Free Riding on Energy Efficiency Subsidies: The Case of Natural Gas Furnaces in Canada.

Author:Rivers, Nicholas

    Home energy efficiency retrofits are generally considered one of the most cost effective options for significantly reducing greenhouse gas emissions and improving energy efficiency. For example, a recent analysis by McKinsey&Company finds that a number of residential energy efficiency improvements including heating, insulation, lighting, and appliance retrofits, have the potential to simultaneously save energy, reduce greenhouse gas emissions, and lower costs of energy services. Even so, uptake of residential energy retrofits is often lower than expected. To address this "energy efficiency gap," a number of countries offer subsidies, grants, or tax rebates to encourage uptake. Focusing on a Canadian program that provided grants to encourage home energy retrofits over a number of years, this study provides one of the few quantitative evaluations of large-scale energy efficiency retrofit programs to date. We find that a significant proportion of the total funds allocated to the program were spent on providing subsidies to households that would have completed the upgrade even without the subsidy. As a result, we conclude that the program was an expensive way to reduce greenhouse gas emissions. We also find that retrofit grants were predominantly received by wealthy households, such that the program had a regressive impact on the distribution of income.

    Economists have long been concerned with the appearance of a so-called energy efficiency gap--the difference between the actual diffusion of energy-saving products and that which would be expected if consumers made all energy-saving investments with positive net present value. Early empirical analysis diagnosed the gap with the observation that implicit discount rates applied by consumers for the purchase of energy-consuming durables were higher than for other investments of similar risk (Hausman, 1979; Train, 1985). More recent studies incorporate elements of behavioral economics to explain the energy efficiency gap. Using modern econometric techniques and large datasets, these studies suggest that other explanations for the gap include consumer heterogeneity, hidden costs, uncertainty, as well as behavioral anomalies such as inattention, habit formation, and biased beliefs.

    Although it is difficult to precisely quantify the magnitude or even presence of an energy efficiency gap, suspicion that it is large has influenced the design and application of energy and environmental policy. Such policies take on a number of forms. Information policies are used by governments and utilities to inform households and firms of cost-effective opportunities for saving energy. Such policies include provision of low-cost audits to firms and households as well as development of energy efficiency labels. Regulations are also widely applied to improve energy efficiency. In developed countries, most household appliances must meet minimum energy efficiency standards in order to receive permission to be sold. Governments also frequently offer financial incentives, including subsidies and tax credits, to encourage consumers to make energy efficient choices regarding new equipment, materials and building alterations.

    One of the more prominent financial incentives for encouraging energy efficiency improvements takes the form of residential retrofit subsidies. In Canada, at the federal level, both Liberal and Conservative governments have implemented subsidy programs designed to encourage homeowners to retrofit their homes to improve energy efficiency. At the provincial level, a number of provinces have offered subsidy programs for residential energy efficiency retrofits, typically piggybacking on the federal programs. In addition, grant programs have been offered by electric and natural gas utilities.

    Our focus in this paper is on evaluating residential retrofit grants offered between 2007 and 2011 in Canada. We concentrate on natural gas furnace retrofits, since these are both a major source of potential energy efficiency improvements as well as being relatively homogeneous, which facilitates analysis. Our analysis uses a large and detailed administrative dataset, which we believe is unique to the evaluation of residential retrofit programs. We have detailed information on demographic and building characteristics for more than 300,000 households that participated in the program, which is the entire population of program participants for the period we consider. We observe retrofit choices for each of these households. We identify the effect of residential retrofit grants on natural gas furnace choice through the significant spatial and temporal variation in retrofit grants in the data.

    We estimate discrete choice models to determine the sensitivity of natural gas furnace choice to changes in subsidy rates and operating costs. Using our estimated coefficients, we simulate a counterfactual scenario in which retrofit grants are not offered. By comparing this scenario with actual furnace choices, we are able to estimate what proportion of high efficiency natural gas furnaces purchased during the sample timeframe were due to the retrofit grants, and what proportion would have occurred regardless. Our simulations suggest that around 50 percent of the subsidies were received by homeowners that would have adopted the high efficiency furnaces during the time period covered by the subsidy even without the subsidy. Our simulations also suggest that over the long term (at the time of eventual furnace replacement), about 80 percent of homeowners would have chosen an identical furnace even without a subsidy. This high level of free-ridership worsens the cost-effectiveness of the program from the government's perspective. (1)

    We extend our analysis in two other directions. First, we couple the administrative database with census information on income. This allows us to estimate the incidence of the program according to household demographics. We find that the program targets middle- and high-income households, and excludes low-income households, primarily because high-income households are more likely to own their home than low-income households. Second, we estimate the impact of the program on energy consumption and greenhouse gas emissions. Our findings suggest that the program reduced greenhouse gas emissions by around 4 Mt C[O.sub.2] in total (over the life of the furnaces). We compare the C[O.sub.2] reductions achieved by the program to the cost of the program, and find that the program had a cost effectiveness of between $70/t and $110/t C[O.sub.2], depending on assumptions relating to the timeframe of the analysis and the costs included in the evaluation. In any case, the program appears to be an expensive way for government to source greenhouse gas reductions.

    Our paper is closely related to a handful of others that analyze the effects of residential energy efficiency subsidies on adoption of new technologies. Cameron employs a nested logit model to assess the determinants of households' choices to undertake home-energy retrofits. Based on a survey of US households, she finds free ridership rates of up to 86 percent when house retrofit subsidies worth 15 percent of the cost of the measure are provided. Similarly Hartman uses a dataset from a US electric utility that includes demographic data and electricity consumption data for a population that includes both participants and non-participants in a utility energy conservation program. With a nested logit model accounting for both the program participation decision and the conservation decision, he estimates that about 45 percent of the electricity savings attributed to the program would have occurred even in the absence of such a program. Revelt and Train estimate a mixed logit model, combining a stated preference choice experiment on consumer choice of appliance efficiency under various levels of subsidy with revealed preference data on actual appliance choice. When simulating the effect of a subsidy program using their estimated model, they find that the proportion of adopters of energy efficient appliances increases by about 8.5 percent, but that subsidies are paid to 15.8 percent of the population, suggesting a free-ridership rate of 46 percent. Train and Atherton use the same dataset to estimate a nested logit model, again combining revealed preference and stated preference data. They estimate free ridership rates of 36 percent for refrigerator programs and 66 percent for air conditioner programs. Grosche and Vance use an error-components logit model and estimate a free ridership rate up to 50% for German households, based on estimates of marginal willingness to pay for energy cost reductions. Based on a similar model and data, Grosche, Schmidt, and Vance use simulation to account for non-marginal changes in willingness-to-pay and obtain free-ridership rates of over 90 percent for a modest home retrofit subsidy and 70 percent for a more substantial incentive program. Overall, these studies suggest that a large proportion of the recipients of this type of subsidy program are consumers that would have adopted the technology even without the subsidy. As a result, the cost-effectiveness of this type of program is generally fairly poor.

    Our paper builds on this prior literature in at least three respects. First, our analysis is based on a large administrative database. In contrast, much of the prior literature uses much smaller datasets, and these are often derived from surveys rather than from program administrative records. Our large dataset helps us provide precise parameter estimates, and the fact that it is from administrative records helps us get around problems of recall that can affect survey data on home retrofits. Second, our dataset covers a program that offered financial incentives of varying size, with variation occurring both between regions and over time, while the basic structure of the program...

To continue reading