Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
CHAPTER 3 THE EFFECTIVENESS OF RESIDENTIAL CONSERVATION INCENTIVES Federal, state, and local governments, utility companies, and community organizations have offered a variety of financial incentives to induce homeowners and occupants to invest in insulation, weather- stripping, furnace improvements, and other energy-cff icient tech- nology. At least in terms of cost, these have been among the lamest energy conservation programs in U. S. history. The federal energy tax credit, for example, may cost the treasury $2.5 billion between 1981 and 1986 (Hirst, Goeltz, and Manning, 1982), and state conservation tax credits have amounted to over $50 million per year in California alone (Randolph, 1984~. The Tennessee Valley Authority has provided 3250 million in interest-free loans to its customers (Berry, 1982), and the Bonneville Power Administration is spending at least that much to hold down residential electricity demand and thus avoid even higher expenses for new power plants. This chapter reviews recent data on the effectiveness of residential incentive programs. It focuses on three of the important behavioral issues affecting incentive programs: the effects of the size of an incentive . the tvue of incentive . and the nonf inancial features of an , , , ~ incentive. Less evidence is available on the other two issues: the incremental effect of incentives on investment and the effect on the pace of investment. The chapter focuses on the first three issues, concluding with a discussion of them in the context of programs for low-income housing. _ It examines the available data, draws some con- clusions, and points some specific directions for research. A behavioral perspective on incentive programs emphasizes a pro- gram's effects on consumers' decision processes. It leads to a focus not only on the size of financial incentives, but also on their form. It also leads to an interest in how programs get the attention of their intended audiences, communicate with them, address energy users' con- cerns, seek and use credible sources for communication, and minimize the effort and risks of investing in energy efficiency. To address these issues, evaluation studies must take into account a wide range of program and client characteristics and outcome variables. The major ones are listed below: 29
30 Prouram Variables Type of incentive Size of incentive Size of target population Type of target population Period studied (months) Qualitative features: credibility of sponsor motivation of sponsor marketing effort consumer protection features other services (audits, bidding) restrictions on participation Client Characteristics Household income Number of household members Education Size of home Type of structure Appliance holdings Type of heating and cooling system Fuels used Home ownership Energy-related attitudes and beliefs Outcome Variables Percent of target population attracted to program (e.g., requesting energy audits) Percent of those attracted who use the incentive Investment per household using incentive Incremental energy savings by households using incentive Incremental savings by those attracted to program Administrative cost Unfortunately, very few evaluations contain the full range of infor- mation listed above. Therefore, our analysis is limited to a restricted part of the phenomenon of response to incentive programs. With this limitation, we give a central place to qualitative factors, such as the form of an incentive and the nonfinancial features of program design, marketing, and implementation. CRITERIA OF EFFECTIVENESS Two classes of criteria can be used to judge a program's effective- ness at improving energy efficiency: the level of induced investment in energy efficiency and measured energy savings. Although energy savings is sometimes considered the only true test of effectiveness, both criteria must be examined for technical and policy reasons. Tech- nically, it is not yet possible to estimate energy savings accurately
31 from given efficiency improvements: estimates are often wrong in the aggregate and for individual homes, and savings usually vary by 50 percent or more from estimates (Goldman, 1984; Hirst and Goeltz, 1984) Thus, one cannot infer energy savings from physical changes in a building; one must measure them. For policy reasons, it is important to pay attention to increased efficiency as well as energy savings because different conservation programs have different goals. For an electric utility sponsoring conservation as an alternative to building new power plants, forestall- ing growth in energy demand is the paramount goal. For such a utility company, economic calculations are not affected by whether the goal is achieved by improving efficiency or in other ways, although it does matter whether the restraint of demand will be robust under short-term changes in the regional economy. For a city's or community group's conservation program, however, increased comfort and health may count as important benefits of energy efficiency even if there is no reduction of total energy use. Thus, both energy efficiency and energy use should be measured to assess the effectiveness of conservation programs. Comfort, health, and other policy goals should also be measured where relevant. Data on measured energy savings from conservation programs are quite limited. Hirst (1984) reviewed such data from studies of Residential Conservation Service (RCS) programs and loan subsidy (mainly zero-interest loan) programs. Participants in six residential conservation service programs saved between 3 and 9 million Btu per household per year {in the median program, 5 million) and participants in six loan programs saved between 10 and 20 million (median, 12 million), compared with savings by nonparticipants in the same programs. Birst concluded that incentive programs save about three times as much energy as RCS programs. Hirst's conclusion should be qualified, however, considering the different ways participation is defined in the two types of programs. For the RCS programs, all households requesting energy audits are counted as participants, regardless of subsequent actions to improve energy ef f iciency. For the incent' ve programs, however, partic i pants are def ined as only those households that took advantage of the incen- tives. In most instances, a minority of the households that request energy audits take advantage of incentives, and they are the ones that made the most extensive investments. For the programs Hirst cited for which data are available, the costs of retrofit per participant ranged from $1,500 to $2,500 (Hirst, Bronfman, et al., 1983; Hirst, Goeltz, et al ., 1983; Puget Sound Power and Light Company, 1984; Weiss et al ., 19831. Thus, a comparison of "participants" exaggerates differences in the effects of the two types of program: if only the households that made major investments were compared, the difference In energy savings between the two types of program might be very small. . 1In PCS programs, capital investment per "participant" probably averaged about $600 (Centaur, 1983). This average might represent an
32 Only limited data are available for comparing energy savings in programs offering different types of incentives, and no clear differ- ences are evident. Hirst (1984) has reported the findings from loan programs, and we have been able to find only three comparable reports of measured energy savings from grant programs. Participants in the Canadian Home Insulation Program, which offers a grant of 60 percent of the cost of recommended conservation measures, cut energy use 12.8 percent per household compared with an econometric estimate of what the same households would have done in the absence of the program (Energy, Mines, and Resources Canada, 1983~. Participants in the Weatherization assistance Program orrerea Free co ~ow-~ncome Domes in the United States saved 14 million Btu, also compared with an econometric estimate (Peabody, 1984~. Seattle City Light's Low-Income Electric Program, which offers free weatherization for households earning less than-90 percent of the median income in the region, saved its average partici- pant about 12 million Btu in the first year (Newcomb and Weiss, 1983a). These outcomes are within the range of savings reported by Hirst (1984) for loan programs. The lack of differences is not meaningful because households counted as participants in these major incentive programs are almost always the ones that made large investments in recommended conservation measures. When programs offer large incentives for major investments, differences in effect are more likely to be due to some programs' success at getting more people to follow audit recommendations than to differences in savings among households who follow the recommendations. The best test of effectiveness is what a program does for its population of potential participants, not what it does for the households that accept its recommendations. ~ _ a ~ _ _ _ ~ _ ~ ~ Consequently, the rate of participation in conservation incentive programs is a useful index of effectiveness. When programs save about the same average amount of energy per participant, as in the case of the major loan and grant programs for which data are available, rate of participation is a good proxy for total energy saved. Rate of partici- pation is also an important variable in its own right because it indicates a program's effectiveness at marketing; when combined with an index of the intensity of participation, such as spending per household on retrofits, it becomes a rough index of improved energy efficiency. (It is only a rough index because expenditures on different retrofits can have very different effects on energy use.) Since more data are investment of $1,500 each by 40 percent of those requesting audits and no investment by the rest or an investment of $2,500 each by 25 percent and no investment by the rest. If one assumes that the most active RCS participants invested $1,500 and saved 13 million Btu, as did households investing the same amount in the Puget Sound Power and Light program (Hirst, 1984) or that they invested $2,500 and saved 20 million Btu, as in Northern States Power's loan program (Hirst, Goeltz, et al., 1983), and if the other households saved nothing, the average savings for all households defined as RCS participants would have been 5 million Btu-- the median savings for RCS reported by Hirst.
33 available on participation than on measured energy savings, the dis- cussion below focuses mainly on participation, measured by the number or percentage of eligible households completing retrofits or by the cost of the retrofits. EFFECTS OF THE SI BE AND TYPE OF INCENTIVE Financial incentives obviously vary in their economic value, but they also vary on several other dimensions. They vary in immediacy: people can collect grants and rebates quickly when they make a retrofit, while tax credits can take a year or more to receive. They vary in their effects on cash flow: a partial rebate requires an immediate expenditure, while some loans can be structured so that the stream of energy savings balances the stream of loan payments. They vary in the requirement to assume debt: loan subsidies require a household to go into debt, while grants and rebates make borrowing optional. They vary in the effort needed to use them: tax incentives require keeping records and filing forms, while some rebates are available at the point of decision without any additional effort by the energy user. They vary in eligibility requirements: screening processes deter some applicants who fear rejection. No doubt incentives vary in other ways as well. To limit the number of dimensions to be considered, we loosely classify incentives as reduced-interest loan subsidies, interest-free loans, or grants or rebates. A few programs combine loan and grant features or include other incentives, such as reduced cost through group bidding. Evidence on the relative attractiveness of different types of incentives comes from surveys of the expressed preferences of actual or potential program participants and from levels of partici- pation in programs offering different kinds of incentives. Reliable evidence on the effect of incentive size comes only from actual participation in programs. Surveys of Preference Surveys of populations eligible for incentive programs measure expressed preferences for different types of subsidy. A few surveys that have asked people to choose among hypothetical incentive packages have had rather mixed results. The Southern California Gas Company found about equal preference for an interest-free loan and a credit of about 50 percent (Berry, 1982~. A survey for Seattle City Light found about equal preference for an interest-free loan with payment deferred 10 years and for a combination of a 10 percent grant and a 6 percent loan; however, one-quarter of the respondents did not rank the interest- free loan option among the five choices given (Berry, 1982~. A national study in Canada found no difference in preference between loan subsidies and grants, but a preference for tax credits over grants when the size of the subsidy was held constant at 50 or 100 percent {Hickling-Partners, 1983~. And in a survey of customers of the Pacific
34 Gas and Electric Company' preference for an interest-free loan increased with income while the opposite occurred for a 50 percent rebate (Berry, 1982~. A survey of people eligible for audit and loan subsidy programs administered by the Northern States Power Company found that over 70 percent of the eligible population preferred a 20 percent rebate to a 7 percent loan with payment deferred until the home is sold. Only those respondents who had already taken advantage of the company's low- interest loan program preferred the loan (Hirst, Goeltz, et al., 1983~. These surveys show no clear overall preference for type of incentive, but they suggest that preference may be segmented by income or other customer characteristics. More consistent results come from surveys of people who have accepted or declined an incentive actually offered them. According to people who accept loan subsidies, they are often essential to action. Only 29 percent of those using zero-interest loans in Pacific Gas and Electric's CAL/NEVA program said they would have made the same retrofits if the loan program had not been available {Moulton, 1984~. The com- parable figure in the Bonneville Power Administration's Weatherization Pilot Program was 45 percent (Hirst, Bronfman, et al., 1983) and for the Northern States Power loan program, 29 percent (Hirst, Goeltz, et al., 1983~. However, many people simply will not borrow for energy conservation. This was the case for 77 percent of people eligible for the Michigan RCS program (Katz and Morgan, 1983) and 52 percent of the people who did not take out a loan even though they were eligible for a Northern States Power loan subsidy (Hirst, Goeltz, et al., 1983~. Among people who received home energy audits in the CAL-NE VA low-income program but who declined the program's interest-free loans, fear of indebtedness was given as the major reason (Moulton, 1984~. Concern about indebtedness was also a major factor in response to the Washington Water Power Company's zero-interest loan program. Four factors predicted 25 percent of the variance in householders' decisions to accept or decline the company's loan offer after receiving an energy audit: how convinced they were by the auditor's description of the program, their willingness to accept debt for weatherization, their willingness to use one's home to guarantee a loan, and the proportion of one's home mortgage left to be paid (Olsen, 1984~. Data for grant programs show a similar pattern: many households that accept grants say they would not have retrofitted without the assistance, but other households find grants an unattractive form of incentive. In the Low Income Electric Program sponsored by Seattle City Light, 68 percent of the participants said that if the free weatherization program were not available, they would not have taken advantage of a loan program (Newcomb and Weiss, 1983a). But many households decline lame Grants when offered. According to a study of the Eugene (or egon) water and Electric soara's weatherization program, one reason is lack of capital. The program's average grant required a household to spend $400 of its own, and many households, especially those declining the subsidy, did not believe they could afford that much. The evaluation concluded that, independent of income, households
35 that were better able to manage their budgets were more likely to take advantage of the grant program (Olsen and Fonseca, 1984~. The most likely interpretation of the survey data is that different types of households have different preferences with respect to incen- tives, partly as a function of income (see below), but also of other factors. Renters, for example, are ineligible for loan programs that require a lien on the home. And other characteristics of a household, such as ability to manage a budget, can also affect the attractiveness of loans and grants. The possibility that different kinds of house- holds prefer different kinds of incentives should be investigated further so that incentives may be designed for their target populations. The best approach to increasing participation may turn out to involve a choice of types of incentive. Participation in Incentive Programs Table 6 summarizes available data on participation in residential incentive programs available to households at all income levels. It should be interpreted with several qualifications in mind. Programs of the same type vary in the terms of the incentives they offer, the maximum level of investment they subsidize, their target populations, and the restrictions or added attractions they offer, and many programs change over time on the above dimensions. Programs also vary greatly in their marketing efforts, which obviously affect participation rates, and participation rates probably also change over time. Evidence from several programs (Moulton, 1984; Newcomb and Weiss, 1983b; Olsen and Fonseca, 1984; Weiss, et al., 1983; Wickman, 1984) suggests a learning curve, with rates of participation increasing over the first two to three years. Thus, participation rates equalized for time, as in Table 6, may not be fully comparable across programs. Although Table 6 shows a range of participation rates within each program type for which several reports were found, there is a clear trend. Grant or rebate programs generate the highest rate of retrofit activity (mean and median, 7 percent per year); interest-free loan programs generate a lower rate of retrofit activity (mean and median, 3 percent per year); and partial loan subsidies have the lowest rate of participation (mean, 1 percent per year; median, less than 1 percent). Assuming that other variations between programs are random, the difference in participation rates is statistically reliable.2 2 Comparison of the three program types (excluding the mixed program) by analysis of variance gives F = 9.62, p = .002 when all programs are considered, and F = 3.32, p = .08 when only programs in the United States are included in the analysis. The nonparametric Kruskal-Wallis test gives comparable values of H = 13.97, p = .0009 for all programs and H = 4.64, p = .10 for United States programs. The success of grant programs outside the United States is apparently responsible for much of the observed difference between types of incentives.
36 TABLE 6 Retrofit Activity in Response to Broadly Based Residential Incentive Programs Program (source ) Loan or Grant Rate (%) Length of Programa Rate of Retro- f ittingb (%) Partial Loan Subsidies Northern States Power (Hirst, Goeltz, et al., 1983) Rhode Islanders Saving Energy (Stern, Black, and Elworth, 1981) 19th Ward, Rochester, N.Y. (Katz and Morgan, 1983 New York Home Insulation and Conservation (New York State Public Service Commission, 1985) Mean Zero-Interest Loans Seattle City Light (Newcomb and Weiss , 1983a ; Weiss and Newcomb, 1981; Weiss et al., 1983) Bonneville Pilot (Lerman, Bronfman, and Tonn, 1983) Portland General Electric (Burnett, 1982) Pacific Power and Light (Hannigan and King, 1982) TVA Zero-Interest (Moulton, 1984 Puget Sound Power Loan (Puget Sound Power and Light Company (1983, 1984 Mean Mixed loans and Grants - Swedish loan and grant (Klingberg and Warkov, 1983: Wickman, 1984 ~ 9 mot 3 11 12 9-11 7 yr o o o o o o 2 yr. 0.2 1 yr. 1 · 0.1 1 f irst year 3 third year 6 2.5 yr. 4 9 mot about 1 1 yr. about 3 4 yr. 7 yr. 3 2 3 9 yr~d ~ ~.
37 TABLE 6 Continued) Loan or Program (source ) Grant Rate (%) Length of Programa Rate of Retro- fittingb (%) Par tial Grants or Rebates Bonneville Interim (Lerman and Bronfman, 1984) Eugene (Oregon) Buyback (Olsen and Fonseca, 1984) Puget Sound Power Grant (Puget Sound Power and Light Company, 1983, 1984) Canadian Home Insulation (Energy, Mines, and Resources Canada, 1983) Danish Grant (Petersen, 1984) Canadian Oil Substitutione (Anderson, 1984) British Home Insulation (Gaskell and Pike, 1983) Netherlands Grant (de Haan, 1985) 22-33 Mean 93 77 72 60 20 50 66 20 mot s 17 mot 4 3 yr. 2 5 yr. 34 mot 11 3 yr. 8 2 yr. 8 3 yr. 7 7 aPeriod covered in evaluation study. bHouseholds using the program's subsidy per eligible household per year. C35 percent grant up to about $400 and 12 percent loan for additional costs. About three-quarters of government payments have been in the form of loans. d For single-family homes. eThis program supported residential fuel switching. fall households eligible throughout 1979-1981; program restricted to renters in 1982 and had a 5 percent rate of retrofitting in 1983.
38 However, the warnings above about comparability must be borne in mind in judging the meaning of the statistical differences. The differences seem to be partly a function of the size of the incentives. Interest-free loans are worth more than partial loan subsidies, and they appear to be worth less than the large grants offered by U. S. programs. The strength of the latter conclusion, however, depends on the discount rate used to calculate the economic value of loan subsidies. Nominal discount rates in the range of 10 to 15 percent are consistent both with data on consumers' expressed preferences for energy conservation incentives (Stern, Berry, and Hirst, 1985) and with the performance of U. S. investment markets over the past few years. At a 10 percent nominal discount rate, the median net present value of the loan subsidy is 57 percent for the four zero-interest loan programs for which data are available; at a 15 percent rate, this value is 71 percent .3 Since these values are close to the 63 percent median subsidy in the eight partial grant or rebate programs in Table 6, the apparent preference for grants over loans may not be due to a difference in their financial value. Reluctance to assume debt for energy conservation may explain the preference, but there is no conclusive evidence on this point. Assumptions about discount rate do not affect the analysis when considering the effect of different sizes of grants or rebates. Among the eight grant and rebate programs in Table 6, the highest partici- pation rates seem to come in programs offering the smallest grants. The explanation for this surprising result can probably be found in the details of how the programs are designed and operated. Grant programs have had much greater success in countries other than the United States. The three U. S. grant programs, which offer a median subsidy of 77 percent, have a median participation rate of 4 percent per year; the five foreign programs, with a median subsidy of 50 percent, have a median participation rate of 8 percent a year. One possible explanation for this finding is a difference in the energy situations of the nations. Denmark has long made energy conservation a national priority, in part because of its great dependence on imported oil. However, that dependence does not exist for- Canada, and the Canadian grant programs have also been highly effective. A second possible explanation is climatic difference between the countries for which program evaluations are available and the areas served by the U. S. grant programs. The former serve homes with much greater needs for heating. However, the only available direct evidence shows that in the These calculations assume that loans payable on the resale of a house will come due, on average, in 10 years. Thus, the value of the subsidy in the Bonneville Pilot or Puget Sound Power and Light zero-interest loan programs is 61 percent at a 10 percent discount rate or 75 percent at a 15 percent discount rate. At the same discount rates, the subsidy value of a partially deferred 10-year Seattle City Light loan is 53 percent or 67 percent and that of a Tennessee Valley Authority 7-year loan is 30 percent or 41 percent.
39 Canadian Home Insulation Program, heating load in degree-days was unrelated to the likelihood of household participation (Hickling- Partners, 1983~. A third possible explanation is the smaller size of the investments made as a result of the foreign grant programs, almost all of which led to a smaller average investment than any of the three U. S. programs cited in Table 6. However, because of the lower rates of subsidy offered in the foreign programs, expenditures per household were not so different in the U. S. and foreign programs.4 A fourth possible explanation for the differences between the United States and other countries is in the procedure for obtaining subsidies. All three U. S. programs require householders to request and receive free energy audits from the sponsoring utility; the subsidy then applies only to those investments the utility determines to be economically justified. In the foreign programs, there is no require- ment for an energy audit (although audits are available privately in Denmark, and the grants can be applied to their cost): a householder purchases goods and services covered by program guidelines and claims the subsidy directly from the government. The O. S. procedure involves an additional step for the household and entails direct intrusion and regulation by an outside entity. Households that will not devote much effort to conservation decisions or that do not trust the local utility company may not request energy audits. The simplicity of the foreign grant programs and their willingness to entrust choices to households may make the programs more attractive. ''his argument is bolstered by the success of one grant program run on the foreign model in the United States: the Pacific Gas and Electric Company's grants for small commercial consumers. The company mails applications to commercial customers that use between 10,000 and 100,000 kwh/yr. of electricity, offering rebates amounting to SO percent of the estimated cost of items listed on the application. Customers simply submit the one-page application with a receipt for purchase and are mailed a check. In its first 14 months of operation, the participation rate averaged nine percent per year (D. Wilcox, Pacific Gas and Electric Company, personal communication), a value very similar to those observed in similarly structured programs in other countries and higher than that reported in any U. S. incentive program (see Table 6~. An audit-based system, despite lower participation rates, has potential advantages. Basing grants on expert audits may better protect against fraud and may save more energy, but the value of these advantages cannot be accurately estimated. The foreign program 4 The value of investments induced by the eight partial grant or rebate programs listed in Table 6 were, respectively, $1,700, $1,760, $1,960, $650Can., 8,870Dkr, $1,180Can., 60 pounds sterling, and 2,780 Builders; the costs to households were, respectively, $129, $400, $550, $260Can., 7,200Dkr, $590Can., 20 pounds sterling, and about 700 guilders. i
40 evaluations provide no data on fraud and few on energy savings.5 Only one foreign program, the Canadian Home Insulation Program, has yet produced data on measured energy savings. Those savings are near the average for U.S. incentive programs, even though the average investment per household in the Canadian program was only about $520 (Energy, Mines, and Resources Canada, 1983), compared with more than $1,500 in the U. S. programs for which data are available. However, differences in measurement methods are great enough to deter strong comparisons. It is worth noting a direct comparison in Table 6 between grant and zero-interest loan programs offered by the same sponsor at the same time. The Puget Sound Power and Light Company in Washington has offered zero-interest loans since 1978 and grants of 72 percent of retrofit costs since 1982. Both programs have yielded retrofits at the rate of two percent per year. However, the loan program is available only to homeowners in electrically heated homes while the grant program, which is not restricted to homeowners, is available to a population almost twice as large. Since renters rarely take advantage of home retrofit~programs, expansion of the program could have been expected to decrease participation rates; thus, the comparison is consistent with the idea that grants are more attractive than loans. Conclusions Despite the multifaceted differences between incentive programs, the data support a few general conclusions about the attractiveness of different incentives. The size of incentives is correlated with retrofit activity for loan subsidies, although data are insufficient to determine the shape of the relationship between subsidy and response. For grant and rebate programs, however, no systematic relationship is evident between retrofit activity and the size of a subsidy. The international comparisons emphasize that the size of the incentive is not the most important determinant of effectiveness in grant programs. Comparisons of types of incentive offer some evidence that house- holds prefer grants over loans. That conclusion, however, depends on the value assumed for the discount rate in comparing the value of grants to loans. More significant is the evidence suggesting that different kinds of households prefer different kinds of incentives. Survey data indicate that grants tend to be preferred by low-income households, those that feel they cannot afford to spend large sums, and those that are averse to indebtedness for other reasons. Preferences specific Gas and Electric Company strongly suspected, however, that customers in its commercial rebate program were taking advantage of the subsidies by installing inferior quality reflective roof coating and claiming full rebates. After roof coating was dropped from the list of to 5 After roof coating was dropped from the subsidized retrofits, the rate of participation dropped from 16 percent on an annual basis (D. Wilcox, Pacific Gas and Electric Company, personal communication).
41 are probably governed by other factors as well, and each type of incentive is probably attractive to some households that would not consider using the other type. Future research should address the question of market segmentation of preferences for incentives. As a first step, data can be collected from participants in programs offering a choice of incentive types. An experimental approach could also be used to see if a group of households offered a choice of incentives has a higher response rate than groups of households offered only one type of incentive program. NONFINANCI~ FAMES OF INCE=I~ PROMS Nonfinancial features of incentive programs are important to their success because nonfinancial motives affect energy users' decisions (Stern, 1984; Stern and Aronson, 1984; Stern, Berry, and Hirst, 1985~. The nonfinancial aspects likely to matter most in incentive programs can be considered under four categories--promotion, simplification, reliability, and trust. Promotion involves gaining consumers' attention through marketing and communication techniques and selling a program with effective communication techniques and extensive contact with potential clients. Simplification involves combining separate choices involved in retrofitting into a single decision. Reliability concerns the common fear of homeowners that contractors' work will fail to live up to its promise. And trust involves potential clients' attitudes toward program sponsors. The importance of nonfinancial factors was suggested by the data in the previous section, and it is illustrated more forcefully by the wide variation in consumer response to programs that are the same in finan- cial terms. Tables 7, 8, and 9 present data on consumer response to incentive programs that offered the same incentive packages through several different utility companies. Table 7 presents results from nine utility companies participating in a New York State-mandated loan subsidy program; Tables 8 and 9 present similar results from zero- interest loan and grant programs administered by the Bonneville Power Administration. In all three programs, there was more than a tenfold variation in program participation rates. Such variation may be due to differences at two points of consumer decision: the initial response to a program, as indicated by requests for energy audits, or the decision to make recommended retrofits after receiving an energy auditor's recommendations. The major source of variation in the New York program was the decision to take loans after receiving audits (Table 7~; in Bonneville's grant program, however, the major source of variation was in rates of requests for audits (Table 8~; and in Bonneville's zero-interest loan program, there was wide variation at both points of decision (Table 9~. The differences may be due in part to the size of the incentives offered. As incentives become more attractive (moving from Table 7 to Table 8 to Table 9), people who request audits become increasingly certain to take advantage of the incentive. But even with a very large incentive, households' initial interest in a program, and therefore the overall rate of
42 TABLE 7 Participation in the New York State Home Insulation and Conservation Program Audits per Loans Eligible Home (~/yr.) Utility Company Eligible Residential Customers per Audit (%) Loans per Eligible Home (%/yr.) Brooklyn Union Gas Co. Centr al Hudson Gas and Elec. Consolidated Edison Long Island Lighting Co. National Fuel Gas Co. N. Y. State Elec. and Gas Niagara-Mohawk Power Orange and Rockland Elec. Co. Rochester Gas & Electr ic Total (omitting double counting) or Average Ratio (highest/lowest) 451,000 155,000 300,000 690,000 357,000 533,000 850,000 105,000 184,000 2.5 1.4 2.6 1.3 1.0 1.2 1.0 2.0 1.8 3,500,000 1.6 6.0 3.5 6.1 1.4 12.6 11.0 11.6 0.7 28.1 0.14 0.05 0.16 0.02 0.12 0.14 0.12 0.01 0.51 8.2 0.13 2.6:1 40.1:1 51.0:1 NOTE: The New York Home Insulation and Conservation Program required regulated utilities in the state to offer reduced-rate loans (9-11 percent) to qualified customers receiving energy audits; the data are for 1978-1984 from New York State Public Service Commission (1985~. participation, is still highly dependent on nonfinancial factors. This finding suggests that the stronger the financial incentives are, the more important the nonfinancial factors, especially marketing, become to a program's success. This point is strikingly illustrated by a marketing experiment recently conducted as part of a shared savings program in Hennepin County, Minnesota {Miller and Ford, 1985~. The county government contracted with a private company that agreed to install energy-saving equipment in homes in return for payment from the homeowners of a percentage of the value of energy saved over a five-year period. The program was marketed only by direct mail, with addressees randomly assigned to receive three forms of solicitation letters. One letter was sent on the company's letterhead with no mention of cooperation with the county; the second went out on company letterhead and mentioned the county's role; the third went out on county letterhead and was signed by the chairman of the county Board of Commissioners. The source of information had a remarkable effect on consumer interest:
43 TABLE 8 Participation in the Bonneville Power Administration's Residential Weatherization Pilot Program Homes Audits per Homes Weatherized Eligible Eligible Weatherized per Eligible Utility Residential Home per Audit Home Company Eligible (~/yr.) (%) (%/yr.) A 3,164 7.0 45.0 2.4 B 1,933 8.2 88.5 7.3 C 1,900 10.8 88.1 9.5 D 1,875 10.5 33.3 3.5 E 5,000 4.0 60.0 2.4 F 944 23.2 28.1 6.5 G 2,520a 9.6a 7.8 o.8a H 13,351 5.1 56.6 3.3 I 6,840 5.5 77.0 4.4 J 517 11.6 66.7 7.7 K 1,073 17.0 66.7 10.4 Total or 39,317a 7.3a 55.4 4.oa Average Range (highest/lowest) 5.8:1 11.3/1 13.0/1 NOTE: The Weatherization Pilot Program, which operated for 2.5 years between 1981 and 1983, offered zero-interest loans, repayable when the house was sold, to single-family electric heating customers. Households receiving energy audits were offered free water heater insulation wraps, shower flow restrictors, and electric outlet gaskets. Data from Lerman, Bronfman, and Tonn (1983~. aUtility G could not estimate the number of its 5,040 residential customers who used electric heat. The figures noted were calculated on the assumption that 50 percent heated electrically, a figure lower than that for any of the other utilities in the sample. The effect of this assumption is probably to overestimate the success of Utility G's program and to narrow the range of participation rates in the last column of the table. requests for energy audits came from 6, 11, and 26 percent, respec- tively, of households receiving the three types of letters. After people had received audits, however, the form of the original solici- tation had no further effect; about 29 percent of all the people who actually met company representatives signed contracts with the company.
44 TABLE 9 Participation in the Bonneville Power Administration's Interim. Residential Weatherization Program Homes Audits per Homes Weatherized Eligible Eligible Weatherized per Eligible Utility Residential Home per Audit Home Company Customers (%/yr. ~ (96) (%/yr. A 62,047 11.8 61.0 7.3 ~ 5,056 14.2 82.9 11.8 C 99,994 23.2 57.9 13.3 D 3,500 23.1 83.4 19.3 E 2,853 1.6a 90.9 1.4a F 10,865 12.1 83.7 10.2 G 267,000 2.4a 77.2 1.ga Total or 433,115 (audit) Average 445,479 {retrofit) Bela 59.6 5.3a Range (highest/lowest) 14.5:1 1.6:1 13.8:1 NOTE: The Bonneville Power Administration Interim Residential Weatherization Program offers a grant to participating homes based on expected energy savings and amounting, on the average, to 93 percent of the cost of installed weatherization measures. The data cover 20 months in 1982-1983, from Lerman and Bronfman (1984~. aUnder previous programs, Utilities E and G had audited 400 and 17,800 homes, respectively, and had weatherized 354 and 5,482, respectively. The noted calculations are based on appropriately reduced figures for the eligible populations. Another recent study adds information about what may be responsible for initial responses to incentive programs. Mark Polich {1984) presented data collected by the state of Minnesota on energy audits in that state's RCS program. Although the program does not provide financial incentives, it is like most incentive programs in the United States in that its effectiveness depends on requests for energy audits. Minnesota had promulgated administrative rules for RCS that had directed utilities to subcontract with local auditors whenever possible, with the result that energy audits in Minnesota have been conducted in three different ways: using utility company employees, subcontracting with a private energy firm, and subcontracting with an existing community group. Table 10 summarizes the findings from the program. Utilities that subcontracted with a private firm cut costs and increased the rate of
45 TABLE 10 Characteristics of Energy Audits in the Minnesota Residential Conservation Service Program Audit Time Auditor Performancea Response Organization Cost Spent Rate Performing Audit (S) (has.) 1 2 3 (~) Utility Company 148 2-2.5 fair- fair fair 3.6 good Private Subcontractor 73 2 poor- good fair 5.7 good Community Group as Subcon- 54 3 fair- exc. good 14.7 tractor exe. aRatings of field observers of audits on quality of: (1) inspection of house and heating system, (2) interaction with homeowner, and (3) presentation of audit results. Data from Polich (1984~. requests for audits, while maintaining about the same quality of audits as utilities using their own personnel. Utilities that subcontracted with a community group cut costs further, and both the rate of requests and the quality of the audits improved. Not reflected in the table are differences in marketing strategy in the three types of programs. Despite a market research project that reportedly showed that response rates to audits in Minnesota might be increased tenfold by including shower-flow restrictors with audit offers, door-to-door marketing, and other aggressive promotional practices, only the community groups adopted such practices. And only the community groups contacted households after completing energy audits to encourage them to take the recommended conservation actions. Polich (1984) interpreted the results in terms of the incentives facing the programs' sponsors. He argued that most of the major utilities in Minnesota were aggressively marketing energy at the time and had little incentive other than customer relations and regulatory requirement to pursue conservation programs. Polich sees the community groups as motivated to enhance the welfare of community residents and to benefit the local economy by keeping energy dollars in the community. Whatever the groups' motives, the higher rate of audit requests in programs operated by community groups was probably due to some combination of aggressive marketing, a reputation gained by doing high-quality audits, and residents' trust in the groups.
46 EFFECTIVENESS OF INCENTIVES IN THE LOW-INCOME HOUSING SECTOR Distributional effects are a major concern in evaluating con- servation programs. Low-income households pay a much greater share of their income for energy than higher-income households (Cooper et al., 1983) and that share has been rising rapidly with increasing energy prices (Energy Information Administration, 1982~. Moreover, low-income housing has great potential for improvement because it is of below- average energy efficiency {see Table 11~. Despite concern about weatherizing low-income housing, most residential conservation programs in the United States have disproportionately attracted higher-income households. This result is consistently reported in evaluations of home energy audit programs (Hirst, Berry, and Soderstrom, 1981) and of the more comprehensive programs offered through the Residential Conservation Service (U.S. Department of Energy, 1984; Hirst, 1984~. However, these programs are not incentive programs. They primarily offer accurate and specific information to those who might invest in energy-efficient equipment; they do little to alleviate the shortage of investment capital among the owners and occupants of low-cost housing. Thus, there is reason to hope that incentive programs may be more successful at reaching the low-income households. In this section we examine data relevant to that hope. We review reports of participation by low-income and higher-income households in TABLE 11 Percentage of Homes Having Weatherization Features in 1980, by 1979 Family Income Family Income Weatherization Poverty National $35,000 Characteristic Level Average or more All windows have storm windows 25 38 40 All doors leave storm doors 23 32 33 All of ceiling is insulated 47 69 82 Ceiling is completely uninsulated 34 15 6 All walls are insulated 37 53 67 All walls are uninsulated 37 20 11 Some or all storm windows some or all storm doors, and ceiling insulation 25 49 56 None of the above types of insulation 27 10 4 Data from Energy Information Administration (1982~.
47 incentive programs available to both groups and reports of incentive programs aimed specifically at low-income or rental housing or at people in low-income communities. We give particular attention to the role of nonfinancial features of incentives, which may be critical for reaching low-income populations (see Gaskell and Pike, 1983; Hirst, Berry, and Soderstrom, 1981; Stern, Black, and Elworth, 1981~. Low-Income Participation in Incentive Programs The largest financial incentive program in federal conservation tax credit, has been used the United States, the predominantly by upper-income households (Hirst et al., 1985~. The limited use of tax credits by low-income households may be due to the fact that many of them do not pay enough tax to be able to claim a credit. In addition, most low-income households have no need to keep detailed tax records except to claim an energy credit and so are unlikely to do so. There are reasons to expect zero-interest loan programs to be more attractive than tax credits for low-income households. They offer the capital that low-income people especially lack, and they do not require recordkeeping. They often defer payment until resale of the home, when cash becomes available, or stabilize household cash flow by keeping the payments small enough to be covered by energy savings. Yet they have not been very attractive to low-income households. For example, the Pacific Gas and Electric Company found that only 2 percent of the households taking zero-interest loans in 1983 had incomes below 125 percent of federal poverty guidelines (Moulton, 1984~. The Tennessee Valley Authority, which offers zero-interest loans to a population of which it defines 32 percent as low-income, gave only six percent of its loans to low-income customers in 1979. Through special efforts to reach that population, TVA substantially increased audit requests from low-income households from 7 percent of requests in 1979 to 21 percent in 1982 (Moulton, 1984), and the proportion of loans to low-income households also increased, to 12 percent in 1982. However, the drop in participation from requesting audits to using loans indicates that low-income households are less attracted to loans than higher-income households, even when they know of the program. Similarly, higher- income households were overrepresented among households taking out zero-interest loans under the Bonneville Power Administration's Weatherization Pilot Program (Hirst, Bronfman, et al., 1983) and Puget Sound Power and Light's loan program (McCutcheon, 1983~. Some grant programs have had better results in the low-income housing sector, possibly because of the success of grants at encourag- ing investment among people who receive energy audits. The limited data show that low-income people are less likely than higher-income people to hear of a grant program, but that once aware of it, they are at least as likely to use it. In the Canadian Home Insulation Program, which offered a grant of up to 500 Canadian dollars to pay 60 percent of the cost of materials and labor for home weatherization, higher- income people were more likely to be aware of the program: the range was from 51 percent to 78 percent with increasing income
48 (Hickling-Partners, 1983~. But unlike the experience of the TVA loan program, low-income people did not drop out disproportionately before deciding to invest. Among the 72 percent of the national sample who knew of the program, those who participated were lower on socioeconomic indicators than those who did not. Participation among people with a high school education was around 50 percent, compared with about 27 percent for people with college and university educations. Lower- to middle-income households participated at rates of 35 to 42 percent, compared with an average of 33 percent for the whole sample. Similarly, among households that requested energy audits in the Eugene (Oregon ) Water and Electr ic Board ' s buy back program, those that used the program' s grants did not have higher incomes, as is the usual experience in loan programs (Olsen and Fonseca, 1984) . 6 Incentive Programs for Low-Income Households The failure of many incentive programs to penetrate the low-income housing market has led some of the programs to increase their efforts. For example, after the Tennessee Valley Authority realized that only seven percent of the energy audits and six percent of the loans in its zero-interest loan program were going to low-income households, it began using local community groups as outreach agents, distributing specially written promotional material by hand, and promoting energy audits with an offer of three free loaves of bread (Moulton, 1984~. noted above, the promotion effort raised the proportion of audit requests from low-income people to 21 percent after three years. The outreach efforts attracted the attention of low-income people, although their rate of investment still lagged, possibly because of their aversion to indebtedness. Table 12 summarizes the features and experiences of ten incentive programs aimed specifically at low-income populations for which recent and reasonably detailed reports are available. The sample is unrepre- sentative of all such programs, but it includes many programs operated by groups that are making serious and explicit efforts to reach low-income populations and that are aware of what similar groups have done. Thus, the sample probably represents the "state of the art" in low-income incentive programs. A few characteristics stand out. All of them are consistent with other available evidence on how to reach low-income households. First, the programs offer very strong incentives: five of the programs offer free weatherization, while none of the programs available to a cross- section of the population does. Second, the programs rely heavily on community groups, usually in a cosponsorship role and, where details on marketing are available, always in a central marketing role. Third, the marketing efforts are labor-intensive, often including door-to-door As 6Households that used the grants had a median income of $22,400 compared with $25,000 for households that did not. ..
49 canvasses, presentations at public meetings, and workshops held off the sponsors' premises. This practice is consistent with evidence that a strong outreach component is important for all residential conservation programs (e.g., U.S. General Accounting Office, 1981), especially those serving low-income households. Fourth, where the programs have had sufficient resources to target a full community, participation is often strong. Four programs, in Rochester, Memphis, Seattle, and a community in the Netherlands, had a mean participation rate of eight percent per year and a median rate of five percent per year even though they were aimed at segments of the housing market that have been consistently unresponsive in the past. 7 The studies of these programs indicate that incentive programs can be made to reach the low-income population with at least the level of success that incentives have shown with the general population. Success seems to require strong incentives, a credible local sponsor, and a stable communication network in the community or intense marketing efforts relying on personal contact with potential clients. CONCLUSIONS The evidence on seemingly obvious facts about incentives is surprisingly inconclusive. Even the size of an incentive does not have a clear and strong relationship to consumers' willingness to use it: other sources of variation in the data are much larger. When the size and type of incentive are held constant, participation rates typically vary by a factor of ten or more. With so much money being spent on incentive programs, the need for further research is clear. On the basis of our previous work and the recent evidence, we believe that such research should emphasize issues of program design, marketing, and implementation, particularly: the credibility and motivation of the sponsoring organization; the use of credible local organizations in the marketing effort; emphasis in marketing on stimulating word-of-mouth communication; and efforts to make successes widely known to past and future clients. Chapter 4 outlines ways to assess the roles of such variables. Below, we summarize the substantive conclusions that can be drawn from available data. The evidence that larger incentives increase participation in conservation programs is surprisingly weak. The data on loan subsidies 7The highest rate of participation by far was in the one program outside the United States, which is consistent with findings about programs for the general public. m e phenomenon in this case probably has much to do with the community, where two small, demographically stable developments were chosen for the program and where a consortium of public and private local institutions developed the program for its benefits to local employment as well as for energy savings (van der Linden and van Eijk, 1985~.
so · - u] ~ o · - u] o :: o u H So o In o · - Er3 En - lU ~ _ O · - ~ ~ ~ So O V O O U] _ In us So 0 Ed · - C) ~ by) _I ·- ~ ~ ~ · - ' 0 · - O · - ~ ~ ~ O ~ ~ O ~ ~ ° ~ ~ ·~ ·~ O '- O ~ U ~ ~ 1 ~ ~ ~ ~ 1 ~ ~ 3 0 ~ ~ O ·- O ·' ~ O ~ ~ 1 Y ~ ~ ~ 1 ~ 3 a) ~ a, s _t ~ v ts dJ ~= ~~ ~ Q O ~ ~ ~ Q ~ ~ O U ~ oU Q ~ ~ · - · O · o~ O3 ~ aQ~ ~ ~= a V q:5 a~ .~ - ~ ~ ~ ~ aJ ' ~ O ~ S~ O ~ ~ ·- ~ S~l _I :~ ~t ~ ~ ~t ·- ·- ~ O ~ ~ ~ ~ ~ ~ ~n U2 ~1 dJ ~ . - ~ e" .~ . - . - ~ ~ aJ, ·rl eth ~ S "d O ~ C~ O P~ ~ E~ ct~ _I Q., e U] JJ = ~ ~= O =~ - -= =' - -= "= ~ ~ ~ ~ · - ~ ^~ - ' - ~ ~- - ~- - O ~ ~ ~ O ~ ~ ~ ~ ~ ° .~ ~ · ~ o ~ o ~ o 7 ~ o ° ~ o~ v o~ ~ o~ ° a ~q~ c'~ Q ~ ~ ~ ~ ~ 3 V ~ V C) ' C, Q Q ~ 0 ~0 o .0 ~ O O ~ O o O ~ U) .,' ~ O O ~ I O ~ O O O N p4 ·- o U) ~~ ~ ,~ ~n O U2 ~ 1 O O 1 ~ ~ ~ ~ O . - ~ ~ ~ · - ~ ~ ~ N .~1 ~ ~ V O U] .rl ~ _1 0 ~ a) ~ ~ ~ ~ ~ . ~ ~ ~ ~ ~ ~ (V ~ =·- LO ~ ~ ~ ~ S ~ ~ O ~ ~ ~ C3 V ~ ~= · o - ~ O ~ O ~ O ~ O ~I H `4 3 N _I ~ _I I LO c0 ·~ ~ U) - · ~ ~ _ ~ ~ . - ~ ~ ~ · - ~ ~ ° ~ '~ O s: ^ · ^ ~ ~ ~ ~ O a, ·,~ a~ ~ ,~ ·~ z ~ c: .y ·, - - ~ 00 a) u' · - r~ ~ ~ {Q te ~ c) ~ cn ~ ~ a, ~ ~ .-, N c) ' ~ ~' _I ~ S~ ~ ~ ~ W ~ ~ m ' ~ ~ - ·~: a~ ~, a S~ ~ ~ C.) Q ' tT5 ~ tt ' ~ ~ ~ ~ ~ ~ O ' ~ G) ~ ~ ~ ~ O ~ ~ ~ N S U2 Q ' ~ k 0, N ~ S ~ ~ Z ~ ~ H P~ 4) ~ s s ~ O~ Q s ~ ~ eG ::~ 1 3 ·~~ 3 4~ O ~ s ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ O Q (L) tQ ~ ~ `4 U ~ O :< 0 3 ~ ~ ~ ~ ~ ~ H ~ ~ ~ ~ ~ ~ X ~ ~ ~ ~
51 to ~ ~ o :, l en o ~ ~ ye ~ ~ en o o ~ ~ m o o o c: ~ 3 o ~ A, ~ 1 0 hi- - O O G) · - £ ' ~ ~ 1 1 0 en ~ a' Q o Sl pod O ~ S ·- · en C) · - ~Q a, . . ~ ~ £ ~ o o . , - _, o ~ - o ~ .` .- o HI, . a, ' . .,, ~ An, ~ c, O ~ ~ ~ . - U as Q. . .. ,' . ~ o ~ ~ ~ ~ ~ ~ ~ · - ~ £ 0-- >-- ~ O c ·^ O ~ ~ o ~ o O v ~ :3 C~ :3 O ~ £ o o o o L~ _, · ~ a ~ ~ · - 1 ~ 1 ~ o ·~ (V ~ · · .^ ~ ~ O s 0 a) ~ s 0 · - ~, ~ ~ =. - a, O ~ ~ aJ . 3 N U~) ~ 1~ 3 N t~ ·~1 ~Q . - ^ S s- ' - ~ a' ~ ~ . ~ oo ~ ' ' 0\ I N ·IJ ey ~ _' ~) _I ~ ~ ~ ~ Q ~ ~ ~ O O t~ ~ ~ ~ ~ H 3 o Se ~ N ~ ~ ' H ~ ~ ~ ~ ~ ~ 3 ~ ~ ~ ~ ~ ~ ~ al. O :< ~ ~ ~ O H Z ~ ~ ~ U] _ ~ ~ ~ ~ ~ ~ _ ~ 1 S _I · - ~ ~Q S 3 S~ ~ O s C) X o o . - Q O ~ s S~ o ~Q ~5 o s U] o s ~Q · - .^ Q ·e · - O ,- v · - s ~ V s ~ a,, O V O ~ ~ a., s~ O e-- a, ~ a ~ ~ s a, ~ eq o~ c0 -- - ~ V s E~ ~ O · V ~ S~ ~ C' O ~ CQ · O ~ ~, - 3 O [Q o U] {Q ~s ~ rms n' Q 3 · - <) a, .,, ~Q X a) V ~r1 S~ 3 ~q o
52 of different sizes supports this conclusion, but not the data on grants and rebates of different sizes. With larger incentives, more households that have made contact with an incentive program (e.g., through an energy audit) seem to make investments. However, larger incentives do not seem to induce house- holds to make initial contact with a program. Among programs offering very large incentives, differences in reaching their clienteles can account for tenfold variations in participation rates. The type of incentive makes a difference. There is some evidence that households prefer grants or rebates to loan subsidies of equal value. This conclusion depends, however, on the discount rate assumed when equating loans and grants. Preference for loans over grants varies across households. Low- income homeowners and households that are pessimistic about their financial futures tend to prefer grants and rebates, while higher- income households and people skilled in managing budgets tend to prefer loans. Programs that of fer a choice of loans or grants may be able to attract people who would reject incentives if only one type were offered. Differences in program marketing and implementation are probably responsible for the widely disparate rates of participation in programs offering identical financial incentives. Potential clients are attracted to programs that have energy audits conducted by local community groups or other organizations that they trust and that have strong motives to make the program work. Aggressive marketing through word-of-mouth and other attention-getting media increases participation. Low-income households can be reached by financial incentive programs. To do so requires a strong incentive, preferably a grant or rebate rather than a loan; implementation by a trusted organization; and a strona market; na efEc~rt that aims: to r-Af~h people by Existing community groups have become the organizations of choice for marketing and often for managing low-income programs both in the United States and elsewhere. word-of-mouth and through local social networks.
