Determinants and Impacts of Demand-side Management Program

Loading...
Portland State University

PDXScholar Dissertations and Theses

Dissertations and Theses

1996

Determinants and Impacts of Demand-side Management Program Investment of Electric Utilities Philipp Degens Portland State University

Let us know how access to this document benefits you. Follow this and additional works at: http://pdxscholar.library.pdx.edu/open_access_etds Recommended Citation Degens, Philipp, "Determinants and Impacts of Demand-side Management Program Investment of Electric Utilities" (1996). Dissertations and Theses. Paper 1173. 10.15760/etd.1172

This Dissertation is brought to you for free and open access. It has been accepted for inclusion in Dissertations and Theses by an authorized administrator of PDXScholar. For more information, please contact [email protected]

DETERMINANTS AND IMP ACTS OF DEMAND-SIDE MANAGEMENT PROGRAM INVESTMENT OF ELECTRIC UTll.JTIES

by

PHILIPP DEGENS

A dissertation submitted in partial fulfillment ofthe requirements for the degree of

DOCTOR OF PlllLOSOPHY in

SYSTEMS SCIENCE: ECONOMICS

Portland State University 1996

DISSERTATION APPROVAL

The abstract and dissertation of Philipp Degens for the Doctor of Philosophy in Systems Science: Economics were presented April30, 1996, and accepted by the dissertation committee and the doctoral program.

COMMITTEE APPROVALS: Kuan-Pin Lin, Chair

Andrew M. Fraser

'----/ James Strathman · Representative of the office of Graduate Studies

DOCTORAL PROGRAM APPROVAL: B[atri;T. Oshika, Director Systems Science Ph.D. Program *****************************************~**************************

ACCEPTED FOR PORTLAND STATE UNIVERSITY BY THE LIBRARY

/996

ABSTRACT

An abstract of the dissertation ofPhilipp Degens for the Doctor ofPhilosophy in

Systems Science: Economics presented April30, 1996.

Title: Detenninants and Impacts ofDemand-Side Management Program Investment ofEiectric Utilities

From the late seventies through the early 1990's electric utilities were facing many different forces that caused them to invest into demand-side management programs (DSM). Roots of the growth ofDSM can be found in the high inflation and energy price shocks of the late seventies and early eighties, spiraling building costs of generation, safety and environmental concerns, increased costs of new capacity with possible exhaustion of scale economies, unexpected high elasticity in the demand for electricity, and public utility commissions that sought alternatives to the resulting high rate increases. This study develops and estimates four equations that look at the more aggregate utility level impacts ofDSM. The goal of two equations is to detennine what factors influence utility investments in DSM and if stock market investment in utilities is affected by DSM. Two additional equations are developed to determine system level impacts of DSM on cost of and quantity demanded of electricity. To estimate these models four

years of annual data were collected for 81 utilities spanning 1990-1993. These utilities have sold over 60% of all the electricity in the US and were responsible for over 80% the national spending in DSM. The DSM investment model indicated that of the major variance in DSM investment is due to the utility's regulatory environment. Both an above average regulatory climate and least-cost planning requirements had major impacts on the level ofDSM investment. The cost of equity capital equation revealed that DSM expenditures had a positive impact on the valuation of utility's stock. Cost and quantity equations were estimated both individually and simultaneously. DSM expenditures seemed to have a negative impact on both average cost and quantity demanded. Although these relationships were statistically significant, the impacts were quite small. To summarize; the regulatory environment seems to have the strongest impact on the level ofDSM investment; DSM spending was associated with an increased stock valuation; as expected DSM investments were found to have a negative relationship with quantity demanded; and finally DSM investment appeared to reduce the average cost.

TABLE OF CONTENTS Page

I. Introduction

1

II. Literature Review

14

11.1.

Demand-Side Management Investment

17

11.2.

Cost ofEquity Capital

25

11.3.

Demand for and Supply ofElectricity

26

III. Methodology

34

Ill.1.

Demand-Side Management Investment

29

Ill.2.

Impact ofDSM: Cost ofEquity Capital

40

111.3.

Impact ofDSM: Demand for Electricity

44

III.4.

Impact ofDSM: Cost ofElectricity

48

111.5.

Scope of Analysis

52

IV. Model Estimation

57

IV. I. Model Summary

57

IV.2.

Demand-Side Management Investment Equation

60

IV.3.

Cost ofEquity Capital Equation

72

IV.4.

Electricity Demand Equation

80

IV.S.

Electricity Average Cost Equation

90

IV.6.

Simultaneous Estimation ofDemand and Cost

100

V.Conclu~on

108

TABLE 01~ CONTENTS (continued)

Page VI. Future Research

115

Appendix 1: Data Sources

117

Appendix 2: Definition of Demand-Side Management •

123

Appendix 3: Hypothesis Tests

131

Bibliography

133

ii

I. Introduction

Demand-side management (DSM) can be con~ide11ed a utility maQagement strategy that allows a utility to more effectively use it~ plant and equipment while allowing more efficient energy use by a customer.

T~s

is accomplished through a set of

programs and policies designed to affect a customer's consumption levels or timing oJf a customer's electricity demand. This contrasts with a ~uppl.y-side strategy, where increased electricity demand will be met by adding generation, transmissic~n, and distribution facilities. In changing the electric power qemand levels, DSM programs and policies are targeted to promote conservation,

re~luce

1

Waste, and increase the use

of more efficient motors, tools, and appliances. These programs affect th€t timing of electricity demand by seeking to smooth fluctuations In demand, so the bt.Jik of electricity provided to consumers can be generated fr~lm nwre efficient b
perfectly defined. However, it can be separated into two major categmie~: load management and conservation (Gellings and Talukdar, 1986). Load management strategies shift the period in which electricity is requir~d. This does not nt;.cessarily reduce the volume of electricity consumed, in part, it 1nay even increase it.. It does

1

result in the more efficient use of plant and equipment, causing a reduction in the total cost of providing the service. Programs targeted only at conserving of existing loads do not move electrical demand from one time to another. Rather they remove the demand entirely 1• Following the oil price shocks ofthe 1970's the conservation of energy resources has become an area of greater interest to the general public. Consequently, energy conservation has come about through government mandates2 and rational economic choices by consumers and producers. In the electric industry, resource conservation also became critically important during this time. Environmental regulation put many electric utilities under tighter constraints when building generation, transmission, and distribution facilities. The threat of further regulation, added costs from proposed "carbon" taxes, and internalization of the costs of other externalities continued to increase the financial risk in the construction of generation, transmission, and distribution facilities. The rampant inflation and changes in safety requirements for nuclear power plants and environmental regulations for coal-fired thermal plants during the seventies and early eighties also caused power plant construction costs to escalate unexpectedly. These rising construction costs combined with concurrent hikes in fuel costs, led to rate increases. Electricity demand proved to be more elastic than expected. Utility forecasts

1

For a more detailed description ofload management and conservation programs, see Appendix 2. 2 The Public Utilities Regulatory Act of 1978 mandate that state regulators adopt marginal cost pricing to encourage economic conservation. 2

of demand for electricity often proved to be overestimated, at times leading to excess power generation corning on line. Faced with the choice of rate increases caused by excess capacity, many public utility commissions (PUCs) denied requests for rate increases in an effort to protect customers from price shocks. However, rates did increase to protect the utilities from financial insolvency. In many cases, PUCs allowed price increases to occur in conjunction with automatic fuel adjustment clauses. Some construction work in progress (CWIP) was also allowed into the rate base, easing the utilities' cash-flow difficulties. The inclusion of CWIP in the rate base went against the previous regulatory practices of allowing "used and useful" items into the rate base. However, some PUCs did not allow unfinished power plants into the rate base, and continued this policy even after the projects were completed, deeming the investments to have been imprudently incurred. Thus, major capital investments, previously considered fairly risk-free, lead to financial instability for many electric utilities. Regardless of the methods used to reduce price shocks, electric utilities' operating environment changed dramatically, putting many companies in financial straits. During this time utilities also began to face increased competition. Price increases made alternative fuels more attractive, causing many customers to switch fuels. In addition, the Public Utilities Regulatory Act of 1978 (PURP A) increased opportunities for non-utility generators to compete with electric utilities. Facilities qualifying under PURPA were allowed access to the existing electric system, and

3

utilities were forced to purchase electricity generated by these non-utility generat(!)rs at the utilities' marginal costs. In cases where marginal costs exceeded average costs, these power purchases would tend to increases prices as utility prices are based o;n average costs. PURP A also created incentives for existing customers to self- or cogenerate electricity. Surplus power could be sold to the utility, while back-up power woult;l be available from the utility. This tended to either erode the customer base or to prqvide added leverage to larger industrial and commercial customers negotiating electri~ity rates. Coupled with consumer advocacy of conservation, these trends increased the electric utilities' interest and activities in what has become known as demand-side management. DSM programs can be utilized for a variety of strategic purposes. Conservation and load management programs can be used to defer investments in new plant. This capital minimization strategy reduces present capital costs and concomitant risks associated with added capacity. Such risks include insufficient demand for the new capacity, cost escalation due to compliance with emerging environmental standards, higher inflation rates that increase capital costs, or prudency reviews that prevent investments from being allowed into the rate base and thus earning a rate of ret~m because a PUC has deemed investments to have been imprudently incurred. DSM programs that aim at deferring capital investments typically target "peak clipping" and

4

load shifting; reducing peak demand or shifting it to off-peak periods, hence reducing the need for added capacity. DSM programs were also developed to increase revenues. This is done by building new, off-peak loads by offering lower rates during those periods. Revenues can be increased through programs that promote strategic load growth. This often takes the form of promoting new electro-technologies, such as electric cars or electric arc fhrnaces in steel mills, or by increasing the overall market penetration of certain appliances and equipment. Some programs use DSM to bolster sales retention and increases in market share. These programs are designed to entice existing customer loads from switching to competing fuels or offer incentives to customers that choose electricity as their fuel for new equipment. DSM programs were believed to be very flexible in their implementation and goals (Hirst, 1989a). Many utilities invested in these programs to counter one or more adverse condition encountered in the new environment. DSM programs were promoted by many PUCs as least-cost alternatives to new construction and therefore a way to keep rates low. PUCs encouraged the implementation ofDSM programs through a variety of incentives, such as higher rates of return or allowing DSM investments into the rate base. In many states PUCs did bring about least-cost planning (Mitchell, 1989, 1992) which forced utilities to at least consider DSM as a potential resource along with conventional generation. Some PUCs actually decoupled electricity sales from the rate of return to prevent discouraging utility investments in DSM programs.

5

Currently milch of the rationale behind these strategies has either diminished in strength, disappeared or 1become secondary to other forces shaping the electric industry. The industry is most influenced by the appearance oflow-cost, flexible I

electricity generation in ~he form of gas-fired combustion turbines, deregulation, competition at all

c~:~stomer

levels (retail wheeling), industry consolidation through

mergers, non-utility pow:er generation, increased system interconnectivity, and increased wholesale interchanges. 3 This has led to an decline in utility interest in pursuing DSM. Part of this cjeclirlle in interest in DSM has also been caused by its success. In many cases, the mo~t cost-effective DSM investments have already been made. Many of these have been realized through state and federal government mandates requiring more efficient appli(l.nces, lighting, HV AC equipment, and motors. For example, product efficiency l(l.bels were required by the Energy Policy and Conservation Act of I

1975. Major home (l.ppliance energy efficiency standards were mandated through the

National Appliance EnefiD' Conservation Act of 1987. Most ofthese standards took effect in 1990, with refrigerator and freezer standards coming into force in 1993. Further legislation,

~stablishing

lighting and plumbing efficiency standards, came into

being with the New Energy Act of 1992. This act furthered the introduction ofvarious

3

For discussions of the issues facing the electric power industry today see Stevenson and Penn (1995), Hempling (1995), Berg and Tschirhart (1995), Hirsch (1991), and Joskow (1989). 6

other efficiency standarcjs, such as increased energy-efficient building codes and more efficient commercial ele~trical ,equipment. Electric utilities promoted many of these moves towards increased standards. EPA's "Green lights"

progra~

could not have happened without electric utilities

developing the initial mqrkets tbr these energy-efficient products. In addition, many utility programs led the way tdwards adoption of these new standards and other market transformations. Such programs geared towards market transformations include the Model Conservation St"ndards for new commercial and residential construction adopted by many states, the "Golden Carrot" energy-efficient refrigerator, and the increased availability of ~om pact fluorescent bulbs. With standards achieving higher efficiency levels the mar~inal cost of a "negawatt" 4 increases, since it is easier and cheaper to develop technology that increases efficiencies at the lower levels (e.g. going from 60 % efficiency to 70% e~fficiency) than to capture the last few remaining percentage points in effi~iency1(e.g. going from a 97% efficiency to 99% efficiency). Another factor npw impacting DSM investment is technological change. DSM has always been touted

fj.S

a flexible resource that can be acquired in piecemeal fashion

without an extensive plarming horizon. With the future electricity demand unknown and base load power plant cqnstrucrtion requiring 10 to 15 years of planning, design, and construction before completion, DSM often proves to be a much more appealing investment. Further, developments in combustion turbine technology (in efficiency,

4

A kWh of energy savin~s. 7

reliability, and size) have allowed thes€!S units to become base-load units (Hirsch, 1991). The combustion turbines can b~ constructed and sited easily and be used to provide incremental 100 MW loads within a year or two. 5 This fact, coupled with fairly stable gas prices, causes DSM to compete with a technology that is also flexible. DSM has been pronounced

de~d

many tim~s. However, tactics such as the

Association for DSM Professionals ch~nging its name to the Association of Energy Service Professionals in 1995, indicat€!S ofhow that the status ofDSM in the electric utility industry is declining. "Regulator driven" DSM is expected to disappear, while "customer driven" DSM is expected t<~ flourish, with utilities offering DSM as a servic;e for their customers. 6 In this way, cust~)mers who benefit from the services pay for them,

,

explicitly, removing issues of cross-subsidization from customers not receiving the services. DSM is presently being repa~kaged as a;value-added service strategy that wlll become part of a utility's competitive strategy (Shick and Hamilton, 1996). It is also expected that more informational interconnectivity will allow many DSM options, suqh as TOU rates, smart homes, utility control of appliances etc., to become increasingly cost-effective to implement. DSM investment has risen slowly since the early eighties, rising from more thim

1

7

$860 million in 1982 to $2.9 billion in 1993. 8 More recently these expenditures, in comparison to new capital investments, have gone from 3% of investment in new plapt

1

5

Electric Utility Week, February 26, 1996, pp. 1!6. Quad Report, vol. 3 no. 7, August, 1995, pp. l1and Shick and Hamilton (1996). 7 Cogan and Williams (1983). 8 EIA Form 861 (1993) ..

6

8

and equipment in 1990 to almost 10% of new investment in electric plant in 1993. As a percentage of total generation, expected savings have risen during this four-year period from 0.6% to 1.5%. Projected demand savings have increased from approximately 17,000 MW in 1990 to 23,000 MW in 1993. Savings produced by these demand programs can be expected to reduce consumption from 2.5% (Prete, 1992) to 3% by the year 2000 and to 6% by the 2010 (Faruqui, Seiden, and Braithwait, 1990). By the turn of the century DSM programs are projected to reduce peak loads representing 73% of new capacity (Prete, 1992). Even though factors thought to spur the growth of utility DSM investment have waned, the impacts of past and current DSM programs are still present. DSM continues to be a flexible resource in which investments fluctuate depending on electricity's costs of supply and price. Many ofthe factors which led to DSM investment may arise again. Inflation may reappear increasing capital costs, and energy costs will not necessarily remain stable over time. Industry behavior may cause regulatory pressures to increase, especially in the face of rising rates. A major factor driving industry restructuring are anticipated lower prices. With regulators seeking to provide rate payers with lower prices, deregulation is considered positive. However, if prices increase, it is possible regulation will reappear. DSM would likely follow programs, especially regulatordriven DSM.

9

Much re&earch has been conducted that models the DSM impacts at the customer and end use: level. 9 In many traditional electric industry models, however little has been done in incorporating the DSM investment impacts into the models' frameworks. At 11 more aggregate level, few models have been developed. None have been estimated tp det1ermine DSM impacts. A few studies (Keelin and Gellings, 1986)(Faruqui, $eideh, and Braithwaite 1990)(Hirst, 1991) have been conducted on the national imp11cts. ':These have primarily used simulation models. None have directly estimated DSM impacts at the utility system level or the national levels. Integrating DSM investments into models at the utility system level proves useful because the decisions by a single economic agent can be more thoroughly analyzed. If a s~fficient number of utilities are analyzed, better forecasts of national level impacts can be made. As part of this study, four models were estimated. The first models DSM investments for factors influencing utility DSM investment levels. Aggregate DSM imp,acts at utility system levels were estimated using a pair of simultaneous equations that model cost of and demanded for electricity. A final model was estimated tp det1ermine how the market values DSM investments. This was accomplished by estimating a simple model for factors influencing equity capital costs, with DSM bein~ one of the factors.

9

DSM program imp:acts have been studied in numerous evaluations carried out by individual utilities. The results of many of these evaluations have been presented in the proceedings of the ACEEE Summer Session, the Energy Program Evaluation Conference, and the1National DSM Conference. Numerous other studies have been sponsored by the El<~ctric Power Research Institute. 10

These models were estimated using four years of data from 81 utility holding companies. In 1990, these utilities sold 60.7% ofthe electricity in the United States, and 78.6% of the electricity sold by investor-owned utilities. That same year they represented 83.5% of total nationwide DSM expenditures, and 92% ofDSM expenditures of all investor-owned utilities. Analysis ofDSM investment by the investor-owned utilities indicated that program expenditures were heavily influenced by the regulatory environment. This suggested that regulatory incentives play a large role in the decision to invest in DSM. Least-cost planning requirements also were positively associated with DSM investments. The estimated model revealed a positive, but not strong, correlation between DSM and capacity constraints. DSM impacts on costs and on quantities demanded revealed a negative relationship. In both cases, the relationship was statistically significant. However, DSM's actual impact on price and quantity was insignificant. When solely examining energy savings, DSM investments did not appear cost-effective unless the persistence of that impact was assumed to last 20 or more years. As DSM investments defer capital investments, the cost-effectiveness ofDSM can not be measured solely by the energy savings. Though DSM is only associated with minute decreases in cost this must be seen in a positive light. Despite utility spending on DSM and the accompanying revenue losses, the model does not indicate that DSM investments raise costs of electricity, that in tum will lead to rate increases.

11

A variety of factors may have led to these low impacts. Government energyefficiency legislation and program spillover 10 may have increased the efticiency of customers, even though the utilities serving them had 1\0t invested heavily in DSM. Further, many DSM programs are load management pmgrams that shift consumption, and possibly even causing an overall increase in energy used~ 11 Other programs are geared towards strategic load growth or load retention. Even if a new or retained load is energy-efficient, the net result remains an increase ill elect:ricity consumption. Impacts ofDSM investments on capital costs were r)ot only positive and statistically significant, but had a fairly large impact Ol\ the v:alue ofthe utility's stock. This large impact on market valuation may be due to the DSM variable acting as a prox-y for other variables. Possibly, utilities investing ip DS~,v1 are considered more innovative and customer services oriented. As the market puts a premium on DSM investment during regulation, this favorable valuation ofDSM may remain as the industry restructures. This study indicates that the substantial financial in\(estments in DSM make it possible to measure the DSM impacts at system level~. The; study also indicates that

10

Spillover can appear in a variety of guises. For example ¢ustomers who obtain program-sponsored appliances move to another utility's service territory. One utility's advertisements affect an adjacent utility's customers' purcl1asing habits. Energyefficiency practices cultivated by a commercial or ind.ustria;l customer are transferred to another site. 11 Energy storage equipment uses more kWh at lower off-peak rates to store heat or cold. 1

12

even though much ofDSM investment appears to be influenced heavily by regulation, DSM services may continue to flourish following industry deregulation.

l3

ll. Literature Review

For ~orne time, a great many detailed studies have been available on the impacts and cost-ef:lbctiveness of individual DSM programs. A fairly large industry has developed

t<~

plan and implement evaluations of specific programs. The detail and

breadth of many of these evaluations arose largely due to PUC's basing shareholder incentives op the1 measured impacts of these programs (Fels and Keating, 1993). With industry restructlllring on the horizon, this area of research has declined. An obvious sign ofthis ~hift was the cessation of the industry newsletter, Evaluation Exchange, in 1995.

DSM evaluation was primarily done on the program level, while less research was concerned with the system-level DSM impacts. Some studies have been carried out on individual nation-wide programs, such as the Residential Conservation Service (Clinton et ~I. 1986) and the Low-Income Weatherization Service (Brown et at, 1993). Early studie~; on overall DSM impacts collected DSM expenditure and savings data using surveys of!Utilities. In 1982, estimated DSM spending in 1982 was $860 million for 120 surveyed utilities (Cogan and Williams, 1983). As these data represented DSM spending for the majority of the largest private and public power producers in the country, the figur:es in the above sum should be fairly close to the total. Expenditures were expect~d to' yield construction cost savings of two dollars for every dollar spent

14

on I;:> SM. An updated study for 1985 and found little, if any increase in DSM spending (Cagan and Williams, 1985). The Electric Power Research Institute (EPRI) has funded numerous studies on all facets ofDSM. EPRI sponsored the first study of overall DSM impacts on the ele~tricity

industry in 1986 (Keelin and Gellings, 1986), which was followed by a more

COQ1prehensive update in 1990 (Faruqui, Seiden and Braithwait, 1990). Both studies for~cast

national. DSM impacts. The earlier study primarily used engineering estimates

to project national impacts. The second study was more comprehensive, and incprporated real world results ofDSM programs and end use simulation models to project DSM impacts. Another EPRI study (Blevins and Miller, 1993) only provided estimated savings and expenditures for the utilities surveyed. In many cases, the utilities did not report either savings or expenditures. On average, the surveyed utilities reported annual1funding levels of$730 million from 1991 to 1993. Trends within DSM sp€1nding indicate a shifting of funds from conservation to load management programs (Prete et al, 1992). A more comprehensive estimate of total DSM impacts has been available from ther Energy Information Agency (EIA), which has collected total utility DSM expenditure andl savings data from all public and private electric utilities since 1990. Aocording to this data, total industry DSM expenditures have grown from around $1.2 billion in 1990 to $2.9 billion in 1993 (Table 1). This is approximately 3% of new plant investment in 1990, and has grown to about 10% of new plant investment in 1993. As a

15

percentage of total expenditures, DSM spending doubled during this time period, growing from just under 0.9% in 1990 to 1.9% in 1993. Expected savings during thi~ time period more than doubled, going from 17 million MWh in 1990 to over 41

milli~m

MWh in 1993. Demand savings increased from around 17,000 :MW in 1990 to 23,000

:MW in 1993. Reductions in peak load were larger than those of savings because many of the programs were geared towards peak clipping (Prete et al, 1992).

Table D.l EIA Reported and Projected DSM Expenditures and Savings for all Utilities

1990

1991

1992

1993

2000

DSM (millions)

$1,201

$1,750

$2,396

$2,888

$3,394

MWh Savings

17,060

23,432

30,029

41,007

78,444

0.6%

0.8%

1.1%

1.4%

2.5%

379

458

1,008

1,143

NA

(thousands) Savings as% of total Generated Number ofUtilities

DSM investments ar~ expected to reduce total electricity demand, over that required in their absence. The reduction in total national electricity consumption wa~ projected to be 1.1% in 1990, 3% in 2000, and 6% in 2010 (Faruqui, Seiden, and Braithwait 1990). Another study (Hirst, 1991) projected even greater impacts ifDSM programs were implemented more aggressively. EIA future forecasts have

estimate~

16

that DSM programs will result in peak load reductions representing 73% of new capacity.

ll.l. Demand-Side Management Investment

DSM expenditures remain quite variable. Some utilities do not have DSM programs while others spend as much as 8% of their total annual expenditures on DSM. No single simple reason explains why utilities invest in DSM. For the electric utility industry, DSM program objectives "are shaped by a utility's reserve margin, fuel mix, expected demand growth, regulatory climate and other exogenous factors" (Cogan and Williams, 1987). The reasons for DSM program investments can be separated into two main groupings: programs embarked upon to achieve profits and those that utilities undertake to fulfill governmental legislation or PUC regulations. In the case of government legislation the reason for DSM investment is quite straightforward. Utilities have had to implement DSM programs that were mandated by the federal and state governments. One of the largest nationwide DSM programs, the "Residential Conservation Service (RCS)," was brought about by the National Energy Conservation and Policy Act of 1978. It lasted until 1989. This legislation required utilities to offer home energy audits at little cost to residential customers. During a single 3 year period of the program's existence, over 3 million audits were performed nationwide at the cost

17

of$360 million (Clint
L~ast-cost

planning incorporates both supply-side and demand-side

alternatives to meet projec1ted demand. Thus, utilities affected by these regulation, must consider DSM programs in their resource acquisition planning process. If deemed costeffective under the le
1. Addition of rever1ues.1 2. Prevention ofrevenue: losses. 3. Risk reductions. 4. Public relations. 5. PUC profit incen,tives.

Revenues ca11 be :increased through a variety of programs including innovative rates that promote new,
18

monopoly behavior, and are explained in the rubric of second- and third-degree price discrimination (Mansfield, 1970), (Scherer, 1970). The success ofthese programs on increasing revenues depends on the customer classes involved and their respective elasticity of demand. To succeed in increasing revenues, the new rates have to be mandatory or they must attract sales that would not have occurred otherwise. Price rebates during off-peak periods or times of excess reserve capacity have to be targeted toward new sales if the utility seeks to realize an increase in revenues. Programs that sponsor strategic load growth can also be seen as profitable. Many of these programs prevent fuel switching, thus retaining market shares. In addition, as yet unserved markets are developed through new electro-technologies such as electric cars, or electric arc furnaces in steel production. Strategic load growth programs can increase the market penetration of certain electric equipment, such as programs promoting the installation of air conditioning equipment. Energy-efficient space heating equipment programs for utilities with a summer peak load or increasing the penetration of space cooling equipment for winter peaking utilities can take advantage of excess system capacity and be profitable. The long-run aspects of these programs are also a consideration, as installation ofthe equipment will provide the utility a steady revenue stream over the lifetime of the equipment. Sales losses can occur either through the customer bypassing the utility's electric system or through entirely leaving the service area. Bypass can take the fonn of self-generation or cogeneration, fuel switching, or non-utility installation of energy

19

conservation measures. This is predominantly a problem for a utility when, large electricity purchasers threaten to do any of the above, because a large revenue Joss will increase costs for customers remaining in the system (Hyman, 1986). Under th~ threat of system bypass, DSM programs or lower rates may be offered to customers t.o decrease their electric power costs and retain them in the system. As long as the cost of DSM programs or lowered rates remain above service costs, the utility and other rate payers benefit because revenue losses and rate increases will not be as great as they would have been had the large customer been allowed to bypass the system. The higher elasticity of demand of many larger customers causing lower rates or conserva~ion programs, results in the DSM programs being a form of third-degree price discrimination. Whereas utility bypass was once seen as a threat primarily coming from larger customers (Plummer, 1990), it may soon be possible to associate this risk with all customer classes. Retail wheeling is a very possible reality (Stevenson and Penn, 1995). This will allow competition at all customer levels, while electricity services are offered from a variety of suppliers. Such competition will place high-cost utilities under pressure. To survive they may have to incur severe financial losses by writing off highcost generation. Bundling DSM with electric services is seen as a method of refllaining competitive at higher prices as DSM services adds value that justifies higher prjces. The risk reduction for capital investments is another reason for investments in DSM. Although this a profit motive, this occurs because of failures in the coun~ry's the

20

regulatory system. During the 1980's, the growth of demand has often failed to keep in step with industry projections. Often electric plant was built for demand that did not materialize. This caused reserve capacity margins to grow in many utilities, resulting in higher fixed costs for thelelectricity sold. To meet these higher fixed cost, utilities filed for rate increases. Often these rate increases were politically unpalatable, due to their size and denied by PUCs.' At the same time, many PUCs did not allow the added capacity into the rate base through a new process of prudency hearings. This caused many utilities: to change tlheir capital investment procedures (Hirsch, 1991 ). Such prudence reviews have led to smaller construction projects with less costs uncertainties (Lyon, 1991). To meet fulture demand growth, DSM programs were favored, as they promoted peak clipping, load shifting, and conservation. Lower cost DSM investments would be made instead ot'building new power plants requiring long lead times that were based on dubious prredictions offuture demand and had a high risk of having all or part of the project's capital expenditures lost due to prudency hearings. Another risk assoeiated with power plant construction are the environmental externalities that are assodated with production, transmission, and distribution. The electric utility industry erriits one-third of the nation's C02, the majority of its S02 , and one-third ofits NOx. (Hiri:;t, 1991). With the Clean Air Act of 1969 and the National Environmental Policy Act of 1970 began an era of environmental control programs in the United States that forced utilities to incorporate environmental costs into their system planning processes. New Source Performance Standards were periodically

21

updated using the "best available control technology," that allowed emissions to be reduced over time. These policies, combined with additional regulations on water pollution and solid waste controls, caused environmental control systems to account for one-fourth to one-third ofthe total costs ofnew coal-fired power plants (Rubin, 1989). Further, increases in electricity generation costs can be expected as more stringent emission standards are enacted. Several state governments and PUCs have already started to incorporate the costs of these externalities in rate hearings (Cohen et al, 1990). Evidence has been found that older capacity is used more intensely (Stanton, 1993), that indicates that the industry is postponing the use and construction of new capacity. To comply with the new standards, ifthey are passed, will put new and even existing generation at risk of higher costs. The potential health risks of electromagnetic fields around transmission lines are also facing higher scrutiny. Possible negative impacts raise the specter of increased transmission costs. This, together with other environmental regulations, should make DSM investments more attractive as they defer the construction of new generation and transmission and distribution systems (Hirst, 1991 ). The reductions of capital and environmental risks are an important facets of DSM that now have competition from supply-side alternatives. Developments in combustion turbine technology and world-wide competition have allowed the prices of gas-fired combined-cycle plants to decrease 30% to 40% over the last 4 years. 12 Cost

12

Electric Utility Week Feb. 26, 1996 pp. 11. 22

savings have notably occurr~d in construction times where the building schedules for such plants have fallen to

ei~hteen

months. New advancements in coal-fired power

plants have also reduced risj(s associated with these plants. Developments in coal gasification and fluidized bed technologies have also reduced these plants' installed costs by 15% to 20% within the same 4 year period. For this coal fired technology construction times have als9 been cut by half, to about three years. Technological improvements have decreased the viable sizes of these coal plants. These shorter lead times and the ability for sm11ller increments in capacity to be installed have reduced the magnitude of capital risks. ln addition, 1using gas-fired or newer coal burning technologies has reduced the environmental risks for these generating plants. DSM investments are also a fonn of public relations. DSM programs that are pursued for such reasons oflen promote conservation and efficient energy by use of advertising, energy audits, installation programs.

~emonstration

Som~ programs~

projects, or small scale retrofit and measure

such as low-income weatherization have a dual

purpose of promoting com1nunity goodwill for the utility and while decreasing the number of unpaid bills. Th(t-se prograrri1s are usually relatively inexpensive and their reductions in electricity demand arc quite small. With the arrival of competition brought about by industry restructuring, DSM :programs can also be seen as a way to promote and differentiate one utility from othe(· energy providers. This can enhance the utility's competitiveness (Flynn, 1992).

23

Since utility profits are regulated by the PUCs, regulatory incentives and disincentives can impact a utility's decision to invest in specific inputs. At the moment, many DSM proponents hold the view that "each kWh a utility sells, no matter how much it costs to produce or how little it sells for, adds to earnings" while "each kWh saved or replaced with an energy-efficiency measure, no matter how little the efficiency measure costs, reduces utility profits" (Moskowitz, 1989). This is certainly true for pure conservation programs that only reduce existing loads, thereby reducing utility revenues. Unless the lost revenue is recouped through a rate increase, there are no incentives for conservation investment. In certain states, attempts have been made to change the incentive to produce electricity rather than conserve electricity. California's Electric Revenue Adjustment Mechanism (ERAM), the state of Washington's larger rate of return on DSM investments, and other DSM incentive plans that have been developed in New York, Massachusetts, and Pennsylvania indicate PUCs trying to foster investments into DSM programs. Greater acceptance levels can occur with the recovery ofDSM program costs, adjustments for the associated revenue losses, and increases in rates of return in these investments countering perceived risks involved with DSM investments (Reid and Chamberlain, 1990). With expected increases in competition, most ofthe regulatory efforts are suspended. Once competition begins, "customer-driven" DSM services may be all that will remain of previous DSM programs. Once the customers are able to change energy service providers, it will no longer be cost-effective for utilities to invest in DSM

24

seiVices that have long-term pay-backs. Capacity and energy savings derived from a customer receiving these services will be shifted if the customer moves to a new electricity service provider. DSM Investments occur for reasons that are dependent on a plethora of utility characteristics, regulatory environments, and characteristics of customers seJVed. Utility investments in production inputs have been modeled by using the framework of a profit maximization model (Averech and Johnson, 1962), (Baumol and Klevorick, 1970). Other models ofutility investments in supply-side resources, such as transmission and generation facilities, were developed by Baughman, Joskow, and Kamat ( 1979). Here the inputs are explained by the characteristics of the firm and its customer base. These models provide an appropriate framework for understanding the DSM investment levels and the factors driving them. These models incorporate a variety of utility characteristics, such as total sales, sales by customer class, total number of customers by customer class, and service area. A model of this type will provide information on what type of utility characteristics are likely to influence the level ofDSM investment.

D.2. Cost of Equity Capital

Various models for capital costs have been developed over time. These include separate models for the cost of debt capital (Berndt, 1979), (Archer, 1981 ), (Dubin and

25

Navarro, 1982), (Prager, 1989) and the cost of equity capital (Trout, 1979), (Archer, 1981), (Pubin and Niavarro 1982), (Gapenski, 1989), (Fanara and Gorman, 1986). For

the mos~ part, these models have been developed to determine if and how regulation impacts ~lectric utilitiies' capital costs. The cost of eiluity models use the market-to-book (M/B) ratio as a dependent variable. This ratio quantifies the relationship between the price that a utility's common stock conunands on the market and the book value of its assets. Financial and managerial factors found to have a strong positive relationship with the MIB ratio include the actual

rate~

of return, the expected rate of return, and the dividend pay-out

ratio (Dt1bin and Nav arro, 1982 ). A negative relationship to the MIB ratio was 1

connecte:d to debt financing and nuclear power plant construction (Gapenski, 1989). The cost of debt capiltal has been found to have a negative relationship with the percenta~e

oftotal costs paid in taxes and for fuel (Dubin and Navarro). Prager (1989)

showed ~hat a large variety of utility characteristics can play an important role in determining the cost of capital.

ll.3.

Ele~tricity

Demand and Price

I1npacts on electricity demand and price have been studied extensively over the years.

El~ctricity

demand is actually driven by the services it offers, such as light, heat,

or powering motors. '1Therefore, it is a derived demand and is associated with a capital

26

good, such as a light bulb, electric appliance, or motor that provides the desired service. Electricity consumption can be modeled using the following equation:

Quantity= f(R,A)

where A is the vector of the stocks of electricity using equipment and R is the vector of the utilization rates of each type of equipment. Lacking exact data on equipment stocks one can create a reduced form equation:

Quantity = f(P. ,P 8 , Y, Z)

where quantity is a function ofPc, the price of electricity, P 8, the price of competing fuels, income, Y, and the vector of other variables, Z. Unless the supply of electricity is perfectly elastic, the identification problem (Pindyck and Rubenfeld, 1991) arises. The identification problem "is essentially one of inferring characteristics of demand alone from market data that reflect the combined influences ofboth supply and demand behavior'' (Bohi, 1984). To deal with this problem a cost function can be modeled separately with the equation:

Cost= f(Quantity,P 1,P,,Pk, W)

27

where the cost of electricity is d~termined by the quantity demanded, the vector of fuel prices Pr, the price oflabor Pt. the cost of capital Pk, and a '~vector of other factors W. These two equations can b~ linked with a linear constr11int: :

Pe *Quantity = Cl)st +1 Return on Investmen,t

Using this constraint, the c;ost Cl!nd demand models caq be €~stimated simultaneously. Bohi (1984) identi~'ies a number of other issue~ inv¢>lved in developing of energy demand models: the aggregation level, measurement issues, functional fonn, and estimation techniques.

Giv~n

the detailed data av~ilabl~e on electricity production

and consumption, numero).ls Stj.ldies have been perfoi111ed ;lt different levels of aggregation. Bohi (1981) has a,rgued that

aggregatin~

derrtand at the utility level is 1

inappropriate, as the basis for demand would differ greatly between commercial, industrial, and residential

~ustqmers.

Residential custpme11s are motivated by utility

maximization, while indu~trialland commercial custorners iare assumed to maximize profits or minimize costs. Beginning with the classic 11rticl1e by Houthakker ( 1951 ), most studies model electricity :demand for specific ra~e cla'fsses, consumers, or finns. Bohi has shown that models operating at detailed levels of aggregation perfonn better than those at greater levels of:aggregation (Bohi,

19~ 1).

1

Developments in demand's functional fonn have e;>ccurred over a long period of time. Linear or log-linear fom1ats were specified in tne ea,rlier models of demand. This

28

changed in the seventies with the advent ofthe translog model (Jorgenson, 1986) (Griffin, 1992). The translog model offers a more flexible functional form which does not restrict the models' elasticities. However, translog estimations are static and thus are unable to capture the temporal properties of demand. Translog model estimation also generates a large number of parameters. To effectively estimate the model only a small number of explanatory variables can be included, which places restrictive assumptions on the model. Measurement issues are also of interest. Electricity consumption is influenced by variables such as income, area of home, appliance stocks, etc., that should be included in the model. Frequently, these variables are correlated with each other, making it difficult to estimate their separate influences. An issue continuously discussed is the choice for the price variable (Taylor, 1977). If the marginal price is considered to be the most appropriate, rate schedules that include items such as block pricing, fixed charges, demand charges, and seasonal and time-of-use rates make its calculation quite difficult. When customers are aggregated over multiple rate schedules, the determination of the marginal price becomes even more complex. The relative merits of including the marginal price are still not established (Taylor, 1977) and one study (Shin, 1985) indicates that customers respond to average cost. Electricity demand and supply studies have been repeatedly enumerated and described in reviews by Taylor (1975), Cowing (1978), Hartman (1979), Bohi (1981, 1984), Jorgenson (1986), Berndt (1991), and Griffin (1992). A few ofthese studies

29

simultaneously estimate price and demand (Halvorsen, 1977, 1978) (Chern, 1978). Using the marginal price instead of the average price is one of the methods employe~! to remove the simultaneity (Berndt, 1991). Estimating the marginal price accurately when aggregating over rate classes, utilities, or over a span of time when seasonal and TOU rates are present proves quite problematic. With larger customers where demand charges apply, actual marginal prices depend on energy prices and demand charges. Interruptible service rates are even more difficult to model. State, regional, and national level models comprised most of the early demand studies. This has changed as more micro-level data have become available allowing studies to focus on industry, firm, and household level data. Hourly demand and demand of specific end uses have also been researched. The leap from regional to micro-level studies has largely avoided the demand at the utility level. One ofthe few published studies performed at this level was undertaken by Lyman (1978). The elasticities estimated through these models have varied significantly. Shortrun residential price elasticities have ranged from -0.03 to -0.54, while long-run elasticities have overlapped slightly, ranging from -0.45 to -2.2 (Bohi, 1981, 1984). Estimated income elasticities have ranged across more unrealistic levels: -0.32 to 2.0 in the short-run and 0.12 to 2.2 in the long-run. Commercial and industrial studies hav~ estimated similar ranges for short and long-run price elasticities. With income elasticities, commercial short-run elasticities have ranged from 0.1 to 0.72 while lon!?run elasticities have naturally been higher, ranging from 0.8 to 1.32. Industrial studies

30

1

have shown output elasticities ranging from 0.08 to 0.87 in the short-run to 0.51 to 0.73 in the long-run. The ranges ofthese estimates are quite great and vary by estimation technique, aggregation level, and data quality. It is possible the elasticities are changing over time as tastes change, the electro-technologies grow, and inter-fuel competition increases. DSM investments have been repeatedly studied within a demand model framework. Numerous billing, load shape, and end use metering studies have been performed at many utilities. Early studies of program impacts began in the seventies (Berry et al, 1981 ). These and later studies have typically been specified as:

Quantity =k (DSM, Price, Customer, Weather)

where the quantity of electricity (hourly, daily, monthly, etc.) is a function of customer characteristics, electricity price, weather, and DSM investment. In earlier studies (Hirst et al, 1982), (Hirst, White, and Goeltz, 1984),the DSM variable used in the estimation was represented by a dummy variable for participation. Eventually, studies used expected savings (Train, 1985) or measure costs (Horowitz and Degens, 1987) to replace the dummy variable, thus allowing the estimation of, respectively, the realization rate of expected savings and the marginal cost of energy savings. Evaluation methods became more and more refined, and methods used came to depend on the type

31

of m(~asures installed, level of' savings, and type of participating customers (Sonlllenblick and Eto, 1995). The cost of electricity has been repeatedly studied at the firm and plant level. Typically, the cost function is estimated by itself and not simultaneously with demand. The reasoning behind this is that in a regulated market the output price is set by the regulatory authority. Therefore, the quantity demanded is exogenous to the cost. With output a function of the regulated price, the output level no longer is under the utility's control. With output fixed, cost minimization becomes a profit maximizing firm's objec:tive from the viewpoint of a regulated utility (Jorgenson, 1986). Most recent electric industry studies specify the cost function as a translog cost function. An early artide by Christensen and Greene (1976) that used the translog cost function to examined the economies of scale in the electric industry. Other studies have researched the effects offuel adjustment clauses (Gollop and Karlson, 1978) and the cost implkations ofvertical integration (Henderson, 1986). These models assumed the utilities in the sample were fairly homogenous, and that most of the variance in cost is explained by input prices. The nature of the translog cost function may have precluded more explanatory variables from being included in these models. If more explanatory varialbles were included, the number of parameter estimates would grow exponentially, maki111g the model more difficult to estimate. Simpler price models have been estimated for the electric industry. These do not use the translog functional form and could also be used to model costs. Halvorsen

32

(1975) estimated a log-linear model of electricity's marginal price as part of a simultaneous model. In a later study by Chern (1978), aggregate industrial prices and demand were estimated in a pooled model for 15 industries. A study by Primeaux and Mann (1986) estimated a set of linear price equations that represented the different rate schedules of the three main customer classes. The research was done primarily to determine if the regulator selection process impacted rates. A one-equation linear model ofprice at the utility level was estimated by Tolley and Bodmer (1990). This model, though simple, explained much of the variance in rates (R 2 =0.64), and indicated that a variety ofutility characteristics and regulatory environments influenced the average price. Berndt, Doane, and Epstein (1995) effectively expanded this study, showing again that many factors associated with state and local government regulation have large impacts on the average electricity price.

33

ill. Methodology

ID.l. Investments in DSM

Many studies have been written about the research on various facets ofDSM. However, a model has not yet been developed to explain DSM investments. Given the substantial investments made in DSM and the variety of reasons for these investments it is of general interest to develop such an investment model. This is especially true with the advent of industry restructuring and ensuing competition. IfDSM investments respond to customer demand, demand will remain despite deregulated competition. If DSM expenditures are partially associated with specific utility characteristics present in a competitive environment, again it is probable that some DSM investments will continue after competition comes about. However, ifDSM is mainly associated with the industry's regulatory climate, deregulation may cause a decrease in utility-funded DSM programs. Simple models explaining investment levels in infrastructure, operations and maintenance expenses have been estimated by Joskow, Baughman, and Kamat (1978). These models can easily be used as a framework to explain DSM investments. They incorporate a set offinanciallmanagerial, operating, and customer characteristic variables. Given that many DSM programs are mandated by regulators, a DSM

34

investment model should also incorporate regulatory variables. The general specification of such a model would be:

DSM =!(Financial, Managerial, Operating, Customer, Re,gulatory)

DSM expenditures can be expected to be caused by a large variety of variables. The prices and costs of electricity is expected to play a large role in determining iDSM investment levels and should be positively associated with DSM investment for tiWO main reasons. First, high electricity costs will translate to higher rates, therefore

1

resulting in greater pressure for lower rates from interest groups. Second, higher costs will also increase the number of cost-effective DSM opportunities

<~vailable

to a ,utility.

The cost of installing new capacity also influences DSM sp
pcrc0iv.;d cost of new capacity cannot be observed as it consists of' more than tHe installed cost. Rather, it is interconnected with the capital and

asso~iated

environmental

risks. A variety of potential proxy variables exist for the need of new capacity and the concomitant costs and risks thereof. Large levels ofCWIP as a percentage ofthe total electric plant, load factor or changes in the load factor, changes in peak load, or changes in capacity can indicate a need for new capacity investmeqts. Also, a low load factor indicates the utility has excess capacity and would be less li~ely to invest in DSM that would reduce total amounts of electricity sold. Wholesale power sales might be a suitable proxy for excess capacity. Available capacity, especially

a~:

peak periods, can

35

spur an utility to market power produced from its remaining idle cap1~city to other utilities. The source of electricity may also impact the DSM investmel)t levels. DSM activity probably will be positively correlated with power purchases
13

Certain conditions may make purchasing power more desirable th"'on DSM investments. Utilities that are not allowed to earn a return on the DSM investment, are not allowed to recoup lost revenues, or have new power plants co~ng into service in the near future may be more inclined to purchase power. 36

also be positively correlated, because the older the stock, the greater the probability that aging plants will be phased out and newe1r, more expensive ones phased in. To reduce the investment in newer, more expensive plants, utilities should be willing to invest in DSM programs. A combined electric and a gas utility would be expected to be associated with higher DSM investment levds. A dual-fuel monopoly would allow the electric utility to remain impartial to fuel switching as a form ofDSM. Fuel switching would not be associated with net lost revenues as sacrificed kWh sales would mean increased gas sales. Weather can lead utilities in areas wiith extreme weather conditions to favor DSM. Extreme weather conditions bringing about higher summer or winter peaks and a consequent rise in excess capacity, increase u1tilities' incentive to undertake DSM investments. DSM programs in these seasonaX!y peaking utilities could garner additional savings as individual customer loads among residential customers would be fairly large, allowing cost-effective installation of energy saving measures. The density of a utility's service territory could play a role in DSM investments, though there are plausible reasons to assume that the relationship could be positive or negative. A lower density could cause many DSM options to be uneconomical due to the greater costs of providing services to a c;ustomer base that is further apart, thus reducing overall DSM investment levels. On the other hand, DSM programs could reduce the need for adding to transmission and distribution systems which could be quite costly in a service territory with greater population density. Greater industrial 1

37

sales are expected to have a negative impact on DSM spending as industrial customers have been known to lobby against DSM, claiming they are subsidizing it and therefore paying higher th~n necessary prices (Clarkson, 1992). Residential customer characteristics may also influence the DSM investment level.s. Many utilities offer DSM programs to low-income customers. Greater numbers oflow income customers may increase the level ofDSM spending. On the other hand, low..,income consumers lack access to the credit necessary to finance energy efficiency mea~ures

and have been found to have higher discount rates (Howarth and Anderson,

199i3). Lower irllcomes therefore may reduce DSM program participation and spending

if programs req!Uire participants to finance some portion ofthe energy efficiency service. Residential customer appliance stocks should also play a role in demand for DSM. Greater appliance stocks should increase the potential savings available to the average customer. As potential savings increase so will the demand for DSM services tha~

will target 1these savings. A utility's regulatory climate remains an important factor in determining

inv~stment.

A J;lUC may have ordered the DSM investment in the first place. A variety

of i.nvestment and research firms evaluate and rank regulatory agencies. These rankings in the past have been typically based on six criteria (Dubin and Navarro, 1982):

1. Allowed ra.te ofretum.

2. Average regulatory lag and use of interim rates.

38

3. Whether a historical or future test year is used. 4. IfCWIP is allowed in the rate base, or if AFUDC 14 is allowed. 5. If tax credits and accelerated depreciation are "flowed through" or "normalized." 6. The presence of an automatic fuel adjustment clause.

It is hypothesized that the less favorable a utility's environment ranks the more a utility

will invest in DSM. For a greater investment in DSM allows investments in more expensive generation alternatives to be delayed until a return on such capital investment will be more profitable. Further, a utility may consider such investments as a way to placate a PUC's least-cost planning demands with the hope of being granted a greater rate of return. The presence ofleast-cost planning requirements and more stringent environmental standards are further variables that come under the rubric of a regulatory environment. State regulations pertaining to emissions and waste disposal, often higher than the federal standards, positively impact DSM programs as the costs of additional and existing generating capacity rise relative to DSM alternatives. However, some stricter state standards, such as building codes and appliance energy use standards, can make many DSM programs redundant or eliminate their cost-effectiveness.

14

Allowance for funds used during construction. 39

ID.2. Impact ofDSM: Cost ofEquity <1apital

As a corporate strateg,r, DSM investment is not rewarded by increases or decreases in price or demand, but in the value ofthe company's stock. Testing the market's acceptance ofDSM programs as this kind of corporate strategy is of great interest with the advent ofindj.Jstry restructuring. lfthe market does not value DSM during regulation, a decline in utility offered DSM services can be expected after deregulation. The cost of equity capital or utility's stock price have been studied through numerous models that can easily be adapted to observe impacts ofDSM expenditures. The simple models estimated l1y Trout (1979) and Dubin and Navarro (1982) can be reestimated with a term for D$M investment included. The adjusted model can be specified as: Market-to-Book Rate= g (DSM, Financial, Managerial, Operating, Regulatory) where the market-to-book rat~ is the dependent variable. The market-to-book (M!B) ratio quantifies the relationship between the utilities' common stock price and the book value of the assets. Variables CJ.SSociated with a strong positive relationship to the MIB ratio are: the actual rate of ret1-1rn, the expected rate of return, and the dividend pay-out ratio (Dubin and Navarro, 1982). Negative relationships have been detected in utilities

40

with high levels of debt, taxes, or fuel costs, or those constructing nuclear power plants. Earlier studies found a negative relationship between the regulatory environment and the cost of capital. In the case ofthe MIB ratio, an improved regulatory environment increased. the MIB rate and, in the case of debt capital, a favorable regulatory environment was associated with a reduction in the rated bond yield. The only study (Gapenslci, 1989) to analyze data from the 1980s found the regulatory environment to have little or no impact on equity capital costs. The regulatory environment has typically been entered into the models as a dummy variable or as a qualitative ranking. Circumstances have changed within the electric utility industry since these studies have been performed. DSM programs and investments have become a multibillion dollar industry and non-utility generation investment has exceeded that of utilities. Further, the arrival of retail wheeling has almost become a reality. With these changed circumstances, it cannot be expected that variables determining the cost of capital will remain fixed. How DSM investments will affect stock values is unknown. It is possible that the market will view them as a successful strategy that minimizes capital expenditures while effectively retaining existing market shares and increasing sales to growing markets. On the other hand, DSM programs may be viewed as a losing strategy that wastes utility funds and results in revenue losses that will result in lower stock values.

41

Finally, DSM expenditures may be so small and unimportant that impacts on stock values will be negligible. Additional variables that may have an impact on the stock values are those associated with increased capital investments, such as CWIP or the age of plant. CWIP was expected to have a negative relationship with the M/B ratio. The greater the amount of this presently unproductive expenditure, the less the market will value the utility. New construction is associated with capital investment risks, given factors such as new environmental rules, inflation, and lack of demand. In addition with industry restructuring, some new construction will become "stranded investments," no longer entitled to be included in the rate base where they can earn a rate of return. Plant age is expected to increase the utility's stock price since greater age will indicate less money invested in more expensive new capacity and possibly more efficient use of existing capacity. Another variable that will be entered into the model is the cost of debt capital. A higher interest rate on short or long-term debt should be associated with higher stock prices as borrowing costs will be higher. The plant age, load factors, changes in peak loads, prices of electricity, and

mak~up

of customers are also important variables that

should be considered when developing the model. The age of plant should have a positive effect as investments in additional plant are being kept low. Average annual load factor should be associated with a higher stock price because more efficient use of existing plant should be associated with higher actual returns on investment and

42

earnings per share, thus increasing stock prices. However, very high load factors may indicate that insufficient generation is available to meet customer demand and will require new construction that, in turn, could be associated with lower stock prices. Increases in peak load will eventually lead to additional capacity or long-term purchases being required. Again, capital expenditures on new plant is expected to reduce the MIB ratio. For utilities with a low load factor this growth in peak load make use of excess capacity. Increases in peak load may be associated with overall load growth (peak and off-peak) that in turn may improve the load factor of the utility with a positive impact on stock prices. The price of electricity is becoming an important factor on the eve of deregulation. It is expected that utilities with lower prices will have lower costs and, therefore, will be able to compete successfully when retail wheeling becomes a reality. These utilities will have their stocks bid upward in anticipation of competition. Distributions of customers should impact stock prices as each class has different costs associated with the services it requires. The share of industrial sales should have a positive impact as they are less expensive to serve and have higher, steadier loads. However, competition may change this. Larger customers with larger loads will have more to gain from switching suppliers for even marginal changes in price. In most scenarios for restructuring the industry, industrial customers first benefit from retail wheeling, while the commercial and residential classes remain captive markets for electric utilities.

43

DSM expenditures will also enter into the model. DSM does not have a predicted sign. A negative sign would indicate that DSM investment depresses the utility's stock value, while a positive sign would indicate the market puts a positive value on DSM investments.

ll.3. Impact of DSM: Demand for Electricity

The factors influencing electricity demand have been studied comprehensively since the fifties. Economists have found the greatest interest in the short-run and longrun, price and income elasticities and cross-price elasticities of competing fuels. This focus is understandable given that this information is necessary in planning electric plant and equipment investments. With the development ofDSM programs, another factor was introduced influencing electricity demand. Much energy demand modeling work has been incorporated into the evaluation ofDSM programs. However, these studies are primarily involved with studying specific program impacts and use microlevel data disaggregated to the customer level or even to equipment levels. Aggregate models ofDSM impacts have not yet been estimated. Detailed aggregated simulation models have been used to estimate DSM impacts (Keelin and Gellings, 1986), (Faruqui, Seiden, and Braithwait, 1990), but actual estimation has not yet been done at a more aggregate level.

44

DSM expenditures in 1993 were close to 10% of new electric plant investments. DSM spending has increased from 1990 through 1993, rising from under 1% to 1.5% oftotal expenditures for investor-owned utilities. At some utilities, DSM represents over 8% of total annual expenditures. In California, DSM was expected to satisfy 60% of the state's growth in electricity requirements for the nineties (Prete, 1992). The overall cost ofthese programs is substantial and is expected to achieve moderate growth levels, despite the oncoming industry restructuring. Therefore, it is timely to look at the more aggregate impacts ofDSM on demand. A simple model ofDSM's affect on electricity demand can be estimated at utility levels. The general specification of the model would be:

Quantity = k (DSM, Price, Fuel, Customer, Weather)

This is the same general model used in countless DSM evaluations. The only difference lies in the level of aggregation. DSM evaluations use customer or end use level data, while the proposed model can be estimated at the system level. The DSM variable that utilities consistently report is the level ofDSM spending. There is no expected sign for DSM investment. This is due to differing impacts ofload management and conservation. In theory, a predominance of conservation in DSM investments will cause the estimated coefficient to be negative. If load management programs comprise the main type ofDSM investment and are structured to shift loads to off-peak periods, the coefficient should be positive.

45

The level of aggregation to the utility level was partially detennined by, as with most studies, the availability of data. Since 1989 the Energy Information Agency has collected data on utility expenditures on DSM. The data for the years 1990 through 1993 has been published and is presently available thus allowing for one consistent source of reported DSM data. Four years ofDSM expenditure data were therefore available for all investor and publicly owned electric utilities 15 . The expected price elasticity will be negative, ranging from -0.45 to -2.0, which approximate the values found in previous residential, commercial, and industrial electricity demand studies. The estimated price elasticity will represent a weighted average price elasticity as demand of all customer classes is aggregated to represent utility system level demand. Aggregating to the system level does not allow the marginal price to be estimated for the average customer, thus the average price is used. Even with disaggregation to the customer class level, it is doubtful that a meaningful marginal price estimate could be estimated. Since the beginning ofthe 1980s, the rate structures of many utilities have become more complex. A variety of rates have been negotiated for commercial and industrial clients. These new rates vary from customer to customer, depending on their ability to switch fuels or become cogenerators, their levels of peak demand, or even their financial stability. Utility rates also vary widely in

15

Even though utility expenditures are reported from a consistent source, expenditures may not be completely comparable. Companies just beginning DSM programs may have all oftheir expenditures in planning and administration costs without actual investments in programs .. Reported DSM expenditures should be viewed more as an index of past and present DSM investments. 46

the residential and commercial sectors, where they may change seasonally, be flat, have increasing or decreasing block rate structures or have fixed charges. Positive cross-price elasticities should be expected given the variety of competing fuels. As the price ofthese fuels rises, customers are expected to increase their electricity demand by partially substituting electricity for now more expensive, competing fuels. Due to supply constraints placed on gas, many electricity customers may prefer gas but cannot obtain the service. Consequently gas' cross-price elasticity will be biased downwards.

Income elasticities have been estimated in many previous studies, but, when aggregation takes place over customer classes the results are not be comparable. Nonetheless, the estimated income elasticity is expected to be positive and within the range of those estimated by previous models. This ranges from 0.2 to 2.2 (Bohi, 1931, 1984), (Taylor, 1975). Electricity demand will naturally be dependent on the population levels and commercial and industrial activity levels in the utility's service territory. The number of residential accounts should provide a good estimate of populations and should have a positive sign. The number of commercial and industrial accounts, however, may not prove a good proxy for energy consumption because in many cases only a few accounts will be responsible for most of the electricity consumption. Thus, employment in these two sectors will be used as a measure of the sector's activity, with an expected positive relationship to electricity demand. General industrial employment may prove to be an

47

inexact measure of industrial demand for electricity due to the differing levels of intensity of electricity use and substitutability for

electri~ity

in industrial processes.

Industrial employment may provide a more exact meas4re in certain SIC groups that use electricity more intensively than others. Climate extremes sho1:1ld increase energy demand duie to increased heating or cooling loads within buildings of all customer classes. Given greater levels of electric space heating, a greater consumption of electricity is to be expected. Appliance stocks may also vary because many states now have more strir1gent appliance and building codes than those of the federal government. These more striingent codes will affect appliance and building stocks, resulting in less electricity consumption for states where they apply.

ill.3. Impact of DSM: Cost of Electricity

There is also interest in estimating the DSM's i1npac:ts on the production of electricity, or specifically on its cost. The proposed co&t eql[lation will be a simple, oneequation model similar to those used to estimate price of electricity used by Halvorsen (1975), Tolley and Bodmer (1990), and Berndt, Doan~, and Epstein (1995). 16 This

equation takes the following form:

16

This specification was used rather than a neoclassical trar~slog cost specification that is typically seen in electric utility cost studies. A large 11umber of factors, other than 48

Cost= j (DSM, Quantity Financial, Managerial, Operating, Regulatory) where the dependent variable cost, is the average cost of producing one kWh. This is a relatively simple model that can incorporate many of the variables affecting the cost of electricity. Quantity is expected to decrease average costs due to economies of scale. The DSM' s variables expected parameter is anticipated to be negative. A negative correlation is expected because the entire point of investing in DSM is to r~duce the cost of electricity. DSM was not promoted as a method of saving electricity at any cost, but was intended as a least-cost alternative. Utilities are investing in DSM to prevent having to invest in more expensive capacity expansion. Nevertheless, it is possible that DSM increases costs if non-cost-effective investments are made. Also, DSM programs that primarily target conservation may, in the short-run, increase costs as conservation results in immediate revenue losses. In the long-run, conservation defers investments in electric plant resulting in lower costs. In the short-run, quantities sold are reduced, resulting is higher costs.

input prices, were expected to influence DSM impacts on costs. A translog model specification that includes numerous non-price variables will tend to have an extraordinary number of coefficients and substantially reduce the degrees offreedom of the regression model. This study includes electric utilities with different power generation technologies. A translog specification typically assumes similar technologies. The neoclassical model assumes that the industry is at its long-run equilibrium. With the market restructuring and the large number of mergers within the industry this assumption is in question. 49

A large number of operating, financial, and managerial characteristics should play a role in determining the average cost of electricity. As with most prior price and cost equations, variables representing capital, labor, and fuel costs will be included. Naturally, increases in the costs of these items are expected to lead to increases in average costs. Financial variables, such as the utility's bond rating, interest rates on long-term debt, and tax payments can be expected to increase the electricity's average cost. Without competitive pressures, utilities may add unnecessary staff, thus increasing costs. For some utilities, taxes account for three-quarters of the variation in price (Thompson, 1985). Operating factors, such as capacity under construction, make up of generating resources, line losses, density of the service territory, and age of the electric plant are also expected affect average costs. Construction work in progress (CWIP) should have a positive impact, either through its inclusion in the rate base or its negative impact on cash flow. Hydroelectric power is an inexpensive form of power generation and a larger share of electricity generated from this source should be associated with a decrease in cost. Electricity generation with nuclear power should increase the cost of electricity. Though initially deemed an inexpensive form of energy, nuclear energy has proven quite expensive due to high construction costs, unexpectedly low plant utilization rates and other reasons. Higher line losses are naturally associated with increased costs. A larger service territory requires more transmission lines per customer. However, as a service territory becomes more dense, additional costs are

50

incurred in laying underground transmission lines. With older plant, much of the capital expenses have been depreciated. Thus, lower costs can be charged. Older generating plant may also be less expensive to operate, even if the embedded costs are the same as in new plant. Older power plants and transmission and distribution facilities may not require the same environmental or safety standards as new plant. For a utility also supplying gas, administrative costs can most likely be kept lower due to synergistic effects such as combined customer records and billing procedures. The more industrial sales are a percentage of total sales, the greater expected electricity cost reductions. Industrial customers are less expensive to serve than residential and commercial customers. On average, industrial customers consume large amounts electricity reducing average fixed costs, and many of them consume high-voltage electricity, which is less expensive to supply. The less the population density of the service territory, the more expensive it is for a utility to supply electricity to its customers, and the greater the expanse, the more transmission and distribution system costs relative to its customer base. An unfiiendly regulatory climate is expected to lower the probability that PUCs

will grant rate increases. This has a dampening effect on cost as utilities will make even more attempts to minimize costs. Though various federal laws already establish permitted emissions and wastes levels, state governments may make these levels even more restrictive. With one-fourth to one-third ofthe total cost for a new coal plant going into environmental control systems, these more restrictive regulations can

51

substantially affect costs (Rubin 1989). Some states now are moving towards requiring utilities to internalize environmental externalities within the utilities' cost structure, adding to the cost of electricity.

ID.S. Scope of Analysis

The EIA has collected DSM investment data since 1989, though these have only been available since 1990. The four years ofEIA data available (1990-1993) were analyzed in this study. DSM investment data were available from other sources, 17 but were not used in favor of staying with a single consistent data set. Also, the EIA surveys all major utilities making it possible to assume that utilities without reported DSM investment actually had little or no DSM expenditures. The EIA collects a large volume of annual data on the financial and operating characteristics of private and public utilities. Fairly detailed information available for, approximately 18 180 largest investor-owned utilities is part of this data. At the end on 1989, these utilities represented 99% of investor-owned utility sales and just under 79% of total electricity revenues from sales to ultimate consumers in the United States.

17

When this study was initially conceived, the analysis period was expected to be the 1980's. During that period DSM investments were inconsistently reported by utilities. DSM expenditures from the decade were obtained from a variety of surveys (Cogan and Williams, 1985, 1987) (Nadel, 1991 ). In 1991, I also collected this expenditure data by surveying utilities and PUCs around the country. Collection methodologies between sources often caused expenditure figures for the same utilities to significantly differ. 18 This is an approximate number as the number reported each year will vary slightly due to mergers and fluctuations in utility sizes. 52

Investor-owned utilities produce the bulk of the electricity used in the Umited States and the majority of these have f;inancial statistics published by the EIA. Using this group ofutilities to estimate the models described above, would give a dear indication ofthe DSM determinants and impacts in the United States. It would be difficult to include in this study municipal, federal, or other types of non-profit generators or electricity distributors. The Federal Energy and Regulatory Commission (FERC) reporting guidelines differ between investor-owned utilities and the non-profit utilities. This causes some variables to be reported by investor-owne:d utilities but not by the non-profits. In addition, the non-profits are not regulated by the state PUCs but by elected boards. Finally, the profit motive, which supposedly spurs the in~estor-owned utilities into DSM program investments to reduce electricity prices and increase stock values, is missing from non-profit firms. Not all ofthe reported utilities can be analyzed as

sc~parate

econ01mic decision

makers. Many utilities are operating units of holding companies and,

th~refore,

do not

operate independently. Also, some reported utilities are actually partially1 owned by other utilities. This is true for a number of the nuclear powc~r plants in the Northeast, such as Connecticut Yankee or Maine Yankee, where up to nine utilities1 own varying shares in the plants. Even if only partial, ownership of othe1r utilities will1impact the operations and investment decisions of a particular utility. Analysis of individual economic agents requires investments and expenditures be aggregated to the holding company levels. In cases where partial ownership of a utility was evident, all variables

53

were weighted by the percentage owned. 19 This weighting was performed for each of the 900 EIA reported variables. Aggregating to the holding company level was particularly important because other utility data were obtained from Value Line, an investment publication which reports at the holding company level. Value Line was the source for a variety of financial data and items such as load factor, peak load, capacity, rate of return, dividend pay-out ratio, and average age of plant. Value Line also reported on merger activities and bankruptcies. Utilities were dropped if they had merged with another utility during the 4 year analysis period, as merger year data might be inconsistent, with some variables representing the utility before and others after the merger. El Paso Electric was excluded from the analysis because of its bankruptcy, which was thought to impact most of its operations and investment decisions. Other utilities were omitted if they were not reported in Value Line or Moody's Utilities. 20 A few utilities were removed from the sample due to significant foreign operations or because most of their business activities were outside the electric utility industry. Texas-New Mexico Power was not included in the final sample because it did not have generating facilities. The final sample of holding companies with complete data for the four-year study period included 81 companies. These 81 actually represent 119 utility operating companies reported by the EIA.

19

A simple example of this would be a generating company owned equally by two electric utilities. If30 MWh was sold by an operating company, the owners would each be associated with 50% of the power produced. 20 Moody's Utilities was used as a source to determine ownership shares. 54

Table ID.l Utility Sample Attrition

Sample

Utilities

Utilities reported on by EIA 1984-1993

190

Utilities dropped

72

Companies used in analysis

119

Number ofholding companies in anaiysis

81

Models estimated using the final sample of electric holding companies should provide a good indication of overall DSM impacts. In 1990, these electric utilities sold most ofthe electricity in the United States representing 60.7% of all electricity sold and 78.6% of electricity sold by investor-owned utilities. In the same year, the utilities represented 83.5% oftotal nationally-reported DSM expenditures and 92% ofDS:M expenditures reported by all investor-owned utilities. Typically, customer characteristics are not available for most utilities except for sales to specific customer classes. Reported population or service area sizes are only available for a portion of the sample. What is known, however, are total sales to end users by state. This breakdown was available for 1993 from data published in EIA's Form 861. Total electric sales by state allow each utility's share of state electric sales to be calculated. The utility's share oftotal state electric sales is assumed to approximate service to the equivalent portion of the state's population and businesses. This assumption means a utility selling 50% of a state's electricity consumption will provide 55

services to 50% of its population and will deliver electricity to companies with 50% of the business activity. This assumption allows various state and regional level demographic data to be used to estimate utility customer characteristics. State and regional level variables include income, employment, weather variables, types of industrial activity, the regulatory environment, energy prices, and appliance stocks. A complete listing of variables and their sources is supplied in Appendix 1. Many utilities do not operate within a single state. For utilities operating in two or more states, customer demographic variables are calculated using the weighted average of each variable. The weight used is the percentage of total sales to end users sold in each stat~ seen as a percentage of total utility sales to end users. The final data set used in the analysis contained information on 81 electric utilities over a four-year period. Variables available were detailed information on the companies' financial and operating characteristics, approximated demographic data on each utility's customer base, information on the regulatory environment of each utility, weather variables, and prices of competing fuels.

56

IV. Empirical Estimation

IV.l. Model Summary

Four equations were estimated as part of this analysis. These models are specified as: 1. DSM =f (Financial, Managerial, Operating, Customer, Regulatory) 2. MIB Ratio = g (DSM, Financial, Managerial, Operating, Regulatory) 3. Cost =j (DSM, Quantity Financial, Managerial, Operating, Regulatory) 4. Quantity = k (DSM, Price, Fuel, Customer, Weather)

Factors influencing DSM investments were incorporated in the first equation. The second equation estimated DSM impacts on the a utility's cost of equity capital, or stock price. The third equation estimated the average cost of producing electricity to determine how DSM influenced the average cost of producing electricity. The fourth equation was used to estimate DSM investments impacts on electricity demand. Variables used in the final models are shown on Table IV. I. They are separated into firm financiaVmanagerial and operating characteristics, customer and demographic characteristics, and regulatory environment characteristics.

57

Table IV.l Variables Used to Estimate Model

Variable

Variable Description

Source

Name Financial and Managerial

ROR

Rate of return on investment.

Value Line

PCTAX

Taxes as a percentage of total expenditures.

FERC Form 1

PCCWIP

CWIP as a percent oftotal electric plant

"FERC Form 1

investment. WAGES

Natural log ofthe average utility wage.

FERC Form 1

PCGASREV

Total gas operation revenues as a percent of

FERCForm 1

total utility revenues. DIVIDE:t-.,TD

Dividend pay-out ratio.

Value Line

EQDBRAT

The equity- debt ratio.

Value Line

MBRATIO

Market-to-book ratio, the market share price

Value Line

divided by the book value per share. Operational

PCIND

Percentage of total sales to industrial customers.

FERC Form 1

PCDSM

DSM percentage share oftotal expenditures.

EIA Form 861 and FERC Form 1

PRICE

Average price of electricity.

FERC Form 1

PCPURCH

Percent oftotal power (purchased and

FERC Form 1

generated) that was purchased.

58

Variable

Variable Description

Source

Name LOAD

The annual load factOi (ratio of average load divided by peak load)?

Value Line

1

COOLDD

Annual cooling degree-days.

BLS

DENSITY

Pole miles per residential customer.

FERC Form 1

QUANTITY

Total quantity electricity sold to final consumers

FERC Form 1

divided by number of residential customers. INVKWH

The investment in plant per kWh generated or

FERC Form 1

purchased. PCNUKE

Percent of generated and purchased power that

FERC Form 1

is generated by nuclear power plants. FUELKWH

Fuel costs per kWh generated.

FERC Form 1

PLANT

Average age of utility plant.

Value Line

PCRESALE

Total electricity resales as a percent of total

FERC Form 1

electricity sold. Customer PC INC

Per capita income.

BEAREIS

RES KWH

Average MWh consumption per residential

FERC Form 1

customer. GAS

Price of natural gas.

Gas Facts Yearbook

PCMFG

Manufacturing as a percentage of total

BEAREIS

employment.

21

This was derived by computing the average annual hourly kWh demanded (total kWh sold divided by 8,760, the number of hours in the year). This was divided by the maximum hourly kWh demanded during the peak period (usually one specific hour). 59

Variable

Variable Description

Source

Name Regulatory PCPAPER

Percentage of all employment income from the

BEAREIS

pulp and paper industry PCPMETAL

Percentage of all employment income from the

BEAREIS

primary metals industry. LCUP

Dummy variable indicating least-cost planning

Mitchell

required within the firm's service territory.

(1989, 1992}

1 =Required 0 =Not required REGOOD

Dummy Variable indicating the regulatory

Value Line

environment is rated above average. 1 = Above average 0 =Average or below average

IV.2. DSM Investment Equation

Large levels of investor-owned utility DSM investment were hypothesized to be caused by specific firm, customer, service territory, and regulatory characteristics. Simple models have been developed explaining investment levels in infrastructure and operations and maintenance expenses. Such a model has been adapted to explain DSM investments. The estimated DSM investment model has the general specification:

60

DSM = f (Financial, Managerial, Operating, Customer, Regulatory) Data were available on total DSM expenditures for 81 utilities over a four-year period of 1990 through 1991. Supplemental infonnation on utility characteristics, service area demographics, and regulatory climate were collected. These data allowed for a set of cross-section time-series models to be estimated. Variable scaling and nonnalization proved important as the utilities analyzed ranged from just under 1 million MWh sales to 119 million MWh sales. The equation's DSM variable was expressed as a percentage of total expenditures. All explanatory variables were either included in the equation as averages, ratios, or dummy variables. The explanatory variables included in the final model are shown in Table IV.2 along with the expected sign ofthe estimated parameter.

Table IV.2 DSM Investment Model Variables

Variable

Expected

Name

Sign

PCDSM

Dependent

Variable Description

DSM as percent of total expenditures

Variable PCIND

-

Percentage of total sales to industrial customers

RES KWH

+

Average MWh consumption per residential customer

PC INC

?

Per capita income

61

Variable

Expected

Name

Sign

PCPURCH

+

Variable Description

Percent of total power (purchased and generated) that was purchased

+

LOAD

The annual load factor (ratio of average load divided by peak load)

ROR

+

Rate of return on investment

LCUP

+

Dummy variable indicating least-cost planning required within the firm's service territory I= Required, 0 =Not required

REGOOD

-

Dummy Variable indicating the regulatory environment is rated above average I = above average, 0 = below average or average

In addition, three dummy variables, Y91, Y92, and Y93, give ihe Jnodel the form of a fixed-effects cross-section time-series model. Given cross-section time-series data, it is possible the actual parameters differed between each times-series and/or cross-section. With 8I different cross-sections with only four time-series observations each, it was not considered viable to estimate varying coefficients for each utility. Varying coefficients were tested for across years22 (Table IV.3). The resulting F-test

22

Two equations were estimated: a restricted model including 8 explanatory variables and the intercept, and an unrestricted model that included annual interaction terms for 8 explanatory variables and different intercepts for each year. The F statistic of0.92 was not statistically significant and was calculated as: ((ESSR- ESSUR)/q)/(ESSUR/(N-k)= ((437-402)/27)/(402/(324-36) Where ESSUR was the error sum of squares of the unrestricted model with varying coefficients, and ESSR was the error sum of squares ofthe model where the parameters 62

rej1 cted the hypothesis that all coefficients varied across all years. Yeanly dummy var ·abies were included in the model because specifYing different jntercepts for each ye« r were found to be statistically significant when compared to the restricted model. The calculated F statistic was 4. 2, which is significant at the 1% l~vel. :Z:l In regression estimation, other statistical problems can oqcur which may result in i iasing the parameter estimates, their variances, or the efficienc;y of the estimate. Tv o problems typically looked for with cross-section time-series models are au· ocorrelation and heteroscedasticity. Autocorrelation arises wh.en en·or terms are correlated with one another. The presence of autocorrelation do~s not ~bias the coefficient estimated using OLS. However, inefficient variances 9fthe1parameter es1 imates typically result. Testing and correcting for autocorrelation is,not really applicable for the data set being analyzed. The data have only fot1r years of data for each firm. Testing for autocorrelation for each cross-sections' fo).lr-year time period w1 uld have dubious results. In addition, correction for even the ~implest first-order autocorrelation model would result in the loss of one-quarter of the observations, le: ding to a serious reduction in the efficiency of the estimation. Heteroscedasticity arises when the variance of the error term is not constant across sample observations. Heteroscedasticity does not bias the parameters'

estimate~.

However, heteroscedasticity can result in inefficient estimators.

did not vary over the years. The number of restrictions placed on the 11nodel number was q, k was the number of parameters in the unrestricted model and N was the m mber of observations. 23 Hypothesis test inputs are reported in Appendix 3. 63

The presence ofheteroscedasticity was tested for using the Breusch-Pagan test24 (Breusch and Pagan, 1979). All explanatory variables were included in the regression used to estimate the test statistjc as oo particular regression variable was suspected of causing the heteroscedasticity. The calculated test statistic was 128, which at 8 degrees of freedom leads to rejection ofthe null hypothesis ofhomoscedasticity. Detecting heteroscedasticity with these itests is straightforward; correction for heteroscedasticity is not. Correcting for heteroscedasticity requires making certain assumptions about the functional form of the heteroscedasticity. As ordinary least squares (OLS) offers unbiased estimates pfthe1parameters, the method developed by White (1980) was used to calculate a het~roscedasticity-consistent variance-covariance matrix (HCV)? 5 This method is extremely usdul as it allows us to make appropriate

24

The Breusch-Pagan test estimates the test statistic through a two-stage procedure. The equation: y = (X. + p:z + E that is suspected of having heteroscedasticity i:s estimated using OLS, where the dependent variable Y is a function of int~rcept. a. and the vector of explanatory variables Z. The error term is E. The esti1nated error term is squared and divided by the asymptotic variance of the error term anll the regression: A2

~ =: a+bZ+e

a

e

is estimated. Where the estimated error t.erm squared 2 is divided by the asymptotic variance of the error term ri defined as t.he mean of the squared residuals. The test statistic is half of the regression (explain~d) SUim of squares. This test statistic is distributed asymptotically as a chi-squar~d distribution with degrees of freedom equal to the number of variables in Z. This is a, geneiral test for the presence of heteroscedasticity and does not require ~~ny prrior knowledge of the functional form that the heteroscedasticity takes. 25 White's HCV is calculated by constm~ting the diagonal matrix G where the diagonal terms are the squared residual terms fro1n the 1original OLS regression. The HCV matrix is then estimated by computing the GyS variance-covariance matrix (X' X)· 64

statistical inferences based on the result of the OLS without specifying the fonn of heteroscedasticity. The parameters' standard errors and t statistics were reestimated for the model using the HCV(Table IV.4).

Table VI.3 DSM Investment Equation Hypothesis Testing

What Was tested

Test and Test

Result

Statistic Varying slopes and

F-test

intercepts for each year

F statistic= 0.92

Varying slopes for each

F-test

year with one intercept

F statistic= 0.85

Varying intercepts for each

F-test

Statistically significant

year

F statistic= 3.88

Fixed-effects model specified

Heteroscedasticity

Breusch-Pagan Test

Presence of

Chi-squared statistic =

Heteroscedasticity detected.

128

White's HCV estimated

Not statistically significant

Not statistically significant

1

X'GX(X'X)"\ where the X matrix consists of all the explanatory variables. The square roots ofthe diagonal tenns ofthis matrix are the HCV standard errors associated with each coefficient. 65

Table IV.4 Parameter Estimates of tlte DSM Investment Equation (N = 324 R 2 = 0.41 log-lik
Variable

Paraml!ter Estimate

Standard Error

t Statistic

INTERCEPT

-4.328546

I

1.137

-3.806

Y91

-0.05171"2

I

0.1280

-0.4046

Y92

0.49656~

I

0.1734

2.864

Y93

0.429587

0.1864

2.304

PCIND

-0.0153;t4

0.005964

-2.573

RES KWH

0.113888

0.03009

3.784

PC INC

0.118443

I

0.02684

4.413

PCPURCH

0.007154

I

0.005477

1.306

LOAD

0.021909

I

0.01162

1.885

ROR

0.0544()0

I

0.02732

1.991

LCUP

0.9601$9

I

0.1913

5.018

REGOOD

1.1878$2

0.2741

4.334

' '

'

' '

The regression e~uatioqs overall performance proved satisfactory because most of the explanatory variables haye parameter estimates with the correct sign and are also statistically significant. The sig~1 of the parameter estimates were, for the most part, as

66

expected, with the exception ofthe income, PCINC, and the regulatory climate, REGOOD. The estimated R2 of0.41 does indicate that much ofthe variation in DSM spending remains to be explained. Utility characteristic variables includeq in the model were the rate of return, ROR, load factor, LOAD, and percentage oftotal power purchased, PPURCH. The estimated rate of return parameter (0.05), wa~ small and positive. This indicated financially healthier utilities tended to invest 111ore in DSM.1 It is possible that the direction of causality was the reverse, since firms investing in DSM may be allowed a higher rate of return. The actual absolute impact was quite1 small given a one-point increase in the rate of return increased the DSM share oft0tal expenditures by just 0.02. The positive relationship between the a1mualload factor, LOAD, and DSM indicates that DSM investments were possibly being used to defer capacity. A lower load factor indicates excess capacity during off-peak periods, but as the load factor becomes higher, the reserve margin shrinks, ~hus producing the need for more capacity. DSM investments that shift load and conserve electricity tend to lower the overall load factor and the need for additional capacity investments. The purchased power parameter estimate (0.007) was positive, indicating that utilities may have been trying to replace purchased power through investment in DSM "1negawatts." However, the parameter was not statistically different from zero, and the impact of the estimated coefficient was quite small, as the model prepicts that a utility purchasing 100% of its electricity would only increase DSM's share oftotal expenditures by 0.7.

67

The percent of electricity sold to industrial customers, PCIND, had the expected negative sign (-0.015). A negative parameter was expected because industrial customers have used their market power to lobby for less DSM. Another interpretation could be that utilities tend to target industrial customers less in DSM programs and, therefore, utilities with greater industrial sales have, in general, less DSM expenditures?6 Again, the absolute Impact of the industrial customers' share of consumption was quite small, as the regression predicted that a utility with only industrial customers would decrease DSM' s share of total expenditures by 1. 5 points. The average residential consumption, RESKWH, had the hypothesized positive impact on DSM expenditures (0.11 ). This positive relationship was assumed because larger per-customer consumption should be associated with more end uses per customer. With more end uses more appliances and equipment are available to target and more areas exist for potential energy savings. Higher consumption levels also allow DSM program implementation to become more cost-effective if there are fixed costs to providing DSM services. The estimated coefficient indicates that a utility with average annual residential customers consuming 15,000 kWh versus one with residential consumption of only 5,000 kWh per year will have a one-point higher DSM share of total expenditures.

26

Lower DSM investments in industrial customers could be due less market imperfections that prevent industrial customers from investing in DSM. Industrial customers will have better access to capital than residential customers and have a profit incentive to invest in energy-efficient processes and equipment. 68

In the case of income, PCINC, utilities in areas with higher per capita incomes were associated with grea'ter DSM expenditures. It had been hypothesized that larger number oflow-income customers would bring about greater levels ofDSM investment. A lower per capita income would be associated with a greater number of low-income customers. However, the parameter estimate (0.12), was positive and indicated that for every thousand dollar increase in average per capita income DSM's share of total expenditures increased by one-tenth of one-point. The positive correlation could have been because people with higher incomes demand more DSM. People with higher incomes could have more to gain from DSM programs due to greater appliance stocks, or lower discount rates will increase their propensity to invest in energy-efficiency measures. lfDSM was viewed as a competing good to electricity of higher quality, it is possible DSM demand increases with income, simply because these people will be willing to pay more for higher quality sources of electricity. The two regulatory variables, LCUP and REGOOD, had a considerable impact on DSM investments. The model predicted that a utility operating within states with least-cost planning and above average regulatory environment were likely to increase DSM expenditures by over 2% of total expenditures. Given that DSM shares of total utility expenditures averaged 1.41% for the four-year analysis period, this is a considerable impact. The least-cost planning variable was obtained from two studies done by Mitchell (1989, 1992). The studies classified the state ofleast-cost planning in each state for the years 1989 and 1991. States were classified into five categories, no

69

progress in LCUP, ifLCUP was under consideration, ifLCUP was under development, ifLCUP was in implementations, and whether LCUP was being practiced. During 1989 seven states had LCUP that was in practice which increased to 14 in 1991. Results for these two years were extended to the adjoining years. The variable was a one ifleast-cost planning had been implemented in states making up over 50% of a utility's service territory and zero otherwise. Least-cost planning compels utilities to consider DSM and supply-side resources when developing resource plans. It is, however, the potential pro-DSM regulatory climate and not just the integration of DSM into resource plans that causes the relationship between LCUP and DSM to be positive. The parameter estimate of0.96 indicated that least-cost planning requirements would increase DSM's share of total expenditures by the same amount. The regulatory climate variable REGOOD was obtained from Value Line. Value Line classified the regulatory climate of utilities as average, above average, and below average. This qualitative rating was converted into a binary variable that was one in an above average regulatory environment and zero otherwise. It was expected that a "PUC driven " DSM would be associated with states with below average regulatory climates. Also, it was assumed that PUCs coercing utilities into implementing DSM programs would receive a below average regulatory climate rating. This proved not to be the case, as the parameter estimate (1.2 ) was positive and statistically different from zero. The model predicted that utilities operating in an above average regulatory climate would increase DSM's share of total spending by 1.2 points. Perhaps in a good

70

regulatory environment the PUC will structure the incentives to invest in DSM in such a ways as to make such an investment profitable. In monetary terms a 1.2% of total expenditures of the average tltility would! represent approximately $14.5 million. The model showed an apparent shift in DSM spending in 1992 and 1993 relative to 1990 and 1991. TWs is seen fi·om the yearly dummy variable coefficients where in the two later years the average utility increased its DSM share of total expenditures by respectively 0.5 and 0.43 of a percentage point. An important result of the model was to show that the regulatory environment

played a very significant role in DSM spi~nding. Given this result, one can expect reductions in the levels ofD~M spending with the onset of industry restructuring and deregulation. Industry restru~turing will not necessarily mean the end of regulator driven DSM. After industry restructuring, PUCs may still regulate transmission and distribution companies and

r~quire

them to offer DSM services perhaps financing this

by assessing a line charge. Many other explanatc~ry variables were considered for inclusion in the model. 27 All of these variables proved not to have: a statistically significant impact on the level of DSM expenditures. The one variable that was expected to have a strong relationship with DSM spending levels WflS price. It had been expected that higher prices would result in the PUC placing greater pressur:e on the utility to invest in DSM. Also, higher

27

Included in these variables were regional dummy variables and a variable for the number of states in which th~ utility operated. Statistically insignificant coefficients were estimated for all of the~e variables. 1

71

prices should result in larger numb~rs ofDSM opportunities becoming cost-effective. However, the average price and price of each cuistomer class all appeared not to be correlated with DSM investments.

IV.3. Cost ofEquity Capital Equation

DSM investment was hypothesized to have an impact on the value of the utility's stock. Successful and unsuccessful corporate strategies should be rewarded (penalized) by an increase (decreas~) in the value of their stock. Earlier studies have investigated the effect ofthe regul~tory climate on the cost of equity capital. The way in which the market values DSM,

~sa

corporate strategy, is of more interest now that

the industry is restructuring. The simple equation that was estimated is quite similar to those used in earlier studies and is ~pecified as:

MIB Ratio = g (DSM, Financial, Managerial, Operating, Regulatory)

where MIB ratio was the market-tq-book ratio, defined as the stock market price per share divided by the book value per share. This ratio quantified the relationship between a utility's common stock price and the 'pook value of the assets?8 Utilities that

28

The book value is an accounting term, calculated as the sum of all assets minus the sum of all debts, liabilities, and preferred share prices. 72

pursue a corporate strategy favored by the market can expect the MIB ratio to rise as the stock price is bid up relative to the book value. Valiables included in the model are presented in Table IV.S

Table IV.S Cost of Equity Capital Equation Variables

Variable

Expected

Name

Sign

MBRATIO

Variable Label

Dependent Market-to-book ratio, the market share price Variable

divided by the book value per share.

PLANT

+

Average age of utility plant

EQDBRAT

-

The equity- debt ratio

PClliD

n

!

Perct':ntage of industrial sales

ROR

+

Rate of return

REGOOD

+

Dummy variable indicating the regulatory environment is rated above average

LCUP

?

Dummy variable indicating least-cost planning required within the finn's service territory

DIVIDEND

+

Dividend pay-out ratio

PCDSM

-

DSM as a percentage of total expenses

PCCWIP

-

CWIP as percent oftotal utility plant and equipment

PCRESALE

+

Percent of total sales that are to other utilities

73

The tests performed on the model are shown in Table V1.6. F-tests indicated that slope parameters and the intercept varied in a statistically significant fashion for all of the four years. The same was true for a regressions that either specified that slopes varied or intercepts varied over all four years. However, after a fixed-effects model was specified, a F-test indicated that there was not statistical grounds for specifying different slope parameters in addition to the varying intercepts since the F statistic was 1.01.

The Breusch-Pagan test indicated that heteroscedasticity was present. As in the DSM investment equations above, instead of correcting for heteroscedasticity, HCV standard errors and t statistics were calculated and presented with the parameter estimates on Table VI. 7.

74

Table VI.6 Cost of Equity Capital Model Hypothesis Testing What Was tested

Test and Test Statistic

Result

Varying slopes and

F-test

Statistically significant

intercepts for each year

F statistic= 5.17

Varying slopes for each

F-test

year with one intercept

F statistic= 5.6

Varying intercepts for

F-test

each year

F statistic= 42.3

Varying intercepts for

F-test .

Not statistically

each year vs. varying

F statistic = 1. 01

significant

Statistically significant

Statistically significant

slopes and intercepts

Fixed-effects model

for each year

specified

Heteroscedasticity

Breusch-Pagan Test

Presence of

Chi-squared statistic =

Heteroscedasticity

40.5

detected. White's HCV estimat€id

75

Table IV.7 Parameter Es.timates of the Cost of Equity Equation Error and t Statistic (N= 324 R 2 = 0.88 log-likelihood= -1,383)

Variable

Parameter Estimate

Standard Error

t Statistic

INTERCEPT

4.011668

8.709

0.460

Y91

8.098673

2.536

3.194

Y92

22.569784

2.624

8.601

Y93

32.480163

3.068

10.59

PLANT

1.234811

0.9181

1.345

EQDBRAT

1.217913

0.1508

8.078

PCIND

0.277331

0.08997

3.083

ROR

3.114928

0.5784

5.386

REGOOD

5.165617

3.486

1.482

-8.930453

2.896

-3.084

1.829703

0.4530

4.039

PCDSM

1.909826

0.9304

2.053

PCCWIP

-0.435635

0.09706

-4.488

-0.136690

0.07167

-1.907

LCUP DIVIDEND

PCRESALE

I

'

'

76

The overall regression .r:esults were quite fruitful in that the explanatory variables included in the regression explain 64% of the variance and had, for the most part, parameter estimates that were significantly different from zero. The parameters estimated in the equation also had the expected signs. The annual intercepts, Y91, Y92, and Y93, captureq much of the steady rise that utility stocks were experiencing over this four-year period?9 Variables ~hat had a very significant impact on the MIB ratio were the rate of return, ROR, and the dividend lpay-out ratio, DIVIDEND. A one-point increase in the rate of return, ROR was predi¢ted by the model to cause a 3.1 point increase in the

MIB ratio. 30 The ~iividend pay-·out ratio, DIVIDEND, had less of an impact as a one pint increase in the dividend pay-out ratio was associated with a 1.8 point increase in the MIB ratio. ln(ireased stock prices should be associated with utilities that have greater earnings qnd provide greater annual income to their stockholders. For every point the equity-debt ratio, EQDBRAT, increased the MIB ratio was expected to increjlse by 1.2 points. This indicated that more leveraged utilities tended to have a lower ~1/B ratio. Tllis in turn showed that the market penalized the use of debt financing.

29

Average utility stock prices 1increased a total of36% over the four-year time period whereas the intercepts account for an average rise of 26%. 30 The average utflity stock had a mean MB ratio of 1.45 during the analysis period. Therefore, a 3.1\ point increase in the MB ratio is the equivalent of2% increase in valuation. 77

The market appeared to favor utilities that do not invest heavily in new construction. Practically a half a point decrease (-0.43) in the MIB ratio was associated with each percentage point increase in PCCWIP, which is CWIP's share of the total utility plant investment. That the market put higher valuation on utilities that deferred capital expenditures was also revealed with the parameter estimate of the average age of plant, PLANT. This parameter estimate indicated that for year increase in mean age of plant the MIB ration increase by 1.2 points. Part ofthis impact on MIB ration is also due to the fact that as plant ages it depreciates and lowers the book value of the firm. If stock value decreases at a lower rate than the asset value due to depreciation the MIB ratio will increase. That DSM defers new construction might be an explanation for the finding that DSM expenditures were treated favorably by the market. In the case of the DSM variable, PCDSM, a one percent increase in DSM's share of total expenditures were associated with a 1.9 point increase in the MIB ratio. That DSM had such a strong impact of the MIB ration was an important finding. This statistically significant parameter estimate indicated that the market puts considerable value on a corporate strategy that incorporated DSM programs and services. It was possible that DSM acted as a proxy for other variables such as utility innovation or the level of customer services the utility offers. If it was DSM investment that brings about the higher stock values then it is possible that after the electric industry is deregulated that DSM services will still be offered by utilities because it is a good corporate strategy.

78

Utilities that have a large industrial customer base could be expected to have I

higher stock prices since industrial customers wiill have lower administrative costs i

associated with them and provide large and stable loads. Also, many industrial customers have large 24 hour electricity loads that utilize the! utilities off-peak, excess capacity. This is expected to change after retail wheeling comes into existence. At that time, industrial customers should have a greater tendency to switch from one supplier to the next, for small marginal changes in price. The industrial customer share of sales, PCIND, had a positive, statistically significant parameter estimate (0.28), that indicated that the market bid up the share price of utilities with a larger industrial customer base. I

A ten point increase in the market share of industrial customers was associated with a 2.8 point increase in the MIB ratio. The regulatory environment variable, REGOOD, had a parameter estimate of 5.17. This indicated that utilities operating in an above average regulatory environment I

had a MIB ratio that was 5.17 points higher than a comparable utility operating in average or below average regulatory environm~mts. This reswlt was not statistically different from zero. The least-cost planning durnmy variable,1 LCUP, had an estimated I

coefficient of -8.9. This indicated that utilities operating in states that required least1

cost planning suffered from decreased stock prices. That thei MfB ratio decreased by i

nearly 9 points with least-cost planning is a good indication that the market does not I

put much stock in mandated least-cost planning. I

79

Comparisons between this and earlier studies may not be appropriate as the electric industry has changed considerably since the time of the earlier studies. One indication of this is that earlier models incorporated f;u fe~er explanatory variables yet had similar explanatory power (Dubin and Navarro, 1982).1That more variables were correlated with the MIB ratio was possibly a sign that factors influencing an electric utility's stock price have become more complex now than ~n the past. Another, reason for this difference was that only a cross-sectional datil was :analyzed in the earlier study. This would tend to reduce much of the variance in stock values since they fluctuate over time. One area in which results are not quite comp11rable:is the impact of regulatory environment. In earlier studies that took place in the seventies a good regulatory environment was positively correlated with cost of Ciipital.: However later studies carried out in the eighties indicated that this regulatory power over the utilities cost of capital had waned. The equation estimated above inqicatecll that regulatory environment still has an impact on the utilities' cost of equity capital.

,

IV.4. Electricity Demand Equation

Since the early eighties, utilities have been investing large sums of money into DSM. DSM has been promoted as a way to meet ex;isting and growing customer

80

demand in electricity without investing in potentially costly generating plant, transmission, and distribution facilities. Though small as a percentage of total expenditures, DSM investments have become quite sizable, with over $2 billion spent in 1993. The effects ofDSM on electricity demand was investigated by estimating a simple demand model. The estimated DSM investment model has the general specification: Quantity

= k (DSM, Price, Fuel, Customer, Weather)

where the quantity of electricity demanded was a function ofDSM investment, price of electricity, Price, customer characteristics, Customer, prices of competing fuels, Fuel, and weather conditions, Weather, that influence electricity consumption31 . As DSM programs will aiTt:ci. a wide variety of cusiumer classes, demand for electricity was analyzed at the utility level. The dependent variable, quantity, was entered into the equation as the natural logarithm of the total MWh sold to all customer classes divided by the number of residential customers. This is to approximate a population -weighted consumption and make the utilities more comparable. All other variables that were not percentages were also entered into the equation as natural logarithms.

31

These are weather conditions that influence, such end uses as electric space heating. space cooling or lighting. 81

Table IV.S Demand for Electricity Equation Variables

Variable

Expected

Name

Sign

QUANTITY

Variable Label

Dependent Natural log oftotal MWhs sold to final Variable

consumers divided by the number of residential customers.

PCDSM

?

DSM percentage share of total expenditures

PRICE

-

Natural log of the average price of electricity

GAS

+

Natural log ofthe price of natural gas

PCMFG

+

Manufacturing as a percentage of total employment

PC INC

+

Natural log ofthe average per capita income

COOLDD

+

Natural log of the annual cooling degree-days

DENSITY

-

Natural log of Pole miles per residential customer

PCPAPER

+

Percentage of all employment income from the pulp and paper industry

PCPMETAL

+

Percentage of all employment income from the primary metals industry

The estimated equation went through a number of tests similar to those described in the DSM investment model section above (Table VI.9). The possibility that coefficients varied over the years for all variables was tested with an F-test. The resulting F statistic of0.69 did not indicate that the slopes and intercepts were different

82

for each year. Additional F-tests were perfonned to determine if only slopes differed, or if only intercepts differed between years. The F-tests revealed that slopes did not appear to vary between years and that specifYing a fixed-effects model with separate annual intercepts was appropriate since the intercepts were statistically different from zero. Using the Breusch-Pagan test again resulted in the detection of heteroscedasticity. The heteroscedasticity was not corrected for, but HCV standard errors and t-tests were calculated.

Table VI.9 Demand for Electricity Model Hypothesis testing

What Was tested

Test and Test Statistic

Result

Varying slopes and

F-test

Not statistically significant

intercepts for each year

F statistic= 0.69

Varying slopes for each

F-test

year with one intercept

F statistic =0.75

Varying intercepts for

F-test

Statistically significant

each year

F statistic = 3. 7

Fixed-effects model used

Heteroscedasticity

Breusch-Pagan Test

Presence of

Chi-squared statistic =

heteroscedasticity

139

detected.

Not statistically significant

White's HCV estimated

83

Table IV.lO Parameter Estimates of the Demand for Electricity Equation 2

(N= 324 R = 0.73 log-likelihood= 136.6) Parameter Estimate Variable

Standard Error

t Statistic

CONSTANT

-3.821700

1.464000

-2.611

Y91

-0.019376

0.025790

-0.7512

Y92

0.070512

0.026360

2.675

Y93

0.032548

0.026490

1.229

PRICE

-0.634480

0.050450

-12.580

PCDSM

-0.030450

0.008292

-3.672

GAS

0.052589

0.055690

0.9443

PCMFG

1.036900

0.277900

3.731

PC INC

0.371980

0.147000

2.530

COOLDD

0.231960

0.019980

11.610

DENSITY

-0.028309

0.009531

-2.970

PCPAPER

2.502300

1.829000

1.368

PCPMETAL

4.393700

1.106000

3.973

This equation for the demand for electricity performed quite well. All of the variables had the expected sign and most of the explanatory variables had parameter estimates that were statistically different from zero. Close to 70% of the variance in the quantity demand is explained by the few variables included in the equation. What was

84

particularly heartening was that for the price of electricity, PRICE, the estimated ownprice elasticity was -.63, which was within the ranges estimated by earlier studies (Bohi 1981, 1984). The price elasticity should be interpreted as the weighted average for all customer classes. The estimated coefficient for per capita income, PCINC, was 0.37. This estimated income elasticity was also within the ranges of earlier studies. Again, a direct correspondence with the values estimated for one customer class was not possible. For this study, the demand of all customer classes is aggregated and therefore the price elasticity is a weighted average for all the customer classes. The estimated cross-price elasticity ofthe gas price, GAS, was 0.05, and had the expected positive sign. However, the estimated parameter was quite small and not statistically different from zero. This would seem to argue that inter-fuel competition between gas and electricity will not be greatly affected by marginal changes in the price of gas.

.

That the estimated elasticities had the correct sign and approximated those estimated in other studies in magnitude bolsters the results of this equation. The equation was estimated, primarily, to determine the impact ofDSM investment. The parameter estimate ofthe DSM variable, PCDSM, was -0.03. The coefficient was statistically significant and indicated that DSM expenditures were associated with reductions in electricity demand. However, DSM's impact on the demand for electricity was quite small as a one-point increase in DSM's share of total expenditures was only associated with a 0.03% decrease in electricity consumption.

85

The investment cost for each kWh saved can be estimated for the average utility. The average utility had $1.2 billion annual expenditures during the analysis period. A 1% share oftotal expenditure was equal to $12 million. This level ofDSM expenditures was associated with a 0.03% decrease in total electricity demand. As the average utility produced approximately 21.4 million MWh of electricity this was a decrease of6,850 MWh for every $12 million dollar spent on DSM. This represented DSM expenditures of approximately $1,750 per MWh reduction or $1.75 per kWh reduction. The relatively modest amount of energy savings that were associated with DSM spending may be cost-effective under certain fairly restrictive assumptions. 32 Other factors that may have obscured the effects ofDSM are spillover, load growth and service retention strategies, or federal and state energy efficiency standards. Spillover refers to many programs sponsored by utilities having impacts that are not just confined to their service territories. Utility advertising that promotes energyefficient products and energy use practices will also be seen and heard by customers of

32

Cost per kWh saved or "negawatt" are presented in the table below, using simple payback calculations that assume the electricity cost escalation and the utility discount rates are the same and only the DSM investment lifetime varies. Simple Pay-back Calculations On DSM Investments DSM Investment Life 10 15 20 30

Cost Per "negawatt" $0.175 $0.116 $0.087 $0.058

86

adjacent utilities. Other direct spillover effects would come through population movements. A person may have received an energy-efficient refrigerator while being the customer at one utility. After changing jobs and place ofresidence the person becomes a customer of a different utility and has transferred the DSM savings embodied in the refrigerator to the new utility. A utility with programs incorporating strategic load growth, off-peak load growth, and service retention strategies will have more electricity demanded from it relative to a utility without DSM programs. Strategic load growth strategies seek to promote electro-technologies and increase electricity's market share in specific end uses. Programs promoting energy-efficient air-conditioners to first-time buyers will only lead to increased demand for electricity. Off-peak load growth would be fostered through lower rates. Many programs promoting hot water or cold storage have been implemented using these rates. Programs that target load retention will cause the utility to lose less load relative to utilities that are not promoting such programs. Programs that target new purchases of electric water heaters will retain the water heating loads of customers that would have otherwise switched to gas. Energy-efficiency standards will result in greater impacts on electricity demand ofutilities that are situated in areas of higher growth. This is because increased customers and income will be associated with greater levels of new construction and purchases of new electricity appliances and equipment. New appliances and new buildings will also be more energy-efficient.

87

DSM should not be judged only by its energy impacts. DSM investments defer other investments in generation capacity. These impacts cannot be known since generation never built is never reported. However, DSM impacts the long-run construction of power plants and investments in transmission and distribution facilities is an area of further research. The level of manufacturing activity, PC.r.AFG, in the service territory had the expected positive sign. The estimated parameter of 1.03 indicated that for every onepoint increase manufacturing has in the share oftotal employment there an associated 1% increase in electricity consumption. The type of industrial activity also played a measurable role in electricity demand. Both the paper and pulp industry, PCP APER, (2.5) and primary metals industry, PCPMETAL, (4.4) as measured by the share in total employment income, had the expected positive parameter signs. These two industries are highly energy intensive (Kahane and Squitieri, 1987), and greater levels of industrial activity in theses areas are expected to increase electricity demand. However, in the case the of the paper industry the parameter was not statistically different from zero. For the paper industry ,a one-point increase in the share of total employment income resulted in an increase in electricity demand of2.5%. The primary metals industry had an even stronger relationship, as electricity demand to increased by 4.4% for each onepoint increase in the share of employment income.

88

The cooling degree-day variable, 33 COOLDD, with the parameter estimate 0.23, was associated with greater levels of electricity consumption. Weather conditions were expected to influence the demand for electricity with cooling degree-days giving an indication of the cooling requirements in any given year. The estimated coefficient indicated that a utility with a service territory in a wanner climate with cooling degreedays 4% greater than average will be associated with almost a 1% increase in annual demand. A service territory density variable, DENSITY, was constructed by dividing the number of residential customers by the number of miles of transmission lines. Demand for electricity was expected to be negatively related to the density inter-fuel competition may increase with greater density given that gas service will be less available to customers in rural areas. Another possible reason for the negative relationship could be that there are less industrial and commercial customers in rural areas, which will also bring down the demand for electricity. The parameter estimate of -0.03 did confonn with this hypothesis and though the parameter was statistically different from zero the actual impact was quite small.

33

Cooling degree-days are a unit measuring the extent that the outdoor mean daily temperature falls above the base (in this case, 65 degrees). The cooling days were population weighted by state, and weighted further by utility sales in each state. 89

IV.S. Electricity Average Cost Equation

DSM has been promoted as a least-cost alternative to adding capacity in both generation plant and transmission and distribution facilities. On the assumption that DSM investment is a lower-cost alternative to other investments in other factors of production, one should expect that utilities that are investing in DSM resources would have lower costs associated with production of electricity. To determine the magnitude ofDSM's impact on cost an average cost equation was estimated with the general form:

Cost

=

j (DSM, Quantity, Financial, Managerial, Operating, Regulatory)

As DSM investment is affected by a variety of factors, a more general form of the cost function is specified. The dependent variable, average cost, was calculated by dividing the total reported electric utility expenditures by the total number of kWh sold to final customers and on the wholesale market. This variable and is entered into the equation as a natural logarithm. The tests that were performed on the model are shown in Table VI.7. F-tests indicated that slope parameters and the intercept did not vary in a statistically significant fashion in for all of the four years. The Breusch-Pagan test indicated that heteroscedasticity was present. As with the earlier models instead of correcting for heteroscedasticity HCV standard errors were calculated, as were t-tests based on them.

90

Table IV.ll Average Cost of Electricity Equation Variables Variable

Expected

Name

Sign

AVGCOST

Variable Label

Dependent Natural log ofthe average kWh cost Variable

QUANTITY

-

Natural log of the total quantity electricity sold to final consumers divided by number of residential customers

PCPURCH

+

Purchased power as percent of total electricity sold.

PCDSM

-

DSM as percent oftotal expenditures.

INVKWH

+

Natural log ofthe investment in plant per kWh generated or purchased.

PC NUKE

+

Percent of generated and purchased power that is generated by nuclear power plants.

FUELKWH

+

Natural log ofthe fuel costs per kWh generated.

PCTAX

+

Taxes as a percentage oftotal expenditures.

PCCWIP

+

CWIP as a percent of total electric plant investment.

PLANT

-

Natural log ofthe average age ofutility plant.

RES KWH

-

Natural log ofthe average annual consumption of a residential customer.

WAGES

+

Natural log ofthe average utility wage.

PCGASREV

-

Total gas operation revenues as a percent oftotal utility revenues.

PCRESALE

-

Total electricity resales as a percent oftotal electricity sold.

91

Table VI.12 Demand for Electricity Model Hypothesis Testing

What Was tested

Test and Test Statistic

Result

Varying slopes and

F-test

Not statistically significant

intercepts for each year

F statistic= 0.69

Varying slopes for each

F-test

year

F statistic= 0.72

Varying intercepts for each

F-test

year

F statistic= 0. 74

Heteroscedasticity

Breusch-Pagan Test

Presence of

Chi-squared statistic =

Heteroscedasticity

40.5

detected.

Not statistically significant

Not statistically significant

White's HCV estimated

92

Table IV.13 Parameter Estim~tes olfthe Average Cost of Electricity Equation (N= 324 R2 = 0.88 log-likelihood 304.3)

Variable

',Parameter Estimate

Standard

t Statistic

Error INTERCEPT

-i.443700

0.320400

-7.627000

QUANTITY

-<).311190

0.024410

-12.750000

PCPURCH

0.308000

0.041060

7.501000

PCDSM

-0.003:209

0.003301

-0.972100

INVKWH

0.1245i50

0.027550

4.520000

PCNUKE

0.004000

0.000445

8.986000

FUELKWH

0.354650

0.023940

14.810000

PCTAX

0.003910

0.001338

2.922000

PCCWIP

(1.248160

0.087630

2.832000

PLANT

~0.176880

0.028050

-6.307000

RES KWH

~0.197410

0.030950

-6.379000

0.024180

0.977300

WAGES

. t).023t526

PCGASREV

~0.153350

0.050010

-3.066000

PCRESALE

~0.270120

0.048620

-5.555000

93

The over~ll md,del predicts average cost of each kWh sold very well with an R 2 of0.88. Also, th~ estimated coefficients had in all cases the expected sign. The DSM parameter

estim~te

of:-.003, had expected negative sign indicating that as DSM's share

oftotal expenditures rose the cost of electricity production went down. The parameter estimate is unfortunat(~ly not statistically different from zero. However, as a point estimate the paq:1mete1r is the best available indicator of the impact ofDSM and as such predicts a only a, very minor decrease in average costs. A one-point increase in DSM's share of total expenditures was predicted to reduce average cost by three thousandth of one percent. Ev~n ifQSM expenditures became quite substantial the absolute impact · on the average c;osts c~fproduction would be quite smal134 . That DSM exj?enditure impacts on average cost were negative rather than positive offered some support for the argument DSM investments are least-cost alternatives to

~upply·,~side

investments. From the demand equation estimated above,

DSM investments res.ulted in reductions in quantity demanded. All other things being equal, a reduction in demand would be associated with a increase in average cost and not a decrease. It seems that in spite of the DSM induced reductions in demand that would tend to increase average costs, DSM expenditures were concurrently associated with reduction~ in av,erage costs.

DSM expenqiture~ averaged 1.4% of total annual expenditures. This was the equivalent of$1.8 billion (1990$) that was associated with a 0.003% reduction in cost or the equivalent of \lpproximately $6 million (1990$).

34

94

The regression results are actually supportive ofDSM since DSM impacts on cost appeared to be negative or in the worst case negligible. This indicated that utilities could use DSM as a customer service or as a tool to support market growth or market retention without great concern that it will lead to increases in cost and rate increases. Three variables that describe the output levels of a utility were included in the model as explanatory variables. The total quantity sold to final consumers, weighted by the number of residential customers, QUANTITY, was -0.3 and had the expected negative sign indicating that as quantity sold increased the average cost per kWh decreased. A ten percent increase in quantity demanded was associated with a decrease in price of about 3%. It was assumed that the average residential demand, RESKWH, would also be negatively correlated since many ofthe costs of transmission and distribution and administration are associated with the residential customer class. This variable proved to be highly significant and did not have problems of multicollinearity with the variable, QUANTITY. The estimated coefficient indicated that a ten percent increase in residential demand would cause a 2% decrease in average cost. The other electricity output variable was, PCRESALE, which was the percent of total electricity sold that was sold on the wholesale market. A negative relationship was assumed since wholesale power sales will make use of excess generating capacity and reduce average costs and also many of the costs associated with selling to the final consumers are not present. The parameter estimate of -0.27 indicated that a three point increase in resales was associated with close to a 1% decrease in average cost.

95

Inputs into the production of electricity were also included in the cost model. Capital is represented by, INVKWH, the ratio of total plant ilnvestment to electricity generated and

pu~chased.

The parameter estimate of0.12, had a positive sign indicating

that when the capital input per kWh went up by one point the average costs went up by a bit more than one-tenth of that. Fuel costs, FUELKWH wtere entered in the equation as the average dollar cost of fuel per megawatt hour generated. The estimated coefficient was 0.36 indicating that a 10% incre~se in fuel prices would be associated with an increase in electricity prices of3.6%. With a parameter estimate of0.023 for WAGES, the changes in the average cost of labor (wages, pension, and benefits), had little impact on average cost. If labor costs incr~ased by 10% average costs would only be expected to increase by 0.24%. The sources of electricity were also foupd to affect the average cost. Utilities that generated all of the electricity demanded appeared to have lower costs of electricity production than those that purchaseCI some portion of the electricity sold. The parameter estimate for purchased power, PCPURCH, t:>f0.31 indicated that is power purchases increased their share of total power sold l:ly ten points, the associated increase in average costs was 3.1%. 35 One of the reasons for this higher price for purchased power may be due to a short-run purchases that 1have a premium price associated with them. Another reason is that

p~rchased

po'fVer will already have profit

embedded in its cost. The share of total electric;ity generated by nuclear power was

35

There may be some question as to the direction of the causal relationship if utilities with higher power costs are more prone to pm·chasing power to reduce their costs. 96

expected to increase costs. The estimated parameter for the variable PCNUKE was 0.004 and indicated that nuclear power was positively correlated with the average cost. The impact of nuclear power appeared to be quite small since regression predicted a 0.4% increase in average cost if 100% of the electricity were generated at nuclear power plants. Financial and operating characteristics ofutilities also influenced the average cost of electricity. Taxes as share of total expenditures, PCTAX, with an estimated coefficient of0.004 was positively associated with cost. The total impact was quite small since a one-point increase in taxes' share was predicted to cause a four thousandths of a percent increase in average cost. New construction had a potentially large impact the average cost since the variable PCCWIP, the ratio ofCWIP to total plant investment, had a parameter estimate of0.24. CWIP represents presently unproductive capital investments which add to the cost of operations. These costs can be from a variety of sources such as increased debt payments36 , and inclusion ofCWIP costs in the general costs. The regression results estimated that a utility where CWIP represented 4% ofthe total electric plant average costs would increase by 1%. Negative relationships with average cost are estimated for the age of plant, PLANT

36

Long-tenn debt payments will increase to finance construction costs and short term debt payments may increase due to cash-flow problems that result from cash-flow problems from incurring the long-term debt. 97

(-0.18), and dual fuel utilities, PCGASREV. The regression results for the variable PLANT, indicated that utilities with older plants have lower costs. This could be due to a variety of factors. The most probable reason for this negative impact was that capital investments of older plants had a greater percentage of their value amortized and the capital investments in the plant no longer entered into the rate base resulting in lower costs. The presence of older plant possibly was also an indication of better operations and maintenance procedures that allow the utility to operate plant longer than other utilities. Putting off the plant retirement in tum allows the utility to put of construction in replacement plants and thus reduce costs. Increased plant age could also indicate that the utility optimized the running of the plants in such a way that load factor is increased again pushing investment in new construction into the future and reducing overall capital costs. Another factor was that older plant may be actually cheaper to run than newer facilities. Older plants may have been grandfathered the right to bum certain cheaper fuels, such as higher sulfur coal, or were not required to comply with certain environmental protection requirements. Dual fuel utilities also appeared to have a cost advantage over those that supply only electricity. The variable, PCGASREV, was the percent of total utility revenue is the result of gas operations and had a parameter estimate of -0.15. The estimated equation predicted that a utility that obtained 10% of its revenue came from gas operations would have average electricity costs that were 1.5% lower than a utility that obtained 100% of its revenue from electricity sales. This should be expected since gas

98

and electric operations will share many costs thus lowering overall average costs to both electric and gas services. An area where high savings can be expected is the administration, where billing and meter reading can be combined, and many administrative tasks overlap.

99

IV.6. Simultaneous Estimation of Demand and Cost

Discussed earlier was the possibility that cost and quantity are both simultaneously determined. Due to the simultaneous nature of the quantity, average cost and price the assumption that the dependent variables are fixed in repeated sample is violated. This causes the OLS estimators to be biased. To correct for possible bias the following set of simultaneous equations were estimated:

Cost

=j (DSM, Quantity Financial, Managerial, Operating, Regulatory)

Quantity

= k (DSM, Price, Fuel, Customer, Weather)

Price

=r

(Cost, Return on Investment)

Where the first two equations are described in the cost equation and demand equation sections above. The third equation was described previously as an equality 37 that connected the cost and demand equations together. It was however, specified as an equation above because the demand being modeled is the demand of final consumers, the price that is entering the equation is the average price that they are faced with while the average cost of kWh sold applies to kWh sold to final consumers and kWh resold to other utilities. The costs associated with kWh sold to final consumers and kWh sold to the wholesale market could not be separated. Average sales to other utilities

37

P, • Quantity

= Cost+ Return on Investment 100

comprised an average of 17% of utility sales during the analysis period. Costs associated with these sales were therefore quite substantial and had to be taken into account when estimating the average costs. The price equation was included in the model to connect the average cost and demand equations. A specification test developed by Hausman (Hausman, 1978) is one method to detect the presence ofsimultaneity. 38 The Hausman test was performed on each ofthe equations. In respect to the tests performed on the cost and quantity equations the residuals that were entered into the test equations had parameter estimates that were statistically different from zero, while the cost residual entered into the price equation was not statistically significant. 39 The tests indicated that average price was simultaneous with quantity and quantity was simultaneous with average cost. 40 To correct for simultaneity a variety of methods have been developed. The method used below is the full information maximum likelihood method (FllvfL). It

38

The test relies on the fact that the endogenous terms are correlated with the error terms. It is a two stage procedure. The first stage estimates the reduced form equations for the price, quantity and cost variables. The residuals ofthese regressions are then entered into the appropriate single equation and the equations are estimated individually. Evidence of simultaneity results in estimation statistically significant coefficients for the residuals. 39 For the quantity equation the price residual had at statistic of 4.8. For the cost equation the quantity variable had at statistic of2.2. In the price equation the cost residual had at statistic of 0.16. 40 A likelihood ratio test was also performed by calculating the test statistic -2*(LR -LtJR). LR is the log-likelihood ofthe restricted model and LUR is the loglikelihood of the unrestricted model. The statistic is distributed in a chi squared with degrees offreedom equal to the number of restrictions. This was calculated as -2(763(136+304+323))=54. Since the chi squared distribution at 1% significance level and with 2 degrees of freedom is 9.21 the likelihood ratio test indicated that that there were statistically significant differences between the restricted and unrestricted model. 101

should also be noted when estimating systems of equations simultaneously the parameters become more sensitive to specification error. 41 This is because specification error in one equation can now affect the parameter estimates of all equations. The three equations were then estimated simultaneously, after specifying PRICE, COST and QUANTITY as endogenous variables. The results of the regression are shown on Table IV.l2.

41

Pindyck and Rubenfeld, 1991. 102

Table IV.12 Price and Quantity Model Full Infonnation Maximum Likelihood (Demand R 2 = 0.71, Cost R 2 = 0.86, Price R 2 =0.89log-likelihood Variable

= 736)

Parameter

Asymptotic

Asymptotic

Estimate

Standard Error

"T' Ratio

DEMAND INTERCEPT

-4.589499

1.11818

-4.10

Y91

-0.031724

0.02277

-1.39

Y92

0.066024

0.02649

2.49

Y93

-0.00226246

0:02807

-0.08

PRICE

-0.529159

0.06433

-8.23

PC INC

0.528366

0.11011

4.80

PCDSM

-0.028933

0.0084108

-3.44

PCPAPER

2.571696

1.11153

2.31

PCP.METAL

5.285998

1.46055

3.62

PCMFG

0.953801

0.36484

2.61

COOLDD

0.225914

0.02837

7.96

GAS

-0.155696

0.06865

-2.27

DENSITY

-0.041138

0.01295

-3.18

103

Table IV.l2 (Continued) Price and Quantity Model Full Infonnation Maximum Likelihood (Demand R 2 = 0.71, Cost R2 = 0.86, Price R2 =0.89)

Variable

Parameter

Asymptotic

Asymptotic

Estimate

Standard Error

"T' Ratio

COST

INTERCEPT

-1.967013

0.30372

-6.48

QUANTITY PCPURCH

-0.457192

0.06365

-7.18

0.247021

0.04702

5.25

PCDSM

-0.010453

0.0050595

-2.07

INVKWH

0.126946

0.02958

4.29

PCNUKE

0.00373310

0.0004410

8.46

FUELKWH PCTAX

0.324953

0.02070

15.70

0.00288132

0.0012720

2.27

PCCWIP

0.237657

0.12637

1.88

PLANT

-0.145528

0.02440

-5.96

RES KWH

-0.157962

0.03501

-4.51

WAGES

0.017129

0.01811

0.95

PCGASREV

-0.072485

0.05163

-1.40

PCRESALE

-0.078066

0.08416

-0.93

INTERCEPT

0.345067

0.08436

4.09

COST

0.886778

0.02233

39.72

ROI 42

0.090960

0.01224

7.43

PRICE

~2

Average income per kWh sold. 104

In general the results of the FIML were similar to those that were estimated for the equations separately. Given that FIML method simultaneously estimated the set of equations and corrects for bias introduced due to simultaneity, the results of the FIML estimation were the preferred results. The signs and the relative magnitudes of the estimated coefficients were for the most part the same. Discussion of the FIML results therefore primarily targeted those estimates where the results differed significantly. In the demand equation the impact ofDSM expenditures, PCDSM, is still as

weak as before with a parameter estimate of -0.03. The estimated price and income elasticities, respectively -.053, and 0.52, changed slightly though were still within the ranges found in earlier studies. One parameter value that changed significantly was the price of gas, GAS. The cross-price elasticity of gas was estimated at -0.15 that was negative and statistically different from zero. This coefficient indicated that an increase in the price of natural gas would bring about a decrease in demand for electricity. This result is not reasonable since a rise in price of a competing fuel should be associated with a reduction in the demand for electricity. The cost equation did not have any variables that changed the direction of their impact. The most significant change was that the estimated PCDSM coefficient, -.01, became statistically significant and had increased its impact by a factor of three. The impact was still extremely small relative to the total spending on DSM and average cost

105

of producing electricity. 43 Two other variables that had their changed parameter estimates with the FIML estimation were the dual fuel utility variable, PCGASREV, and the utility's wholesale activity variable, PCRESALE. The dual fuel utility coefficient, -0.07, was no longer statistically different from zero and the actual impact on average cost was much smaller. With the new coefficient a utility that obtained 10% of its revenues from gas would only experience a 0.7% reduction in the average cost of electricity compared to the 1.6% reduction measured in the single equation regression. The large reductions to average costs that had been associated greater levels of wholesale activity were also no longer estimated with the simultaneous model. The estimated coefficient for the share of resales was -0.08 compared to the -0.27 estimated in the single equation regression and was not statistically different from zeros. This indicated that increased bulk power sales did not lead to substantial decreases in average costs. Possibly bulk power sales are made at the short-run marginal cost and primarily cover only variable and not fixed costs. The price equation performed quite well as was expected. Average costs and average income were assumed to be the primary components of the average price paid by retail customers. The parameter estimate of average cost, COST, was 0.89 and highly significant indicating that every one percent increase in cost led to a 0.89% increase in rate. The return on income parameter, ROI, was 0.09 and indicated that a

43

Even increasing the associated reduction in average cost savings by a factor of 3 would only result in average total annual cost savings of approximately S18 million ( 1990$) compared to $1.8 billion (1990$) in average annual total DSM spending for the sample (see footnote 30). 106

1% increase in the investment income led to a 0.1% increase in price paid by final consumers.

107

V. Conclusion

The results of this study indicated that it was possible to measure the DSM system level impacts. During the period analyzed DSM investments constituted a large enough share oftotal expenditures and investments of investor owned utilities as to allow their impacts to be measurable. If future DSM spending on programs continues at its present rate, this type of an evaluation can be useful in providing national DSM impacts. Even with immanent industry restructuring, there are a sufficient number of plausible scenarios that include continued spending on DSM. The estimated DSM investment model indicated that the utility's regulatory environment had a major impact on the variances in DSM investment. An above average regulatory climate and least-cost planning requirements were found to have large impacts on the level ofDSM investment. As a favorable regulatory climate was associated with higher DSM spending, it was reasonable to assume that profit incentives, such as increases in the rates of return, allowing DSM investments into the rate base, or allowing utilities to recoup lost revenues were a PUC's methods for promoting DSM. The argument that PUC profit incentives promoted DSM investments was bolstered by the rate of return being positively correlated with DSM investments. Higher rates of return may have been due to the PUC profit incentives. Another important regulatory factor proved to be PUC-mandated least-cost planning requirements. Utilities must consider DSM as a viable alternative to supply-

108

side electricity generation to meet future electricity demand when using least-cost planning. In theory, this should allow DSM resources to be placed on a level play:ing field with supply-side resources. Forcing a utility to consider DSM in its resource planning raises the probability that DSM will be included in the mix of utility resource investments. Some indications exist that capacity-constrained utilities inves~ed more in1DSM than did others. The model also indicated that power purchases were possibly substituted with DSM resources. Higher incomes and greater resident.ial consumption of electricity were also associated with increased DSM spending. The positive relationship with income could indicate that the income elasticity for ~he demand for DSM is such that demand for DSM services rises with income.

Great~r

DSM

investments also occurred when average residential electricity consumption is higher. With greater electricity consumption, customers may have a greater d.emand for

1

services to reduce energy expenditures that will increase demand for PSM programs. Also greater consumption will increase the average customer energy ~;avings potential that will also increase the ability for a utility to offer cost-effective to DSM services. The estimated DSM investment model explained 40% of the variance in IDSM expenditures. Many other variables were entered into the equation, bl,Jt were not found to be correlated with DSM expenditures. However, many of these variables, sudt as those used to represent capacity constraints, may have been too general. FurtherJ using only dummy variables to describe the regulatory environment grossly simplifies tlhe

109

relationships between the utilities and one or more PUCs. A model using more detailed information !ihould improve the estimation. The estimated model indicates areas, such as the regulatory environment and utility capacity requirements, where more research may be usefi.ll. The direction ofDSM impacts on average costs and quantities demanded were not unexpected. A negative correlation between DSM investments and quantity actually provided "negawatt" cost estimates, that under certain assumptions were in the vicinity of actual prices of electricity. However, estimated reductions in demand were only a fraction of the reported expected energy savings. It is possible that the demand equation regression underestimated the impact of

DSM investments. DSM programs have spillover effects into other utilities' service territories. DSM program participants may move the energy-efficient equipment sponsored through the program into the service territories of other utilities. Utility programs that bring about market transformations also have a spillover effects as it is not only the market in their service territory that is being transformed. Many DSM programs also result in 111et load growth. By targeting load retention, utilities investing in DSM prevent reducti1ons in quantity demanded relative to utilities that do not pursue such a strat,egy. TherefQre, the reported energy savings will be masked when energy savings are associated with new and retained loads. The estimated average cost equation did verify that DSM's impacts on average costs was nelatively small and negative. To put this impact in perspective it should be

110

considered that short-run average cost impacts of energy conservation programs resulting in lost sales give an upward push to average costs. Even with these reductions in demand, DSM impacts on average costs were negative. This indicates that, on average over the four-year period, DSM has tended to reduce costs relative to utilities not instituting DSM programs. It is possible that even within the this study's time frame, DSM has deferred capacity and thereby lowered capital costs. The estimated cost of equity capital equation results were quite surprising. Initially, DSM's theorized impact on the M/B ratio was without an expected sign and potentially could not have been correlated with the utilities' stock prices. DSM expenditures were found to have a positive, statistically significant, and substantial impact on the cost of equity capital. The model predicted that spending 5% oftotal expenditures on DSM would be associated with a 10 point increase in the MfB ratio. This represents an utility stock valuation, approximately 7% greater than that predicted if the utility had not made DSM investments. Part ofthis increase could be due to DSM

deferring investments in capacity, as utilities with older plant and equipment, and less current new construction were also associated with higher stock prices. Part ofDSM's impact may also be due to its acting as a proxy for other factors, such as improved customer relations and services or better operations and maintenance procedures, which should be associated with higher stock prices. The DSM investment model indicated that a large factor influencing DSM spending levels was the regulatory environment. As the industry restructures and

111

becomes more competitive, less regulatory involvement could mean reductions in DSM investments. The demand, cost, and equity capital models all indicate that utilities may continue offering DSM services after the industry's restructuring. DSM investments were not associated with large reductions in demand, which is an anathema for utilities. DSM was also associated with decreases, not increases, in average costs. IfDSM programs were really not cost-effective by not deferring expensive capacity and purchasing "negawatts" at inflated prices, an increase in electricity costs should have been measured. The cost of capital equation produced results that present DSM's future in a favorable light. DSM spending is associated with reduced capital costs. Thus, DSM investments may continue after the electric industry restructures because these investments can decrease the cost of equity capital. As the market places a premium on DSM investments during regulation, this favorable valuation of DSM may remain after the industry restructures. It appears that regulator "driven" DSM constitutes a large part ofDSM

investment. If regulators and other interested parties do not wish DSM to decline after the industry restructures, provisions for DSM funding must be made, either through line charges, taxes, or other forms of funding. The findings also have implications for DSM practitioners in that the large and possibly excessive savings attributed to DSM are not in evidence. These savings may have been masked by programs that expand demand or retain loads. These types of programs will increase net demand regardless of

112

how energy-efficient the new and re:tained load is. Savings from energy efficiency should not be equated with reductions in total load. The study shows that utiliti~ investing in DSM may be following an appropriate corporate strategy that 1rewards its shareholders. The estimated demand equation did not predi•~ large revenue losses whilt~ the average: cost equation associated DSM inves~ments with ~ small decreas'e in average costs. Given these small impacts DSM has on ~1oth electricity cost and elec:tricity demand, DSM investments could be pursued as a method to in~rease stock prices. In the future, DSM services may be used as a way to di.fferentiate one utility from its competitOJrs. The years 1990 through 1993 were the years used for this analysis. However, the future ofDSM is no longer what it was then. Industry widle restructuring is seen by many as the death kn~U of utility fi1r1anced DSM. Retail wheeli1r1g will e11able customers to choose and change electricity s~rvice providem. Without captive customers electric utilities will have no QlCentive to invest in demand-side programs. These DSM investments will be at risk if a customer were to c:hange electriicity providers. This does not mean that DSM sl!rvices will

n~ot

be offered by utilities. If utilities were to charge

the customer receivin~ the DSM services directly for this service DSM would be less risky. Retail-level cornpetition may in fact put ele:ctricity servi•:e providers under greater pressure to pmvide quality DSM services that customers are willing to pay for. Retail wheelir1g will bring about greater price competi1tion between energy service providers. TNs will eventUally cause the Jprofit margin in selling electricity to

113

narrow. Energy services that include DSM will be used by utilities to compete for customers and to

dev~lop

areas business with greater profitability. Retail wheeling may

also cause the consuJrJwr to chang\C their patterns of energy use ifTOU pricing is adopted. The consuml'!r will obtain the correct price signal from which to determine their consumption. These signals (may induce the consumer to invest in energy efficient appliances, energy elfflcient homes or home improvements, or obtain DSM services from the energy provider. Another possible scenariO! that includes DSM programs is a continuation and expansion of federal ~d state go•,vernment funded DSM programs. PUCs could. mandate or provide ir1centives fotr local distribution companies to provide DSM services if these sen~~es are not offered by actual electricity service providers. Such programs could be financed using a transmission charge.« From the mo~lels estimated in this analysis, it seems inevitable that factors influencing DSM invt;stment patterns will change in the future. Industry restructuring will include deregulation,

tendin~~

to reduce DSM spending, since regulatory factors

were found to influcmce investm(tnts in DSM. However, DSM was not found to induce large revenue losses pr increase costs. These results, combined with the finding that DSM spending may ~ncrease stoiCk prices, indicates that utility DSM spending will probably continue in some form or another in the deregulated future.

-----

-------------·-~.

. _. Line charges, billte~ as "distribl:ution benefits charges", are promoted to fund energy efficiency, renewabl~ energy so~rces and assistance to low-income customers. They exist in 3 states and life being pr1oposed in 8 others (E News, January/February 1996, pp. 2). 114

VI. Future Research

This study reveals several important areas offurther nesearch. One ofthe.se is , the impact ofDSM on utility investments on new capacity investments. DSM investments were made, in part, to defer investments in the e
1

on deferred capacity can be undertaken. Estimating DSM investment impacts at the customer class level would a.llow for better specification of demand equations. DSM impacts may also vary depending on 1 the customer class. Disaggregating DSM investments to the level where1 specifi<; types of expenditures can be groups will enable the impacts of conservation and load management programs to be measured separately. This will remove some of th~ ambiguity in the expected direction ofDSM impacts, as conservation pr:ograms are expected to reduce demand, while load management programs will primarily sh.ift electricity demand to another period.

1115

DSM interactions with national and local legislation on energy efficiency remain to be measured. The National Appliance Energy Conservation Act 1987 and the New Energy Act of 1992 should have substantial impacts on energy consumption over time. How DSM programs change in the wake of industry restructuring remains another rich area for future research.

116

Appendix 1: Data Sources Variables Used to Estimate Model Variable Description

Variable

Source

Name Financial and Managerial

ROR

Rate of return on investment

PCTAX.

Taxes as a percentage oftotal expenditures. Sum FERCForm 1

Value Line

of total local, state, federal, and other taxes divided by total electric utility expenditures. PCCWIP

CWIP as a percent of total electric plant

FERC Form 1

investment. Total CWIP divided by total electric plant. WAGES

Average utility wage. The sum of total wages,

FERC Form 1

benefits, and pensions divided sum offull-time employees and half of the part-time employees. PCGASREV

Total gas operation revenues as a percent of

FERC Form 1

total utility revenues. Total gas revenue divided by total utility revenue. DIVIDEND

Dividend pay-out ratio

Value Line

EQDBRAT

The equity-debt ratio. The mean of end of year

FERC Form 1

and beginning ofyear total proprietary capital minus the mean value of end of year and beginning ofyear preferred stock divided by end ofyear and beginning of year long term debt.

117

Variable

Variable Description

Source

Name MBRATIO

Market-to-book ratio, the market share price

Value Line

divided by the book value per share. Annual earning per share multiplied by the average annual price earning ration divided by the book value per share. Operational PCIND

Percentage of total sales to industrial customers.

FERCForm 1

Total MWh sold to industrial customers divided by total MWh sold to final consumers and for resale. PCDSM

DSM percentage share oftotal expenditures.

EIA Form 861

Total DSM expenditures divided by total electric and FERC PRICE

company operating expenses.

Form 1

Average price of eiectricity. Total revenues from

FERC Form 1

final consumers divided by the total kWh sales to final consumers. PCPURCH

Percent of total power (purchased and

FERC Form 1

generated) that was purchased. Total power purchased divided by sum of total power purchased and generated. LOAD

The annual load factor (ratio of average load

Value Line

divided by peak load) COOLDD

Annual cooling degree-days.

BLS

118

Variable

Variable Description

I

Source

Name

DENSITY

Polle miles per residential customer. Total pole milc~s

FERCForm 1

divided by the total residential customers.

Av2illable for 1990 and 1991. Value from 1991 used for .1992 and 1993. QUANTITY

Total quantity electricity sold to final consumers

FERC Form 1

divided by number of residential customers INVKWH

Investment in plant per kWh generated or

FERC Form 1

purc::hased. Total electric plant divided by total kWh sold to final consumers and for resale. PCNUKE

Percent of generated and purchased power that

FERC Form 1

is gt~nerated by nuclear power plants. Total electricity generated through electric power divided by sum oftotal generated and purchased. FUELKWH

Fuel costs per kWh generated. Total fuel costs

FERC Form 1

divided by kWh generated. PLANT

Average age of utility plant by holding company.

Value Ljne

PCRESALE

Total electricity resales as a percent of total

FERC Fprm 1

electricity sold to final consumers and for resale. Customer

I

PC INC

Per capita income by state and year.

BEAREIS

RES KWH

Average MWh consumption per residential

FERC Fprm 1

customer. Total MWh sold to residential customers divided by number of residential customer.

119

Variable

Variable Description

Source

Name GAS

Price of natural gas by state and year.

Gas Facts

Yearbook PCMFG

Manufacturing as a percentage oftotal

BEAREIS

employment by state and year. PCPAPER

Percentage of all employment income from the

BEAREIS

pulp and paper industry by state and year.

PCPlVIETAL

Percentage of all employment income from the

BEAREIS

primary metal by state and year..

Reguiatory LCUP

Dummy variable indicating least-cost planning

Mitchell

required within the firm's service territory by

(1989, 1992)

state for 1989 and 1992. 1 =Required 0 =Not required REGOOD

Dummy Variable indicating the regulatory

Value Line

environment is rated above average by state and year. 1 = Above average 0 =Average or below average

120

Data :Sources 1. Fin:ancial Statistics oflnvestor-Owned Electric Utilities (1989-1993).

Data on electric utility financial and operating variables, contained on FERC Fonn I, were obtained from the EIA. This included a total of900 variables or approximately 45% of the data contained on the fonn. Practically all of the data contained in the annual EIA publication, Financial Statistics of Selected Investor-Owned Electric Utilities, for the years 1989-1993 was included in this database. Included in this data were the distribution of hydropower, nuclear, steam, and other power generation, fuel costs, total costs of power purchases and sales, residential, commercial, and industrial consumption, revenue by type of customer class, electricity sold by customer class, labor costs, construction work in progress (CWIP), taxes paid, environmental protection investment and costs, and rate of return.

4. Value Line {1988-1994) Electric utility financial characteristics such as book value per share, dividend pay-out percentage, percentage of AFUDC in profits, rate of return, price earning ratio, average earnings, regulatory environment, plant age, load factor, capacity, merger activity. ~l.

Moody's Public Utilities ( 1985). Electric utility merger activity, ownership percentages, and bond ratings.

4. EIA Form 861 (1990-1993) This data provided information on annual DSM expenditures and expected savings. The distribution of sales by state was also available for the year 1993. The 1993 distribution was used to represent that of the previous three years. 5.

R(~gional

Economic Information System (1989-1993)

The Bureau of Economic Analysis maintains this database which is available on the Internet. This database contains wage and employment data at the state level for the year 1969 through 1993. Employment and wage data is available at the two digit SIC code level.

121

6. Upited States Oommerce Department Web Site Regional annual temperature data was obtained from the Commerce Department Web Site. This was for each of the 9 regions and had annual data. An averag~~ was used for 1993. 7.

Gf1S Facts Yearlbook (1989-1993}, American Gas Association Gas prices by state.

8. Bureau ofLabor Statistics Web Site Producer ailld regional CPis and Weather data were obtained from this Web site. 9. Mitchell (1989,11992) and Chamberlin, Fry. and Braithwait (1988)

Least-cost planning by state. 10.

~-lousing

Characteristics 1990, Energy Information Agency (1992)

Electric heat penetration rate. 11. Statistical Yea:rbook of the Electric Utility Industry,

Institute

,(annu~).

Edison Electric

I

Total electric sales by state.

122

Appendix Q: A Definition of Demand-Side Management

DSM ranges over a large set of policies and programs and has not been perfectly defined. The se~ of strategies pursued through these programs is shown on Figure 1. In general DSM can be separated into two major categories; loadmanagement and conservation. These are discussed separately below. Load-management

The typical electric utililty does not generate a constant amount of electricity. Demand varies by time of day ~o time of year (Figure 2, 3, and 4). Most utilities in the south are summer peaking due 1to cooling demands while areas such as the Pacific Northwest are winter p~g due to large heating loads. Utilities must by law meet demand and for this reaspn wiUI have excess generating capacity and an under used distribution and transmis~ion system so as to be able to meet the peak demands on the system. As a peak perio~ progtesses more and more ofthe available capacity is used. The additional power thEj.t is g~nerated is more costly since less efficient plant and equipment will be brought into1 use. Due to the practice of average cost pricing the consumer does not obtai.n the c:orrect price signal, which is the marginal cost of production. Therefore at peak periods customers are paying below marginal costs and during off-peak periods the price is above the actual marginal cost. Large difference~ between peak and off-peak periods will generally lead to higher than needed rates due t
demand charges). Direct load ~.ontrot Control of q.~stomer appliances. Energy stor&ge. Fuel switchi~1g. 1

2. 3. 4. 5.

TOU rates, in effect, are treating electricity demanded over a daily period and over seasonal periods 3.$ diffe~ent products for which a different price is charged in each period due to differences: in generation, distribution, and transmission costs. In 123

PE.AK CLIPPIN?· or the re
Tune

V ti1Ll.EY AWHG is the second du.sic form of load lllaiQgement..

Valley filling enc:ompuses building otf-pe2k lo3cls. This m::ay be particularfy desir:~bte where the long-nm ~tat cost is tess than thee average price of electricity. Adding property priced off-peak load under those c:in:umst:ances decreases the average price. Vattey filling ca'n be accompfr.shed in sevem wa-p.. one ot the most popular of which is new thermal energy storage (water ~ting and/or space heating) that cfu.places lo:ads served by fo:uil fuels..

TIITie

LCIAO SHIFTING is the last clas..sic1onn of load management. This in"olves shifting load froro on-peak to oil-peak periods. Popular applications include use of storage water heating. SIOfCige space heating, coolness storage. and customer load shills.. In t11is case. the load sllil! from storage devices involves di~placing what would have been conventional appliance~ served by elee1rici!y. ST:RA IEGIC CONSERVATION is the load shape change thai results fro1m utility-stimulated programs direc1ed at end use consumption. Not nonnally considered load management the change reflects
Time

~

S

Time

ST'R:.. TEGIC LOAD GROWTH is the load shape

-g .3

FLEXIBLE LOAD SHAPE is a COflcepl related to reliability, a planning constraint Once the anticipated load shape. including demand-side activilies. is forecast over ttle corporate planning horizon. the power supply planner studies the final optimum supply-side options. Among lh1e many criteria he uses is reliability. Load shape can be flexible - if customerS' are presented with options as to the variations in quality of service that they are willing to allow in e.:cllange for various incen~ lives. The programs involved can be variatio~ of interruptible or cur- 0 tailable load: concepts of pooled. integrated energy management --' systems: or individual customer loa<;~ control devices oflering service constraints..

Time

Time

Figure 1 Demand-Side Management Strategies (Gellings and Chamberlin, 1988) 124

Total Industrial

u

cu 0

_J (l)

> cu Q)

Commercial

0:

Residential

00:00

. 06:00

12:00 18:00 Hour of Day

24:00

Figure2 Typical Summer Daily Load Curve (Gellings and Talukdar, 1986) 125

Sc:n.

Mon.

Tue.

Wed.

Thur.

;:-~.

Set.

Figure 3 Typical Weekly Load Curve Margins for US Utilities (Gellings and Talukdar, 1986)

Available capaory

Demand

Jan.

Apr.

Jul.

OcL

Figure 4 Monthly Peak Demand: Capacity and Operating Margins for US Utilities (Gellings and Talukdar, 1986) 126

theory TOU rates would be more efficient rates to charge since they would better reflect the marginal cost of the electricity demanded. In practice they have seen only limited use in the United States while in France and later in England their use has become more widespread. In the United States the rate structure prior to the 1980's was one in which declining block rates were prevalent. Increasing block rates and TOU rates were viewed as ways to constrict the demand for electricity, which would not have been in the utility's interest. Prior to the eighties, typical utility strategy to meet future demand had been the construction of more generation. The passing of the Public Utilities Regulation and Policy Act of 1978 encouraged the implementation of decreasing block and TOU rates, mainly in the form of a seasonal rate structure. There has been limited experiments in TOU pricing for most customer dasses but actual implementation of such rates has only been used for selected larger industrial customers. At present metering costs for smaller customers have not been conducive to adopting utility-wide TOU pricing. One could predict with the lowering of cost of automatic metering devices, through either the projected fiber optic networks or small radio controlled meters, that real time pricing will become a reality by the end of the decade. This event would reduce the need for utility DSM programs since consumers would be obtaining the correct price signals and then be able to adjust their consumption accordingly. Demand charges have been in use since the creation of the electric industry in the United States Their implementation is a two part tariff, usually referred to as the Hopkinson electricity tariff, where actual consumption of kWh and maximum instantaneous demand in kW are priced separately. Demand charges are based on the instantaneous demand in kW and have the desired effect of reflecting marginal costs associated with capacity usage. This is under the assumption the customer peak demand is coincident with system 'peak demand (Bonbright et al, 1988). At present demand charges apply mainly to industrial customers and to larger commercial users. Their peak kW demand is calculated either actual measurements of the customer demand or using customer dass averages. The additional four load-management program types have only seen limited use, though in many areas they should prove to be a cost effective alternative to additional generation. Direct load control applies mainly to industrial clients who have facilities to generate their own electricity. A peak clipping strategy that is employed is to contract with large customers the right not to supply them with electricity during the peak period. By offering them lower electric rates in exchange for lowering the quality of service utilities can achieve these service curtailments. These interruptable rates would be set through a bargaining process to be less than the normal rate, though how much less would depend on the lead time needed to notify the customer of service interruption, the total number of hours interruptions are allowable, and other factors. In a survey ofutilities in 1985 over 75% of the sampled utilities had interruptible rates (Cogan and Williams, 1987). In theory programs allowing 127

customers to buy power at a lower th~ normal reliability could be extended to the commercial and residential classes but as yet little has been done in this area. Some programs result in the u~ility hiaving direct control of customer appliances such as air conditioning sy~tems,1hot water heaters, or boilers. Again customers are offered power at n low,r ratel in exchange for losing some control over their appliances. In many cases the prQgrams are designed not to interfere greatly with the workings of the appliances. In the case of hot water heaters homes are targeted in which little hot water is used during tl1e day! The water heater is turned off by the utility during peak periods and then ~rned back on before the client returns home. With air-conditioning units the utility would1 use its control to let only a certain fraction of the units under its control cycle on for the same fraction of an hour. This would not necessarily reduce total corlSumption of electricity, but would remove the possibility of aU units cycling on at the same time; causing s micro-peak. Also, in the case of a potential black or brown out the appliances under direct control could be turned offreducing this potential problem. By 1990 these types of programs have been planned to increase the number c,fcontrolled appliances from the estimated 2.5 million in 1985 to a total of7 million t'Limaye and Rabl, 1988). DSM energy storage can be thought! of as storing heat or cold at the place of usage and not being related to pumpCI;i storage. In the case of heating, heat is generated during an off-peak period ~d stred in a solid storage such as bricks or a boiler. During the peak hours this stor·ed energy is used to heat or produce hot water. In the case of cooling, ice or cooled water are created in the off-peak period and used to cool a facility during the peak periqd. ObMously more electricity is consumed using this type of alternative and decreased rates must be used to encourage the installation of such facilities. Fuel switching is generally not, a preferred alternative to many utilities. In the case of an all electric utility, electric ~es will suffer if customers switch from electricity to gas, oil or biomass. Where the1electric reduction will only appear at peak periods a fuel switching program could be a1 cost effective alternative to supplying electricity from peaking facilities or h~ving to add additional capacity. Some utilities have instituted programs which foster use of dual fuel appliances where electricity is the primary fuel and the secondary fu~l is only used during peak periods. A possible problem that could arise from having custoriners switch from electricity to gas, is that gas supply problems could be created or aggravated.

128

Conservation Two program types which do not fall into the category of load-management are: 1. Conservation 2. Efficient Load Growth When added demand on the system is forecast, a utility should be indifferent between adding new generation and instituting conservation programs which would make the added demand unnecessary if they are granted an equal return on their investment. This would oo true if prices were allowed to change reflecting the loss in revenue resulting from the reduction in sales brought on by conservation programs. At present most such programs are looked upon as variable costs. Only slowly are PUCs allowing for conservation investments to go into the rate base and recompensing the utilities for their revenue shortfall. In the advent of retail wheeling investments in such programs will become more problematic since many of these investments will be viewed as stranded costs. These DSM investments are typically owned by the customer and not the utility. With customers being able to switch service providers, any investments made for this customer will not benefit the sponsoring utility. Existing conservation programs have targeted most customer classes and have a variable amount of direct utility involvement in conservation activities. The least direct involvement is the use of advertising by the utility to promote conservation and provide conservation information to consumers. A more direct form of utility promotion is the provision of utility sponsored energy audits with a varying degree of utility subsidization. Further programs offer installation of energy reducing features either through reduced interest loans or utility sponsored energy efficient appliance rebates. Another variant is buying electricity consumption reductions due to conservation activities from customers directly or from firms which provide conservation measures to customers. Other programs involve different types ofutility and industry partnerships where incentives are paid to dealers to promote and stock items, manufacturers to design and produce efficientf'smart" appliances/equipment, or construct buildings that are energy efficient or "smart". Recently the market transformation has become the goal of many utilities programs. Programs with the goal of transforming the market seek to make fundamental changes in the demand for energy efficiency services and energy efficient products. Successful programs can be phased out without causing the demand for the programs products and services to disappear. These programs are geared to change the buying habits of consumers, the stocking practices of dealers, and the product development and design practices of manufacturers. Programs such as these have made compact fluorescent light bulbs, high efficiency heat pumps and gas furnaces, motors, and appliances more available. It is only after the markets for these products 129

have b~n developed and consumer acceptance is high that government legislation of efficiency standards becomes possible. This in tum makes the program impacts perm~1ent. A clear example ofthls was the development ofthe energy efficient building cod€:s that are now in plaCe in Oregon and Washington. Utility programs that sponsqred changes in building practices transformed the residential new construction marke~ in suc:h a way that code changes could be passed without great resistance from the ho1ne building industry. Stratc~gic load growth can fall into both conservation and load shaping strategies. This strategy targets energy use in new or renovated buildings and facilitie,es, the replacement of old appliances, and new markets involving electrotechnologies. Programs encouraging energy efficient designs and the placement of energy effici€mt appliances by architects, construction companies, appliance dealers and ~ufae1turers, and contractors can fall into this category and can culminate in reduced futu1re energy use and the targeting load impacts of new or changed demands. Often limes the goal of these programs is to reduce fuel switching and retain market share i;n certaun lucrative areas or to develop a toehold in potential new markets. Many peatpump and water heater programs are geared to retain market share of an important pa:rt of a utilities load. Air conditioner programs are often times used to boost the penetration rate of not only energy efficient air conditioners but the air condit~oners in general.

130

Appendix ill: Hypothesis

T~ts

DSM Investment Equation

1. Regression with 8 explanatory variables, interaction tenns for each of the three years (24 variables), and different intercepts for each year (4 variables). ESS = 402 N ~ 324 2. Regression with 8 explanatory variables, interaction tenns for each of the three years (24 variables), and one intercept. ESS =408 N =324 3. Regression with 8 explanatory variables and different intercepts for each year (4 variables). ESS = 420 N = 324 4. Regression with 8 explanatory variables one intercept. ESS = 437 N = 324

F statistic 1. Regression 4 tested against Regression 1 0.92 2. Regression 4 tested against Regression 2 0.85

3. Regression 4 tested against Regression 3 3.89 Cost of Equity Capital Equation

1. Regression with 10 explanatory variables, interaction terms for each of the three years (30 variables), and different intercepts for each year (4 variables). ESS = 87,426 N = 324 2. Regression with 10 explanatory variables, interaction tenns for each of the three years (30 variables), and one intercept. ESS = 87,763 N = 324 3. Regression with 10 explanatory variables and different intercepts for each year (4 variables). ESS = 96,874 N = 324 4. Regression with 10 explanatory variables one intercept. ESS = 140,729 N = 324 F statistic

I. Regression 4 tested against Regression 1 2. Regression 4 tested against Regression 2 3. Regression 4 tested against Regression 3 4. Regression 3 tested against Regression 1

5.17 5.63 42.3 1. 0 1 131

Demand for Electricity Equation

1. Regression with 9 explanatory variables, interaction tenns for each of the three years (27 variables), and different intercepts for each year (4 variables). ESS

= 7.9 N = 324

2. Regression with 9 explanatory variables, interaction tenns for each of the three years (27 variables), and one intercept. ESS = 7.91 N = 324

3. Regression with 9 explanatory variables and different intercepts for each year (4 variables). ESS = 8.16 N = 324

4. Regression with 9 explanatory variables one intercept. ESS = 8.48 N = 324 F statistic

1. Regression 4 tested against Regression 1 0.70 2. Regression 4 tested against Regression 2 0. 75 3. Regression 4 tested against Regression 3 3. 71

Average Cost of Electricity Equation

1. Regression with 13 explanatory variables, interaction tenns for each of the three years (39 variables), and different intercepts for each year (4 variables) ESS = 2.614 N = 324 2. Regression with 13 explanatory variables, interaction tenns for each of the three years (39 variables), and one intercepts ESS = 2.622 N = 324 3. Regression with 13 explanatory variables and different intercepts for each year (4 variables) ESS = 2.875 N = 324 4. Regression with 13 explanatory variables one intercept ESS = 2.90 N = 324 F statistic

1. Regression 4 tested against Regression 1 0.70 2. Regression 4 tested against Regression 2 0.73 3. Regression 4 tested against Regression 3 0.75

132

Bibliography American Council For fUl Energy Efficient Economy, (1990), ACEEE Proceedings 1990. American Gas Association,(annual), Gas Facts Yearbook. Anderson, Dennis, et al, (1995),"Roundtable on Energy Efficiency and the Economists- Six Perspe~ctives1 and and Assessment," Annual Review of Energy and the Environment, 1995, val. 2,~0. pp. 493-573. Archer, Stephen H., (1981),"'1The Regulatory Effects of the Cost of Capital," Public Utilities Fortnightly, FebruSI]', 26, pp. 36-39. Atkinson, Scott and Robert HaliJerson, (1980), "A Test ofRelative and Absolute Price Efficiency in Regulated Utilities," Review of Economics and Statistics, Feb., pp. 81-88. Atkinson, Scott and Retbert Halverson, ( 1986), "The Relative Efficiency of Public and Private Firms in a RegUilated lEnvironment: The Case ofUS Electric Utilities," Journal ofPublic Economics, Apr., PIP· 281-294. :".vcrech, H~.rvcy :!.nd Ley!a.11d Johnson, (1962), "Behl!.vior of the FLrm Under Regulatory Constraint,'' Ame1rican Economic Review, vol. 52, Dec., pp. 1052-1069. Balestra, Pietro and Marc Nerlove, (1966),"Pooling Cross Section and Time Series Data in Estimation of a Dynamic Model: The Demand for Natural Gas," Econometrica, val. 34, no. 3,1 pp. 585-612. Baughman, Martin L., PaulL. Joskow and Dilip P. Kamat, (1979), Electric Power in the United States: Modlels and Policy Analysis, MIT Press, Cambridge, MA. Baumgartner, Tomas and Att:e Midtun (Editors), (1987), The Politics ofEnergy Forecasting, Clarendon Press, Oxford, UK. Baumol, William and ..dJvin I
Berndt, Ernst R., G. May and G. C. Watkins, (1980),"An Econometric Model of Alberta Electricity Demand," W. T. Ziemba et al (Editors), Energy Policy Modeling: United States and Canadian Experiences, Martinus NtihoffPublishing, Boston, pp. 103-116. Berndt, Ernst R., Michael Doane, and Roy Epstein, (1995),"Electric Utility Rates and the Evaluation of Management Performance," The Electricity Journal, August/September, pp. 69-77. Berg, Sanford V. and John Tschirhart, (1988), Natural Monopoly Regulation: Principles and Practice, Cambridge University Press, Cambridge, UK. Berg, Sanford V. and John Tschirhart, (1995),"Contnoutions ofNeoclassical Economics to Public Utility Analysis," Land Economics, vol. 71 no. 3, pp. 310-330. Berry, Linda, Jon Soderstrom, Eric Hirst, Bobbie Newman and Rose Weaver, (1981 ), Review of Utility Home Energy Audit Programs, Oakridge National Laboratory, Oakridge TN, ORNUCON-58. Blevins, R. P. and B. A Miller, (1993), 1992 Survey of Utility Demand Side Management Programs, Electric Power Research Institute Palo Alto, CA, EPRI TR102193.

Boru, Douglas R., (1931), Armlyzing Demand D;;;~~vior. }'~esources for the Future Inc. Washington, DC. Bohi, Douglas R. and Mary Beth Zimmerman, (1984),"An Update on Econometric Studies ofEnergy Demand Behavior," Annual Review ofEnergy, vel. 9, pp. 105-54. Bonbright, James C., Albert L. Danielsen, and David R. Kamerschen, (1988), Principles of Public Utility Rates, Public Utilities Reports, Arlington, VA Brennan, Michael J. and Eduardo S. Schwartz, (1982), "Regulation and Corporate Investment Policy," Journal ofFinance, May, pp. 289-300. Canine, Craig, (1990),"The Second Coming ofEnergy Conservation," Utne Reader, no. 37, Jan./Feb., pp. 114-120. Chamberlin, John, Terry Fry, and C. L. Mahany Braithwait, (1989), Status ofLeastCost Planning in the United States, Electric Power Research Institute, Palo Alto, CA, EPRI (EM-6133).

134

Chern, WenS., (1978),"Aggregate Demand for Energy in the U.S.," in Maddala, G.S., Wen S. Chern, and Gurmukh S. Gill (Editors), Econometric Studies in Energy Demand and Supply, Praeger, New York, NY, pp. 5-43. Chow, G., (1960),"Tests for Eq!Jality Between Sets of Coefficients in Two Linear Regressions," Econometrica, vol. 28, pp. 591-605. Christensen, Laurits R. and William H Greene, (1976),"Economies of Scale in U.S. Electric Power Generation," Journal ofPolitical Economy, vol. 84, no. 4, pp. 655676. Clarkson, James A (1992)," Industrial Energy Efficiency via Centralized Planning is a Bad Idea," Strategic Planning for Energy and the Environment, vol. 12, no. 2, pp. 4453. Clinton, Jeanne, Howard Geller, and Eric Hirst, ( 1986), "Review of Government and Utility Energy Conservation Programs," Annual Review ofEnergy, vol. 11, pp. 95142. Cogan, Douglas and Susan Williams, (1983), Generating Energy Alternatives, Investor Responsibility Research Center Inc., Washington DC. Cogan, Douglas and Susan Williams, (1987), Generating Energy Alternatives, Investor Responsibility Research Center Inc., Washington DC. Cohen, Sam D. et al, (1990),"Emironmental Externalities: What State Regulators are Doing," The Electricity Journal, July, pp. 24-35. Consumer Energy Council of America Research Foundation, (1986), A Comparative Analysis of Utility and Non-Utility Based Energy Service Companies: A Case Study Approach, Washington DC. Cowing, Thomas G. and V.Kerry Smith, (1978),"Estimation of a Production Technology: A Survey of Econometric Analyses of Steam Electric Generation," Land Economics, vol. 54, no. 2, pp. 158-168. Cowing, Thomas G. and Rodney E. Stevenson (Editors), (1981), Productivity Measurement in Regulated Industries, Academic Press, New York, NY. Crandall, Robert A., {1985),"The Use ofEnvironmental Policy to Reduce Economic Growth in the Sun Belt: The Role of Electric-Utility Rates," Michael A. Crew (Editor), Analyzing the Impact ofRegulatory Change in Public Utilities, Lexington Books, Lexington, MA, pp. 125-140. 135

Crew, Michael A. and Paul Kleindorfer, (1986), The Economics ofPublic Utility Regulation, The MIT Press, Cambridge, MA. Danielsen, Albert A. and David R. Kamerschcn, (1983), Current Issues in PublicUtility Economics, Lexington Books, Lexington, MA. DeSouza, Glenn R., (1981), Energy Policy and Forecasting, Lexington Books, Lexington, MA. Donnelly, William A., (1987), The Econometrics ofEnergy Demand: A Survey of Applicatio~ Praeger Press, New York, NY. Dubin, Jeffrey, A. and Peter Navarro, (1982),"Regulatory Climate and the Cost of Capital," Michael A. Crew (Editor), Regulatory Reform and Public Utilities, Lexington Books, Lexington, MA, pp. 141-166. Eckel, Catherine C. and Theo Vermaelen, (1989),"The Financial Fallout from Chemobyl: Risk Perceptions and Regulatory Response," in Deregulation and Diversification of Utilities, edited by Michael A. Crew, Kluwer Academic Publishers, Boston, MA, pp. 165-181. Edison Electric Institute, (annual), Statistical Yearbook ofthe Electric Utility Industry Edison Electric Institute, (1994), Nonutility Business ofinvestor-Owned Electric Utilities, Edison Electric Institute, Washington D.C .. Electric Power Research Institute, (1993), 1992 Survey of Utility Demand-Side Programs, Electric Power Research Institute, Palo Alto, CA, EPRI TR-1 02193. Electric Power Research Institute, (1993), Principles and Practice ofDemand-Side Management, Electric Power Research Institute, Palo Alto, CA, EPRI TR-102556. Electric Power Research Institute, (1993), Drivers of Electricity Growth and the Role ofUtility Demand-Side Management, EPRI TR-102639. Energy Information Agency, (annual), Financial Statistics of Selected Investor-Owned Electric Utilities. Energy Information Agency, (1992), Housing Characteristics 1990, DOE/EIA0314(90). 136

Energy lnfonnation Agency, (1996), Annual Energy Outlook 1996, DOEIEIA0383(96). . Energy Program Evaluation Conference, (1989, 1991, 1993), Energy Program Evaluation: Conservation and Resource Management, Proceedings. Enholrn, Gregory B., and McOmber, Roger B., (1991), The Salomon Brothers 100 Electric Utilities: Company Summaries, Salomon Brothers, New York, NY. Fanara, Philip Jr. and Raymond Gonnan, (I 986),"The Effect of Regulatory Risk on the Cost of Capital," Public Utilities Fortnightly, :March, 6, pp. 32-36. Faruqui, Ahmad, Ken Seiden, and Steven D. Braithwait, (1990),"Impact ofUtility Demand-Side Management Programs on US Electricity Demand: 1990-2010," ACEEE Proceedings 1990, vol. 5, pp. 83-94. Fels, Margeret F. and Kenneth M. Keating, (1993),"Measurement ofEnergy Savings from Demand-Side Management Programs in US Electric Utilities," Annual Review ofEnergy and the Environment, vol. 18, pp. 57-88. Fenn, Scott A., (1983), America's Electric Utilities under Siege and in Transition, Investor Responsibility Research Center, Washington DC. Fenn, Scott A., (1988), Mergers and Financial Restructuring in the Electric Power Industry, Investor Responsibility Research Center, Washington DC. Ferrar, Tery A., Frank Clemente, and Robert G. Uhler, (1979), Electric Energy Policy Issues, Ann Arbor Science, Ann Arbor, Ml Fisher, Franklin M., and Carl Kaysen, (1962), A Study in Econometrics: The Demand for Electricity in the United States, North Holland, Amsterdam. Flynn, Jack, (1992),"How to Put an Environmental Spin on Your Energy Product, Service, Program, or Project," Strategic Planning for Energy and the Environment, vol. 12, no. 1, pp. 73-79. Fromm, Gary (Editor), (1981 ), Studies in Public Regulation. MIT Press, Cambridge, MA. Gapenski, Louis C., (1989),"An Hypothesis on the Impact ofRegulatory Climate on Utility Capital Costs," Public Utilities Fortnightly, March 30, pp. 28-31.

137

Geller, Howard and Steven Nadel, {1994),"Market Transformation Strategies to Promote End-Use Efficiency," Annual Review ofEnergy, vol. 19. Gettings, Clark W. and John H. Chamberlain, {1988), Demand-Side Management: Concepts and Methods, The Fairmont Press, Lilburn, GA. Gellings, Clark and Sarosh Talukdar, (1986),"Load-management Concepts," Sarosh Talukdar and Clark Gellings (Editors), Load Management, IEEE Press, New York, NY, pp. 3-28. Genzer, Jeffrey C., {1993),"Highlights of the New Energy Act of 1992," Strategic Planning for the Energy and Environment. vol. 12, no. 4, pp. 8-22. Gollop, Frank M. and Stephen Karlson, (1978),"The Impact ofthe Fuel Adjustment Mechanism on Economic Efficiency," Review of Economics and Statistics, November, 1978, pp. 574-584. Gordon, Richard, (1982), Reforming the Regulation ofEiectric Utilities, Lexington Books, Lexington, MA. Greene, William H., {1993), Econometric Analysis, Macmillan Publishing Company, New York, NY. Greenwald, Bruce C., {1984), "Rate Base Sdection and the Structure ofRegulation," Rand Journal ofEconomics, Spring, pp. 85-95. Griffin, James M, (1992),"Methodological Advances in Energy Modeling: 19701990," The Energy Journal, vol. 14, no. 1, pp. 111-124. Halvorsen, Robert, (1975),"Residential Demand For Electric Energy," Review of Economics and Statistics, vol. 57, no. 1, Feb., pp. 12-18. Halvorsen, Robert, (1978), Econometric Models ofUS Energy Demand, DC Heath, Lexington, MA. Halvorsen, Robert and Palmquist, ( 1980),"The Interpretation of Dummy Variables in Semilogarithmic Equations," American Economic Review, vol. 70, pp. 474-475. Hausman, Jerry A., (1978), "Specification Tests in Econometrics," Econometrica, val. 46, no. 6, pp. 1251-1271. Hartman, Raymond S., (1979), "Frontiers in Energy Demand Modeling," Annual Review ofEnergy, val. 4, pp. 433-466. 138

Hayashi, Paul M. and John M. Trapani, (1985), "Rate ofReturn Regulation and the Regulated Firm's Choice of Capital-Labor Ratio: Further Empirical Evidence on the Averch-Johnson Model," Southern Economic Journal, Jan., pp. 384-398. Hempling, Scott, (1995),"Electric Utility Holding Companies: The New Regulatory Challenges," Land Economics, vel. 71, no. 3, pp. 343-353. Henderson, Stephen J., (1985), "Cost Estimation for Vertically Integrated Firms: The Case ofEiectricity," in Analyzing the Impact ofRegulatory Change in Public Utilities, :Michael A. Crew (editor), Lexington Books, Lexington, MA. Herriott, Scott R., (1989),"Competition and Cooperation in Deregulated Bulk Power Markets," in Deregulation and Diversification ofUtitities, :Michael A. Crew (editor), Kluwer Academic Publishers, Boston, MA, pp. 143-163. Hilton, George W., (1972), "The Basic Behavior of Regulatory Commissions," American Economic Review, Papers and Proceedings, May, pp. 47-54. Hirsch, Richard F., (1989), Technology and Transformation in the American Electric Utility Industry, Cambridge University Press, Cambridge, UK. Hirsch, Richard F., (1991),"Regulation and Technology in the Electric Utility Industry: A Historical Analysis oflnterdependence and Change," in Regulation: Economic Theory and History, Jack High (editor), The Uruversity ofMichigan Press, Ann Arbor MI, pp. 147-177. Hirst, Eric, ( 1989a),"Flexibility Benefits of Demand-Side Programs in Electric Utility Planning," The Energy Journal, vol. 11, no. 1, pp. 151-163. Hirst, Eric, (1989b), Electric-Utility Energy-Efficiency and Load Management Programs: Resources for The 1990s, Oak Ridge National Laboratory, Oak Ridge, TN. Hirst, Eric, (1990), Balancing The Scales: Data on Electricity Supply and Demand Resources, Oak Ridge National Laboratory, Oak Ridge, TN. Hirst, Eric, (1991), Possible Effects ofEiectric-Utility DSM Programs 1990 to 2010, Oak Ridge National Laboratory, Oak Ridge, TN, ORNUCON-312.

139

Hirst, Eric, Benson Bronfinan, Richard Goelts, John Trimble, and David Lennan, (1983), Evaluation ofthe BPA Residential Weatherization Pilot Program, Oak Ridge National Laboratory, Oak Ridge, TN, ORNUCON-124. Hirst, Eric and Charles Goldman, (1990),"Review ofDemand-Side Data Needs for Least-Cost Utility Planning," Energy, vol. 15, no. 5, pp. 403-411. Hirst, Eric, Dennis White, and Richard Goelts, (1984), Energy Savings Due to the

BPA Residential Weatherization Pilot Program Two Years after Participation, Oak Ridge National Laboratory, Oak Ridge, 1N, ORNUCON-146. Hallas, Daniel and Stanley R. Stansell, (1984)," An Examination of the Effect of Ownership Fonn on Price Efficiency: Proprietary, Cooperative and Municipal Electric Utilities," Southern Economic Journal, vol. 55, no 2, pp. 236-350. Horowitz, Marvin and Philipp Degens, (1987), Evaluation ofthe Utility Weatherization Assistance Program~ International Energy Associates LTD, Portland, OR, lEAL P0-21. Houthakker, H. S., (1951),"Electricity Tariffs in Theory and Practice," The Economic Journal, vol. 61, no. 241., pp. 1-25. Howarth, Richard B. and Bo Anderson, (1993),"Market Barriers to Energy Efficiency," Energy Economics, October, pp. 262-272. Hsiao, Cheng, (1986), Analysis of Panel Data, Cambridge University Press, London. Hyman, Leonard S., (1988), America's Electric Utilities: Past. Present and Future, Public Utilities Reports, Arlington, VA. Hyman, Leonard S. and Ernst R. Habicht Jr., (1986),"State Electric Utility Regulation: Financial Issues, Influences, and Trends," Annual Review ofEnergy, vol. 11, 163-85. Hyman, Leonard S. and Heidimarie West, "Diversification , Deregulation and Competition: Cost of Capital Implications for Electric Utilities," in Deregulation and Diversification ofUtilities, Michael A. Crew (editor), KJuwer Academic Publishers, Boston, MA, pp. 165-181. 1m, Bae-Geun, David L. Kaserman and Francois Melese, (1989),"Endogenous Regulation and the Firm's Regulatory Expenditures," Applied Economics, vol. 21, March, pp. 375-385. 140

Jorgenson, Dale W., (1986),"Econometric M~thods for Modeling Producer Behavior," in Handbook ofEconometric~. Volume ill, Z. Griliches and M.D. Intriligator (editors), Elsevier Science P4blishers BV, pp. 1842-1915. Joskow, PaulL., (1974), "Inflation and ~~nvitonmental Concern: Structural Change in the Process of Public Utility Price Regul~tio111," Journal of Law and Economics, Oct., pp. 118-140. Joskow, PaulL., (1988),"The Evolution of Competition in the Electric Power Industry," Annual Energy Review. vol. 13, pp. 215-238. Joskow, PaulL., {1989),"Regulatory FaUureJ regulatory Refonn, and Structural Change in the Electric Power Industry," Bro(l)kings Papers on Economic Activity, pp. 125-208. Joskow, PaulL. and Roger G. Noll, (19~1), 1'Regulation in Theory and Practice: An Overview," Gary Fromm (editor), Studi~s in ,public Regulation. MlT Press, Cambridge, MA, pp. 1-65. Joskow, Paul L. and Richard Schmalens~e, (1983), Markets for Power: An Analysis of Electric Utility Deregulation, MIT Pr~ss, ~ambridge, MA. Joskow, PaulL. and Richard L. Schmal~nsee, (1986), "Incentive Regulation for Electric Utilities," Yale Journal ofRegu!ation, Fall, pp. 1-49. Kafoglis, Milton Z., (1986),"Tax Policy and 'Public Utility Regulation," John C. Moorehouse (Editor), Electric Power, Pacifi~ Research Institute for Public Policy, San Fransisco, CA, pp. 97-134. Kahane, Adam and Ray Squitieri, {1987),"mectricity use in Manufacturing," Annual Energy Review, vol. 12, pp. 223-251. · Kahn, Alfred E., {1988), The Economic~ ofRegulation: Principles and Institutions, MIT Press, Cambridge, MA. Karlson, Stephen H., {1986),"Multiple-Output Production and Pricing in Electric Utilities," Southern Economic Journal, July~ pp. 317-32. Keelin, Thomas W. and Clark W. Gellings, (11986), Impact ofDemand-Side Management on Future Customer Electricitv Demand. Electric Power Research Institute, Palo Alto, CA, EPRI EM-4815-SR.

141

Kennedy, Peter, {1993), A Guide to Econometrics, MIT Press Cambridge, MA. Khazzoom, J. Daniel, (1986), An Econometric Model Integrating Conservation Measures in the Residential Demand for Electricity, JAI Press Inc. Greenwich, CT. Kolbe, A Lawrence and James A Read Jr., (1984), The Cost of Capital: Estimating the Rate of Return ofPublic Utilities. MIT Press, Cambridge, MA. Labys, Walter C., (1973), Dynamic Commodity Models: Specification. Estimation. and Simulation, Lexington Books, Lexington, MA. Lirnaye, Dilip R. and Veronika Rabl (Editors), (1988), International Load Management: Methods and Practices, The Fairmont Press, Lilburn, GA Lyman R. Ashley, (1978),"Price Elasticities in the Electric Power Industry," Energy Systems and Policy, vol. 2, no. 4, pp. 381-406. Lyon, Thomas P., {1991),"Regulation with 20-20 Hindsight: 'Heads I Win, Tails You Lose'?," Rand Journal ofEconomics, vol. 22, no. 4, pp. 581-595. Maddala, G.S., {1992), Introduction to Econometrics, MacmillanPublishing Company, New York, NY. Mansfield, Edwin, (1970), Microeconomics, W. W. Norton, New York, NY. Marino, Anthony M. and Joseph Sicilian, (1982),"Residential Energy Conservation and Public Utilities: Regulatory Strategy," in Regulatory Reform and Public Utilities, Michael A Crew (editor), Lexington Books, Lexington, MA, pp. 183-212. McCormick, Robert E., (1986),"Inflation, Regulation, and Financial Adequacy," in Electric Power, John C. Moorehouse (editor), Pacific Research Institute for Public Policy, San Fransisco, CA, pp. 135-152. McFaddden, Daniel and Jeffiey Dubin, (1984),"Econometric Analysis of Residential Electric Appliance Holdings," Econometrica, vol. 52. pp. 345-362. Millhome, John P., John Beldock, Bamewy Capehart, and Jerome LaMontagne, (1992), "United States Energy Efficiency Policy: Strategies for the Future," Strategic Planning for the Energy and Environment, vol. 2, no. 3, pp. 9-32. Mitchel, Bridger M., Rolla Edward Park, and Francis Lab rune, ( 1986), Projecting the Demand for Electricity: A Survey and Forecast, The Rand Corporation, Santa Monica, CA, R-3312-PSSP. 142

Mtchell, Cynthia, (1989), "Lagging in Least-Cost Planning: Not as Far Along as We Thought We Were," The Electricity Journal, vol. 2, no. 10, pp. 24-31. Mtchell, Cynthia, (1992), "Integrated Resource Planning Survey: Where the States Stand," The Electricity Journal, vol. 5, no. 5, pp. 10-15. Morse, M. F. and J. L. Small, (1987),"Deriving Electricity Demand Elasticities From a Simulation Model," Energy Journal, Mar., pp. 51-76. Moskovitz, David, (1989), Profits & Progress Through Least-Cost Planning, National Association ofRegulatory Utility Commissioners, Washington DC. Mount, T. D. and L. D. Chapman, (1978),"Capacity Planning for Electric Utilities: An Econometric Approach," in Energy Policy, J. S. Aronnofsky et al (Editors), North-Holland, Amsterdam, pp. 77-94. Nadel, Steven, ( 1991 ), "Achievable Electric Conservation Potential: The Role of Utility and Non-Utility Programs," Proceedings 5th National Demand-Side Management Conference, Electric Power Research Institute, Palo Alto, CA, pp. 1115. Nadel, Steven and Howard Geller, (1995),"Market Transformation Programs: Past Results and New Initiatives," Energy Services Journal, vol. 1, no. 2, pp. 93-108. Navarro, Peter, (1986), "The Performance ofUtility Commissions," Electric Power, John C. Moorehouse (editor), Pacific Research Institute for Public Policy, San Fransisco, CA, pp. 337-358. Navarro, Peter, ( 1996),"Eiectric Utilities: The Argument for Radical Deregulation," Harvard Business Review, January-February, pp. 112-125. Nemetz, Peter N. and Marilyn Hankey, (1984), Economic Incentives for Energy Conservation, John Wiley & Sons, New York, NY. New York State Energy Research and Development Authority, (1990), Lessons Learned: A Review ofUtility Experience with Conservation and Load Management Programs for Commercial and Industrial Customers, Report 90-8. Nordhaus, William D. (editor), (1977), International Studies ofthe Demand for Energy, North-Holland, Amsterdam.

143

Parmeano, Hethie, William Bridgeman and Virginia Perry-Failor, ( 1987), The Role and Nature of Marginal and Avoided Costs in Ratemaking: A Survey, National Economic Research Associates, White Plains, NY. Payne, F. William, (1991),"EPA's 'Green Lights' Program Promotes Environmental; Protection, Energy Savings, and Profits!," Strategic Planning for the Energy and Environment, vol. 2, no. 1, pp. 19-29. Peirce, William Spangar, (1986), Economics ofthe Energy Industries, Wadsworth Publishing Company, Belmont, CA. Petersen, H. Craig, (1975), "An Empirical Test ofRegulatory Effects," Bell Journal ofEconomics, Spring, pp. 111-126. Pinyck, RobertS., (1979),"The Characteristics ofEnergy Demand," in Energy Conservation and Public Policy, J. Sawhill (editor), Englewood Cliffs, NJ, PrenticeHall. Pindyck, Robert S. and Daniel L Rubenfeld, ( 1991 ), Econometric Models and Econometric Forecasts, McGraw-Hill Inc., New York, NY. Plat, Harlan D. and Jonathan B. Welch, (1989),"Factors that Influence Public Utility Perfonnance," Energy Economics, July, pp. 219-227. Plummer, James, (1990),"Bypass Concepts and Transmission Access," in Competition in Electricitv: New Markets and New Structures, James Plummer and Susan Troppmann (editors), Public Utilities Reports Inc., Arlington, VA, pp. 43-60. Prager, R. A., (1989), "The Effects ofRegulatory Policies on the Cost of Debt for Electric Utilities: An Empirical Investigation," Journal ofBusiness, vol. 62, no. 1, pp. 33-53. Prete, Lawrence, Janet Gordon, and Linda Bromley, (1992),"Electric Utility DemandSide Management," Electric Power Monthly, April, pp. 19-33. Primeaux, Walter J. Jr. and Patrick C. Mann, (1986),"Regulator Selection Methods and Electricity Prices," Land Economics, vol. 62 no. 1, pp. 1-13 Rapp, Donald, (1981), Solar Energy, Prentice-Hall Inc., Englewood Cliffs, NJ. Ray, William J., (1995), Glasgow's Fully Interactive Communications and Control System, Glasgow Electric Plant Board, Glasgow, KY. 144

Reeb, Donald J. and Edward T. Howe, (1985),"State and Local Taxation ofEiectric Utilities: A Study of the Record," in Analyzing The Impact ofRegulatory Change in Public Utilities, Michael A. Crew (editor), Lexington Books, J....exington, MA, pp. 5974. Regional Economic Research, Inc. and Barakat and Chamberlin, Inc., (19931 Drivers of Electricity Growth and the Role ofUtility Demand-Side Management, Electric Power Research Institute, Palo Alto, CA, EPRI TR-102639. Reid, Michael W., (1988), "Ratebasing ofUtility Conservation and Load Management Programs," Alliance to Save Energy, Washington DC. Reid, Michael W., (1990),"Financia1 Incentives for DSM Programs: A Review and Analysis ofThree Mechanisms," ACEE Proceedings 1990, vol. 5, pp. 157-166. Rubin, EdwardS., (1989),"Implications ofFuture Environmental Regulation of CoalBased Electric Power," Annual Energy Review, pp. 19-45. Scheidell, John M., (1976), "The Relevance ofDemand Elasticity for Rate ofReturn Regulation," Southern Economic Journal, Oct., pp. 1088-1095. Scherer, F. M., ( 1970), Industrial Pricing, Rand McNally, Chicago, IL. Sherman, Roger, (1985),"The Averech and Johnson Analysis ofPublic Utility Regulation Twenty Years Later," Review of Industrial Organization, pp. 178-193. Schick, Harold and Mark Hamilton, (1996),"Encouraging Competitive and Societal Energy Efficiency Opportunities in a Restructured Utility Industry," Barakat and Chamberlin Inc., (unpublished manuscript). Shin, Jeong-Shik, (1985),"Perception ofPrice when Price Information is Costly: Evidence from Residential Electricity Demand," Review of Economics and Statistics, vol. 67, no. 4, pp. 591-598. Sonnenblick, R and J. Eto, ( 1995), A Framework for Improving the CostEffectiveness ofDSM Program Evaluations. Lawrence Berkeley Laboratory, University ofCalifornia, Berkeley, CA, LBL-37158. Spann, Robert M., (1974),"Rate ofRetum Regulation and Efficiency in Production: An Empirical Test of the Aver~h-Johnson Thesis," Bell Journal of Economics, Spring, pp. 38-52. 145

Spulber, Daniel F. (1989), Regulation and Markets, MlT Press, Cambridge, MA. Stanton, Timothy J., (1993),"Capacity Utilization and New Source Bias: Evidence from the US Electric Power Industry," Energy Economics;, Jan 1993, pp. 57-60. Stein, Jerome L. and George H. Borts, (1972),"Behavior ofthe Firm Under Regulatory Constraint," American Economic Review, December, pp. 964-970. Stevenson, Rodney E. and David W. Penn, (1995),"Discretionary Evolution: Restructuring the Electric Utility Industry," Land Economics, vol. 71, no. 3, pp. 354367. Stewart, John F., (1982),"Econornic Efficiency and Automatic Fuel-Cost Adjustment Mechanisms: Theory and Empirical Evidence," in Regulatory Reform and Public Utilities, Michael A. Crew (editor), Lexington Books, Lexington, MA, pp. 167-182. Taggart, Robert A., Jr., (1985),"Effects ofRegulation on Utility Financing: Theory and Evidence," Journal oflndustrial Economics, Mar., pp. 257-276. Talukdar, Sarosh and Clark W. Gellings (Editors), (1986), Load Management, IEEE Press, New York, NY. Taylor, D. L., (1975),"Demand for Electricity: A Survey," The Bell Journal of Economics and Management Science, vol. 6, no. 1, Spring, pp. 74-110. Taylor, D. Thomas and Russel G. Thompson,, (1995),"The Efficient Utility: Labor, Capital, and Profit," Public Utilities Fortnightly, September 1, 1995. Tempchin, RichardS., A. Joseph Van den Berg, Vera B. Geba, CurtisS. Felix, and Marc W. Goldsmith, (1990),"Ernission Impacts ofDemand-Side Programs: What Have We Achieved so Far and How Will Recent Policy Decisions Change Program Choices," ACEEE Summer Session, Environment, pp. 4.237-4.248. Thompson, Howard E., (1985),"Estimating Return Deficiencies of Electric Utilities 1963-1981," in Analyzing The Impact ofRegulatory Change in Public Utilities, Michael A. Crew (editor), Lexington Books, Lexington, MA, pp. 17-38. Tolley, GeorgeS. and Edward C. Bodmer, (1990),"Utility Rate Comparisons and Management Efficiency," Public Utilities Fortnightly, Jan. 4, pp. 29-34. Tempchin, RichardS. et al, (1990),"Emission Impacts ofDemand-Side Programs: What Have We Achieved so Far and How Will Recent Policy Decision Change Program Choices?," ACEE Proceedings 1990, vol. 4, pp. 237-248. 146

Train, Kenneth, (1985),"Statistically Adjusted Engineering (SAE) Models ofEnd-use Load Curves," Energy ,vol. 10, no. 10, pp. 1103-1113 Trout, Robert R., (1979),"Regulatory Factor and Electric Utility Common Stock Investment Values," Public Utilities Fortnightly, November 22, pp. 28-31. Varian, Hal R., (1984), Microeconomic Analysis, W. W. Norton and Company, New York, NY. Watson, Donald, Sand Malcom Getz (editors), (1981), Price Theory in Action, Houghton Mifflin Company, Boston, MA. White, Hal, (1980)," A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity," Econometrica, vol. 48, pp. 817-838. Uri, Noel D., (1981), Dimensions ofEnergyEconomics. JAI Press, Greenwich, CT. Woo, Chi-Keung, Debra LLoyd-Zannetti, Ren Orans, Brian Horii and Grayson Heffuer, (1995),"Marginal Capacity Costs of Electricity Distribution and Demand for Distributed Generation," The Energy Journal, vol. 16, no. 2. , pp. 111-130. Zakem, Alexander J., (1986),"Comparing Utility Rates," Public Utilities Fortnightly, November 13, pp. 30-32.

147

Loading...

Determinants and Impacts of Demand-side Management Program

Portland State University PDXScholar Dissertations and Theses Dissertations and Theses 1996 Determinants and Impacts of Demand-side Management Pro...

10MB Sizes 1 Downloads 8 Views

Recommend Documents

Tourism Impacts, Planning and Management
and maintain the attraction, as well as enhanced interest from visitors in the importance of the natural environment and

Analysis of Energy Efficiency Program Impacts Based on Program - EIA
May 3, 2015 - spending allocations ('model portfolios') at the regional level in the NEMS building modules for lighting,

Determinants of consumer buying behavior | Management Education
Sep 8, 2012 - It is the family values, beliefs, perceptions and preferences affect the consumer buying behavior. 2) Subc

Determinants Implementation of Management Information System in
[6] Kemenkes 2014, Permenkes Nomor 75 Tahun 2014 Tentang Puskesmas. Kemenkes, Jakarta. [7] Alwan Sri Kustono,2011, Penga

Performance Management: Impacts and Trends White Paper
and implementing a new performance management system and equally beneficial to those trying to prove the value of effect

Summer environmental awareness program 1981 energy impacts
Urbanization brings about the crowding of large numbers of people in areas of probable hazards, such as earthquake, floo

determinants test of the effectiveness of fixed asset management in
functions of local government as a provider of public services to the society. General issues that always exist in Indon

The Determinants and Consequences of Remittances - Staffordshire
aim at attracting remittances and shifting them from the informal into the formal sector, notably in Mexico ..... are of

Driver Assessment and Management Program
Driver Assessment and Management Program. If you have a physical disability or suffer from combined physical and cogniti

negative impacts of television program (private television - TIJOSS
PJTV Program Pelatihan Jurnalistik Televisi, ABC. Program Berita TV, FISIP UI, Jakarta, 2001. Piliang, Yasraf Amir, “H