Rebound and Rigor: NRDC’s Entry Into Rebound Effect Debate Stuck in the Past

Image: A large gulf stands between the work of serious energy analysts and a recent essay published by NRDC's analysts, David Goldstein (left) and Sierra Martinez (right), which stubbornly assert that "rebounds at the economy-wide level are trivially small."

A recent article in Electricity Policy by Natural Resource Defense Council (NRDC) analysts (David Goldstein et al.) purports to offer a fresh look at the question of energy consumption rebound resulting from cost-effective efficiency improvements. But rather than advancing the ongoing discussion about rebound among serious energy analysts, NRDC attempts to turn back the clock, relying on outdated and recycled citations dating from as far back as the early 1990s and asserting that conclusions about rebound effects must be testable against "rigorously framed hypotheses" while failing to apply that standard to their own claims regarding the historic success of efficiency policies in reducing energy use.

In reviewing their article, it is difficult to escape the feeling that Goldstein and his colleagues simply ignore any recent work that is inconvenient to their premise, including a rich trove of literature and inquiry into rebound effects published in recent years. It is particularly revealing that the authors restrict their analysis to those sectors of the global energy economy where rebound effects appear to be least significant--end-use consumption in rich, developed economies. In so doing they ignore both the productive sectors of the economy responsible for two-thirds of the global energy use and the emerging economies driving the vast bulk of global energy demand growth--in short those sectors of the global energy economy in which the vast majority of current and future energy demand is concentrated and in which the rebound literature suggests rebound effects are likely to be greatest.

In fact, NRDC's contention that "rebounds at the economy-wide level are trivially small" is controverted by virtually every review of the evidence for energy efficiency rebound conducted in recent years. So while NRDC attempts to re-litigate a decades-old debate, for serious analysts and policymakers, particularly in Europe, this debate, about whether rebound exists and is non-trivial, is effectively over. The focus now is on developing a richer understanding of when and where such effects operate, at what scale, and, increasingly, a focus on what, if anything, can be done to mitigate such phenomena. The UK government, for example, now explicitly considers at least one rebound mechanism when planning efficiency policies. And the European Commission funded a large study in 2011 that begins from the consensus that rebound effects are real and significant, and explores what can done about it.

In the United States, the tone seems to be shifting and signs are appearing that energy researchers are beginning to realize they need to deal forthrightly with this issue. Many who before quite adamantly denied the rebound phenomenon now treat it more cautiously as the academic substantiveness of multiple recent studies becomes apparent. Small working groups of scholars are forming to address the gaps in our knowledge. The Center for Climate and Energy Decision Making at Carnegie Mellon University will soon host a gathering of scholars to define the research questions that call out for serious inquiry. And the latest Stanford Energy Modeling Forum study on energy efficiency (EMF, 2011), while it still overlooks much of the recent literature (perhaps because it was framed a few years ago) nonetheless acknowledges key rebound mechanisms.

The critical question really isn't whether or not rebound effects exist -- as basic economic theory dictates, they most certainly do -- but rather how large they may be in various contexts.

Truly cost-effective energy efficiency measures lower the effective price of the services derived from fuel consumption - heating, cooling, transportation, industrial processes, etc. We know that economic actors react in complex ways to changes in the relative and absolute prices of various goods and services, and in particular, that when prices fall, consumers and industry alike demand more of these services, all else being equal. Other indirect and economy-wide effects can result from efficiency improvements as well, as consumers re-spend money saved through efficiency on other energy-consuming goods and services, industrial sectors adjust to changes in the relative prices of final and intermediate goods, and greater energy productivity causes the economy as a whole to grow. Collectively, these various mechanisms are known as "rebound effects" as they drive a rebound in demand for energy services that significantly erodes reductions in total energy use otherwise expected from efficiency improvements, along with much-hoped-for reductions in greenhouse gas emissions. In rough terms, for every two steps forward we may take through efficiency, rebound effects take us one (or more) steps backwards.

(Read an introductory FAQ on rebound here)

Unfortunately, conventional forecasts of energy use and the reductions possible through efficiency measures routinely ignore many (if not all) of the various rebound mechanisms. To the extent rebound phenomena are non-trivial, the implication is that the traditional forecasts of global energy use on which so much of climate change policy is reliant may seriously understate the scale of the challenge by ignoring or improperly treating rebound, meaning we have less time than we think to devise climate solutions.

NRDC's entry into this high-stakes debate disappoints on the methodology side, as we discuss in detail below. But the article also reads like an effort to turn back the clock to a time five to ten years ago when many still dismissed the rebound phenomenon as irrelevant, the province of a few fringe theorists, perhaps. This finds its reflection in the outdated citations the analysts rely on, with the most frequently cited report dating from 2005 (IEA/Geller) and reliant in turn upon Greene (1992), itself a survey of even older literature.

The field has progressed substantially since then--especially in Europe.

Perhaps triggered by the exhaustive UK Energy Research Center study of rebound led by Steve Sorrell (2007, 2009), inquiry into rebound effects has since seen noteworthy advances overseas. Significant funding in Europe is now going to researchers examining the problem through multiple analytic methods, and the fruits of these labors are appearing monthly in the literature.

In a statement that NRDC's analysts clearly did not take to heart, Sorrell concluded his rigorous assessment of the literature in 2007 with this statement:

"It would be wrong to assume that, in the absence of evidence, rebound effects are so small that they can be disregarded. Under some circumstances ... economy-wide rebound effects may exceed 50% and could potentially increase energy consumption in the long-term. In other circumstances ... economy-wide rebound effects are likely to be smaller. But in no circumstances are they likely to be zero."

In 2011, one of your authors led another comprehensive survey of the field (Jenkins et al.), which concludes:

"Rebound effects are real and significant and combine to drive a total, economy-wide rebound in energy demand with the potential to erode much (and in some cases all) of the reductions in energy consumption expected to arise from below-cost efficiency improvements."

While both literature reviews put the state-of-the-art in the field at NRDC's fingertips, their analysts unfortunately opt to merely cite selectively from both works, while ignoring the broad consensus that has developed in the academic literature.

We suggest the NRDC would be better advised to instead climb on board and move without delay up this learning curve. The bright and committed staff and analysts at NRDC have much to contribute to the understanding - and management - of rebound effects.

But as it stands, there are several methodological difficulties with the current NRDC analysis. Two are central:

  • First, the NRDC paper hangs on an effort to construct and examine "rigorously-testable hypotheses" of rebound, a method they fail to appropriately utilize, while eventually falling afoul of their own requirement for testability of hypotheses in their effort to prove the historic success of efficiency policies in reducing energy use.
  • Second, the authors make the common error of focusing their arguments on the smallest part of the energy economy--end-use consumption in rich, developed economies. This means they ignore both the productive sector of the economy responsible for two-thirds of the global energy use and the emerging economies driving the vast bulk of global energy demand growth--and the different-in-kind rebound mechanics in play in both places. In such sectors, the shadow of Jevons still lurks (see Jenkins et al. 2011 for survey of key literature).

Other non-methodological difficulties arise in their portrayal of the positions of rebound analysts, possibly due to a failure to undertake the hard work of examining the rich and burgeoning recent literature. Whatever the cause, it leads them to falsely portray key elements of the debate, and to apparently lay claim to new insights that are in fact old ones.

1. Testability of Hypotheses

As the authors make much of the theoretical rigor of their approach, we begin there.

Their basic method is to construct several versions of rebound hypotheses, and to subject each to varying degrees of scrutiny in an effort to determine their testability and veracity. This is an appropriate method, but a few basic errors impair its application.

First, it should be noted that a hypothesis is not invalidated simply because the authors themselves are unable to construct a methodology to test such a hypothesis. A hypothesis is not disproven or invalidated until it can be tested, so further effort must be expended either to refine the hypothesis or to devise methods of inquiry that can properly validate or invalidate a standing hypothesis. This is basic to the scientific method.

Yet NRDC's analysts appear ready to reject out of hand the idea that rebound effects may "increase energy consumption above where it would be without these [efficiency] gains" - as in their Hypotheses A (and its variants), which is equivalent to a 'backfire' scenario where rebound exceeds 100% of initially-expected energy savings - simply because they cannot determine a way to test "what energy consumption would have been if efficiencies had remained constant..."

The authors are correct to identify this "with and without" problem, that is, the difficulty of knowing what would have happened to energy use in the absence of energy efficiency gains. Here however, NRDC's analysts would have done well to delve into the literature on rebounds to date, where this observation is central to recent work by one of your authors (e.g., Saunders, 2010) and to discussions of the literature led by another of your authors (see Jenkins et al. 2011 at page 29).

First, we note that this very problem also invalidates numerous past efforts to use historical trends showing declining energy use per unit of economic activity as "proof" of the absence of significant rebounds (e.g. Schipper and Grubb 2000), since these many past analyses have failed to ask the key question, "What would energy use have looked like in the absence of rebound?" We therefore applaud the overdue recognition of this problem by NRDC's analysts, but are baffled as to why they ignore the challenge of constructing methods to rigorously test the hypothesis that historic energy intensity trends would not have been much different had rebound effects been entirely absent--i.e., the hypothesis central to claims that rebound effects are "trivially small."

In short, this methodological challenge cuts both ways, against proponents of hypotheses that rebound effects are "trivially small" as well as against the large body of analysis that sees such effects as "significant" and important.

Testing Hypothesis C

In their Hypothesis C, the NRDC analysts propose to test whether energy efficiency gains induced by policy measures result in energy consumption being different from what the policy measures were expected to produce. They approach the analysis of this proposition by comparing California's experience with electricity sector demand to that of other states.

The difficulty with their analysis is that in the five scant paragraphs they dedicate to 'disproving' their Hypothesis C, they fail to circumscribe their test with sufficient precision, and indeed, present data that actually tell us little about the operation of rebound effects in this historic case study - save perhaps that such rebound are not 100% or greater (just as one would expect for the kinds of efficiency policies pursued in this case).

On their own, historic energy trends tell us little about the operation of rebound effects, since so many other factors are at work affecting energy use across any historic period. As we've noted previously (Saunders 2010, Jenkins et al. 2011), the key to determining the scale of rebound evident in historic trends is to determine appropriate "with and without" rebound scenarios against which to compare historic energy use trends. That is, one must identify what energy use would have been with all of the savings from efficiency measures enacted over the period eroded due to rebound, and what energy use would have been without any efficiency savings eroded by rebound. Then one can compare actual energy use trends to both the "Zero Rebound" and "100% Rebound" scenarios, taking care throughout this process to properly account for other exogenous factors unrelated to efficiency gains (relative factor prices, industrial shifts, fuel switching, income effects etc.). This method is described in Saunders (2010).

Goldstein and colleagues fail to perform any such rigorous "with and without" analysis, and instead present a tangle of historic data and unexplained California Energy Commission (CEC) estimates as alleged 'proof' that rebounds were insignificant.

The NRDC analysts first cite a 2005 CEC estimate that California efficiency policies have saved 1,134 kWh per person in 2003, or roughly 15% of actual 2003 electricity consumption. They then compare California electricity usage per capita to the electricity use trends of the other 49 states, noting that the divergence in electricity use between California and the other states is more than four-times as large as the CEC-projected energy savings. The implicit conclusion the reader is apparently supposed to draw from this 'evidence' is that rebounds must not be significant, otherwise California electricity usage per person would more closely match the rest of the United States.

This logic and methodology is flawed, and amounts to little more than an apples to oranges comparison.

Trends for the rest of the U.S. states are an inappropriate stand-in for a "100% Rebound" comparison scenario. As Goldstein et al. acknowledge in citing the work of Anant Sudarshan, "the majority of this difference" between California and the other 49 states can be explained by a host of other structural, demographic, and economic factors unrelated to energy efficiency policies (i.e. fuel switching to natural gas, changes in household incomes, structural shifts in industrial sectors, etc.). Sudarshan (2010) and Sudarshan and Sweeney (2008) show that such factors account for 77% of the difference between electricity use trends in California and the rest of the U.S. states.

Importantly, as Sudarshan and Sweeney note, their analysis simply says the remaining 23% is likely due to policy measures, whose effects go beyond efficiency gains and include the impacts on electricity demand of the much-higher (policy-induced) electricity rates in California, relative to other states. Policy effects also include any impacts of stringent efficiency regulations that force businesses or consumers to take non-cost-effective efficiency actions, which are welfare reducing and depress economic activity and thus electricity use (in contrast to the cost-effective, welfare enhancing efficiency savings that principally drive rebound effects). Both factors - higher electricity prices and welfare-reducing efficiency regulations - would further contribute to the difference between electricity use in California and the other 49 states, making it impossible to determine, given the evidence presented by Goldstein et al., what California electricity use would have been had no savings due to cost-effective efficiency measures been achieved - i.e. if rebound were 100%.

Furthermore, the authors make no attempt to provide a "Zero Rebound" comparison at all. That means that even if we were able to determine that electricity usage in California is lower than it would have been if rebounds were 100%, Goldstein and colleagues leave us with no way of knowing if rebounds were insignificant or significant. Consider this question: if rebound effects had been entirely absent, would California electricity usage have been even lower than it is? Answering that question would require the ability to compare actual trends to a rigorously constructed scenario in which all energy efficiency savings "take" on a one-for-one basis (e.g. rebounds are entirely absent), yet the authors provide no such comparison.

Goldstein et al.'s use of historic trends therefore falls short of the kind of rigorous test necessary to truly validate or invalidate their Hypothesis C. And in a rather large irony, by confining themselves to trends analysis and dismissing modeling as a means to conduct such a test--thereby abandoning hope of constructing the "what it would have been otherwise" cases--the authors run afoul of their own requirement that the hypothesis be testable.

Rebound effects, or their absence, are not discernable from the test as they have constructed it.

Testing Hypotheses B and A3

Hypothesis B (which effectively encompasses Hypothesis A) is the hypothesis most analysts have recently engaged, representing a test of non-zero rebound. As these various analyses have shown, this hypothesis is by no means undefinable or untestable.

Nor is it the case that most predictive models used for climate change analysis already incorporate rebound effects, a brash claim the authors use to excuse themselves from squarely addressing Hypothesis B. Nor is it even true that the very few models that purport to incorporate rebound do so in anything resembling an appropriate way. To avoid a long, highly technical sidetrack, for now we will simply state here for the record that the authors' assertion about these models is not just incorrect, but widely off the mark

A different methodological problem undermines their analysis of their "Hypothesis A3." Hypothesis A3 is cast as being in response to what they claim is an assumption made by "many analysts," as in "...many analysts assume that without any policy, energy use would grow proportionally to GDP." One wonders who such analysts are. No serious rebound analyst we know of makes such a claim. Energy intensity trends have been generally declining. But intensity trends are driven by a host of determining factors and it requires a challenging "with and without" analysis to extract rebound information. Their Hypothesis A3 is testable, as they have shown, but it is the wrong hypothesis.

Modeling and Econometric Analysis: In Pursuit of a Workable Test

Oddly, despite recognizing the inherent challenge of devising methods to test the "what would have been" scenarios necessary to truly validate or invalidate rebound hypotheses, NRDC's analysts cavalierly dismiss the economic modeling methods that offer the best route forward. In so doing, the NRDC authors put us all in a much tighter straightjacket than they likely imagine.

Climatologists face a problem similar to economists. Climatology is science, but it is not science like particle physics is science. That is, climatologists cannot, say, run the global climate for five years under one set of conditions and then go back and re-run it under a different set of conditions--just as economists cannot re-run history. This lack of any scientifically controlled experiment opens the door to critics who feel free to ignore the growing cumulative weight of indirect evidence for climate change--and, on the same grounds the NRDC authors dismiss economic modeling, to dismiss evidence coming from climate change models, since in the authors' words, they " to meet scientific standards because they cannot be tested."

We need to do better than this. And we can. And have. No need to throw our hands up in despair. Just as modeling has yielded innumerable insights into climate science, economic models can be designed to look at history and compare actual energy use against meticulously and clearly-devised control trajectories, adhering to rigorous theoretical principles, thereby providing methods to test and explore rebound hypotheses. As Steve Sorrell (2009) writes:

"While rebound effects are difficult to study, they are not necessarily any more difficult than well-researched issues such as price-induced technical change. Their continued neglect may result as much from their uncomfortable implications as from a lack of methodological tools. Too much is at stake for this to continue."

The uncomfortable reality for NRDC's analysts is that when rigorous modeling and analysis is conducted, substantial rebounds become visible in various national economies (including the US economy [1]), and the global economy as a whole (the scale and scope of analysis most pertinent to climate change mitigation concerns) [2]. Once again, such analysis was clearly presented in recent literature reviews, including Jenkins et al. (2011), and available to Goldstein et al., had they opted to delve into such methods [3].

2. Narrow Scope Ignores Where Rebound Lurks

In addition to failing to construct sufficiently testable methods and ignoring the wide range of research attempting to do just that through various methods of economic analysis and inquiry, the NRDC analysts also make the common error of focusing their arguments on the small portion of the energy economy where rebound effects are likely to be smallest, and extrapolating such trends to the far larger portions of the energy economy outside their immediate scope.

Perhaps no better example of this is that found in the authors' analysis of Jevons. They employ a narrow foundation for the (surprisingly common) claim that "[a]s energy costs decrease as a share of total costs, sensitivity to energy prices decreases, as does the rebound effect." On this basis, they argue that today's world is much less rebound-prone than Jevons' world.

This is a theoretical argument. It is also incorrect. They rely on observations narrowly confined to personal transportation in rich nations (EPA/NHTSA 2010), and boldly extrapolate these to a very broad theoretical conclusion. The correct theoretical picture, in the large, looks like this:

Energy is a unique and sine qua non input to economic activity. There is virtually no good or service, and no industry, that does not rely on energy as a vital input. This distinguishes it from every other raw material. All economic activity, of whatever kind, requires energy.

While most people acknowledge and understand this, they often overlook that this means a lower energy share actually makes energy a more vital input. Theory says that the lower the relative share of an input, the higher is its "marginal productivity," or "value of use," in lay language. If energy is this input, theory says reducing energy's share makes it more sensitive to energy price decreases or improvements in the supply of high-quality energy services, not less, and rebound effects become morepronounced, in stark contrast to these analysts' claims [4].

Second, the authors assert boldly that the world of today looks little like the industrial revolution economy of Jevons. This may be true for the wealthy American economy that NRDC's analysts inhabit. But in a world where roughly one third of the global population lives in energy poverty, and where the vast bulk of the world's inhabitants live in rapidly emerging economies, the world as a whole may not look quite so different from Jevons' England after all.

This failure of perspective is actually quite endemic to Goldstein et al.'s article.

The Rutherford Experiment: What You Find Depends on Where You Look

On this 100th anniversary of Ernest Rutherford's great discovery, it is entirely fitting that the authors choose to make allusion to it.

Legend has it that Rutherford conscripted a graduate student looking for something to do (either Geiger or Marsden), and, at a loss for anything more constructive to propose, suggested he set up detectors well in front of the gold foil target, rather than behind it as they had done to date. (Graduate students, while they are essentially free labor, need to be kept occupied.) Rutherford, the story goes, expected nothing from this exercise. But of course, the rest is history...

Whether or not this story is apocryphal, the parallel to the NRDC analysis is striking and rather amusing. The authors, in making their case for the absence of rebound, focus exclusively on what we might dub "behind the gold foil target"--energy consumption by end-use consumers in rich nations. In doing so, they ignore the majority of the rebound action, which instead lies "in front" of the target. Their frame, like Rutherford's initial frame, is poorly constructed.

Globally, some two-thirds of energy consumption occurs in the productive part of countries' economies. Only about one-third is consumed directly by end-use consumers. The great majority of energy we all consume is embedded in the goods and services we consume. While it is much easier to "see" (and relate our personal energy efficiency experiences to) the energy we consume directly in our households and for personal transportation, the future of global energy consumption, and the impact of efficiency gains, will be instead largely driven by what happens in this preponderant sector of the global energy economy.

Furthermore, virtually all of projected energy demand growth in the coming half-century will be driven by the world's emerging economies (IEA World Energy Outlook 2010), not the industrialized economies that NRDC's analysts are far more familiar with. In the world's emerging economies, the cost and availability of energy services is often a key constraint on their wellbeing. Demand is thus far more elastic (responsive to changes in price), and rebound effects much larger than in the developed economies. That in turn means rebound effects are much larger, even for end-use consumer energy uses.

It is in both of these realms - the productive sectors of the economy and the world's emerging economies - that analysts see strong evidence for large rebound effects [5].

The NRDC authors do not find strong evidence for large rebounds in substantial part because they do not look in the right place, keeping their scope of analysis narrowly focused on the end-use energy services for which demand has already been largely saturated for the world's wealthy consumers. In reality, alpha particles do in fact "rebound" backwards from gold foil.

3. "New" Claims and False Claims

As discussed above, the authors reveal an apparent lack of diligence in looking at the academic literature on rebound.

Rather than basing their conclusions on recent contributions to the field [6], the authors instead rely on older studies by the IEA (2005) and EPA (which both rely heavily on a 20-year-old study by Greene (1992)) that do not account for the latest in scholarly research on rebound.

This may explain why the authors attribute to themselves what they (at least seemingly) offer as "new" insights.

First, the authors place great emphasis on their insight that rebound effects could be "positive" (by which they mean "negative" in the parlance of rebound analysts--we'll use the standard terminology here), and they imply that traditionally constructed hypotheses restrict the sign of rebound. In fact, it is only their own hypotheses that they interpret as doing this. A negative sign for rebound is allowed for in numerous rebound analyses, and this was recognized long ago [7]. This is not a new insight, and analyses incorporating no sign restriction add even further credibility to measurements showing rebounds to be worrisomely large and positive [8].

The second of these "new" observations is the recognition that energy efficiency gains improve economic welfare and so should be advantageous, especially to the developing world.

But of course! Rebound analysts have long recognized this and have repeatedly emphasized that this is part of the conundrum created by rebound. As we both write in a forthcoming essay in the UN Industrial Development Organization journal Making It:

"Unlocking the full potential of efficiency may very well mean the difference between a richer, more efficient world, and a poorer, less efficient world. The former is clearly the desirable case - even if the world uses more or less the same amount of energy in either scenario. The pursuit of any and all cost-effective efficiency opportunities should thus continue as a key component of an efficient course for global development, even as we reconsider the degree to which these measures can contribute to climate mitigation efforts."

For this reason, we can't think of a rebound analyst who "disparages" energy efficiency. Yet this recognition of the welfare-enhancing nature of rebound creates a thorny ethical dilemma for those who propose to mitigate rebound effects by various policy measures. Efficiency increases welfare and should be vigorously pursued, even as we recognize that efficiency can be far less effective at mitigating climate-destabilizing greenhouse gas emissions or conserving finite resources than it may seem at first blush. This is the very dilemma at hand.

Furthermore, no one seriously claims we would be better off without energy efficiency gains, even in a case where rebound becomes backfire, a view Goldstein and colleagues incorrectly attribute to rebound analysts. This position is in fact a faulty inference drawn by those, such as Amory Lovins, who seem to hold out reduced energy consumption as the sole goal, with no account taken of economic welfare. By such logic, the best and surest way to delay climate change is to create a global depression - hardly a tenable policy prescription.

The deep problem here is that rebound means one cannot have it both ways: one can no longer simultaneously claim, as the authors do, that "energy efficiency provides a solution that allows us to reduce energy consumption" and that this can be done "without stifling the standard of living for many poor and developing populations around the world." This is an extremely seductive conclusion--we would all like to believe this--but unfortunately rebound makes it problematic. Blindness to rebound effects means we risk over-reliance on efficiency measures to reduce climate risks and conserve energy resources, and in so doing, condemn future generations to the consequences of our false hopes.

4. The Meaning of Rebounds

At root, the confusion over rebound effects lies in a common misconception of how energy use actually unfolds. More specifically, it lies in a faulty mental model of what people think "should" happen to energy use when energy efficiency gains are introduced.

It at first seems reasonable to assume that an x% gain in energy efficiency will result in an x% reduction in energy use below where it would otherwise be. Technically speaking, this is the behavior one theoretically expects if the economy is a so-called "Leontief," or "fixed factors" economy: improving the efficiency of one input results in a one-for-one reduction in its use, and everything else--other inputs, output, prices--stays the same.

But economists do not think this is at all descriptive of the real economy. None of us would expect, for instance, that a 10% improvement in labor productivity will result in a 10% decrease in labor employment. Any economist knows improving labor productivity drives economic growth, creates new profitable ways to utilize labor, and generally increases overall employment, rather than decreases it.

So what many efficiency advocates argue "should happen" is a condition readily thought of as the zero rebound condition. But it is simply a false characterization of what will happen, however alluring it may at first appear to the rational mind.

The real economy is much more complex than this simplistic picture. Producers discovering new energy efficiency opportunities can adjust all their inputs, and their outputs, in a highly flexible way; resulting reductions in their output prices move through multiple pathways to other producers and to the final consumer, affecting energy use all along the way. The "should happen" economy is rigid and passive in how it accommodates energy efficiency gains, while the real-world economy is flexible and creative in how it accommodates energy efficiency gains.

And the result we must all come to terms with is energy consumption rebound.


The National Resources Defense Council is widely, and appropriately, viewed as a voice of our collective conscience, speaking in defense of our invaluable national and global resources, including those natural resources embodied in the fragile and complex ecological balance of this planet we call home. Their analysis provides a potentially useful starting point, but does not yet live up to its own promise of correctly evaluating testable hypotheses.

We call upon the NRDC to apply their formidable skills and talents to more seriously and substantively analyzing the threat to our legacy wielded by this specter of the energy consumption rebound phenomenon.

Corrigendum appended 9/2/2011

In this article we made the claim that when energy share goes down, "theory says...rebound effects become more pronounced, in stark contrast to these analysts' claims." This claim is incorrect. Here we correct this error and explain further.

Significantly for those familiar with the work of the late Berkeley/Stanford energy researcher Lee Schipper, this means he was correct, according to current theory, in asserting that rebound effects should become less pronounced as energy share declines. For one of us, Harry Saunders, such corrections to his work by Lee have arisen in the past, unfortunately for his pride (but in the end fortunately for the field of energy efficiency.) Ironically, Harry's own derivations of this relationship, referenced in footnote [4] of our article, support Lee's position. Lee would have been amused by this.

On a personal note, Lee's passing is a great sadness, He made such a voluminous-in-content contribution, and at an incredible output rate. A humorist, musician and scholar the field will sorely miss. Data-driven, and absolutely relentless in the pursuit of a better-functioning energy-economic system that would deliver more to the people it serves by way of environmental pollutants-reduction through better efficiency, at the same time improving their economic lives.

However, it remains true as stated in our article that the lower energy's value share, the higher will be its value to the economy, and that it becomes more vital, because it is now fueling an economy many-fold greater than its absolute magnitude as an input to production. Theory predicts this result from marginal productivity considerations. The Hogan-Manne article cited in footnote [4] shows why.

This all should be treated with some caution as the theory as it stands also relies on explicit assumptions about functional forms. The above conclusion holds true for a so-called CES function. More complex (but more useful) "flexible" functions (e.g., the highly-popular Translog function) show behavior that appears similar, but the mathematical expressions include complex combinations of own-and-cross price elasticities and factor value shares, making the task of understanding and interpreting the rebound mathematically more challenging (see again footnote [4]). Nonetheless even from the CES function we are quite likely getting reliable indications of what the real energy economy looks like and how it behaves rebound-wise when energy efficiency is introduced. And the CES function indicates that energy's value goes up as its input share is reduced. But the magnitude of rebound effects goes down as share goes down, just as Lee always predicted.

Perhaps most importantly as to Lee's prediction, it stands as a lesson to others of us the field. His prediction was based on observed reality, supported by his famously-avaricious appetite for data and for understanding the real world, physicist-at-heart as he was.

Lee was a man of deep intellectual integrity and productivity whose departure leaves a gaping hole in the field of energy efficiency.

Jesse Jenkins is Director of Energy and Climate Policy at the Breakthrough Institute, an Oakland-California based public policy think tank. He is the lead author, along with Ted Nordhaus and Michael Shellenberger, of "Energy Emergence: Rebound and Backfire as Emergent Phenomena," a comprehensive review of the literature and evidence for rebound effects published in February 2011.

Dr. Harry Saunders is Managing Director of Decisions Processes Incorporated, a corporate management and decisions consultancy, and a Senior Fellow at the Breakthrough Institute. He holds a PhD in Engineering-Economic Systems from Stanford University and is a pioneering and oft-published author in the field of rebound effects.

End Notes

[1]. Saunders (2010). In this study, the control trajectories represent what would have happened had there been no energy efficiency gains over the historical period, and what would have happened had the measured efficiency gains "taken" on a one-for-one "engineering" basis, representing the expectations of traditionalists and accordingly depicting the case of zero rebound. When actual energy use is compared to these two trajectories, the magnitude of rebound becomes evident. (And, it is worthwhile noting, energy prices are not held fixed in real terms but reflect actual history, notwithstanding the formal statement of Hypothesis A.). See also Turner (2009) showing backfire in the energy producing sectors of the Scottish economy, Allan et al (2007) and Barker and Foxon (2008) each showing substantive rebounds for the UK economy.

[2]. Barker et al (2009) examines rebound at the global scale due to cost-effective efficiency policies identified by the International Energy Agency, concluding that rebounds would erode 31% of projected savings by 2020 and 52% by 2030.

[3]. See Jenkins et al. (2011) pages 30-42.

[4]. Hogan and Manne (1977) first showed the relationship between energy price elasticity and energy share. The relationship between rebound and energy share is given in Saunders (2008).

[5]. Saunders (2010) reports rebound aggregated across 30 US sectors of approximately 50%. Note that this analysis measures so-called direct effects only. Other, broader, rebound effects identified and measured by other researchers (e.g., Druckman et al., 2011, Barker and Foxon, 2006, Turner et al., 2009) will augment those reported there.

[6]. For instance Druckman et al. (2011), Tsao et al. (2010), Barker and Foxon (2006), Turner et al. (2009). A richer and more complete list can be found in Jenkins et al. (2011).

[7]. The term "super-conservation" was introduced in Saunders (2005). See also Turner (2009), Wei (2010) and Turner et al (2010).

[8]. e.g. Barker et al (2009), Druckman et al (2011), Saunders (2010).


Allan G, Hanley N, McGregor P, Swales K, Turner K. The macroeconomic rebound effect
and the UK economy, Final report to the Department Of Environment Food and Rural Affairs, Department Economics, University of Strathclyde, Strathclyde, UK. 2006.

Allan G, Hanley N, McGregor P, Swales K, Turner K. The impact of increased efficiency in the industrial use of energy: a computable general equilibrium analysis for the United Kingdom, Energy Economics 2007 29(4): 779-798..

Barker T, Foxon T. The macroeconomic rebound effect and the UK economy, Report to the
Department of the Environment, Food and Rural Affairs, Cambridge Center for Climate Change Mitigation Research (4CMR), Cambridge, UK. 2006.

Barker T, Foxon T. The Macroeconomic Rebound Effect and the UK Economy. A Research Report for the UK Energy Research Center, February 2008.

Barker T, Dagoumas A, Rubin J. The macroeconomic rebound effect and the world economy,
Energy Efficiency 2009 2(4): 411-427..

Brookes L. A low energy strategy for the UK by G. Leach et al: a review and reply. Atom 1979; March; 269.

___________ The greenhouse effect: the fallacies in the energy efficiency solution. Energy Policy 1990a; March; 199-201.

___________ Energy efficiency and the greenhouse effect. Energy World 1990b;18(2); 199-201.

___________ The fifth fuel and the greenhouse effect. Energy & Environment 1990c;1(4); 318-333.

___________ Energy efficiency and economic fallacies: a reply. Energy Policy 1992; 390-392.

___________ Energy efficiency fallacies revisited. Energy Policy 2000; 28(6,7) 355-366.

___________ Energy efficiency fallacies--a postscript. Energy Policy 2004; 32(8) 945-947

Druckman A, Chitnis M, Sorrel S, and Jackson T. Missing carbon reductions? Exploring rebound and backfire effects in UK households. Energy Policy 2011; 39; 3572-3581.

Energy Modeling Forum. EMF 25 Summary Report: Energy Efficiency and Climate Change Mitigation. 2011.

Goldstein D, Martinez S, Roy R. Are there rebound effects from energy efficiency? - An analysis of empirical data, internal consistency, and solutions. Electricity 2011. Available online at:

Hogan, W., Manne, A., 1977. Energy-economy interaction: the fable of the elephant and the rabbit? Energy and the Economy, EMF Report 1 of the Energy Modeling Forum (Stanford University). Available online at .

Jenkins J, Nordhaus T, Shellenberger M. Energy emergence: rebound and backfire as emergent phenomena. 2011; The Breakthrough Institute.

Saunders H. The Khazzoom-Brookes postulate and neoclassical growth. The Energy Journal 1992;13(4); 131-148.

Saunders, H. Fuel conserving (and using) production functions. Energy Economics 2008; 30; 2184-2235.

Saunders, H. A calculator for energy consumption changes arising from new technologies. Topics in Economic Analysis and Policy 2005; 5 (1) Article 15;

Saunders, H. Historical evidence for energy consumption rebound in 30 US sectors and a toolkit for rebound analysts. 2010 (under review; posted on BTI website

Sorrell, S. The Rebound Effect: an assessment of the evidence for economy-wide energy savings from improved energy efficiency. UK Energy Research Centre. 2007.
Available online at:

Sudarshan A, Sweeney J. Deconstructing the 'Rosenfeld Curve.' 2008.

Turner K. Negative rebound and disinvestment effects in response to an improvement in energy efficiency in the UK Economy, Energy Economics 2009 31: 648-666.

Turner K, Allan G, McGregor P, Swales K. Energy Efficiency Improvements and Rebound
Effects: some lessons from the Scottish case, Welsh Economic Review, 21: 36-41. Winter 2009.

Tsao J, Saunders H, Creighton J, Coltrin M, Simmons J. Solid-state lighting: an energy-economics perspective. Journal of Physics D 2010; 43(35); 1-17.

Wei T. Impact of energy efficiency gains on output and energy use with Cobb-Douglas production function. Energy Policy 2007 35(4): 2023-2030.