Assessing the Energy-Efficiency Gap

Global energy consumption is on a path to grow 30-50 percent over the next 25 years, bringing with it, in many countries, increased local air pollution, greenhouse gas (GHG) emissions, and oil consumption, as well as higher energy prices.  Energy-efficient technologies offer considerable promise for reducing the costs and environmental damages associated with energy use, but these technologies appear not to be used by consumers and businesses to the degree that would apparently be justified, even on the basis of their own (private) financial net benefits.

For some thirty years, there have been discussions and debates about this phenomenon among researchers and others in academia, government, non-profits, and private industry, typically couched in terms of potential explanations of the so-called “energy efficiency gap” or “energy paradox.”

Thinking About the Energy-Efficiency Gap

I wrote about this some two years ago at this blog ().  I  noted then that Professor Richard Newell of Duke University and I had just launched an initiative – sponsored by the Alfred P. Sloan Foundation — to synthesize past work on potential explanations of the energy paradox and identify key gaps in knowledge. We subsequently conducted a comprehensive review and assessment of social-science research on the adoption of energy-efficient technologies.

We worked with leading social scientists — including scholars from economics, psychology, and other disciplines, at a workshop held at Harvard — to examine the various possible explanations of the energy paradox and thereby to help identify the frontiers of knowledge on the diffusion of energy-efficient technologies.  As materials became available, we posted them at the project’s Harvard website and the project’s Duke website.

Releasing a New Monograph

I’m pleased to inform readers of this blog that we have now released a major monograph, Assessing the Energy Efficiency Gap, co-authored with Todd Gerarden, a Harvard Ph.D. student in Public Policy and a Pre-Doctoral Fellow of the Harvard Environmental Economics Program (HEEP).  The monograph draws in part from the research workshop held at Harvard (in October 2013), in which most of the U.S.-based scholars (primarily, but not exclusively, economists) then conducting research on the energy-efficiency gap participated. HEEP co-sponsored a second such research workshop with the Centre for European Economic Research (ZEW) in Mannheim, Germany in March 2014, where European economists explored the same topic. Closely-related research was presented by panelists at the annual conference of the Allied Social Science Association in January 2015.

In the new monograph, Gerarden, Newell, and I examine both the “energy paradox,” the apparent reality that some energy-efficiency technologies that would pay off for adopters are nevertheless not adopted, and the broader phenomenon we characterize as the “energy-efficiency gap,” the apparent reality that some energy-efficiency technologies that would be socially efficient are not adopted. The contrast is between private and social optimality, which ultimately has important implications for the role of various policies, as well as their expected net benefits.

Four Key Questions

We begin by decomposing cost-minimizing energy-efficiency decisions into their fundamental elements, which allows us to identify four major questions, the answers to which are germane to sorting out the causes (and reality or lack thereof) of the paradox and gap.

First, we ask whether the energy efficiency and associated pricing of products on the market are economically efficient. To answer this question, we examine the variety of energy-efficient products on the market, their energy-efficiency levels, and their pricing. Although the theory is clear, empirical evidence is—in general—quite limited. More data that could facilitate potential future empirical research are becoming available, although firm-level data are much less plentiful than data on consumers. We do not see this area as meriting high priority for future research, however, with the exception of research that evaluates the effectiveness and efficiency of existing energy-efficiency information policies and examines options for improving these policies.

Second, we ask whether energy operating costs are inefficiently priced and/or understood. Even if consumers make privately optimal decisions, energy-saving technology may diffuse more slowly than the socially optimal rate, because of negative externalities. So, even if the energy paradox is not present, the energy-efficiency gap may be. As in the first realm, the theoretical arguments are strong. Empirical evidence is considerable, and in many cases data are likely to be available for additional research. Existing policies appear not to be sufficient from an economic perspective, suggesting that further research is warranted. Indeed, we ascribe high priority to the pursuit of research in this realm.

Third, we ask whether product choices are cost-minimizing in present-value terms, or whether various market failures and/or behavioral phenomena inhibit such cost-minimization. We find that the empirical evidence ranges from strong (split incentives/agency issues and inattention/salience phenomena) to moderate (heuristic decision-making/bounded rationality, systematic risk, and option value) to weak (learning-by-using, loss aversion, myopia, and capital market failures). Importantly, here, as elsewhere in our review, the bulk of previous work has focused on the residential sector and much less attention has been given to the commercial and industrial sectors. Some areas merit priority for future research, such as empirical analysis of split incentives/agency issues in areas where efficiency standards are not present, and much more work can be done in the behavioral realm.

Fourth, we ask whether other unobserved costs may inhibit energy-efficient decisions. We find that the empirical evidence is generally sound, and that data needed for more research are available. We assign a relatively high priority to future research, particularly to aid understanding of consumer demand for product attributes that are correlated with energy efficiency, thereby informing policy and product development decisions.

Three Categories of Potential Explanations of the Gap

Finally, we ask what these findings have to say about the three categories of explanations (reviewed in detail in my 2013 essay at this blog) for the apparent underinvestment in energy-efficient technologies relative to the predictions of some engineering and economic models: (1) market failures, (2) behavioral effects, and (3) modeling flaws.  In brief, potential market-failure explanations include information problems, energy market failures, capital market failures, and innovation market failures. Potential behavioral explanations include inattentiveness and salience, myopia and short sightedness, bounded rationality and heuristic decision-making, prospect theory and reference-point phenomena, and systematically biased beliefs. Finally, potential modeling flaws include unobserved or understated costs of adoption; ignored product attributes; heterogeneity in benefits and costs of adoption across potential adopters; use of incorrect discount rates; and uncertainty, irreversibility, and option value.

It turns out that all three categories of explanations are theoretically sound and that limited empirical evidence exists for every category as well, although the empirical research is by no means consistently strong across all of the specific explanations.  The validity of each of these explanations—and the degree to which each contributes to the energy-efficiency gap—are relevant for crafting sensible policies, so Gerarden, Newell, and I hope that our new monograph can help inform both future research and policy.  Given the many energy-efficiency policies and programs that are already in place, high priority should be given to research that evaluates the effectiveness, cost-effectiveness, and overall economic efficiency of existing energy-efficiency policies, as well as options for their improvement.


Thinking About the Energy-Efficiency Gap

Adoption of energy-efficient technologies could reap both private and social rewards, in the form of economic, environmental, and other social benefits from reduced energy consumption. Social benefits include improvements in air quality, reduced greenhouse-gas emissions, and increased energy security. In response, governments around the world have adopted policies to increase energy efficiency.  Still, there is a broadly held view that various barriers to the adoption of energy-efficient technologies have prevented the realization of a substantial portion of these benefits.

For some thirty years, there have been discussions and debates among researchers and others in academia, government, non-profits, and private industry regarding the so-called “energy efficiency gap” (or “energy paradox”) — the apparent reality that many energy-efficiency technologies are not adopted even when it makes sense for consumers and businesses to do so, based on their private costs and benefits. That is, decision makers appear to “under-invest” in energy-efficient technologies, relative to the predictions of some engineering and economic models.

What causes this gap?  The answer to that question could presumably help inform the development of better public policy in this realm.

Possible Explanations for the Energy-Efficiency Gap

Potential explanations for the energy efficiency gap tend to fall into three broad categories: (1) market failures, such as lack of information or misplaced incentives; (2) behavioral effects, such as inattentiveness to future energy savings when purchasing energy-consuming products; and (3) modeling flaws, such as assumptions that understate the costs or overstate the benefits of energy efficiency.  In this essay, I simply want to outline the types of hypothetical explanations of the gap that have been posited within these three broad categories.

Market-Failure Explanations

First, various Innovation Market Failures have been posited, including:  research and development (R&D) and learning-by-doing spillovers; inefficient product quality and differentiation due to market power; and inefficient introduction of new products due to consumer taste spillovers (for example, consumers becoming comfortable with a new technology).

Second, another set of potential market-failure explanations for the gap may be characterized as Information Problems.  These include:  lack of information on the part of consumers (learning-by-using or so-called experience goods; energy prices; energy consumption of products; and available substitutes); asymmetric information (the “lemons problem”); and split incentives and principal-agent issues (such as the frequently-discussed renter/owner dichotomy).

Third, there are Capital Market Failures and Liquidity Constraints, which may be a particularly significant issue in developing-country contexts.

Fourth, there are Energy Market Failures, including various externalities (environmental, energy security, congestion, and accident risk), as well as average-cost pricing of electricity.

Behavioral Explanations

The rise of behavioral economics has brought to the fore another well-defined set of potential explanations of the energy efficiency gap.  A variety of alternative taxonomies could be employed to separate these explanations, but one such taxonomy would categorize the explanations as:

Model and Measurement Explanations

The third category of possible explanations of the energy efficiency gap consists essentially of a set of reasons why observed levels of diffusion of energy-efficiency technologies may actually be privately optimal.

First, there is the possibility of unobserved or understated adoption costs, including unaccounted for product characteristics.

Second, there may be overstated benefits of adoption, due to inferior project execution relative to assumptions, and/or poor policy design.

Third, an incorrect discount rate may be employed in an analysis, when the correct consumer and firm discount rates should vary with:

  • opportunity cost of and access to capital
  • income
  • buying versus retrofitting equipment
  • systematic risk
  • option value (see below)

Fourth, there is frequently heterogeneity across end users in the benefits and costs of employing energy-efficiency technologies, so that what is privately optimal on average will not be privately optimal for all.  This can refer either to static (cross-sectional) heterogeneity or to dynamic (intertemporal) heterogeneity, that is, technology improvements over time, which raises two possibilities:  the reality of some potential adopters being short of the frontier, and the presence of option value to waiting.

Fifth and finally, there is the possibility of uncertainty (real, not informational, as above), irreversibility, and option value.  This could be due to uncertainty regarding future energy prices, or can be linked with option value that arises for delaying investments that have only minimal if any salvage value.

Public Policy and Next Steps

Determining the validity of each of these possible explanations — and the degree to which each contributes to the energy efficiency gap — are crucial steps in crafting the most appropriate public policy responses.

To inform future research and policy, Professor Richard Newell of Duke University and I have launched an initiative – sponsored by the Alfred P. Sloan Foundation — to synthesize past work on these potential explanations of the energy paradox and identify key gaps in knowledge.  We are conducting a comprehensive review and assessment of published and ongoing social-science research on the adoption of energy-efficient technologies, including scholarly literature, industry case studies, reports from national and sub-national governments, and, to the extent possible, consulting reports evaluating specific programs.

We are working with leading social scientists — including scholars from economics, psychology, and other disciplines — to examine the various possible explanations of the energy paradox and thereby to help identify the frontiers of knowledge on the diffusion of energy-efficient technologies.  We hope the products of this initiative will help decision makers in industry and government better understand the energy efficiency gap, and will thereby contribute to decisions that maximize the potential economic, environmental, and other social benefits associated with optimal adoption of energy-efficient technologies.  As materials become available, we will post them at the project’s Harvard website and the project’s Duke website, and I will alert readers of this blog.  In the meantime, please stay tuned.


Both Are Necessary, But Neither is Sufficient: Carbon-Pricing and Technology R&D Initiatives in a Meaningful National Climate Policy

For many years, there has been a great deal of discussion about carbon-pricing – whether carbon taxes or cap-and-trade – as an essential part of a meaningful national climate policy.  It has long been recognized that although carbon-pricing will be necessary, it will not be sufficient. Economists and other policy analysts have noted that policies intended to foster climate-friendly technology research and development (R&D) will also be necessary, but likewise will not be sufficient on their own.

Some recent studies and press accounts, which I reference below, have identified these two approaches to addressing CO2 emissions as substitutes, rather than complements.  That is fundamentally inconsistent with decades of research, and so my purpose in this essay is to set the record straight.

Carbon Pricing:  Necessary But Not Sufficient

First of all, why is there so much talk among policy analysts and policy makers – not simply among academics – about carbon‑pricing as the core of a meaningful strategy to reduce CO2 emissions?  Why, in fact, is this approach so overwhelmingly favored by the analytical community?  The answer is simple and surprisingly pragmatic.

First, there is no other feasible approach that can provide meaningful emissions reductions, such as the 80 percent reduction in national CO2 emissions by 2050 that was part of the legislation passed by the U.S. House of Representatives and proposed in the Senate and part of the Obama administration’s conditional pledge under the Copenhagen Accord.  Because of the ubiquity and diversity of energy use in a modern economy, conventional regulatory approaches –standards of various kinds – simply cannot do the job.  Only carbon pricing – either in the form of carbon taxes or cap-and-trade – can significantly tilt in a climate-friendly direction the millions of decentralized decisions that are made in our economy every day.

Second, carbon-pricing is the least costly approach in the short term, because abatement costs are exceptionally heterogeneous across sources.  Only carbon-pricing provides strong incentives that push all sources to control at the same marginal abatement cost, thereby achieving a given aggregate target at the lowest possible cost.

Third, it is the least costly approach in the long term, because it provides incentives for carbon-friendly technological change, which brings down costs over time.

For these reasons, carbon-pricing is a necessary component of a truly meaningful national climate policy.  [I’ve written about this in many previous blog posts, including on June 23, 2010, “The Real Options for U.S. Climate Policy.”]  However, although it is a necessary policy component, carbon-pricing is not sufficient on its own. This is because there are other market failures that dilute the impacts of price signals on decision makers.

Technology R&D Policies:  Also Necessary, Also Not Sufficient

The most important of these “other market failures” is the public good nature of information.  Companies carrying out research and development (R&D) incur the full costs of their efforts, but they do not capture the full benefits.  This is because even with a perfectly-enforced system of intellectual property rights (such as patents), there are tremendous spillover benefits to other firms.  Decades of economic research – much of it by my former colleague and co-author, Professor Adam Jaffe, now Dean of Arts and Sciences at Brandeis University – has analyzed with empirical (econometric) analysis the remarkable degree to which inventions and innovations by one firm provide valuable information that leads to new inventions and innovations by other firms.

So, firms pay the costs of their R&D, but do not reap all the benefits.  The existence of this positive externality of firms’ R&D – or put differently, the public-good nature of the information generated by R&D – means that the private sector will carry out less than the “efficient” amount of R&D of new climate-friendly technologies in response to given carbon prices.  Hence, other public policies are needed to address this “R&D market failure.”

New path-breaking technologies will be needed to address climate change, and public support for private-sector or public-sector R&D will be crucial to meet this need.  But, at the same time, to address the climate-change market failure itself (that is, the externality associated with greenhouse gas emissions), carbon pricing will be necessary, for all of the reasons I gave above.  This is an application of an important and fundamental principle in economics:  two market failures require the use of two policy instruments.

Empirical analyses have repeatedly verified this crucial point – that combining carbon-pricing with R&D support is more cost-effective than adopting either approach alone.  Included in this set of studies are the following:  Carolyn Fischer (Resources for the Future) and Richard Newell (U.S. Energy Information Administration, on leave from Duke University), “Environmental and Technology Policies for Climate Mitigation”; Stephen Schneider (late of Stanford University) and Lawrence Goulder (Stanford University), “Achieving Low-Cost Emissions Targets”; and Daren Acemoglu (MIT), Philippe Aghion, Leonardo Bursztyn, and David Hemous (Harvard University), “The Environment and Directed Technical Change.”

Complements, Not Substitutes

An interesting, recent column, “Next Step on Policy for Climate,” by David Leonhardt in the New York Times (October 13, 2010, p. B1) might give some people the mistaken impression that technology policies are an adequate, even sensible substitute for carbon-pricing.  That was not the intended message of the column.  In fact, Leonhardt – perhaps the leading economic journalist writing today in the United States ­– indicates clearly in his column that he is skeptical of the notion of thinking of technology subsidies as an adequate substitute for carbon-pricing (in particular, cap-and-trade).  And in a follow-up post at the New York Times’ Economix, he makes clear that “these two policies are not mutually exclusive.”

Nevertheless, Leonhardt’s original column (which included a very nice profile of my colleague, Professor Michael Greenstone of MIT) focused attention on a recent report –  a report that could give the false impression that technology policies would be a sensible substitute for serious carbon-pricing.  The report in question – “Post-Partisan Power” – received significant coverage, primarily because of its sponsorship:  a combination of a prominent Republican-oriented Washington think tank, the American Enterprise Institute (AEI), and an equally prominent Democratic-oriented Washington think tank, the Brooking Institution (and a third partner, the Breakthrough Institute, a California-based environmental think tank).

The report may well garner some bi-partisan political support, because it promises a free lunch of painless, win-win solutions, a promise that will resonate with many elected officials.  Indeed, the report’s sub-title is “how a limited and direct approach to energy innovation can deliver clean, cheap energy, economic productivity, and national prosperity.”  What’s not to like? And the authors are presumably smart and politically shrewd.  I know that’s the case with the AEI author, Steven Hayward, who I debated last year in the pages of the Wall Street Journal.

To its credit, the report lays out a menu of policies intended to stimulate carbon-friendly technological change, ranging from $500 million of Federal government funding of K-12 curriculum development and teacher training to $25 billion annually of direct Federal funding of energy innovation.

For the reasons I explained above (the “R&D market failure” and the “carbon emissions externality”), both direct technology R&D policies and serious carbon-pricing are necessary, but neither is sufficient on its own.  Unfortunately, this new report ­­– and some of the press coverage surrounding it – makes the claim that such direct government funding of technology innovation is a sufficient and sensible substitute for meaningful carbon-pricing.  That claim is both unfortunate and wrong, as it is supported neither by sound reasoning nor empirical research, as I have described above.

Again, many of the individual technology policy recommendations offered by the AEI-Brookings-BI report are worthy of serious consideration (as a complement, not a substitute for an economy-wide carbon-pricing policy).  But the specifics – indeed, much of the meat – are missing.  “Reform the nation’s morass of energy subsidies” – yes, but exactly which subsidies (all of which have important political constituencies behind them) will be eliminated?  “Recognize the potential for nuclear power” – yes, and both the House and Senate carbon-pricing schemes would have provided tremendous incentives for nuclear power investment.

Overall, there should be concern about how all of this will be funded.  Where will the $25 billion per year come from?  The report appropriately states that this should not come from general revenues, and thus add to the Federal debt.  “Phasing out current subsidies for wind, solar, ethanol, and fossil fuels” could be meritorious on its own, but how much does this generate, and does it even pass a political laugh-test?  Interestingly, beyond this, despite considerable rhetoric about moving beyond debates about carbon-pricing, the report recommends that in order to avoid adding to the Federal debt, it would be necessary to impose new taxes, including increased royalties for oil and gas extraction, a tax on imported oil, a tax on electricity sales, and a “very small carbon price” (presumably from a modest carbon tax or unambitious cap-and-trade system).

The actual numbers would be helpful, and the political feasibility remains a serious question.  The political challenges that emerged in the effort to pass cap-and-trade climate legislation will not magically disappear if there’s an attempt to induce Congress to approve $25 billion in funding.  As Tom Friedman noted on October 12th in the New York Times, Congress has not come close to fully funding the outstanding requests for about $4 billion for ARPA-E (energy) research.

More broadly, despite the attraction of the AEI-Brookings-BI proposal as a potential complement to carbon-pricing (and I am serious that the proposal is of value in that context), one has to be very careful about comparing proposed new policies in idealized form (for example, precisely the right subsidies eliminated and precisely the right new subsidies introduced) with real policies with all their warts (for example, the cap-and-trade bill that was passed by the House last year).  Making such comparisons can lead to flawed analysis and misleading results.

This is not a new issue.  Robert Hahn and I wrote about this generic problem nearly 20 years ago in an article (“Economic Incentives for Environmental Protection:  Integrating Theory and Practice”) which appeared the American Economic Review Papers and Proceedings (May 1992).  At the time, our concern was that this mistake was being made not by the opponents but by the supporters of cap-and-trade and other (then essentially untested) market-based instruments.  We worried that “many analysts use highly stylized benchmarks for comparison that ignore likely political realities,” and suggested that an appropriate “comparison would be between actual command-and-control policies and either actual trading [cap-and-trade] programs … or a reasonably constrained theoretical … program.”

Likewise today, when carrying out comparisons of policy alternatives, it is fine to compare two theoretical, idealized alternatives, or to compare two real-world policies, but it is problematic and usually misleading to compare a theoretical, idealized policy of one type with a real-world example of another type of policy.

The Bottom Line

Carbon-pricing – whether carbon taxes or cap-and-trade – will be an essential part of any truly meaningful national climate policy.  Likewise, to address the “R&D market failure,” direct technology innovation policies will also be required.  Both are necessary.  Neither is sufficient.  These are complements, not substitutes.


Postscript: Four years ago, the U.S. Congressional Budget Office (CBO) — widely recognized for its non-partisan, first-rate research — produced a study on the same topic as the AEI-Brookings-BI report, but did so with rigor and without ideology.  The CBO report — Evaluating the Role of Prices and R&D in Reducing Carbon Dioxide Emissions (September 2006) — was prepared by Dr. Terry Dinan, a long-time, respected CBO economist, and was peer reviewed by an impressive set of academic and other experts.  Sadly, the CBO paper received little press coverage, despite its high quality and its relevance.  For anyone interested in the topic of this post, particularly those who disagree with my theme, I hope you will read the CBO report.

Also, a reader of this blog post sent me a paper by David Hart and Kadri Kallas (from MIT’s Energy Innovation working paper series) that examines “Alignment and Misalignment of Technology Push and Regulatory Pull.” It’s worth reading in the context of combining carbon pricing and technology R&D policies.