The consumer's energy analysis environment

Energy Policy 1994 22 (10) 857-866

The consumer's energy analysis environment

Willett Kempton and Linda L Layne

This article describes how residential energy consumers measure and analyze their own energy consumption and energy costs. Using in-depth interviews, we find more extensive data collection and analysis by residential energy consumers than has been previously documented in the energy literature. However, the conclusions consumers can draw from their analytical efforts are restricted by the form in which they receive price and consumption data and their limited analytic capabilities. The relative information processing strengths of consumers are compared with those of institutions such as energy utilities, leading to the conclusion that many of the analytic tasks are currently assigned to the less efficient parties, degrading decision quality and creating a market barrier to energy conservation. We suggest a more efficient allocation of data collection and analysis between the consumer and energy utility.

Keywords: Energy efficiency; Consumption data; Folk analysis

Efficient market functioning requires that the consumer know the prices, quantity and quality of goods - and use that information to make purchase decisions. This may be a reasonable assumption in a retail store, where the price is marked on the shelf or directly on the product itself. We will argue that retail energy purchases are very different, with price and consumption data difficult to acquire and expensive to analyze. The buyer receives energy services (light, heat etc) but is billed via the easy to meter, but irrelevant to the buyer, measure of electron flow (kWh). The cost of individual services is further obscured by the quaint practice of manual meter reading, which dictates temporal aggregation - monthly in most of the USA, quarterly or annually in many parts of the world. The resulting bills in kilowatt hours meet the seller's need for revenue flow but, as we will demon-

Willett Kempton is with the Center for Energy and Environmental Policy, University of Delaware, Newark, DE 19716, USA; Linda L Layne is with the Department of Science and Technology Studies, Rensselaer Polytechnic Institute, Troy, NY 12180, USA.

strate, poorly serve buyer decisions about consumption and efficiency investments. Given the monopolistic nature of retail electric and gas utilities, potential consumer data improvements are not stimulated by market competition. Regulators and consumer advocates have concentrated on rates, with rarely any consideration of how price and consumption information reaches the consumer.

For readers whose familiarity with current energy billing has dulled their appreciation of its absurdity, we ask them to contemplate parallel examples. For example, telephone bills might give a single dollar figure for monthly long-distance service without calls itemized (this is in fact the case in many European coun- tries). Or, consider groceries in a hypothetical store totally without price markings, billed via a monthly statement like 'US$527 for 2362 food units in April'. j How could grocery shoppers economize under such a billing regime? Prior qualitative research has shown that energy consumers encounter precisely these types of data analysis and evaluation problems. 2 Energy conserv- ers receive no report on savings from past actions, which makes evaluation of energy savings very difficult. Failure to report achieved savings removes an incentive to further conservation measures and presumably impedes the diffusion of effective methods across households) In this context, the normal economic assumption of 'perfect information' is not only inaccurate, but has led energy analysts to completely ignore essential components of the policy mix needed to address barriers to energy efficiency.

A normal economic research paradigm would focus on the relationship of price and demand, ignoring the mechanisms by which buyers know the price and their own consumption. A few researchers, fortuitously unschooled in normal economic assumptions, have instead experimented with different means for convey- ing price and consumption information to the consumer. Two types of system would be logically desirable: a method of billing per end use rather than aggregating all energy uses in the household, and a method for report- ing the results of prior conservation efforts.

0301-4215/94/10 0857-10 © 1994 Butterworth-Heinemann Ltd 857

The consumer's energy analysis: W Kempton and L L Layne

We briefly review three experiments which measured whether better consumption and price information changed energy use. One study in the USA provided daily feedback on whole house electricity consumption as guidance for more efficient air conditioner management, resulting in a 10.5% reduction in energy use. 4 A second experiment was carried out in Norway, where meters have traditionally been read only annually. Shifting to bimonthly meter reading and billing resulted in a 8% drop in energy use the first year, increasing to a 10.4% drop the second year. 5 A similar experiment in Finland yielded 4.9% savings. The third experiment used meters which indicated how much each appliance consumed, a crude form of billing for end uses. The experiment showed that such a tool stimulated customer initiated investigation of costs of individual end uses, and subsequent change in consumption patterns, resulting in a drop of 13% in electric use. 6 A more detailed review of these and similar experiments is available separately. 7

The above experiments suggest that current energy markets do not deliver adequate decision making information to buyers, since energy use dropped 5% to 10% when consumption data were improved. 8 Since these experiments treat somewhat different information and analysis problems, we estimate that the sum of these information problems results in at least 10% more energy consumption than well informed consumers would choose. The aggregate effects are substantial. In the USA, for example, residential energy purchases total US$103 billion. This market failure is therefore costing at least US$10 billion in excess expenditures in the USA annually (plus resulting environmental extern- alities).

The above studies, since they provide quantitative savings from controlled experiments in improved information, allow an estimate of the scope of the market failure. To devise solutions, we need to understand how consumers analyze energy information and how consumer analysis affects decisions. We consider this empirical basis to be necessary for an effective redesign of the current retail energy information and analysis system.

The energy analysis environment The evidence above leads us to ask more generally about the analysis environment of consumer economic choices in complex domains. By analysis environment we mean the total set of data sources and processing capabilities a consumer draws upon to make some economic decision. Although we discuss utility billing and energy efficiency here, our approach can be applied to any area of consumer choice in which consumption,

cost, or product characteristics are difficult for consumers to measure and compare.

Our concept of analysis environment has some over- lap with the choice environment of market research, 9 but the latter deals more with analysis in stores and in advertisements used for choices among brands. Compared with choice environment research, our discussion covers more information sources and also covers more parties, since we compare consumers' analytic capabilities and limitations with those of energy analysts and utilities. We will sometimes speak of folk analysis to denote the energy analysis done by consumers, in informal contexts and without benefit of professional energy training.l°

Utility bills have historically been constrained by available technology (some early computerized bills were provided on Hollerth cards), and are molded by the needs and concepts of the utility and its state regulators. Current bill formats result from negotiation among internal utility departments (customer service, data processing etc), with some constraints set by public regulators in conjunction with consumer advocates and other stakeholders in public regulation. When bill design has been considered at all, readability and use evalua- tions have been based on reduction of complaints and maintaining the ability to compute the billed total from legally approved rates. This design process is in contrast to industries such as news media, which have explicitly designed their product to optimize consumer information processing. For example, people process large amounts of news data quite well, by skipping (newspaper) or not attending to (television) stories that are repetitive or less relevant, and by selective forgetting.ll News providers help with prominence clues for significant stories and by the newspaper inverted pyramid writing style which leads with the most important information and adds successively greater detail.

Our data collection first documented the current energy analysis environment, then introduced a new element into that environment to observe its effects. The new element was a printed energy report derived from a set of computerized analytic tools developed for organ- izations to make energy decisions, for example, to decide whether or not a conservation program has achieved sufficient savings to justify its cost (the analytic tool is the Princeton scorekeeping method). 12 We sought to evaluate which, if any, of those tools would be useful if provided - with interpretation - directly to consumers. Our sample consisted of 56 all electric households in central New Jersey, selected to have com- plete billing data and a clear pattern of electric heating. We also included a subsample of 13 who had participated in a utility energy conservation programs, so we could examine any household analysis of those program

858 Energy Policy 1994 Volume 22 Number I0

savings. The first 10 interviews were carried out in the informants' homes; subsequently, we conducted 46 interviews by telephone.

We used a combination of semistructured interview- ing methods and fixed questions. 13 We describe the consumer analysis environment both through percent- ages of informants who use various information resources and through illustrative individual cases with direct quotations demonstrating their own reasoning and data analysis. This article deals primarily with the first part of our interviews, the current analysis environment. Our experimental energy report is described elsewhere 14 and we discuss it here only as it relates to our exposition of the consumer's analysis environment.

Analysis now carried out by energy consumers

We find that sources of energy consumption information are more varied, and folk analysis is more rich, than is generally acknowledged in the energy literature. For example, the bill, managed by the utility primarily as a revenue generating mechanism, is for many consumers heavily utilized to extract as much information as can be gained from it.

The bill

Our interviews began by asking about the context in which the bill is read. For example, we asked how much mail informants get on an average day, how they catego- rize or sort it and what they do with the utility bills. We asked each respondent to get a current bill and to read it during the interview, itemizing the main things they pay attention to and why.

The utility bill is processed first as a piece of mail, then as a bill. Our informants report an average of five pieces of mail received on a normal day, with ten pieces on a maximum day. This mail is a diverse mix, including personal correspondence, bills, junk mail etc. People develop strategies for minimizing the time to process this daily flow. Two-thirds sort their mail and treat different types in different ways (66% of the 56 responding to this question). Some sort their mail by throwing out what looks like junk before ever opening the envelope, others open everything. Some open and pay their bills immediately upon receipt (42% of the 52 who answered), more set them aside (56%), often paying all of their bills at a set time of each month. This prepro- cessing and batch treatment of bills suggests to us that for the majority of consumers, energy conservation messages included with utility bills will be ignored because they are injected into an activity primarily concerned with verifying dollar amounts and writing checks.


Figure 1 is a sample bill of the type received by our informants. Note that it shows energy units consumed (kWh) and price per unit, intermixed with numerous other quantities making up the total amount due. In the lower left is a usage comparison with the same month the prior year.

Our interview asked the open-ended question 'What are the main things you pay attention to on the electricity bill?' to which we recorded all responses. Such open-ended questions reduce the interviewers' bias but also lower reported frequencies. This methodological phenomenon can be inferred from Table 1, since only 89% report that they looked at the dollar total due, whereas the true figure is surely 100%. As shown in Table 1, about 40% mentioned the usage comparison table and the kilowatt hours used for the current month. The only other item noted by as many as one-fourth was the number of days in the billing period.

The usage comparison table, a recent addition to this bill and many others around the country, compares the current month's kilowatt hour consumption with the corresponding month the previous year (see the lower left comer of Figure 1). Our finding that just as many look at it as the current units used in kilowatt hours (41% look at each) supports the idea that comparative information has value in this analysis environment. ~5 We also note the surprising discrepancy between monthly and daily figures: 41% report looking at the actual monthly kilowatt hours, only 4% at the (estimated) kilowatt hours per day. During the 1980s, a number of utilities began to use the daily figure in such comparisons across time periods, as it simply adjusts for uneven meter reading intervals. However, the daily figure is artificial, generated by dividing monthly kilowatt hours by days between meter reads, and unlike the monthly kilowatt hour figures, it has no price information associated with it.

An overwhelming majority (91% of the 56 answering this question) keep their old bills, on average retaining them four years; some reported keeping bills for the

Table 1. Information looked at on the bill (n = 54, open-ended question).

Item on bill Percentage mentioning

Dollar total due 89 Usage comparison table 41 Energy used this month (kWh) 41 Days in hilling period 28 Balance or credit 22 Due date 7 Energy adjustment charge 7 Meter reading 6 Arithmetic 4 Average kWh per day 4 Other 17

Energy Policy 1994 Volume 22 Number 10 859

The consumer's energy analysis." W Kempton and L L Layne

JERSEY CENTRAL POWER & UGHT COMPANY

WHEN PAYING BY MAIL, USE ENCLOSED ENVELOPE. A Member COml~ny of the BILL MAY ALSO BE PAID AT ANY COMPANY OFFICE OR AUTHORIZED AGENCY. General Public Utilities System

B I L L FORHAT 15 MAIN STREET ANYTOHN NJ 0 1 2 5 q - 9 8 7 6

MAKE CHECK PAYABLE TO JOP&L

DEC 3 0 , 1 9 8 5 5 2 . 1 8

JERSEY CENTRAL P&L CO. P . O. BOX 2 0 3 ALLENHURST, NJ 0 7 7 0 9

~EI tYz¢: [ TO

BILL FORt~T 1S HAZN STREET ANYTQNN NJ 01Z34

RAT( CLASSIFICATION RG -- RESIDENTIAL SERVICE

703256381013 00000052180000052187

PLEASE RETURN THIS PORTION WITH YOUR PAYMENT

9 0 P.][OQEDALE AVE HORRZ$'rcNH~ NJ 07960

ACTUAL

T H I S I S YOUR CURRENT B I L L CALCULATION

CUSTOMER CHARGE t ~. 75 ENERGY CHARGE 455 KNH ;) 1.1010 45.96 ENERGY ADJ CHG 4...~S IQ4H ;) $.OO3ZZ8 1.47 CURRENT PERZOD CHARGES # SZ.18 11/13/85 to 1Z/17./8.~

FOR Z9 DAYS

PAYNENTS/CHARGES S INCE LAST B I L L

PREVZOU~ BALANCE t ¢~). gS PAYMENT NOV ZO THANK YOU ~,o.gscR BALAI~E AT BZLLZHG t .00 CURRENT PERZOD cHARGES F,Z.18 AHOUNT OUE t 5Z. 18

HETER HETER READING NUt'BE R CURRENT PREVZOU~

3 Z17.Y4~78 O 48719 4)BZ6e~

HULTZPLZER

1

KZLOIIA~ HOURS USED

455

USAGE CONPARZSON

BZLL'rNG ~ ~,ZLLZNG KHH PERIOD ~ O DAYS PER DAY

DEC 85 &SS Z9 16

DEC 84 SlO 30 17

THTS NIEN BILL FORIIAT HAS BEEN DESTGIdED TO PROVIDE YOU NZTH HOItE TNFORtIATZGN REGARBZN8 YOUR C(]¢SUHPI"ZON PAT'IrlERNS AND TO SHON YOU HCN YOUR BILL HAS BEEN CALCULATED. HE HOPE YOU FZND THE HEN BZLL TO BE ZNFORFU~.TZVE AND HELPFUL ZN YOUR CQNSERVATZON EFFORTS.

. . . . . . . . . . . . . . . - " _ _ _ L . . . . . . . . . . . . . . . . . . . . . . . . I JERSEY CENTRAL POWER AND LIGHT CO. '

~.E~SE eAv T~IS Ak~JNI

5 2 . 1 8

Figure 1. Current bill for consumers in this study.

8~0 Energy Policy 1994 Volume 22 Number 10

entire occupancy period of a residence. When asked why they kept bills, informants gave a diversity of responses, with the two most common being to compare energy use or cost in previous years (34% of the 53 answering), and because they generally keep receipts (21%).

It is instructive to note which bill information is not used. Beyond the necessary components of any bill (prior balance, current charges, payments and balance due), typical utility bills include account number, date and meter reading for the current and prior period, and 'fuel cost adjustment' or 'energy adjustment charge' (usually given to six or more significant digits, an indi- cator that these figures are provided for regulators, not for the consumers who receive the bills). Both the utility 's customer service department and the state regulators are driven by complaints, from a minority of customers, toward putting each separate charge and rate on the bill so the billed amount can be computed to the exact cent. Our data suggest that the complaining minority has unduly focused utilities and regulators on tabulation to the exact cent, since only 4% of our sample reported even looking at the arithmetic on the bill. The provider's cost of the extra information is near zero (extra ink), but it may be costly to the consumer if additional irrelevant information obscures or deemphasizes what is useful. One particular source of confusion is the fuel cost adjustment. 16 Since this is listed as a line item on a par with the cost of energy and the monthly service charge, it seems to be an item which the customer has consumed. In fact, the customer has no control over it, and it exists as a separate line item only due to artifacts of the regulatory process. ~7 We would argue that the fuel cost adjustment should be folded into the rate per energy unit in order to facilitate readability and compre- hension of bills.

Comparisons based on bills

Our study indicates that bills are used for at least two purposes beyond making monthly payments: checking when consumption is unusual and evaluating conservation actions.

The most frequently reported purpose of comparing bill information was to check on unusual bills. As we found in several cases reported in detail, a bill which seems unusually high begins a process by which the bill payer asks other members of the household whether they know any behavioral changes which could explain it. If so, household members decide whether corrective action should be taken. The cost or benefit of the inferred behavioral change is computed simply as dollars this billing period minus dollars last billing period. Such cost estimates can be very inaccurate due to fluctuations in rates, the weather, other simultaneous


behavioral or technical energy changes, and the days per billing period. 18 If the consumer does not identify any likely cause, either the matter is dropped (possibly with a mental note to make another check the subsequent month) or a call is made to the utility company.

A second purpose of comparing bill information is to evaluate conservation actions. Several of our respondents had participated in a utility weatherization program, and a few of them tried to analyze bill information to infer whether it was effective.

How did they make the comparisons? We find examples of four methods: (i) summing each yearly total; (ii) this month versus recent months; (iii) this month versus same month last year; and (iv) highest bill this year versus highest last year. The most systematic of these was to sum up all the monthly bills for the year. More than a third of our informants (39% of 54 answering this question) reported having computed an annual total at some time. A few did so every year, typically around tax time, and compared energy with other annual expenses. Others did so at life decision points, for example, to decide whether they could afford to retire. Other comparison methods are described in more detail elsewhere. J9

Bill inserts

Most US utilities periodically include preprinted inserts with their monthly bill, some of which include energy conservation information. Many people look at the inserts (38% said they always read them and 59% said they sometimes read them), but some respondents volunteered that they were of less interest because they are not house specific and the information may not apply to them. Furthermore, the same inserts are mailed out at periodic intervals, so the information is often repetitive. These features may make sense from a marketing point of view, ie in mass mailing not all of the information will be relevant to everyone, repetition is a common marketing technique, and people respond differently at different points in their lives, z° However, this redun- dancy and lack of house specificity contribute to the consumer information glut and hence reduce attentive- ness to such messages. 21

In addition to conservation suggestions, utility bill inserts sometimes include rate schedules. Although several people mentioned saving these rate schedules, only one person in our sample of 56 actually used them to compute the bill (a case which we describe subsequently).

Analytical comparisons with neighbors

Fully 70% of our respondents reported discussing electricity bills with other people. Some reported this

Energy Policy 1994 Volume 22 Number lO 861


when they moved into a new home, especially if the move entailed a change in heating system. Others con- sult with their neighbors or friends regularly, using their added knowledge of the neighbor's family and behavior to adjust their interpretation of energy differences. For instance, several residents in a retirement community mentioned neighbors who they knew from conversations to have lower bills during the coldest months of every year. They attributed the lower bills to the neighbors' trips to Florida during just those months. In such comparisons, fluctuations in weather are controlled for, since the effect of weather conditions on the neighbor's bills would be comparable to that on one's own bills, and the meter read date is likely to be the same. However, our respondents sometimes incorrectly identified small end uses as primary causes of the observed differences.

The utility meter

A surprisingly high number of people reported having read their own meters, although we felt many of them did not fully utilize the information. Several reported going out to look at their meter just to verify that it was working, especially if they had a particularly high bill. When asked specifically about reading the meter, a majority (58% of the 55 answering) claimed to have attempted it, 53% to have succeeded in reading the amount, and 24% to have done so in the past year. Many of the meters in this area were electronic and had digital readouts rather than the more typical array of counter- rotating dials, which may partly explain what we consider an unexpectedly high percentage of respondents reading their own meters. A substantial majority (78%) felt that the utility's meter readers are accurate. In response to our questions, some people volunteered that reading one's own meter would not do any good since one had to pay for what the utility said had been used no matter what.

A few respondents made unusual uses of their meters, uses that illustrate potential applications of better energy data. One low-income woman had a neighbor read her meter for her towards the end of every month. She then calculated her expected bill from the rate schedule flyer, so she could budget energy costs with other anticipated expenses. Others used the meter to gauge the consumption of individual appliances. A few admitted reading their neighbor's meter, in complexes where the meters of neighboring units were adjacent. One person, described subsequently, even reads his meter daily and correlates these readings with the weather.

Weather

Another important element in folk calculations regarding home energy consumption is the weather. Nearly

everyone (89% of the 56 answering) considered the weather to have a very important effect on their utility bills. We asked how they would determine which winter was colder; their reported methods include the feel of the winter, the number of days that snow had to be shov- eled, the length of time winter clothing was needed, the frequency with which the heater switched on, and if weather reports declared particular days to be of record temperature.

Summary of information sources and analysis

Overall, our reaction to studying the current analysis environment was one of surprise at how many information sources people used and the variety of analytical uses to which they put them. To summarize, they examine non-financial information on the utility bill, including kilwatt hour usage (41%) and the comparison table with the same month the previous year (41%). A minority always read bill inserts (37%, while 59% do sometimes), 70% discuss utility bills with other people, 89% acknowledge the effect of weather on bills, 39% had computed their annual total, and a surprising 24% reported reading their own meters within the past year.

A case of unusually extensive analysis We describe elsewhere two cases of consumers who do much more energy analysis than average. 22 One of these two, whom we call the 'home weather analyst', takes daily high-low temperature and electric meter readings, computes heating degree days per month, and compares the degree days with his monthly kilowatt hour consumption from his utility bill. When the consumption is unexpected, he and his wife investigate possible explanations, and take corrective action.

The home weather analyst exhibits greatest strength in data collection and interpretation, rather than in computation. His use of heating degree days is a far more sophisticated weather calculation than that of anyone else in our sample. However, energy analysts would regard his computations as deficient, in ways which seem - to analysts - easy to correct. He has no way of quantitatively adjusting the energy use to correct for temperature, but simply 'eyeballs' it. As examples cited in the full case study show, this can work effectively if the two are qualitatively different, for example, if energy use goes up while degree days go down. However, inferences cannot be made when changes are of different sizes but in the same direction. In addition, he uses a reference temperature (also called base temperature) of 70°F (19.4°C) rather than the accepted 65°F (16.7°C) or a house specific one. 23 He does this, he says, because ' I 'd like to have the thermostat up to 70, it

862 Energy Policy 1994 Volume 22 Number 10

makes me feel good to look at the 70 during the computation. I realize it's not very scientific.' This method will cause his degree day figures to overestimate heating requirements in the 'shoulder months' of fall and spring, when outdoor temperatures are between 70°F and his true reference temperature.

Considering the amount of effort which he puts into data collection, interpretation and corrective action, his analytical computation seems incongruously deficient. A slightly better formula, and as simple an analytic tool as a graph of degree days against energy units with a line through them, would greatly improve the interpretive powers of his methods. But for him to discover these methods would require research or peer review, an effort unlikely to be cost-effective since the methods would benefit only a single household.

Cost of analysis

To illustrate the impracticality of having individual households each do weather record keeping and analysis, we compute the cost of computation for the home weather analyst case. Ignoring his modest start up costs, his total time expenditures are about 1 hour and ten minutes per month. This activity is in part a hobby, so one cannot directly compare it to his wage labor; however, at his salary of US$26 000/year, 1 hour 10 minutes is US$14.50/month, or US$174.00/year labor. This is the closest we have yet to a real estimate of the time cost for consumers to perform energy analysis themselves.

For comparison, we estimated the cost for a utility to print weather data and perform the additional computations of an annual summary and weather correction. For a moderately large utility, say 500 000 to one million customers, the cost of printing one sheet and sending it out separately is approximately US$0.05 per customer, plus postage. Estimates of setting up the logic for a new printing vary widely; the closest comparison is revising the bill format, for which we have been quoted costs varying from US$50 000 to US$1 million. The highest cost figures include extensive reviews, surveys, solicita- tion of comment and negotiation with regulatory bodies, for a large (million-plus customers) company. Most utilities already collect weather data and enter it in their computers, for the purpose of computing estimated bills; thus collecting weather data would not represent an increased expense. In sum, we would estimate set up costs as ranging from 10¢ to US$1 per customer, and ongoing costs at approximately US$0.30 for a separate mailing, under US$0.05 if a second sheet were enclosed with the bill, and close to zero if printed on the existing bill. These are rough figures, but since the highest figures are at least two orders of magnitude below the costs for an individual making the computations him-


self, they make our point that bill summaries, rate adjustments and weather adjustments are carried out more economically via computerized mass production techniques than by households.

The division of labor in energy analysis

The above cases of consumers' energy analysis reveal both strengths and weaknesses of folk energy analysis. The explanation for the coexistence of both good and poor analysis is that households are well adapted to some forms of analysis and poorly adapted to others. For example, computation of weather effects is clearly not an area of strength. Even our weather recorder computed heating degree days inaccurately and 'eyeballed' the kWh/degree day relationship rather than making the incrementally small effort of a graph or numerical calculation. In a second case study of careful record keeping and analysis, 24 a person we called the financial planner used methods drawn from his experience as an accoun- tant. For example, he measured monthly energy use in dollars, since that allowed him to compare energy with other expenditure categories. However, his methods dis- torted energy comparisons through time due to price changes and weather fluctuations. In both cases, these individuals' methods were not well suited to making certain inferences about energy.

On the other hand, we were impressed by these and other consumers' ability to interpret anomalous months by reference to their intimate knowledge of household events. During the interviews, when people could see our analytical report, they readily made sense of some of the months that our standardized weather procedure saw as anomalous, for example, identifying months of low energy use as vacations and high months as times of visitors or unusual appliance use. It is possible that some of their interpretations were inaccurate; for example, lay people tend to overestimate energy use arising from activities which are perceptually salient or socially significant (eg Christmas lights or baking the Thanksgiving turkey). Nevertheless, occupants, unlike analysts or utilities, have large amounts of household data immediately available for recall, and they use it to interpret months of unusual energy consumption.

These findings suggest that many households would make good use of higher quality, higher resolution energy consumption data. Supporting conclusions derive from other studies. In manufacturing, cost savings were found to be greatest when high quality information was provided, not to managers but to line operators having greatest familiarity with, and control over, processes. 25 In service industries, productivity and quality go up when the customer is more involved in production. For example, customers have a strong interest in accuracy



Table 2. Comparison of the economics of analysis for households versus large institutions such as electric or gas utilities.

Cost for Cost for Processing task institutions households

I Collect data on Expensive Cheap or intrinsic household activity

2 Computation Very cheap Expensive (unless by rule-of-thumb)

3 Development of Cost can be spread across Cost must pay off in single methods, research, many households household system development

4 Interpretation of data Cheap at aggregate level, Cheap to relate to household data; and analysis expensive at casual level impractical if training required

5 Decision making after Requires expensive Cheap if analytical interpretation analysis consultation with customer is clear

when filling out bank deposit slips or providing medical histories. 26 Similarly, energy consumers have the greatest familiarity with household activities and are motivated to minimize costs.

Comparative costs of analysis

Table 2 compares the analytic strengths and weaknesses of analysis by institutions and households. The table is based on the data from our interviews on current consumer energy analysis, as well as our considerations of systems like that in our experiment, in which the utility could send out additional information. Consider the first item. Household data such as occupancy, visitors and appliance usage are expensive for an institution such as a utility to acquire, whereas members of the household themselves have these data immediately available. 27

For item 2 in Table 2, the costs of information processing are the opposite of item 1. Computation is cheap for institutions because they can amortize the costs of planning, research and capital (computers and software) against many users. In the case of utilities, computation may include the bill calculation as done now (subtract- ing two meter reads and multiplying by the rate per energy unit) or more complicated processing such as weather adjustment. Typical electric and gas data processing systems are input/output bound and could easily

perform thousands of compute operations per customer bill, rather than the dozen or so they do now, with no detectable difference in performance. The related item 3 in Table 2 notes that much more investment in data and systems is possible for institutions because investment can be amortized across all customers. Items 4 and 5 are the interpretation and decision making which follow from the analysis. Institutions can interpret some standardized variables; for example, that energy use went up (or down) even after controlling for weather, but they cannot easily interpret in terms of household events (vacations, new baby etc) because they do not have access to the necessary data, as noted in item 1. Similarly, decision making is a function that proceeds already within the household; the utility cannot enter that process (even if it were invited) without expensive individual consultation.

Data flow and optimum division of labor

From the perspective of the relative data processing strengths of individuals and institutions, we examine current systems of information flow and energy analysis in Figure 2. The figure illustrates that the steps are the same in all energy consumption data systems - the systems differ only in which steps of the process are done by the institution and which are done by the individual

Read Weather Appliance and meter data behavior data

Compute use i ~ :: Adjust for Decide if use Interpret i ~ Choose corrective and bill i weather ~ is abnormal ~ cause ' action & carry out

Traditional Advanced Optimum break Conservation billing billing in processing program evaluation

Figure 2. Energy data flow. Institutions begin processing at left of figure. Dotted lines represent the break between data processing by institution (to left of line) and processing by household (to right of line).

864 Energy Policy 1994 Volume 22 Number lO

household. In all the systems shown in Figure 2, the institution processes initially, and the household performs final processing.

For example, in traditional billing the utility only col- lects meter data, computes consumption and cost, and sends the bill out to the customer. Figure 2 indicates the end of utility data processing by the dotted line labelled traditional billing. From that point on to the right, customers may use recollection of prior bills and weather data to determine if the current bill is high, low or normal. If they believe the bills are anomalous, they may try to explain them with reference to their extensive knowledge of household activities.

The problem we see with traditional billing is that it leaves to the consumer the intermediate steps of collecting weather data, adjusting energy use for weather, and deciding whether the month is anomalously high or low. These types of standardized data collection and analysis are more easily done by large institutions.

A second example of where the division of information processing could be improved is an evaluation of a conservation program. Here the institution (typically a research group or conservation analyst) goes beyond its area of strength in carrying analysis all the way to collecting and interpreting household behavioral and occupancy data. This is shown in the figure as 'Conser- vation program evaluation'. If the only household change was the conservation program, interpretation would be straightforward. But households do not stand still during conservation programs. Since analysts typically lack detailed behavioral histories, they can only speculate about why the savings differ from predictions. Was the conservation work done poorly? Did the residents 'take back' savings in higher thermostat settings? And so on. Institutional data collection in this case - blind to within house events - looks as amateurish as does consumer estimation of weather data based on days of shoveling snow. Analysts might better tap household data resources by presenting anomalous data to the members of the household who produced it and asking for explanations based on household events.

Based on the strengths of each party, we propose an optimum break in the processing sequence of Figure 2; a point at which each party performs the tasks which it does best and most efficiently. The institution (presumably the utility company) would carry processing further than current bills but not as far as energy program evaluation. Specifically, the institution would add weather data, adjust energy consumption for weather and price changes, and decide whether to report a statistically significant change. This point in the processing sequence of Figure 2 is shown as the optimum break in processing. From that point on, households are better at processing; they would collect household behavioral and retrofit

The consumer's energy analysis." W Kempton and L L Layne

data, attribute causes to energy changes and make any decisions as to corrective action. Based on this reasoning, we have developed and tested a prototype consumer energy report fitting into the point we identify as optimum, as described in more detail elsewhere. 28 Our proposal is contrary to stereotypes about computing trends, since we recommend that more computer power be delivered to consumers not by purchasing microcom- puters, but - far more efficiently - by additional processing with a mainframe at a central facility and distributing the results on paper. As noted earlier, this service can be carried out in the range of US$0.05 to US$0.30 per customer per year, whereas our example consumer doing the same analysis consumed US$174 per year worth of his time. Our interview asked the amount consumers would be willing to pay for this type of analysis, yielding a mean of US$1.59. Since the cost to the utility is less than one-fifth of the value to the customer, and less than 1% of the cost for customers to do it themselves, utilities might be motivated to provide such analysis for reasons of customer service as well as energy efficiency.

Conclusion

This article has attempted to comprehensively examine the consumer's analysis environment regarding energy consumption. Data sources in this environment include utility bills and inserts, discussions and comparisons with neighbors, and the utility meter. We find that consumers do more processing and analysis of these data than would be expected from the energy literature. However, consumer weaknesses in data collection and analysis distort market decisions. We outline the strengths and weaknesses of consumer analysis in comparison with that of institutions such as utilities and energy research groups. Based on this analysis, we propose a more rational point in the data processing sequence for the termination of utility data processing and the beginning of consumer data processing. Current programs suboptimally make the processing division on either side of this point, degrading market decisions regarding energy conservation. Our more general point applies to non-energy examples as well: careful exam- ination of methods by which consumers collect and analyze data on consumption, cost and product quality can suggest improved allocation of data acquisition and analysis among consumers and the institutions that serve them.

The research described here was funded by the New Jersey gas and electric utilities and the New Jersey Department of Commerce, through Princeton University's Center for Energy and Environmental Studies. This research was conducted in collaboration with Alfredo Behrens, Richard C Diamond, Margaret Fels and Cathy Reynolds. We



are grateful to Jersey Central Power & Light, and especially to Eugene McCarthy and William Zdep, for assistance in contacting their customers who participated in our study. For comments and suggestions on this manuscript, we are grateful to Gautam Dutt, Meg Fels, Hillard Huntington, Paul Komor, Loren Lutzenhiser, Lee Schipper, Robert Socolow and Ed Vine. Some sections of this paper draw from our presentation at the ACEEE Summer Study on Energy Efficiency in Buildings.

i The example of a grocery store without prices originally derives from Willett Kempton and Laura Montgomery, 'Folk quantification of energy', EneJ~gy: The International Journal, Vol 7, No 10, 1982, pp 817-827. Merrilee Harrigan completed the analogy by suggesting monthly billing. 21bid; also Richard Wilk and Harold Wilhite, 'Why don't people weatherize their homes? An ethnographic solution', Energy: The International Journal, Vol 10, 1985, pp 621---631. 3John M Darley, 'Energy conservation techniques as innovations and their diffusion', in Robert H Socolow, ed, Saving Energy in the Home: Princeton's Experiments at Twin Rivers, Ballinger, Cambridge, MA, 1978, pp 255-262. 4Clive Seligman, John M Darley and Lawrence J Becker, 'Behavioral approaches to residential energy conservation', in Robert H Socolow, ed, Saving Energy in the Home: Princeton's Experiments at Twin Rivers, Ballinger, Cambridge, MA, pp 231-254. 5Harold Wilhite and Rich Ling, 'The effects of better billing feedback on electrical consumption: a preliminary report', 1992 ACEEE Summer Study on Energy Efficiency in Buildings (Proceedings), Vol 10, 1992, pp 173-175. 6John K Dobson and J D Anthony Griffin, 'Conservation effect of immediate electricity cost feedback on residential consumption behav- iour', Proceedings, ACEEE Summer Study on Energy Efficiency in Buildings, 1992, pp 10.33-10.35. 7This review, with recommendations, is being prepared as a joint pro- ject of the Alliance to Save Energy and the University of Delaware's Center for Energy and Environmental Policy. 8The 5% to 10% we cite does not exhaust the range of studies. Some yield no savings, others, by combining consumption information with incentives, have yielded above 10%. We judge 5% to 10% to be repre- sentative results from the better designed programs. 9For example, see James R Bettman, An lnfotwtation Processing Theory of Consumer Choice, Addison-Wesley, Reading, MA, 1979. ~°For comparison, see the discussion of folk quantification by Kempton and Montgomery, op cit, Ref 1. tlDoris A Graber, Processing the News: How People Tame the Information Tide, Longman, New York and London, 2nd edn, 1988, p 249. 12Margaret F Fels, ed, Scorekeeping, Special issue of Energy and Buildings, Vol 9, Nos 1 and 2, 1986. 13H Russell Bernard, Research Methods in Anthropology, Sage, Newbury Park, London and New Delhi, 1994. 14L Layne, W Kempton, A Behrens, R C Diamond, M F Fels and C Reynolds, Design Criteria far a Consumer Energy Report: A Pilot Field Study, PU/CEES Report 220, January 1988. Available from Center for Energy and Environmental Studies, Princeton University.

tSThe importance of comparative information is also emphasized by others: Harold Wilhite and Teresa Ribiero, Feedback Information and Residential Energy Billing: Towards a Better Nordic Bill, GRS-714, Nordic Council of Ministers (English Translation) 1988. Also Steven C Hays and John Cone, 'Reduction of residential consumption of electricity through simple monthly feedback', Journal of Applied Behavioral Analysis, Vol 14, 1981, pp 81-88. ~q'he fuel cost adjustment factor was also noted as the most confusing element of the bill in an earlier study; John Byrne, Daniel Rich, Francis Tannian and Young-Doo Wang, Energy Policy Options: The Billing Ptw.ess and Consumer Energy Information in Delaware, Report, College of Urban Affairs and Public Policy, .pp 55-59. Available from Center for Energy and Environmental Policy, University of Delaware, Newark, DE, 1982. a7David Moskovitz, 'Why regulatory reform for DSM?', in S M Nadel, M W Reid and D R Wolcott, eds, Regulatory Incentives for Demand-Side Management, ACEEE, Washington and Berkeley, 1992, p5. ~SAt least one utility has recognized that customers compare current consumption with recent months and has created bills to facilitate this process. Public Service Electric & Gas (New Jersey) provides energy usage (kilowatt hours or therms) for current and previous two months in comparison with the corresponding three months of the previous year. This provides a larger window of comparative months to support existing customer analysis methods. In an incident which illustrates both the value of comparisons and the failure of specialists to recog- nize processes customers are actually using, this comparative information was removed during bill improvements encouraged by state regulators. Subsequently, so many customers complained about its loss that it was promptly restored. 190p tit, Refs 1 and 14. 2°For example, one study has shown that people are more open to energy-efficiency improvements when they buy a new home; op tit, Ref 2. 2~For a discussion of information overload, see op cit, Ref 9. 22Willett Kempton and Linda Layne, 'The consumer's energy information environment', Proceedings of the 1988 ACEEE Summer Study on Energy Efficiency in Buildings, 1988, pp I 1.50-11.66. 230p tit, Ref 12. 24This case is described in detail in op eit, Ref 22. 25See pp 270-283 in Shoshana Zuboff, In The Age of the Smart Machine: The Future of Work and Power, Basic Books, New York, 1988. 26 Victor Fuchs, The Service Economy, Columbia University Press, New York, 1968. 27The meter reading is one household datum which institutions are better at collecting. This is an exception to the overall pattern, because the meter has been explicitly designed for the utility's use. The meter is placed outside the building for efficient access by the utility, and the traditional utility meters are so difficult for consumers to read that they have literally become a textbook example of poor user design; see Victor Papanek, Design for the Real World, Pantheon Books, New York, 1971. 280p tit, Ref 14. Also see Cathy Reynolds and Margaret F Fels, 'Reliability criteria for weather adjustment of energy billing data', Proceedings of the 1988 Summer Study on Energy Efficiency in Buildings, American Council for an Energy-Efficient Economy, Washington DC, 1988.

866 Energy Policy 1994 Volume 22 Number 10

The consumer's energy analysis environment

Documents