The Energy Rebound Battle

By Ted Nordhaus

In the early 1990s, the resource economist Harry Saunders started asking hard questions about energy efficiency programs. Climate change at that time had only recently come to wide public attention. But already, dramatic improvements in energy efficiency figured centrally in most estimations of what to do about the problem.

Two factors conjoined to push this view. One was that energy efficiency represented a seemingly costless path to lower emissions, a way for politicians to reduce emissions without imposing high energy costs on their constituents. The other was that energy efficiency already figured prominently in the environmental agenda; in the late 1970s, green energy guru Amory Lovins had bundled radical efficiency improvements together with wind and solar energy technologies in what he dubbed the “soft energy path,” the alternative to both fossil and nuclear energy.

The problem, as Saunders saw it, was that costless energy savings—say, an energy-efficient light bulb that could light a room using half as much energy as a less efficient one—functionally reduced the cost of lighting a room. Having written his dissertation on how economies had responded to energy price changes after the Arab oil embargo, and subsequently building a consulting practice advising manufacturers how to deploy capital investments to maximize their productivity, Saunders knew that when the cost of a service or commodity declines, consumption tends to go up.

This phenomenon is known today as the rebound effect. Energy efficiency promotes a rebound in energy use, thus eroding the reductions in consumption that more efficient technologies would otherwise be expected to yield.

Saunders’s insight wasn’t a new one. The great nineteenth century economist Stanley Jevons had made the same observation about coal and more efficient steam engines. Jevons argued, correctly, that improving the efficiency of steam engines wouldn’t result in less coal use, but rather would reduce the cost of using coal to operate steam engines, resulting, ultimately, in higher use. But in the midst of the energy crises of the 1970s, Jevons’s paradox had largely been forgotten.

Still, Saunders assumed that more efficient technologies would, to some degree at least, result in lower energy use and hence lower carbon emissions. The question was how much lower, and Saunders set out to figure that out.

Saunders created a simple model of the global economy and started fiddling with the economic productivity of energy. This, in economic terms, is what more efficient energy technologies represent: an energy productivity improvement. What Saunders found surprised him. When he increased energy productivity in his model, global energy consumption went up, not down. As the effective cost of energy declined thanks to more efficient technologies, firms found more ways to use it. Saunders’s finding was due, in part, to the fact that the model he had built assumed that all inputs to economic production could be substituted for one another with no additional effort or cost.

But the economy doesn’t actually work that way. Saunders had used a production function, an equation that economists use to describe and constrain how firms swap out one input for another in response to prices, that assumed no limit to how much energy could substitute for other inputs. The more energy productivity improved, the cheaper energy inputs became and the more energy firms substituted for other inputs. This sort of substitution could in theory go on until energy was the only input into production.

In the real world, there is nothing that can be made with pure energy and no machines, no raw materials, and no labor. So Saunders started using different production functions that assumed that inputs could more or less easily be substituted for each other. And though the results varied, the underlying driver of the results did not. The effectiveness of energy efficiency improvements as a means of reducing energy use hinged entirely on the question of how easily energy could be substituted for labor, capital, and materials.

In 1992, Saunders published a paper in Energy Journal introducing what he called the Khazzoom-Brookes Postulate. Saunders generously named the postulate after two fellow economists who had raised similar concerns during the 1980s. But Saunders, in his postulate, was the first person to state the proposition in the formal mathematical language of neoclassical economic theory. If energy could be easily substituted for a range of other inputs to economic production, improvements in energy efficiency could over the long term result in higher global energy consumption. It all depended on what economists call the elasticity of substitution—that is, how easily one input, in this case energy, can substitute for others in response to changes in their cost. What at first might have seemed an obscure academic question about the proper production function to use to estimate energy savings from energy efficiency improvements turns out to have rather momentous implications for how difficult it will be to mitigate climate change. Saunders would spend the next two decades trying to quantify those implications.

He’s lost that Lovins feeling

Saunders’s postulate was not well received among those who had touted energy efficiency as a costless remedy to the nation’s energy challenges. Energy efficiency was, in the words of Lovins, “a lunch you get paid to eat,” and nobody, from policy makers looking for a quick and easy way to address climate change to companies selling energy-efficient technologies to environmental groups opposed to new energy development, was much interested in learning that the energy and emissions reduction benefits might be less than advertised.

In the years after the publication of Khazzoom-Brookes, a series of studies seemed to suggest that the rebound effect wasn’t worth worrying about. When consumers insulated their homes and installed more efficient appliances and lighting, they appeared to use those amenities a little bit more, but not a lot more. Most people, it seemed, weren’t going to leave the lights on all night or turn their thermostats up to 90 in the winter just because it was cheap to do so.

The critical studies weren’t terribly definitive. They had looked at a very small number of energy end uses, mostly in the home and almost exclusively in affluent developed economies. But they did offer sufficient evidence for what most people paying attention to the issue already believed—that energy efficiency was a key pathway, maybe the key pathway, to reducing energy use and fighting climate change. For the next two decades, the debate about rebound effects quietly raged on among energy analysts, occasionally drawing broader attention from journalists and politicians, but mostly playing out in obscure peer-reviewed journals and in assessments by government tribunals such as the International Energy Agency (IEA).

For efficiency advocates, the issue was mostly viewed as a nuisance, something that needed to be swatted away so that the world could get on with the business of radically reducing demand for energy. After articles appeared in the late 1990s in the New York Times and New Scientist discussing the rebound debate, the academic journal Energy Policy commissioned Lee Schipper, an energy economist at the Lawrence Berkeley National Laboratory, to edit a special issue on the matter. It was published in 2000. In its introduction, Schipper, a rebound skeptic, compared rebound to the Loch Ness monster, a mythical beast that reappeared from time to time but whose existence could not be confirmed.

That world-weary posture has been the default position of efficiency advocates ever since. “Every few years,” David Goldstein and his colleagues at the Natural Resources Defense Council wrote, in response to a review of the peer-reviewed literature on rebound by my organization in 2011, “a new report emerges that tries to resurrect an old hypothesis: that energy efficiency policy paradoxically increases the amount of energy we consume.”

But as concern about climate change has grown, attention to rebound effects has increased. Forty percent or more of greenhouse gas emissions reduction in most climate mitigation scenarios are predicated on lower energy use due to more efficient technologies. If those energy savings are substantially eroded due to rebound effects, the scale of the emissions reduction challenge becomes much larger.

For environmental groups, which almost universally subscribe to the soft energy path, the stakes are higher still. Without dramatic reductions in global energy use through radically more efficient technology, a return to a world entirely powered by renewable energy sources, the holy grail of the green environmental and energy agenda and a debatable prospect to begin with, becomes completely implausible.

And though the growing literature on rebound effects remains deeply contested—there is often little agreement on what even counts as rebound and even less on how to calculate it—estimates of how large the rebound effect may be have risen over time, as more studies have been conducted across a broader range economic conditions. Summarizing the evidence in his introduction to the special Energy Policy issue, Schipper suggested rebound of 10-40% depending on the sector and economy in question. Since that time, assessments by the United Kingdom, the Organization for Economic Cooperation and Development, the Intergovernmental Panel on Climate Change, and the IEA have all concluded that rebound effects are likely significantly larger.

By 2015, Gernot Wagner, at the time the top economist at the Environmental Defense Fund, a long-time efficiency champion, had acknowledged that rebound effects probably ranged from 20% to 60%, a level that he judged to be encouraging insofar as it seemed likely, to him at least, that 50% or more of the engineering-based estimates of energy savings due to efficiency improvements might ultimately be realized.

The rebound debate has also become more acrimonious as the stakes have risen. Efficiency advocates have characterized rebound scholarship as an “attack” on energy efficiency and suggested that those who believe rebound effects to be significant are in effect arguing that reducing energy efficiency must therefore be part of the solution to climate change. One well-known efficiency advocate has gone so far as to label rebound proponents “efficiency deniers,” a characterization seemingly designed to echo the polarizing language of “climate denial.”

Here be monsters

While “Nessy,” as Schipper had it, continued to make occasional appearances, Saunders kept plugging away. Having established theoretically the factors that would determine the extent to which energy productivity enhancements would save energy rather than rebound in the form of new production and consumption, Saunders set about devising a method to answer the question empirically.

Saunders’s idea was to build an econometric model of every production sector of the US economy and then crank in real data on prices, inputs, and outputs, looking back over 45 years (1960-2005). With that data in hand, he would model two counterfactual scenarios, one in which there was no rebound effect and one in which rebound was 100%, meaning that all of the energy savings from more efficient technology was taken back in the form of new production. He could then compare these calculations with actual energy use in each sector in order to estimate how much of the savings due to energy-efficient technology had been taken in the form of lower energy consumption versus increasing production.

It was a clever approach to the problem. But before he could undertake that analysis, Saunders would need to identify a production function that was flexible enough to accommodate a range of potential behaviors by firms and that wouldn’t predetermine the result of the analysis. As Saunders had discovered in his early work on Khazzoom-Brookes, an analyst wanting to show high rebound could simply choose a production function in which substitution was easy, while an analyst wanting to show low rebound could choose a function in which substitution was very hard. Saunders wanted to find a function that didn’t overly constrain or unduly allow substitution of energy for other inputs, so that he could empirically derive substitution elasticities from the data.

Over a number of years, Saunders tested different production functions to see how they affected the outcome of rebound simulations. In 2008, he published a summary of that work, suggesting two functions for analyses of rebound that appeared to allow for a wide range of substitution elasticities across a broad range of heterogeneous sectors of the economy. Saunders could finally set about the work of building an econometric model that would be capable of testing Khazzoom-Brookes empirically.

It took Saunders four more years to publish his results. Using data painstakingly assembled by Harvard economist Dale Jorgenson from a variety of government and industry sources going back many decades, Saunders estimated that about 60% of energy savings from more efficient technologies had been plowed back into the production process. Six sectors had seen outright backfire in short order, meaning that all of the energy savings associated with technical efficiency improvements were lost to higher energy use. These included energy-intensive sectors such as electric utilities, primary metal, and mining.

Saunders published the disquieting results in 2013. His article, in the journal Technological Forecasting & Social Change, provided the first definitive and carefully quantified estimates of long-term rebound in production sectors of the US economy. Most early empirical studies had looked at end-use energy consumption in the United States and Europe, focusing on consumer energy-intensive energy uses, such as home heating, residential lighting and appliances, and driving. But two-thirds of energy consumption occurs in the production sectors of the economy: to construct the homes we live in and the buildings we work in, to grow and transport the food we purchase at the supermarket, to manufacture the goods we purchase at the shopping mall, and to build and operate the infrastructure that allows us to move people and goods among all those places.

It was here that Saunders had consistently found levels of rebound that were much higher than was typically found among end-use consumers in rich countries, where demand for energy services such as lighting, transport, and heating quickly saturates, meaning that consumers have little desire to consume more of them. By contrast, energy productivity improvements in production sectors create all sorts of new production possibilities—to substitute energy for labor or other resource inputs; to produce goods at lower cost, thereby allowing higher consumption; and to invent new products and services that are made possible only by greater efficiency.

Energy-saving technological change enabled not only more efficient provision of existing energy services but also new and expanded uses of energy. LED lighting allowed us to put lights in all manner of places we couldn’t put them before. Liquid crystal display screens allow us to put video screens onto skyscrapers, inside taxicabs, and into our pockets. A family that once had a single inefficient refrigerator might now have an efficient one in the kitchen, a freezer in the basement, a wine cooler in the bar area, and a portable electric cooler for the car.

All those LEDs and LCDs and mini refrigerators might still not raise the user’s energy consumption due to their much higher efficiency. But the production side of the equation is a different story. Manufacturers produced more of all those things, here and abroad, and further efficiencies at the production level meant that costs to consumers could be further reduced, making new production more profitable and new consumption more affordable.

It wasn’t necessarily that energy productivity improvements alone drove these developments. Often, new and more energy efficient technologies in the production sector brought other productivity factors along for the ride, raising labor and resource productivity in a variety of ways, making everything else more productive, too. This is called total factor productivity, and when you raise it, costs go down while output and consumption typically go up. That brings with it a general benefit to economic welfare, but also higher energy consumption, all else being equal.

The implications of Saunders’s findings are all the more significant globally, where demand for energy is much less saturated than it is in wealthy economies such as the United States. More efficient lighting, heating, cooking, and refrigeration allow poor populations living in energy poverty to consume much more energy. Beyond the household, energy-intensive production sectors such as steel, cement, chemical manufacturing, and refining are expected to grow enormously across the globe over much of this century, as emerging economies worldwide build the basic infrastructure of modernity. These are precisely the sectors that both historical analysis and Saunders’s studies suggest are most prone to backfire.

The soft path plays hardball

Even before Saunders published his 2013 analysis, efficiency advocates went to work attempting to discredit it. Saunders had shared a prepublication copy widely with other analysts and presented a version of it at a 2011 workshop on rebound effects hosted by Carnegie Mellon University. In 2012, two efficiency consultants from an outfit called CO₂ Scorecard demanded that Saunders’s paper be retracted, claiming, incorrectly, that he had based his analysis on the monetary value of energy inputs, not the quantities of energy being consumed. The claim was based on criticisms originally made by two prominent efficiency researchers, Jon Koomey of Stanford University and Danny Cullenward, then a Stanford PhD student and now a research fellow at the University of California, Berkeley.

Koomey and Cullenward were subsequently forced to concede that Saunders did in fact utilize primary data on physical quantities of energy inputs. But their concession did not come before the supposed “debunking” had been widely disseminated on blogs at the liberal Center for American Progress and at UC-Berkeley.

Undeterred, Koomey and Cullenward created a new pretext on which to dismiss Saunders’s findings, claiming in a 2016 response published in the same journal as the original analysis that because Saunders had not accounted for regional differences in energy prices, his results were invalid. The pair had conducted no independent analysis. They simply asserted that having failed to include price differences, Saunders’s conclusions were false. Writing at Koomey’s website, the pair went further, characterizing Saunders’s findings as “aggressive and unsubstantiated” and “wholly without support.”

So Saunders reran his analysis with a wide range of energy price sensitivities. In early 2017, he published his new findings, again in Technological Forecasting & Social Change, demonstrating that marginal price variations were in fact immaterial to the earlier result.

This sort of give and take might, generously, be chalked up to the normal processes of scientific progress. Researchers find problems in existing scholarship, and subsequent scholarship then addresses those shortcomings. But it is hard to read the public attacks on Saunders’s work and conclude that Koomey, Cullenward, and other efficiency advocates were acting in good faith.

Rather, they seized on one purported shortcoming after another in an effort to discredit Saunders’s findings. The criticisms weren’t constructive. They made no suggestion as to how the alleged shortcomings in the Jorgenson data set might be rectified, or even to ascertain, as was easily done, that the data set did contain primary data on energy input quantities. Nor did they contemplate undertaking their own independent modeling exercises to determine whether regional variances in marginal energy prices might suggest different levels of rebound. The intent, it would appear, was not to advance better understanding of rebound effects but to suppress that understanding.

Modeling backfires

The long-running debate about rebound might not be otherwise settled, but it would almost certainly be less contentious were the issue not so tied up with persistent debates about climate change and the energy future. Saunders’s modeling of rebound in the production sectors of the US economy is an impressive analytical accomplishment, and it adds to a growing literature suggesting that rebound in the aggregate is likely to take back a very substantial portion of the emissions savings that many energy analysts and climate advocates have long counted toward climate mitigation. But even putting aside the rear-guard sniping about Saunders’s data and methods, basic questions of causation remain, questions that are more a reflection of the limits of knowledge about the future than the methods of econometric modeling.

Because energy productivity is so tied up with other factors of production and consumption, no clever econometric model can tell us whether, had incandescent light bulbs not come along, the Earth at night, when viewed from outer space, would instead be illuminated with hog fat lanterns and wax candles, to take a particularly absurd example. If the alternative to electric lights had been a world lit by hog fat, then incandescent light bulbs and subsequently LEDs would have resulted in enormous energy savings. Or to take another example, if you think that without the development of LCD screens we would all have 50-inch cathode ray television sets on our walls and cathode ray smartphones in our pockets, then the development of vastly more efficient LCD technology has also been a huge energy saver.

The problem is that whereas the LCD screen was one of a series of enabling technologies that made smartphones possible, it didn’t cause us to invent them, exactly. And that, for the most part, has been the story of energy productivity improving technology for over two centuries. More efficient technologies often initially provide benefits to existing economic activities and forms of production. Better steam engines initially reduced the amount of coal that was needed to pump water out of mines. But the more important benefits were ultimately all sorts of new uses for the technology not envisioned initially. James Watt had no idea that the technological revolution that he unleashed with his newfangled steam engine would ultimately power trains and electrical generators. Neither had been invented at the time.

That history continues today. LCDs might not be the cause of smartphones, but the existence of smartphones erode the savings that an engineering-based estimate of the energy savings associated with replacing cathode ray televisions with LCDs in, say, the year 2000 would have arrived at. The same is true at the macroeconomic level. Implicit in long-term projections of economic growth, and the energy use that comes with it, are broadly recognized but impossible-to-predict interactions among energy productivity, multifactor productivity, and economic growth.

Climate mitigation models attempt to account for these dynamics by using historical trends to project economic growth and energy intensity decline in baseline scenarios. These projections ostensibly account for rebound because it is a component of both economic growth trends and energy intensity trends. But the mitigation scenarios also include additional and specific efficiency improvements that fail to account for either the growth effects or the substitution effects associated with those improvements.

The consequences of this oversight are considerable. The International Energy Agency continues to tout energy efficiency as the “first fuel” available to member countries to constrain energy use, while the Intergovernmental Panel on Climate Change follows suit, presenting forecasts showing energy efficiency to be the best lever for reducing emissions of greenhouse gases in the coming decades. However, both organizations rely on energy models that assume highly rigid productive economies with minimal flexibility to accommodate energy efficiency gains. The IEA, for example, assumes rebound will be no more than 10% in the coming decades.

To illustrate the implications of this IEA assumption, if one were to instead assume rebound will be 50%, a figure that can be easily supported by the growing literature on rebound effects, meeting the carbon emissions targets contemplated in the IEA “New Policies Scenario” would require global clean energy deployment about one-third higher than the agency’s already ambitious targets, about 4.7 Terawatt-hours of additional clean energy by 2035, or slightly more than total US electric power production in 2016.

How efficiency matters

While the rebound debate rages on, Saunders continues to pull on the thread that he first began to unravel with the Khazzoom-Brookes postulate 25 years ago. In 2014, Saunders published a paper in Ecological Economics that extended his analysis to the long-term evolution of market economies. Using the same sort of theoretical framework that he used to establish Khazzoom-Brookes and many of the same well-established economic principles, Saunders demonstrated that under highly plausible conditions—namely, well-functioning markets, population stabilization, and saturating demand for further consumption—market economies would, in theory, evolve toward zero growth and declining demand for natural resources, an unexpected conclusion from a researcher widely criticized as believing that there is no alternative to endless unchecked economic growth, and hence energy demand.

Those conclusions should remind us that intuitions and assumptions, not to mention political orthodoxies, about cause-and-effect relations for incredibly complex problems such as climate change should continually be subject to critical analysis. And ideas that challenge present orthodoxies, such as those around the costless emissions reductions that might be achieved through energy-efficiency programs, often open up new possibilities and frameworks for making progress.

The rebound debate mostly obviously shows how our beliefs about how the world ought to work influence our willingness to accept some scientific findings and our inclination to reject or ignore others. But more importantly, resistance to evidence of the limits of pursuing energy efficiency as a strategy for addressing climate change has blinded many scientists and advocates to more fundamental understandings of the relationship between energy use and human development that, after all, is the reason we care about climate change in the first place.

Recognizing that quite significant levels of energy-efficiency rebound are a likely result of efficiency gains in many cases and in the global aggregate is not an argument against energy efficiency, as some on both sides of the debate have suggested. Nor will improving energy efficiency inevitably result in higher energy use. Rather, rebound is a crucial indicator of long-term progress toward a more equitable and sustainable world.

Rising energy productivity and rising energy use are inexorably entwined with broader ecological modernization processes. As populations become wealthier around the world—thanks in no small part to increasing energy productivity—fertility rates decline, population growth slows, and population stabilizes. As those populations achieve modern living standards, material consumption begins to saturate, as it has in the industrialized world. The low levels of rebound measured in end-use sectors of the economy in wealthy economies are evidence of this dynamic. As material demands saturate, the structure of economies shifts, from output that is skewed toward agriculture, manufacturing, and other energy intensive forms of production toward knowledge and service sectors that have much lower energy intensities.

Counter to commonly held intuitions, it could even turn out that the faster energy consumption grows in the short- and medium-term, the sooner energy use and emissions will peak and the lower that peak will be in the long-term. But this also means that efficiency won’t turn the tide of rising energy use anytime soon, and it won’t likely make the difference in allowing us to meet mid-century emissions targets. Ultimately, progress toward mitigating climate change will primarily depend not on how quickly we boost energy efficiency, but on how quickly we are able to replace fossil-based sources of energy with carbon-free energy.

What is most important about improving energy efficiency is that it will help create the conditions necessary to both better mitigate climate change and manage the impacts that can’t be avoided. That’s because improving energy efficiency is welfare-enhancing irrespective of its climate benefits. A wealthier global population is a healthier population and one that will be more resilient to climate impacts. It will also be better able to bear the costs necessary to reduce emissions by building a low-carbon energy system. That should be cause for optimism, not pessimism.