring adifferent level of the socially efficient incentive rate to be established.

The work presented in this paper does not address uncertainties associ-ated with energy savings supply and demand. In reality, utility firms face un-certainties about the penetration of individual energy-efficient products andservices into the market and their associated energy savings impacts. Thesesupply-side uncertainties may emerge from market barriers for efficient energyuse, including agency problems and cognitive limitations of customers. Evenwith energy efficiency programs in place, these market barriers can makereal-world adoption of energy efficiency measures not as flexible or logical aswould be indicated by the model proposed in this work. With regard to thedemand for energy savings, particularly over the long term, utility regulatorsare confronted with uncertainties about future energy demand, future fuelprices, forthcoming carbon emission regulation, and emerging energy tech-nologies of unknown cost and performance. We ignored these uncertaintiesto focus on the basic implications that the shared-savings incentive mecha-nism has for energy efficiency programs. Future research may incorporatesuch uncertainties into a more realistic design and implementation of utility-delivered energy efficiency programs.

References

Bachrach, D., Carter, S., and Jaffe, S.(2004), “Do Portfolio Managers Havean Inherent Conflict of Interest with Energy Efficiency?” The ElectricityJournal, 17(8), 52-62.

Baron, D. P. and Myerson R. B. (1982), “Regulating a Monopolist withUnknown Costs,” Econometrica, 50(4), 911-930.

CPUC. (2004), “Interim Opinion: Energy Savings Goals for Program Year2006 and Beyond,” Decision 04-09-060.

41

CPUC. (2005a), “Interim Opinion on the Administrative Structure for En-ergy Efficiency: Threshold Issues,” Decision 05-01-055.

CPUC. (2005b), “Interim Opinion: Updated Policy Rules for Post-2005 En-ergy Efficiency and Threshold Issues Related to Evaluation, Measurement,and Verification of Energy Efficiency Programs,” Decision 05-04-051.

CPUC. (2005c), “Interim Opinion: Energy Efficiency Portfolio Plans andProgram Funding Levels for 2006-2008 - Phase I Issues,” Decision 05-09-043.

CPUC. (2005d), “Test Year 2006 Return on Equity for the Major EnergyUtilities,” Decision 05-12-043. The California Public Utilities Commission.December.

CPUC. (2006), “Interim Opinion: 2006 Update of Avoided Costs and RelatedIssues Pertaining to Energy Efficiency Resources,” Decision 06-06-063.

CPUC. (2007), “Interim Opinion on Phase I Issues: ShareholderRisk/Reward Incentive Mechanism for Energy Efficiency Programs,” De-cision 07-09-043.

Energy Information Administration, 2006. “Class of Ownership, Number ofConsumers, Revenue, Sales, and Average Retail Price by State and Utility:All Sectors, 2006.”

Eom, J. (2009), “Incentives and Politics of Utility-based Energy EfficiencyPrograms in California,” Unpublished Ph.D. dissertation, ManagementScience and Engineering, Stanford University.

Eto, J., S. Stoft, S. Kito. (1998), “DSM Shareholder Incentives: RecentDesigns and Economic Theory,” Utilities Policy, 7, 47-62.

Holmstrom, B. (1979), “Moral Hazard and Observability,” The Bell Journalof Economics, 10(1), 74-91.

Hurwicz, L. and Shapiro, L. (1978), “Incentive Structures Maximizing Resid-ual Gain under Incomplete Information,” The Bell Journal of Economics,9(1), 180-191.

Joskow, P. L. and Schmalensee R. (1986), “Incentive Regulation for ElectricUtilities,” Yale Journal on Regulation, 4(1), 1-49.

42

Kushler, M., York D., and Witte, P. (2006), “Aligning Utility Interests withEnergy Efficiency Objectives: A Review of Recent Efforts at Decouplingand Performance Incentives,” Report Number U061, American Council forand Energy-Efficiency Economy.

Loeb, M. and Magat, W. A. (1979), “A Decentralized Method for UtilityRegulation,” Journal of Law and Economics, 22(2), 399-404.

National Action Plan for Energy Efficiency. (2007), “Aligning Utility Incen-tives with Investment in Energy Efficiency,” Prepared by Val R. Jensen,ICF International.

Stoft, S., Eto, J., and Kito, S. (1995), “DSM Shareholder Incentives: RecentDesigns and Economic Theory,” LBL-36580, Lawrence Berkeley NationalLaboratory.

Stoft, S. and Gilbert, R. J. (1994), “A Review and Analysis of Electric UtilityConservation Incentives,” Yale Journal on Regulation, 11(1), 1-42.

43