Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 75
s
Behavioral Study of
Appliance Efficiency Decisions
Energy consumption is determined in part by consumers'
choices about purchasing, and then about using new and
replacement technology. Choices about the energy effici-
ency of appliances are a significant instance of technol-
ogy choice for several reasons. A large part of the long-
run ability of consumers to modify energy consumption
patterns is embodied in the energy-using technology they
choose and is fixed for the lifetime of that equipment.
Two ways of affecting consumer choice, standards for
appliance efficiency and incentives for selected technol-
ogies, are often considered or used as energy policy
tools. And consumer information and marketing programs
may also be effective tools for energy management by cor-
recting consumers' misperceptions and encouraging purchase
of more efficient appliances. Thus, wherever possible,
energy policy analyses and forecasting systems should
incorporate adequate descriptions of equipment technology
and of choices about the purchase and use of technologies.
To capture the effects of information and consumer percep-
tion that may be important in the design of policies and
programs, these descriptions should incorporate a sub-
stantive account of the behavioral process by which con-
sumers assess the alternatives and choose among technolo-
gies.
Many current policy simulation models assume that
appliance technology is chosen using a criterion of life-
cycle cost minimization. This assumption excludes con-
sideration of factors that may be behaviorally important,
including appliance capacity, safety, convenience, brand-
name loyalty, marketing of different models, dealer char-
acteristics, and so forth. It also ignores what may be
an important interaction between efficiency and utiliza-
tion decisions: consumers may use more efficient appli-
75
OCR for page 76
76
ances more heavily, and may anticipate this when assessing
technological alternatives. Furthermore, even if the
life-cycle cost criterion were an acceptable description
of behavior, its realization in models now often fails to
include factors that may also be important, such as con-
sumer expectations about future fuel prices and attitudes
toward risk.
In examining appliance choice, this chapter differs
from Chapters 2, 3, and 4--each of which explored a class
of factors that influence demand--by looking at a class
of demand-related behavior that may be influenced by any
of the factors discussed in the previous chapters--prices,
incentives, and information. This chapter, then, is
closer to the typical work of energy demand analysts:
focuses on energy demand in a particular sector of the
energy economy. We first consider some of the issues
raised by current models of appliance choice and identify
particular areas where there are gaps in available infor-
mation and in behavioral theory. We then outline the
experiments and field data collection efforts that would
be required for an ideal and comprehensive behavioral
tuav Of cho ices of appliance technoloav.
~ ~ ~ Finally, we
discuss a few practical data collection efforts that
would provide useful incremental information on appliance
choice.
ISSUES AND LIMITATIONS OF CURRENT ANALYSIS
There are a very limited number of major empirical bases
for formal modeling of appliance fuel choices, efficiency
choices, and decisions about utilization. There are a few
market studies of the initial costs and technological
characteristics of purchased appliances:
A ~
room air con-
ditioners (e.g., Hausman, 1979; Brownstone, 1980); heating
and central air conditioning systems (e.g., Dubin and
McFadden, 1984); and refrigerators (e.g., Meter and
Whittier, 1983). There are also laboratory bench studies
and engineering projections of the costs of appliances of
various efficiencies (Hirst and Carney, 1978) and thermal
models used to calculate the sizes and levels of use of
heating and cooling systems in buildings with well-
specified structural characteristics. In addition, there
are statistical analyses of appliance purchases or hold-
ings using household survey data--the National Interim
Energy Consumption Survey and the Residential Energy Con-
sumption Survey (RECS) of the Energy Information Adminis-
OCR for page 77
77
"ration; the Annual Housing Survey; and national surveys
conducted by the Midwest Research Institute (MRI), the
Washington Center for Metropolitan Studies, and numerous
utilities--or state and regional cross-sectional, time-
series data. However, none of these data sources contains
sufficient information on appliance efficiencies to permit
direct study of choices about efficiency.) Consequently,
attention has concentrated on appliance fuel choices.
One research approach has been to assume some version
of the life-cycle cost-minimization hypothesis and to
develop estimates of the implied discount rates that
accord with observed choices (e.g., Hirst and Carney,
1978). Another approach has been to fit discrete choice
models that allow that consumers sometimes do not minimize
cost, possibly because they respond to factors other than
initial and operating costs (see McFadden, Puig, and
Kirschner, 1977; Berkovec, Hausman, and Rust, 1983; Dubin
and McFadden, 1984). The latter studies permit testing
of the life-cycle cost-minimization hypothesis, albeit in
models in which the functional forms of equations are
chosen for convenience rather than on the basis of data
or established theory. These studies generally reject the
cost-minimization hypothesis in its simpler forms. Fur-
ther, the choice models they estimate meet some of the
criteria for behavioral explanation. For example, esti-
mated parameters are reasonably stable across data sets,
and parameters estimated in some of the studies have gen-
erated accurate accounts of fuel shares among appliances
purchased in different regions and time periods (Goett and
McFadden, 1984).
But these models are still considerably short of an
adequate behavioral theory, even for fuel choice. The
models offer inadequate treatment of consumer information
and expectations, a critical limitation in light of the
policy need to predict the effects of marketing programs.
There is considerable error in measuring consumer charac-
teristics and appliance attributes. Finally, the models
contain a large number of parameters and impose restric-
tions that are neither implied by behavioral theory nor
supported empirically. This does not in itself imply that
these models are useless for policy purposes. Even imper-
fect models may help account for physical and economic
1A partial exception is the study by Hausman (1979) of
room air conditioner choice that used MRI data on indivi-
dually metered appliances in a small sample of households.
OCR for page 78
78
constraints and may identify variables to which consumer-
response is sensitive. However, policy analysts should
avoid making assertions that are not supported by behav-
· . .
ioral evidence, and improved congruence between nolicv
models and behavioral theory should be a constant objec-
tive.
In short, existing empirical studies of appliance tech-
nology and fuel choice have important limitations, and
there are virtually no studies of choices of appliance
, ~
efficiency. This state of knowledge is partly due to lack
ot data and partly to weaknesses of theory. Questions on
four topics must be addressed if analysis of appliance
efficiency is to improve: (1) the analytic representation
of appliances; (2) analytic representation of energy price
expectations; (3) determinants of appliance use; and (4)
the behavioral processes that affect consumers' decisions.
Asking these questions focuses attention on the assump-
tions and limitations of the hypothesis of life-cycle
cost-minimization as a foundation for policy modeling.
Analytic Representation of Appliances
Many questions need to be addressed in considering how
appliances are represented analytically: How is the set
of feasible technological alternatives defined? Is there
a continuum of idealized alternatives or a finite list of
brands and models? IS efficiency to be measured by manu-
facturer's rating, bench testing, or field testing? What
dimensions of technology other than fuel and efficiency--
such as capacity, noise level, size, durability, safety,
convenience, and so forth--need to be included as behav-
iorally relevant attributes? How is the initial price of
equipment defined: by list price, observed transaction
price, or engineering calculation? If engineering studies
are used to estimate the costs of alternative designs,
what relationship holds between fabrication cost and
prices? How should the prices of appliances supplied in
new construction be imputed in life-cYcle cost calcula-
· . . .
tlons:
by using equipment costs to contractors or by
regression analysis of building prices as a function of
attributes of the equipment in the building?
Most current models are derived from the early model
of Hirst and Carney (1978), which considers discrete fuel
alternatives and represents the trade-off of efficiency
and initial cost in terms of only three parameters.
Appliance attributes such as capacity are incorporated
OCR for page 79
79
into the analyses only at fixed levels; the model does not
subject them to behavioral control. The estimates of
technical efficiency are determined from engineering
principles, and there is only limited validation of the
link from idealized fabrication cost to market price.
Future refinements of such models might include: repre-
sentation of efficiency alternatives in terms of the
brands and models actually available to consumers; direct
measurement of purchase prices; representation of effici-
encies as rated by manufacturers; field testing of real-
ized efficiencies; measurement of attributes other than
purchase price and efficiency; data collection on the
degree of discounting by distributors from list prices,
particularly in new construction; and measurement of the
"perceived value" of equipment alternatives in new con-
struction by hedonic regression. 2
Analytic Representation of Energy Price Expectations
In considering energy price expectations, two major ques-
tions stand out: How are future energy prices, as antici-
pated by consumers, to be estimated? Which assumption is
most justified for behavioral models of expectations:
static prices at current levels, perfectly anticipated
prices, or prices changing at historical trend rates?
Most current studies assume a static expectation that
current real fuel prices will continue to prevail in the
future. Recent econometric investigations of expectation
formation (primarily of financial markets, for which con-
siderable data are available to market participants and
to observers) suggest that consumers are less naive and
use information to form "models" of future market behavior
(Eichenbaum and Hansen, 1983; Mishkin, 1983). The limited
evidence on energy markets is consistent with an
adaptative-expectations hypothesis for gasoline prices
(see Chapter 2), but this conclusion may not be applicable
to electricity and natural gas markets, in which con-
2 In this procedure, detailed data on actual equipment
choices or laboratory simulations of equipment choices
are used to regress prices paid for appliances on a
number of appliance characteristics measured or simulated
in the study. The resulting regression equation repre-
sents the "perceived value" of each of the character-
istics.
OCR for page 80
80
sumers' decisions are less routine and information is less
systematic. Still, the importance of expectations for
policy affecting appliance purchases makes the behavioral
process of forming expectations a critical topic for
study.
Determinants of Appliance Use
In studying appliance use analysts should address at least
two major questions: How is anticipated future use of
appliances estimated? If the level of use is not assumed
to be fixed and beyond behavioral control, how does one
incorporate feedback from appliance operating costs into
level of use in estimating life-cycle cost?
Current models using the criterion of life-cycle cost-
minimization assume that, at least from the standpoint of
a purchaser at the time of an appliance choice, the inten-
sity of appliance use is fixed and beyond behavioral con-
trol. While some of the models allow for changes in use
once appliances are in place, this approach implies com-
plete lack of planning by consumers. The possibility of
feedback from anticipated operating cost to utilization
to choice of technology poses a fundamental problem for
the life-cycle cost-minimization hypothesis, since it
implies that consumers may trade initial costs against the
benefits of greater future utilization: for example, by
buying an energy-efficient air conditioner, a household
may be able to afford to keep cool more of the time.
Assessing this possibility requires study of the joint
choice of technology and utilization, with a theory of
choice that allows trade-offs and with survey data that
encompass both purchases and subsequent appliance use.
Consumers' Decision Processes
Analysts should address several questions on decision
processes: How are future costs to be discounted under
the cost-minimization hypothesis? How are appliance
durability and consumers' replacement strategies taken
into account? Are there perceptual elements that lead to
different discount factors in fuel choice and in effici-
ency choice? What is the impact of uncertainty about
appliance performance or about resale market value? What
is the impact of credit constraints?
OCR for page 81
81
A number of studies have estimated discount rates by
assuming the weights given to initial and operating costs
in discrete choice models of appliance fuel choice. These
studies have usually ignored the effects of expectations
about fuel prices, interest rates, and equipment life and
of feedback from utilization. The empirical results are
quite erratic, reflecting these factors as well as prob-
lems in accurately measuring initial and operating costs
and limitations of the life-cycle cost-minimization
hypothesis as a behavioral model.
Refinements require a recognition that the life-cycle
cost-minimization criterion must be modified under at
least three conditions: (1) when the alternatives vary
not only in energy efficiency but in service attributes,
such as capacity and convenience; t2) when there is feed-
back from anticipated levels of appliance use; (3) when
consumer behavior is likely to be more complex than the
criterion suggests. Some of the objectives of a behav-
ioral theory may be achievable by adopting a psychophysi-
cal version of the traditional economic model of prefer-
ence maximization, in which consumer behavior is assumed
to vary around optimal choice according to some mathe-
matical distribution. This is a relatively mechanical
approach to accounting for deviations from life-cycle cost
minimization. It is easily adapted to the task of quanti-
tative simulation (see Thurstone, 1927; Tver sky, 1972;
Luce, 1977; McFadden, 1981), but it does not provide a
useful framework for studying the relationships of infor-
mation processing to choice or the social aspects of the
transmission or use of information.
A second approach is to conduct experimental or mar-
keting studies of specific behavioral phenomena. Two
examples are a controlled field experiment to test the
effects of alternative cable television advertising
"treatments" on consumer awareness of appliance costs and
the use of focused group discussions to investigate the
structure of consumers' information networks. Such
studies can potentially provide information critical to
the design of marketing programs for energy-efficient
appliances. However, they are not well suited to develop-
ing general purpose quantitative policy simulations.
THE IDEAL DATA FOR BEHAVIORAL STUDY
OF APPLIANCE CHOICES
A comprehensive behavioral study of choices affecting
appliance efficiency would concurrently examine the
OCR for page 82
82
effects on appliance choice of fuel, efficiency, expected
utilization levels, and a range of features of appliances
and their marketing. It would also address the efforts
of government and utility companies to influence consumer
choice. This comprehensiveness imposes requirements on
both data collection and on the method of analysis. The
ideal data base must contain descriptions of all appliance
alternatives, chosen and unchosen, as well as of con-
sumers' actual choices and their subsequent intensities
of use of the appliances. The efficiencies of alternative
appliances should be measured both as perceived by con-
sumers (e.g., manufacturers' labels, the content of adver-
tising, and attitudes about the appliances) and in terms
of field performance. The initial prices of the appli-
ances should be measured as actual transaction costs.
Appliance use should be observed under a sufficient vari-
ety of conditions to reliably assess behavioral response
and that information should be used in describing any
feedback from utilization to appliance choice.
To carry out the collection of such data in a con-
trolled setting, with availability and attributes of
appliances, prices, and other factors under experimental
control, would be a formidable undertaking. Less defini-
tive, but nevertheless informative studies, could be car-
ried out with field data from natural experiments. Col-
lection of such data by the U. S. Department of Energy
would require a different orientation from that of annual
surveys such as RECS that provide census-like data on
energy consumption. Such studies would necessarily be
smaller and more specialized, with intensive data collec-
tion. Most of the elements of such surveys are already
in use by various electric utilities, which collect
appliance serial numbers, develop panel surveys, and
attach meters to individual appliances. The U. S. Bureau
of Labor Statistics routinely collects data on appliance
transaction prices and uses hedonic regression techniques
to relate price to equipment attributes. Without a
national framework for simultaneous collection of such
data, however, it is unlikely that enough standardization
and comprehensiveness will be achieved to give good an-
swers to questions about how consumers make choices about
the energy efficiency of purchased appliances.
Any general analysis of these questions based on field
surveys will certainly leave some key questions unan-
swered. Specialized experiments and marketing studies
may shed light on specific behavioral phenomena that are
critical to particular policy problems and may help make
OCR for page 83
83
sense of the relationships observed in field data. For
example, research by cognitive psychologists shows that
people's choices can depend on the way in which the
alternatives are presented to them (Tversky and Kahneman,
1981). Small-scale experiments could be conducted to
explore the implications of this phenomenon for the mar-
keting of energy efficiency in appliances.
With regard to methods of analysis, there will always
be a conflict between researchers who hesitate to extrapo-
late from weak empirical evidence and policy analysts who
seek the least inadequate method of extrapolating as far
as is required by an immediate question or problem. Given
the current state of behavioral theory, the limits to
experimentation with human subjects, and the problems of
understanding complex behavior by use of survey data, any
practical policy tool will almost inevitably severely
compromise scientific standards. Consequently, policy
modeling should be a continuing process of scientific
attack, invalidation, and improvement of interim simula-
tion tools. With specific reference to the problem of
analyzing appliance efficiency standards, an initial goal
might be to carry out the data collection and analysis
necessary to describe efficiency choice at the level of
empirical precision that has already been achieved for
fuel choice, while at the same time developing the experi-
ments and behavioral knowledge necessary to refine the
descriptions of both fuel choice and efficiency choice.
PRACTICAL ALTERNATIVES FOR DATA COLLECTION
Most of the quantitative policy models currently in use
at the U. S. Department of Energy and elsewhere are based
on the assumption that equipment efficiency choices are
governed by life-cycle cost-minimization criteria or on
some variant that admits exceptions in which additional
aspects, such as capacity, may influence choice. The
most narrowly defined cost-minimization models would be
rejected by most social scientists as behaviorally
unrealistic, particularly because of their insensitivity
to such factors as consumer information, which may be
important foci of policy. As a result, most existing
models would be judged inaccurate for forecasting. Such
models, if modified to make them consistent with classical
psychophysical and economic ~laws" and fitted to survey
data, may offer reasonably accurate bases for simulations
of policies that operate primarily through economic vari-
OCR for page 84
84
ables, but even these models still have to be validated
for such applications.
One task for those people who challenge current models
is to propose improved alternatives that are practical
enough to be adopted by policy makers and planners. A
recommendation to scrap the current generation of models
may be sound, but is unlikely to be followed (as indicated
in Chapter 1). What may be easier is to move away from
the current models toward directed experiments and field
studies when possible and to continue to try to improve
the behavioral foundations of current simulations. One
way to implement such an approach would be for the U.S.
Department of Energy (DOE) to cooperate with the private
sector whenever feasible to promote standardization and
sharing of data sources and to further develop data col-
lection methods, such as automatic metering of individual
appliances. Such coordination might involve industry-wide
groups such as the Electric Power Research Institute and
the Gas Research Institute. The DOE could also cooperate
with other federal agencies in data collection. It would
be helpful, for example, if DOE promoted joint collection
with the Bureau of Labor Statistics (BLS) of data on the
transaction prices of appliances and encouraged BLS to
refine and publish hedonic regressions accounting for
housing prices as a function of building attributes. The
DOE should also review its use of large-scale policy-
simulation models, adopt an ongoing program of upgrading
and evaluation, and redirect policy analysis of appliance
choice toward using specialized experiments and surveys
as much as possible.
The central element in a program of data collection on
appliance efficiency would be a household survey that
covers appliance choice, including fuel, efficiency, and
other features, and that subsequently meters the chosen
appliance. The Electric Power Research Institute is
developing the technology for the last of these, appliance
metering. A random survey sample could be drawn from the
population with further selection of a sample of recent
appliance purchasers. Alternatively, information could
be collected retrospectively on old appliances, although
this possibility is limited by the ability to obtain
retrospective data on price and rated efficiency. Another
design would measure appliance use among a panel of pur-
chasers. This last approach would require careful statis-
tical analysis and would lack useful observations on
utilization before appliance purchase. Measurement of
appliance efficiencies and ratings could be carried out
OCR for page 85
as
in a manner similar to Consumer Reports bench studies or
EPA fuel efficiency measurements for automobiles; some
utilities are now collecting such data with the coopera-
tion of manufacturers, using appliance serial numbers.
By cooperating with the private sector, DOE may be able
to improve its own capacity to forecast the quantitative
effects of energy policies on appliance efficiencies.
Representative terms from entire chapter:
fuel choice
