Utilities Policy, Vol. 6, No. 1, pp. 43-55, 1997
Pergamon
S0957-1787(96)00012-4
© 1997 Elsevier Science Ltd
All rights reserved. Printed in Great Britain
0957-1787197 $17.00+0.00
The theory and practice of
decoupling utility revenues from
sales
Joseph Eto, Steven Stoft and Timothy Belden
Decoupling has emerged in the US as an important
regulatory strategy for insulating utility revenues
from sales fluctuations. Breaking the link between
revenues and sales, it is argued, is an important
prerequisite for transforming utilities from sellers of
an energy commodity to providers of energy
services. We characterize the cost and regulatory
conditions that underlie these arguments and,
thereby, provide guidance on the applicability of
decoupling to other regulated utilities. We describe
how decoupling works in practice and then, using
historic information on utility costs, examine the
cost-tracking assumptions inherent in traditional
rate-making and current decoupling approaches.
Finally, we report on the actual rate impacts of
decoupling examining the three US utilities with the
longest history of decoupling. © 1997 Elsevier
Science Ltd. All rights reserved
Keywords: Electric utilities; Regulation; Demand-side management
The social benefits of public utilities actively managing
demand for their commodities are becoming increasingly
clear. In the US, the idea of demand-side management
(DSM) for electric utilities has been embodied in a host
of least-cost or integrated resource planning (IRP)
regulations. These regulations direct utilities to pursue
demand-side management whenever the social cost of
helping customers use electricity more efficiently is less
than the social cost of producing more electricity (Krause
and Eto, 1988). Benefits of DSM include lower energy
costs for consumers and reduced need for new power
The authors are with the Lawrence Berkeley National Laboratory, 1
Cyclotron Road, MS 90-4000, Berkeley, CA 94720, USA.
The work described in this study was funded by the Assistant
Secretary of Energy Efficiency and Renewable Energy, Office of Utility
Technologies, Office of Energy Management Division of the US
Department of Energy under Contract No. DE-AC03-76SF00098.
plants with their attendant environmental problems.
Evidence suggests that many regulated utility services
(e.g. natural gas, water, solid waste, etc.) could be
provided at lower total societal cost through active
demand-side management by utilities (Hirst et al., 1991;
Winpenny, 1992). However, under most regulatory
schemes, a utility's best course of action from a financial
perspective--to sell more of its regulated commodity-can be at odds with the socially efficient, least-cost
planning outcome--to maximize net resource benefits,
which may mean reducing sales (Moskovitz, 1989).
Decoupling revenues from sales has emerged as an
innovative and controversial approach used by US
electricity regulators to address this dilemma (Hirst and
Blank, 1994). Decoupling refers to a class of automatic
or semiautomatic annual rate-making adjustments that
insures that utilities collect an agreed-upon level of
revenues independent of actual sales between rate cases.
To appreciate the controversy generated by decoupling
and to understand decoupling's applicability outside the
US and to utilities other than electricity, it is important to
understand the current price-regulation mechanics and
utility cost conditions that underlie decoupling in the US.
This paper uses aggregate information on the current cost
structure of the US investor-owned electric utility
industry to characterize the disincentive to reducing sales
of electricity (e.g. through utility-sponsored customer
energy-efficiency programs) that decoupling is designed
to mitigate. We also make clear that this disincentive is
not present everywhere; its strength depends on a host of
circumstances, many of which are not uniform throughout the electric utility industry.
We next describe how various decoupling schemes
operate by introducing them as a modification to existing
forms of rate-making. We rely on historic data on US
electric utility performance to explore the adequacy of
the cost-tracking assumptions that underlie various
decoupling schemes and compare them to the assumptions that underlie traditional rate-of-return regulation
without decoupling.
43
Decoupling utility revenues from sales
In the US, concerns have been expressed about rate
impacts and the shifting of business risks from utility
shareholders to ratepayers when revenues are decoupled
from sales. In our final section, we report on the actual
rate impacts of decoupling by examining the rate history
of the three California electric utilities with the longest
history of decoupling.
Why decouple?
Traditional rate-of-return regulation discourages utilities
from pursuing customer energy efficiency programs
because: (1) utilities may not recover demand-side
management (DSM) program expenses when these
expenses have not been included in some previous ratesetting process; (2) utilities may lose revenue from sales
not made because of the success of customer energyefficiency programs; and (3) utilities may forego
earnings opportunities because resources are devoted to
DSM programs rather than to other profit-making
activities (Nadel et al., 1992). Lost revenues that are the
primary disincentive addressed by decoupling have
received the most attention because they are often the
largest negative financial consequence of a successful
energy efficiency program in the short run. The lost
revenue disincentive is seen most clearly when we
examine utilities' incentives to sell more electricity.
Utility incentives to sell more electricity
Utilities have an incentive to sell more of their product
and a disincentive to sell less whenever the marginal
revenue (MR) from a sale exceeds the marginal cost
(MC) of production. These conditions (MR>MC) are
generally reflective of the current revenue stream and
cost structure of most US electric utilities today.
Table 1 shows a representative income statement for a
composite utility that allows us to quantify the effects of
incremental sales revenues on a utility's profits. In the
income statement in Table 1, revenues, are simply sales
multiplied by an average price. Average price has been
fixed at 70 mills/kWh, and sales have been derived to
yield an arbitrary total revenue of $100. Costs consist of
fuel, nonfuel operation and maintenance (O&M), depreciation, interest, and taxes. Although these costs have
been normalized to be consistent with a total revenue of
$100, the fractions of revenues that they represent reflect
a composite of US investor-owned utilities, as reported
in the Energy Information Agency's most recent annual
survey of utilities (Energy Information Administration,
1993).
Net income or profit is the difference between
revenues and costs. Net income can be expressed in two
ways: as a percentage of total revenues, which can also
be thought of as a profit margin; or as a return on equity
if the capital structure of the utility is specified. In this
Table 1. Profitability of 1% sales increase without decoupling--examples
($ unless noted otherwise)
No Fuel Adjustment
Clause
Base Case
Revenue
Sales (kWh)
Price (S/kWh)
Total Revenue
Cost
Nonfuel O&M
Fuel
Depreciation
Interest
Total Costs (before taxes)
Gross Income
Taxes
Net Income
ROE (%)
Profit Margin
Variable Cost/Total Costs
Marg. Variable/Avg. Variable Cost
•
•
•
•
Change from
Base Case
1429
0.07
100.00
1.00%
0.00%
1443
0.07
101,00
25.40
33.30
9.70
8.60
77.00
23.00
i 3.00
10.00
12.00
0.89%
0.57%
0.00%
0.00%
77.42
25.63
33.49
9.70
8.60
10.0%
58.7%
23.58
13.23
10.35
12.42
10.3%
71.6%
90.7%
Marginal income tax rate = 40%
Profit Margin = Net Income / Total Revenue
Variable Cost = Nonfuel O&M + Fuel + Taxes
Marginal Variable Cost = Change in Variable Cost divided by Change in
Sales
• Average Variable Cost = Base Case Variable Cost divided by Base Case
Sales
44
Decoupling utility revenues from sales
example, the profit margin is 10%, and the return on
equity is 12%.
To illustrate a utility's incentive to increase sales, we
change the situation in Table 1 to consider how profits
are affected by a 1% increase in sales. Marginal revenue
is assumed to equal average revenue; in other words, the
price of electricity is fixed and is assumed to be linear in
the short term (i.e. before the next rate case). As a result,
a 1% increase in sales leads to a 1% increase in
revenue.
However, marginal cost is not equal to average cost. In
the short run (primarily, between rate cases), not all costs
will be affected by changes in sales. Interest, depreciation, and some portion of nonfuel O&M are all unlikely
to vary in the short run as a result of changes in sales, so
they are in this sense fixed.
Fuel and some portion of nonfuel O&M costs, on the
other hand, are likely to be affected by changes in sales
and are in this sense variable. If gross income changes,
taxes will also be affected. Based on aggregate US
electricity industry performance, our example shows that
these variable costs (fuel, nonfuel O&M, and taxes)
account for nearly 60% of total costs.
The way in which variable costs change in response to
changes in sales in the short run varies. Marginal variable
costs can either exceed or be less than average variable
costs. For the two most recent, consecutive years of the
US utility financial information available for our study
(1987 and 1988), we find that marginal variable costs
(MVC) resulting from a 1% increase in electricity sales
have been slightly more than 0.70% of average variable
costs (AVC). In other words, marginal variable costs are
less than average variable costs. Referring to Table 1, we
see that 0.70% represents the weighted average of three
changes: an 0.89% change in MVC to AVC for nonfuel
O&M costs, a 0.57% change in MVC to AVC for fuel,
and an increase in taxes calculated using a 40% marginal
tax rate.
Marginal profitability is the difference between marginal revenues and marginal costs. In our example, net
income and return on equity increase by almost 4%.
Expressed as a change in basis points from an initial
return on equity of 12%, the effect works out to be about
40 basis points or less than $0.03/kWh of incremental
sales. (See Eto et al., 1994 for additional examples
including the effect of a fuel adjustment clause.)
Clearly, this result is a reflection of the various cost
assumptions we have made regarding the magnitude of
the affected cost elements as fractions of total cost, the
rate of change of these cost elements compared to
changes in sales, and the level of profits at the start.
Although our assumptions are based on recent, aggregate
US electricity industry performance, individual utility
performance can be expected to vary considerably.
Fortunately, it is straightforward to generalize from
these specific assumptions to treat other situations.
Between rate cases, which are the forum where rates are
determined, a utility's profit depends on: (1) the utility's
authorized profit margin prior to any incremental sales,
(2) the fraction of their total costs affected by the
production expenses of making incremental sales, and
(3) the way this fraction is affected by increased
production (i.e. the relationship between marginal variable and average variable costs)
Figure 1 illustrates the relationship between these
three items for a 1% increase in sales. Results are
presented for three levels of variable costs as a fraction
of total costs (40%, 60% and 80%), which are represented by three downward-sloping horizontal lines. A
range of possible marginal-to-average variable cost
relationships is represented along the horizontal axis.
The resulting change in profit, expressed as a change in
return on equity (normalized to an initial 12%), is seen
on the vertical axis. The example described above is
indicated on the figure as case 1.
Figure 1 indicates that the profitability of an incremental increase in sales goes up: (1) as the variable cost
component of total costs decreases because more costs
are fixed, and (2) as the responsiveness of these costs (i.e.
the ratio of marginal variable costs to average variable
costs) to increases in sales decreases.
In Figure 1, we can also see that, when prices are fixed
and linear, increased sales almost always lead to
increased profits. Only a fraction of the total cost of
production is affected by increases in sales, and the
degree to which these costs are affected must greatly
exceed the percentage increase in sales in order to offset
the increase in revenues from sales. For the cost structure
of the most US electric utilities today, in which 40-80%
of total costs are affected in the short run by changes in
sales, costs must increase by a factor of two to three
times the percentage increase in sales in order for the
increase in sales not to be profitable. Stated another way,
if costs do not increase this sharply in response to
changes in sales (and there are few instances in which
this appears to be the case), increased sales will always
lead to increased profits. In other words, US electric
utilities have a powerful motivation to increase sales
between rate cases. So, if decoupling is to successfully
encourage energy efficiency, it must successfully mitigate a powerful incentive to increase sales that is deeply
embedded in the current way in which rates are set by
traditional rate-of-return regulation.
The rate case is a limit to the incentive f o r incremental
sales
The profitability of increased sales described in the
previous section depends on two critical assumptions: (1)
retail rates are fixed and linear so that marginal revenue
is equal to average revenue, and (2) some fraction of
45
Decoupling utility revenues from sales
i ........ Variable Cost = 4 0 % of Total
B ase Profit Margin = 10%
I
100,. •
I
............
qr--
Variable Cost = 6 0 % of Total
I ............. Variable
Cost = 80*/, of Total
II
:~
O"
U.I
tO
80
E
"".
o
iv"
~
o
..........
1
Region 2
"".... !
....................
-40%
40
E
•~-
~Case
""'• Region ~
Region 5
"-..
"""., Regi°n4
2O
n
•
._
"'-.
~
60%
m
113
0
@
'- - 2 0
80%
0•5
I
I
1•0
1•5
Marginal Variable Cost / Average Variable Cost
F i g u r e 1. Profitability of 1 % sales increase•
costs is fixed and therefore marginal costs are usually
less than average costs• A rate case calls both assumptions into question.
During rate cases, fixed and variable costs are
considered simultaneously. Adjustments are made to the
rate base, a rate-of-return is determined, operating and
other expenses are considered, and an estimate of sales is
used to set rates• Although there are important procedural
differences between states that rely on historic test years
and states that use future test years for this process, the
outcome is similar: rates are established that apply until
they are revised. In other words, rate cases limit the
continuing efficacy o f the conditions described above
that make incremental sales profitable• Therefore, the
profitability of incremental sales is currently a direct
consequence of regulatory lag.
Reviewing 160 rates cases covered in 10 years of
Public Utilities Fortnightly to determine the historic
frequency of rate cases, we find that the average time
between rate cases from 1984 to 1992 has been about
three years, with the median being slightly less (see
Figure 2).
The implication for decoupling is clear: if the
46
incentive to sell additional electricity described in the
previous section is the primary incentive addressed by
decoupling, the usefulness of decoupling depends on the
frequency of rate cases• Because rate cases can, in
principle, fully address all issues underlying the shortTime between most recent rate case for
major investor-owned utilities
Compiled from Public Utilil~es Fortnightly's
Annual Review of Rate Cases lg84-1992
3O
2S
///:
/ / /
20
///
/ .,"/t
/ //u
10
5
0
/ / /
/ / /
i / .,i
/ / /
//~
JJ~
J
/J
/~
A
]JJ|
JJ~
JJ~
i i .4
jjj
jj]
j
/]
/j
/
jjj
f ]]
jjj
0 to 1
1 to 2
~
2 to 3
3 to 4
4 to 5
5 to 6
>7
N u m b e r of years between rate c a s e s
F i g u r e 2. The rate case as a limit to the profitability of sales
increases.
Decoupling utility revenues from sales
run profitability of incremental sales, the value of
decoupling as an additional regulatory intervention
diminishes as rate cases become more frequent.
Nevertheless, more frequent rate cases, simply to
address the profitability of incremental sales, are unlikely
because rate cases are time-consuming and expensive.
Moreover, utilities are often reluctant to subject their
businesses to frequent, detailed scrutiny by regulators,
particularly if their businesses are excessively profitable.
In fact, cost changes that in the past initially led to more
frequent rate cases have, more recently, led to the
creation of automatic adjustment clauses (fuel adjustment clauses are the most well-known example) that try
to deal directly with specific cost changes without
requiring a rate case. Recent regulatory practice has
created a variety o f out-of-rate-case procedures precisely
to ensure that rate cases will not be held more frequently.
Decoupling may therefore be desirable because it can
address changing costs that would otherwise lead to rate
cases.
Another incentive to sell electricity
Regulatory lag is not the only incentive for incremental
sales between rate cases. Another potential incentive
comes from the relationship between increasing the rate
base and a utility's ability to earn a regulated return on
this rate base.
Understanding the strength of this
incentive is important because it is not addressed by
decoupling.
Rate-of-return regulation creates a shareholder incentive for utilities to build their rate base whenever the rate
of return exceeds the cost of capital. This feature of
regulation is known as the Averch-Johnson thesis (Train,
1991). One purpose of the rate case is to provide a
periodic check on a utility's activities to ensure that
additions to the rate base are prudent. However, the
purpose of the rate case is not to question the wisdom of
basing utility rates on formulas that link authorized
earnings to a fixed percentage of undepreciated assets. If
building rate base to meet increased loads leads to
increases in authorized revenues and also increases in
profits, then the very formulation of rate-of-return
regulation creates a distinct incentive for incremental
sales.
Decoupling is neutral on the issue of how big a
utility's rate base and sales base should be, so decoupling
makes the utility indifferent to incremental sales or
losses between rate cases. Where decoupling is practiced, questions about the appropriate level of sales and
size of the rate base must then be addressed in rate cases
or by some other means.
We cannot treat these level of sales and size of rate
base issues adequately here, but we think it is important
to understand that utilities may have other incentives to
build load besides the short-run incentive created by
regulatory lag. We have identified rate-of-return regulation as being one such incentive; there are probably
others. A systematic treatment of decoupling requires
consideration of these incentives. If their influence is
small, they may be ignored. If their influence is large,
then whether they reinforce or mitigate the incentives
created by regulatory lag becomes more important.
How does decoupling work?
The critical differences between traditional rate-making
and decoupling are in the focus and frequency of the
rate-making process. Traditionally, rate-setting takes
place in the context of a rate case cycle, which usually
spans many years. Decoupling does not change this basic
process but adds an explicit means for setting revenues
during the period between rate cases. Therefore, decoupling eliminates the incentive to increase sales between
rate cases because it insures that revenues will be
unaffected by actual sales.
In traditional rate-making procedures, the revenue
requirement used to set rates almost always differs from
actual revenue because of fluctuations in sales. Decoupiing ensures that actual revenues exactly match an
established revenue requirement, regardless of the sales
level. For this reason, we will refer to all decoupling
schemes as revenue adjustment mechanisms or RAMs.
We will also refer to the revenue requirement established
under decoupling as the authorized revenue.
Every decoupling RAM consists of two parts. First, all
decoupling RAMs use balancing accounts to guarantee
the exact collection of authorized revenue over time.
Second, all decoupling RAMs work in conjunction with
an explicit method for changing the level of authorized
revenue during years between general rate cases.
Breaking the link between sales and revenue using a
balancing account
The use of a balancing account to ensure exact collection
of authorized revenue is consistent in all revenue
decoupling RAMs and is central to removing bias against
energy conservation. We begin our explication of the
different decoupling RAMs by describing a simplified
decoupling mechanism that only involves use of a
balancing account. We assume that this decoupling
mechanism, which we call the Basic Rate Adjustment
Mechanism, operates in a state with a two-year general
rate case cycle and no other between-rate-case revenue
adjustments. Table 2 illustrates the Basic RAM. The
basic R A M requires three sets of numbers to track
revenue and price. Columns A - C in Table 2 are
established in the general rate case and remain fixed until
the next general rate case. Columns D - F represent what
actually occurs during each year. Columns G - I represent
the numbers that the utility reports in its income
47
Decoupling utility revenues from sales
Table 2. Basic RAM example
GRC 1
GRC 2
Yr 1
Yr 2
Yr 3
A
Expected
Price
S/kWh
B
Expected
Sales
kWh
C
Authorized
Rev
$
0.100
0.100
0.110
1000
1000
1010
100.00
100.00
111.10
D
Price
S/kWh
E
Collected
Sales
kWh
0.100
0.090
0.111
1100
990
1010
statement and balance sheet. The changes in these
numbers from year to year illustrate how the Basic RAM
(BRAM) operates.
Year 1. General Rate Case no. 1 (GRC 1) authorizes
revenue of $100 based on expected sales of 1,000 kWh.
During the year, the utility sells 1100 kWh at $0.10/kWh,
resulting in a Collected Revenue of $110. The BRAM
ensures that the utility can only keep the Authorized
Revenue of $100. Thus, Reported Revenue equals $100
and - $ 1 0 is placed into a balancing account. Negative
values in the balancing account indicate money that the
utility owes the ratepayers (accounts payable); positive
values indicate money that ratepayers owe the utility
(accounts receivable).
F
H
Revenue
$
G
Reported
Revenue
$
+/$
I
Balance
Account
$
110.00
89.10
112.00
100.00
100.00
111.10
10.00
(10.90)
0.90
(10.00)
0.90
0.00
in the balancing account ($0.90) means that the utility
has recovered the previous year's shortfall. 2
The need for changes in authorized revenue between
rate cases----a taxonomy of decoupling mechanisms
Year 2. Authorized revenue of $100 and expected
sales of 1000 kWh are still in effect from GRC 1. In
addition, the utility must return $10 to ratepayers from
the previous year's overcollection. Accordingly, if the
utility collects $90 this year, it will be even with the
ratepayers. So, the Year 2 Price of $0.09/kWh is
calculated by dividing the total revenue that the utility
needs to collect ($90) by expected sales (still 1000 kWh).
However, in this case, the utility sells less electricity than
expected, resulting in a Collected Revenue of only
$89.10. The utility still reports revenue of $100, which
covers the $89.10 collected from ratepayers this year, the
$10 extra that was collected from ratepayers last year,
and $0.90 that appears in the balancing account,
representing money that the ratepayers will now owe the
utility in Year 3.
In our simple example, we showed how balancing
accounts ensure that authorized revenues are collected
over time. However, our example assumes that authorized revenue remains fixed between general rate cases.
This is an unrealistic assumption if expenses increase
from year to year while revenues remain fixed. The
problem may become more severe as the time between
general rate cases increases. Under traditional rate-ofreturn regulation, additional revenue associated with
increased sales offsets growth in expenses. Decoupling
regulations address the problem of increasing expenses
by making specific changes to the authorized revenue in
years between rate cases. Although balancing accounts
operate the same way in all decoupling mechanisms,
each decoupling mechanism has a unique method
for making between-rate-case changes to authorized
revenue.
Decoupling revenue adjustment mechanisms are currently used in the states of California, New York, Maine,
and Washington. California and New York developed
decoupling RAMs that rely on already established
procedures for adjusting the revenue requirement
between general rate cases. In contrast, Maine and
Washington developed new procedures for adjusting
authorized revenue between general rate cases (see Table
3). The precise formulation of these procedures is
described below.
Year 3. As a result of General Rate Case no. 2 (GRC
2), authorized revenue has increased to $111.10 based on
expected sales of 1010 kWh. In addition, the utility is
allowed to collect $0.90 from ratepayers because of the
previous year's shortfall. Accordingly, if the utility
collects $112 this year, it will be even with the
ratepayers. Thus, the Year 3 Price of $0.111/kWh cents is
calculated by dividing the total revenue that the utility
wants to collect ($112) by the expected sales (now 1,010
kWh). As it turns out, actual sales match expected sales,
resulting in collected revenues of $112. The utility
reports revenue of $111.10 for Year 3, and the difference
California ERAM. Revenue decoupling was implemented in California in 1982 by the California Public
Utilities Commission (CPUC), Decision 82-12-055
(1981). The stated purpose of California's Electric
Revenue Adjustment Mechanism (ERAM) "is to adjust
base rate (nonfuel) revenues for changes in revenues due
to unexpected fluctuations in sales during the test
period." Advantages of ERAM are said to be: (1) it
affords a utility a better opportunity to earn its authorized
rate of return, (2) it eliminates disincentives for the
utility to promote conservation, and (3) it stimulates
innovative rate design. 3 Currently, all regulated electric
48
Decoupling utility revenues from sales
Table 3. Comparison of ratemaking approaches
Traditional Ratemaking
California's ERAM
New York's RDM
Maine's RPC
Washington's RPC
Decouples Revenue
From Sales
Authorized Between-RateCase Revenue
Adjustments
Fuel Adjustment
No
Yes
Yes
Yes
Yes
Limited attrition, in a few states
Detailed attrition procedures
Broad attrition procedures
No. of customers
No. of customers
Yes, in most states
Yes
Yes
Yes
Yes, reintroduced with RPC
and gas utilities in California are subject to ERAM,
including Pacific Gas and Electric (PG&E), Southern
California Edison (SCE), San Diego Electric and Gas
(SDG&E) and Southern California Gas (SCG).
California sets rates and revenue using a future test
year and a three-year general rate case cycle. Accordingly, authorized revenue is based on assumptions about
what will happen in subsequent years. When ERAM was
implemented, California was already using a variety of
between-rate-case revenue adjustment techniques that it
continues to use with ERAM, including an attrition rate
adjustment (ARA), an annual cost-of-capital proceeding,
and a fuel adjustment clause. Under ARA, authorized
revenue is escalated using both recorded and forecast
escalation factors for labor and nonlabor operation
expenses. These escalation factors assume that cost
increases associated with sales and customer growth are
offset by increased productivity. Additions to the rate
base also are addressed by the ARA. Changes in the
adopted rate of return are addressed separately in the
annual cost-of-capital proceeding. California has also
used a number of ad hoc between-rate-case adjustments
associated with major construction projects such as the
Diablo Canyon and San Onofre nuclear generating
stations.
New York revenue decoupling mechanism. In 1988,
the New York Public Service Commission ordered New
York utilities to propose rate-making innovations that
would align the interests of utility shareholders and
customers. The Commission's goal was to provide
customers with the benefits of least-cost planning and
DSM using a mechanism that would also benefit utility
shareholders. As part of this reform process, Orange and
Rockland Utilities in 1991 adopted an ERAM-Iike
Revenue Decoupling Mechanism (RDM) to remove bias
against energy conservation (DiValentino et al., 1992).
Since that time, some form of decoupling has been
adopted by all New York utilities except one.
Like California, New York uses future test years and
has a tradition of multi-stage revenue filings in which
base rates are set and adjusted periodically to reflect
changes in specific costs. RDM was implemented in
conjunction with a provision for annual changes in
authorized revenue to recover increases in the cost of
Clause
providing services during a three-year rate plan. Adjustments are provided for fuel, operation and maintenance
expenses, rate base investment, and the cost of senior
capital. Most O&M expenses are subject to an inflation
attrition allowance based on a forecast gross national
product (GNP) price deflator index. Authorized revenue
is updated annually to reflect forecast additions to net
utility plant investment and related increases in depreciation. Changes in the utilities' capital structure, and the
costs of debt and preferred stock are updated annually.
These changes are reviewed by the New York Public
Service Commission through petitions and other required
filings.
Although the exact techniques used to adjust authorized revenue are different in New York and California,
both provide for between-rate-case adjustments to reflect
changes in fuel expenses, O&M expenses, rate base,
capital structure and cost of senior capital. One difference is that California adjusts the adopted return on
equity annually while New York fixes it between general
rate cases. Despite this difference, the decoupling
mechanisms used in the two states are essentially the
same.
Maine and Washington revenue-per-customer In
1991, Puget Power (Moskovitz and Swofford, 1991) and
Central Maine Power (Goldfarb and Spellman, 1993)
adopted decoupling revenue adjustment mechanisms.
According to the agreements authorized by utility
commissions in Washington and Maine, general rate
cases would proceed using existing methods, and the
timing of rate cases would continue to be on an "as
needed" basis. The new regulations decoupled revenue
from sales and recoupled revenue to the number of
customers. This decoupling revenue adjustment mechanism, called Revenue-Per-Customer (RPC), requires two
calculations. First, authorized revenue per customer,
which remains fixed until the next general rate case, is
computed by dividing allowed revenues (Rh) by the
number of customers (AP), as determined in a historic
test-year rate case:
g h
RPC h= _ _
N h"
49
Decoupling utility revenuesfrom sales
Second, authorized revenue for a given year t is
computed by multiplying the authorized revenue per
customer times the number of customers (N'):
R~'=R P C • iV'.
After each year, the difference between collected revenue
and authorized revenue is placed in a balancing account.
The following year's rates are adjusted to refund/collect
the over/undercollected balance.
Maine and Washington's RPC mechanisms are nearly
identical because both decouple revenue from sales and
recouple revenue to customers. However, prior to using
RPC, Puget Power did not have an adjustment clause
(with which this hydro-based utility would recover costs
for a variety of resources, not just fuel). Now, Puget
Power recovers fuel, purchased power (including hydro)
and conservation costs through an annual adjustment
mechanism that operates in conjunction with RPC.
Maine, in contrast, already had a fuel adjustment clause
that remained in effect after the implementation of RPC.
Evaluating the cost-tracking assumptions underlying
traditional rate-making and revenue per customer
decoupling
A fundamental principle of rate-making is to set rates
equal to the cost of service, which includes an allowance
for reasonable return on equity. In this section, we
examine empirically the cost-tracking assumptions
underlying both traditional rate-making practices and the
RPC decoupling approach. 4 If RPC decoupling does not
improve cost-tracking compared to traditional ratemaking practices, then decoupling revenues from sales
may hinder a utility's ability to earn its authorized
return.
Since most utilities operate with some form of fuel
adjustment clause, which passes fuel and purchased
power (i.e. variable) costs through to consumers, the
generic issue for cost recovery is how, between rate
cases, various rate-making practices allow revenues to
change in response to changes in nonfuel (i.e. nonvariable) costs. Traditional rate-making, by fixing prices
between rate cases, links the recovery of nonfuel costs to
changes in sales. The RPC approach used in Maine and
Washington, by establishing a balancing account, recoupies revenues to the number of customers and thus links
the recovery of these costs to changes in the number of
customers.
Traditional rate-making and revenue-per-customer
decoupling can be modeled as:
S,
N,
R'=RhSh and R,=R~~,
where R is revenue, S is sales, and N is the number of
customers. The subscript t refers to the current period,
SO
while h refers to the test year. Because we believe that
both S and N influence nonfuel costs, we need incorporate both into a single equation for purposes of
estimation. By algebraically rewriting these relationships, we see how to proceed:
s,
R,=Rh+Rh ~ - 1
andR,=Rh+R h ~ - 1
Now we can adjust R, separately for percentage changes
in S and N. Finally, let us also include the possibility that
R, may be a weighted average of S and N and that R, may
grow autonomously. We now have a more general
model:
Rt=Rh" /3"h+Rh" /31~h--1 +Rh" fl2 ~'h--1
We can simplify this model to the single period case
(t=h+ 1), subtract Rh from both sides and divide by R h.
Note that the terms in brackets are just the annual or
year-to-year percentage changes in S and N, respectively;
when t=h+ 1, we will call these %AS and %AN:
R , - Rh_Rh •/3+Rh "/31%AS+Rh •/32%AS - Rn
Rh
Rh
Remembering that nonfuel revenues should equal nonfuel costs because we have defined costs to include the
allowed rate of return, s we rewrite the last equation in
terms of nonfuel cost and add the standard regression
error term, ~:
%AC=fl0+/3~ • %AS+t2" %ANe,
(1)
where %AC indicates the percentage change in nonfuel
cost for one year and /30=/3- 1. This equation will be
referred to as Equation (1).
We now run several regressions, most of which are
specific cases of Equation (1) above: two with sales, two
with customers, and one with both plus an autonomous
trend. The estimated coefficients measure the strength of
the cost-tracking relationships embodied in both traditional rate-making and RPC decoupling. They also allow
us to comment on a final, potentially less biased,
Revenue Adjustment Mechanism. Each regression is run
on approximately 3300 data points from a data set
consisting of year-to-year changes in nonfuel costs from
25 years of aggregate financial statistics from 160
investor-owned utilities.
The sales regression yields:
%AC=0.399 • %AS
(24.5)
R2=0.15,
where the number in parentheses refers to the t-statistic
Decoupling utility revenues from sales
associated with the regression coefficient estimate. This
result says that only 0.40% of a change in sales is
correlated with a 1% change in nonfuel costs. This
should be compared to traditional rate-making, which is
based on the implicit assumption that
%AC= 1.0 • %AS.
This assumption says that every change in sales perfectly
correlates with changes in nonfuel costs. Though it
appears that traditional rate regulation provides significant rewards in the short run to utilities that have
typical sales growth, this model suppresses any effects of
increased sales on long-term costs.
We need to run one more regression with sales in order
to find the true incentive for load building. This
regression includes an intercept or constant term, as
follows:
%AC=0.032+0.099 • %AS
(20.1) (4.6)
R/=0.01.
This regression shows that the change in cost of
0.399%AS discovered in the previous regression was not
caused solely by %AS but was simply associated with it.
Thus, if a utility deliberately achieved a higher %AS, it
would probably expect this change to be associated with
a cost increase 10% as great instead of 40% as great as
the extra change in %AS. Thus, the cost of load-building
is quite low, and the compensation from traditional ratemaking is 90% in excess of this c o s t . 6
Next, we turn to a regression involving the number of
customers in order to examine the basic assumption
underlying
the
revenue-per-customer decoupling
approach. We begin again with the regression without a
constant term. The estimated version of this regression
is:
%AC=0.725 • %AN
(23.3)
R2=0.14.
We see that, on average, RPC over-rewards by somewhat
more than can be observed in the historic data (compare
1.0 to 0.72).
We need to run one more regression with customers in
order to find the true relationship between customer and
nonfuel cost growth. This regression includes an intercept or constant term, as follows:
%AC=0.030+0.294 • %AN
(22.7) (8.5)
R/=0.02.
This second customer regression shows that the change
in cost of 0.725%AN from the previous regression was
mostly not caused by %AN but was simply associated
with it. The cost of serving additional customers is
substantially lower, as evidenced by the second regression's coefficient of 0.294%AN.
Finally, we present a comprehensive regression
reflected in Equation (1), which considers all three
influences--sales, number of customers, and autonomous change--simultaneously. In addition to the
inclusion of a number of customers and a constant, we
specify a sales-related term in the form of sales-percustomer (SPC). 7 This regression yields the following
coefficients, standard errors and R2:
%AC=0.029+0.035 • % A S P C + 0 . 3 0 5 • % A N
(18.5) (1.6)
(8.7)
R2=0.02.
We can see that the effect of a 1% change in the number
of customers is roughly nine times larger than the effect
of a similar change in sales-per-customer (compare 0.305
to 0.035).
Although the effect of customers (compared to salesper-customers or the constant term) is substantial on
average, it is important to note the extremely low R2 of
this regression. Such a low R2 does not indicate that the
effect of customers or sales-per-customer is either poorly
estimated or small; instead, it simply indicates that other
strong effects have been omitted from our regression.
Some of these omitted effects are undoubtedly idiosyncratic; others may be factors that might be addressed
explicitly through attrition adjustments (such as interest
rates).
Our results show that one-year changes in the number
of customers have a fairly strong one-year impact on
nonfuel costs but that one-year changes in sales have a
rather weak effect. So the proponents of RPC are correct
in arguing that RPC does no worse than traditional ratemaking in tracking nonfuel costs (it actually does slightly
better). Nevertheless, even after accounting for the
effects of these two variables and the autonomous trend
or constant term, the vast majority of the year-to-year
variation in nonfuel costs remains unexplained. In other
words, neither the traditional basis for adjusting revenues
to account for changes in nonfuel costs nor that
embodied in RPC does a very good job of tracking these
costs. Thus, as long as cost of service is an important
rate-making principle, a periodic rate case will be needed
under both traditional rate-making and RPC decoupling.
The historic rate impacts of E R A M in California
Much of the current controversy surrounding decoupling
has centered on its rate impacts that arise as a
consequence of the balancing account required to
implement decoupling. The issues range from philosophical implications of risk-shifting to pragmatic concerns
regarding the magnitude of accrued balances and their
potentially dramatic impacts on rates. The historic record
of decoupling from the state with the longest experience,
California, provides a context for discussing these
issues.
As has been well-documented in Marnay and Comnes
(1990), California rate-making is a complicated process.
51
Decoupling utility revenues from sales
Rates are adjusted both through triennial general rate
cases and through a variety of annual adjustments, of
which ERAM is only one. Annually, there is also a fuel
adjustment clause and an attrition adjustment, in addition
to several other less-well-known or less-systematicallyused adjustments. Retail rates reflect the net impact of all
of these adjustments.
We obtained revenue requirement and rate data for
California's utilities for the entire time that ERAM has
been in effect. The primary challenge for documenting
the rate impacts of ERAM was identifying a consistent
set of records to use for our analysis. Wherever possible,
we relied on publicly available rate decisions on file at
the California Public Utilities Commission (CPUC).
Nevertheless, we were not able to locate decisions for
several years and have relied on company-supplied data
instead (Eto et al., 1994).
The data clearly indicate that, in the overall context of
California rate-making, the clearing of ERAM balances
has accounted for only a small portion of the total change
in revenue requirements in the last 10 years. Adjustments
resulting from ECAC have been, by far, the main source
of changes to revenue requirements. The compound
effect of multiple, annual adjustments to revenue requirements is highlighted by the relatively small role played
by the GRC in adjusting revenue requirements.
Electricity rate changes with and without ERAM
The annual rate changes depicted in Figure 3 include
ERAM adjustments to revenue requirements. In order to
determine the effect of ERAM on rates, we compared the
actual rates, as reported above, to hypothetical rates,
(a) Pacific Gas & Electric Company
35
30
California electricity rate history
In the first stage of our review, we disaggregated the rate
changes for each of the utilities into four types of
changes, resulting from: (1) general rate cases or GRCs;
(2) fuel-adjustment clauses, which California labels as
energy cost adjustment clauses or ECACs; (3) the
decoupling mechanism, known as the electric revenue
adjustment mechanism or ERAM; (4) and all others. The
GRC includes primarily nonfuel revenue changes that
are determined in years when a complete rate case is
held. The ECAC consists primarily of fuel cost changes
resulting from retroactive adjustments that 'true up'
previous rniscollections and prospective adjustments that
are based on expected future fuel expenditures. ECACs
also include payments to Qualifying Facilities (QFs) and
recovery of utility DSM incentives. The ERAM balance
should only contain income from sales-related mismatches between authorized revenues and actual
revenues collected. In many cases, the ERAM balance
included items not related to over- or undercollections
resulting from sales fluctuations. Usually, we were able
to identify these other items and move them from the
ERAM category to the 'other' category. The 'other'
category includes a wide range of revenue requirement
changes, mainly, attrition adjustments. However, many
one-time adjustments that relate to particular construction projects or changes in tax laws are also included in
this category.
Figure 3 shows the changes in revenue requirements
and retail rates from 1983 to 1993, for California's three
state-regulated electric utilities: Pacific Gas & Electric
Co. (PG&E), Southern California Edison (SCE), and San
Diego Gas & Electric Co. (SDG&E), respectively. The
net effect of positive and negative revenue requirement
adjustments does not equal the change in retail rates in
this figure. The reason is that the sales forecast also
changes in the annual adjustments.
52
25
J~
o 20
~ 15
0
P 10
0
O.
m
5
0
~4
"m
_
<: - 1 0
-15
I
I
I
' =
I~1
I
I
I
I
I
I
19831984 19851986 196719881989 1990 19911992 1993
(b) Southern California Edison
25
~
ZU
15
_~ 1o
-10
I
-15
I
I
I
I
I
I
I
I
I
1984 1985 1986 19871988 1989 1990 1991 1992 1993
(c) San Diego Gas & Electric
ao _
m 25 --
20
•~
15 --
I
r~
~
I EcAc
ERAM
fill rrrrrnG . c
Illl
O.,er
II I I
-,Avg. Retail R==te6
-10 1
_lS i
I
1983 1984 1985 1986 1987 1988
I
1989 1990 1991
I
I
1992
Figure 3. Changes in authorized revenue requirement and
average retail rates.
Decoupling utility revenues from sales
which exclude ERAM. To do this, we subtracted the
ERAM balance from each year's revenue requirement
and divided by authorized sales. Figure 4 shows annual
changes in retail rate levels, both with and without the
ERAM, for PG&E, SCE, and SDG&E, respectively.
Table 4 summarizes these findings. The data indicate
that, just as the magnitude of the ERAM adjustments has
been a small factor influencing changes to authorized
revenues, the rate impacts of ERAM have also been
small. For PG&E, ERAM adjustments actually reduced
rate volatility, as evidenced by a reduction in the standard
deviation of annual rate changes from 9.6% to 7.5%. For
SCE and SDG&E, ERAM has led to a small increase in
volatility.
The history of decoupling in California suggests that:
(1) decoupling has had a negligible effect on rate levels
(a)
30
m
20
°
15
o
lO
®
0
¢-
0
25 - -
,"
o
Pacific Gas & Electric Company
B
/I
A
tl
II
D _ ...t1
"~'%
~ {, "'"-,/'oN
~I
0 I
I~1J"
-6-
o.x,,//
<-1o
I
I
"~
I
I
I
I
I
I
I
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993
(b)
Southern
California Edison
_
¢e~
"
o
10
"Q
5
e)
0
t-x
,
-5 I
<
i' V
-t9
I
-15
I
I
I
I
I
I
I
I
1984 1985 1986 1987 1988 1989 1990 1991 19921993
(c)
15-0
10
""
(3
5
cQ
0
0
-tO
<
I
S a n Diego Gas & Electric
--
,=.
•
o
- % ° \./
7
,'', l "
_
I
-15
--2O
Rates with ERAM
R a t e s without ERAM
I
I
I
P
1963 1984 1985 1988 1987
I
I
I
1988 1989 1990
1
1
1991 1992
Figure 4. Changes in average retail rates with and without
ERAM.
and has, for PG&E, actually reduced rate volatility; and
(2) rate changes resulting from California's fuel adjustment clause have had far more dramatic effects on rates
and, consequently, on the shifting of business risk from
utility shareholders to utility ratepayers. In our opinion,
the utility policy issue is that we must consider
decoupling in the context of a comprehensive framework
that jointly considers all sources of rate risk and rate
volatility.
Concluding
thoughts
We believe that utilities and regulators who are considering decoupling should consider three key issues. First,
the importance of lost revenues and therefore of
decoupling depends strongly on pre-existing features of
regulation; foremost among these is the frequency of rate
cases and the design of fuel adjustment clauses because
they directly influence the size and persistence of the
disincentives that decoupling seeks to address. At the
same time, we also believe there are other incentives
(and disincentives) for utilities to build load that are
distinct from the lost revenue problem. Regulatory
reforms, therefore, should not focus exclusively on lost
revenues but instead take a broad perspective when
trying to align utility incentives with the objectives of
integrated resource planning.
Second, adoption of a decoupling mechanism requires
consideration of the means by which revenues are set
between rate cases, especially the means for allowing
revenues to change in response to changes in nonfuel
costs. Our examination of the empirical record suggests
that, for short periods of time, neither sales growth
(which underlies traditional rate-making) nor customer
growth (which underlies RPC) provides a very powerful
explanation for changes in these costs. In other words,
the revenue-per-customer approach (in addition to
decoupling sales from revenues) will, on average, do no
worse than traditional rate-making in recovering these
costs. Thus, if cost-recovery is an important rate-making
objective, it is a separable concern from decoupling, and
other approaches should be considered to address it, such
as attrition mechanisms or future test years. Hirst and
Blank (1994) offer a promising approach.
Third, the record in California suggests that the issue
of the additional rate volatility introduced by decoupling
has been overemphasized. We further believe that
discussions of the additional rate volatility and riskshifting associated with decoupling should consider all
sources of rate volatility and risk-shifting in rate-making.
Then, what the risks are and who is best suited to bear
them can be made explicit and their treatment made
comprehensive rather than piecemeal.
While restructuring in the US and around the world
will change the utility industry dramatically, we expect
53
Decoupling utility revenues from sales
Table 4. Annual percent changes in revenue requirements and retail rates
PG&E
Re~
Req.
(%)
SCE
Rates
~
ERAM
(%)
Rat~
w/o
ERAM
(%)
Re~
Req.
(%)
SDG&E
Rates
~
ERAM
(%)
Rat~
w/o
ERAM
(%)
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
2.0
5.3
24.3
-5.6
7.6
0.7
15.9
7.7
14.8
2.2
3.5
4.0
0.4
22.1
-2.4
-7.9
4.8
9.4
7.1
11.7
i.5
3.0
1.7
3.8
26.9
-4.3
-10.0
-0.2
13.7
12.5
7.8
-1.9
3.0
-4.4
21.8
2.5
18.1
-9.2
15.5
4.8
12.7
2.4
-4.6
-3.5
9.6
-0.2
15.1
-11.5
12.6
3.0
6.9
2.5
-2.8
-4.0
11.7
-2.3
16.1
-10.0
10.2
5.1
3.4
3.1
-0.9
Mean
SD
7.1
8.0
4.9
7.5
4.8
9.6
6.0
10.1
3.2
7.7
3.2
7.5
Rates
~
ERAM
(%)
Rat~
w/o
ERAM
(%)
7.3
4.0
-7.9
15.2
-2.2
-2.9
-8.8
12.4
7.9
-0.8
8.4
-1.7
-10.5
10.1
-7.3
-9.9
-10.7
3.1
4.0
-0.3
7.2
-2.7
-3.7
8.3
-18.3
2.1
-11.4
-1.7
5.8
-1.7
2.4
7.8
-1.5
7.4
-1.6
7.9
Re~
Req.
(%)
t h a t s o m e utilities w i l l c o n t i n u e to a d m i n i s t e r r a t e p a y e r funded customer energy-efficiency programs (Eto and
Hirst, 1996). D e c o u p l i n g c a n p l a y an i m p o r t a n t r o l e in
transforming utilities from sellers of a least-cost energy
c o m m o d i t y to p r o v i d e r s o f l e a s t - c o s t e n e r g y s e r v i c e s , b u t
it is n o p a n a c e a . A l t h o u g h it c a n s u c c e s s f u l l y e l i m i n a t e
a n i m p o r t a n t d i s i n c e n t i v e f o r utility D S M p r o g r a m s , it
m u s t b e d e s i g n e d c a r e f u l l y to t a k e e x p l i c i t a c c o u n t o f
o t h e r r e g u l a t o r y o b j e c t i v e s , s u c h as c o s t - r e c o v e r y a n d
rate volatility.
of California ERAM.
*l'he RAMs used in California and New York recouple revenues to
detailed annual adjustments that are difficult to characterize precisely.
Thus, they are not amenable to this type of analysis.
5Recall that we are interested in marginal, not total, profitability.
Marginal profitability is thus measured by deviations from the allowed
rate of return.
6Even this estimate of the incentive is too high because we should have
looked at the effect of sales-per-customer on costs instead of the effect
of total sales on costs.
7Simply including sales would be inappropriate because some sales
growth is already accounted for by the number of customers. Using
sales-per-customer allows us to estimate the residual sales-driven costs
separately from those that are driven by the number of customers.
tHowever, increases in authorized revenues may not translate automatically into increases in profits. Building rate base generally requires
new capital; the increased cost of debt may not be fully covered by the
authorized increase in earnings. In other words, the basic premise of the
Averch-Johnson thesis, that the rate-of-return exceeds the cost of
capital, may not always be true. If additional shares must be sold to
raise capital, shareholder equity will be diluted and, other things being
equal, earnings per share will drop. In addition, returns from each
project to build the rate base as well as the size of the utility will
influence the profitability of individual rate base additions. Fundamentally, if additions to generating capacity cost more than historic average
costs, rates will increase. Depending on the options available to utility
customers (i.e. their price elasticity of demand), rate increases could
have disproportionate effects on future sales and thus on earnings.
Finally, in a world where utilities do very little of the building of new
generation, the continuing relevance of an incentive to build load needs
to be re-examined.
2In order to make the Basic RAM simple to understand, we have
suppressed the interest component of the balancing account and
matched Year 3's expected and actual sales. The balancing account's
interest rate is usually pegged to the cost of short-term debt (although
some states use the utility's weighted average cost of capital). To the
extent that the two rates differ, the utility could be motivated to increase
or decrease the decoupling balance. In our analysis, we assume that the
interest rate on the balancing account and the cost of funding the
balancing account are the same, eliminating the motivation to game the
balancing account.
3The history, mechanics, and policy issues of California's ERAM have
been well documented. See, for example (Marnay and Comnes, 1990).
Our object in this discussion is to review and contrast it with other
decoupling approaches. In Section 4.0, we summarize the rate impacts
References
54
California Public Utilities C o m m i s s i o n (CPUC) (1981) Decision 93887. Application 60153.
DiValentino, L. M., Dittrich, T. L., Cuccaro, J. E. and
Freedman, A. M. (1992) R e v e n u e decoupling plus incentives
m e c h a n i s m . In Regulatory Incentives for Demand-side
Management, eds S. Nadel, M. Reid and D. Wolcott,
pp. 125-139.
Energy Information Administration (EIA) (1993) Financial
Statistics
of Major Investor-Owned Electric
Utilities.
D e p a r t m e n t o f Energy/Energy Information Administration,
Washington, DC.
Eto, J. and Hirst, E. (1996) What kind o f future for energyefficiency programs, The Electricity Journal 9(5), 76-82.
Eto, J. Stoft, S. and Belden, T. (1994) The Theory and Practice
ofDecoupling. LBL-34555. Lawrence Berkeley Laboratory,
Berkeley, CA.
Goldfarb, L, K. and Speliman, R. E (1993) D S M Incentives
+ E R A M = B o t t o m - L i n e Impact. Proceedings of the 6th
National Demand-Side Management Conference. Electric
P o w e r Research Institute (EPRI), 259 pp.
Hirst, E. and Blank, E. (1994) Solutions for regulatory
disincentives for utility D S M programmes. Utilities Policy,
4(2), 105-112.
Hirst, E., Goldman, C. and Hopkins, M. E. (1991) Integrated
Decoupling utility revenues from sales
resource planning: electric and gas utilities in the USA.
Utilities Policy 1(2), 172-186.
Krause, E and Eto, J. (1988) Least-cost Planning Handbook for
Public Utility Commissioners: Volume 2--The Demand Side:
Conceptual and Methodological Issues. National Association of Regulatory Commissioners (NARUC), Washington,
DC.
Mar'nay, C. and Comnes, G. A. (1990) Rate-making for
Conservation: The California ERAM Experience. LBL28019. Lawrence Berkeley Laboratory, Berkeley, CA.
Moskovitz, D. (1989) Profits and Progress Through Least Cost
Planning. National Association of Regulatory Commis-
sioners (NARUC), Washington, DC.
Moskovitz, D. and Swofford, G. (1991) Decoupling sales and
profits: an incentive approach that works. The Electricity
Journal 4(6), 46-53.
Nadel, S., Reid, M. and Wolcott, D. (ed.) (1992) Regulatory
Incentives for Demand-Side Management. American Council
for an Energy Efficient Economy, Washington, DC.
Train, K. E. (1991) Optimal Regulation, The Economic Theory
of Natural Monopoly. MIT Press, Cambridge, MA.
Winpenny, J. (1992) Powerless and thirsty? The outlook for
energy and water in developing countries. Utilities Policy
2(4), 290-295.
55