January 23, 2013
Understanding Energy Efficiency Rebound: Interview with Harry Saunders
The "Godfather" of rebound, Harry Saunders is widely known as an international expert on energy efficiency and consumption and has also published articles in the fields of evolutionary biology and legal theory. Following Daniel Khazzoom and Leonard Brookes' work on energy consumption and behavior, Saunders coined the "Khazzoom-Brookes Postulate" which broadly states that increased energy efficiency leads to increased energy consumption. Most recently, Saunders surveyed and analyzed thirty sectors of the United States economy for historical evidence of rebound.
What is your training in?
I received my undergraduate degree in physics at the University of Alberta; a master’s in resources, environment and planning at the University of Calgary; and a PhD in engineering-economic systems from Stanford University.
What drove your interest in energy?
In 1968 I saw the famous Hubbert [peak oil] curve of oil production and got really curious about what that meant for future economic growth in the world and became obsessed with understanding the relationship between energy and economic growth.
Were you convinced that peak oil was a reality?
I was an early environmentalist and worried about Third World poverty. I saw the peak oil curve and at that time everyone was predicting infinite economic growth from infinitely available energy and I wanted to understand what the relationship was between energy and economic activity.
Did you stay compelled by the notion of resource scarcity, or did your views change?
My views changed but I was still interested in OPEC oil supply and oil shocks, and how they affect the economy. I came to understand that demand for oil (and energy generally) is more complex than anyone seemed to realize. And the link to economic activity was equally complex.
How did you become interested in energy efficiency?
I realized that many of the energy forecasts that introduce energy efficiency into their modeling, and that predict that energy efficiency will reduce energy consumption, are flawed. These forecasts don't describe how energy efficiency actually works in the economy. I came to realize that energy efficiency is not as easy or straightforward a solution to scarcity and climate change as many claimed it to be.
I began a long journey to establish a formal theoretical foundation for understanding energy efficiency rebound, and was interested in creating a method for the empirical measurement of rebound.
Isn't rebound predicted by basic economics as something that you get when you increase the productivity of an input (factor) of production?
When you increase the productivity of a factor, consumption can increase or decrease. It depends on what happens at the margin (i.e. what level of consumption will equate its so-called marginal productivity to the price you pay for it). Theoretically, this consumption level can be lower or higher than before the efficiency gain. Nothing fundamental prevents energy use going up in response to an energy efficiency gain, although the prevailing expectation is it will typically go down somewhat.
In the case of the non-energy factors, however, according to standard economic theory, increasing their efficiency increases the marginal productivity of energy (i.e. makes that factor more valuable to use) and thereby leads to increased use of energy.
Can you give me an example of rebound as it relates to labor productivity, and an example as it relates to materials productivity?
Labor productivity: Assume I own a warehouse and decide to buy a forklift to replace the handcarts that my workers have been using to move and load freight. Unlike the handcard, the forklift requires fuel. This is a substitution of energy (and capital) for labor. If this is a new fuel efficient forklift that my calculations show has now made forklifts affordable for me to operate, this substitution toward energy has been enabled by the energy efficiency gain.
Materials productivity: Assume I am a primary metal producer who has just implemented new energy efficiency technologies that allow me to produce the same amount of metal with less energy. But what I notice is that it has now become profitable for me to use lower grade ores to produce that metal. But lower grade ores require more energy per ton of ore used (due to a more energy intensive extraction process). I will increase my use of energy compared to what I would have needed to produce from high quality ores. This is a substitution of energy for materials inputs due to the energy efficiency gain.
In both these cases, simple profit-maximization principles lead to energy use being higher than a simple engineering efficiency calculation would indicate. Further, in both cases, it is likely that the profitable scale of production will be larger as well, leading to a so-called “output” rebound effect. In these examples we’ve restricted ourselves to “direct” rebound effects only.
Don't energy efficiency advocates understand that?
This is where many energy efficiency advocates fail. They look at energy in isolation, with out considering rebound. They do not account for the fact that productivity gains for one factor like energy affect the use of all the factors.
How does Rocky Mountain Institute’s Amory Lovins treat rebound?
Amory Lovins looks at energy efficiency gains he has seen implemented by particular firms and extrapolates such gains over the entire economy, and then assumes one-to-one energy savings without accounting for rebound. He completely ignores the potential for industry to substitute among factors in response to efficiency-related cost savings and increased profits. Lovins sees production as rigidly accommodating to energy efficiency gains; in reality, firms are flexible, adaptive, and creative in their quest for increased profits. This mental model leads him to grossly understate future energy needs.
Do the International Energy Agency or Intergovernmental Panel on Climate Change understand rebound?
The vast majority of energy modelers whose work feeds into IPCC and IEA reports are using models that are fundamentally limited in several ways. Even the best among these models use oversimplified and arbitrary functional forms, assume (rather than measure) key parameter values, ignore non-energy efficiency gains, and don’t introduce energy efficiency in any properly defensible way.
I eagerly await the day when energy consumption forecasters pay closer attention to basic microeconomic principles, sensibly link engineering efficiency gains to real-world economic decision making, undertake proper econometric measurements of the key driving parameters, and squarely address energy efficiency rebound in a way that gives policy makers what they sorely need.
Is the computer revolution a case of rebound?
Yes. Today I use a laptop computer like millions of others. But when I was an undergraduate student, we used to go to a computer center. There was a big mainframe machine in a climate controlled room. It was a huge power hog and its floating point operations per second and bytes moved per second were trivial compared to a modern laptop. Your laptop is way more powerful than those old mainframe computers, and you probably recognize that your laptop is more efficient. It uses only a tiny fraction of the energy that an old mainframe did. But then you look around and ask yourself how many laptops are there in the world compared to the mainframes that used to be there, and what their total energy use is, compared to the old mainframes? The total energy use for computing is probably at least an order of magnitude greater.
Efficiency advocates respond that you can't blame energy efficiency for the proliferation of laptops.
But ask yourself, “How many people would use a laptop today if it used the same amount of energy as those old mainframes did?” Energy efficiency gains have enabled the computing revolution. The computer revolution has been made possible by making chips smaller and smaller, but smaller chips is only possible if they are highly efficient in the use of energy (they have to be more and more efficient because otherwise the heat they release would defeat their functioning). To date, nobody has tried to analyze the effects of energy efficiency on creating whole new products, applications, industries, or expanding the frontier of where economic production is now.
So screens everywhere – smart phones, tablets, billboards, larger TVs — are caused by increasing efficiency in energy and lighting?
Yes. Jeff Tsao and his team looked at the history of lighting over 300 years and observed enormous gains in energy efficiency. They also found that, because they reduce the effective price of lighting, efficiency gains have led to an enormous expansion of energy use for lighting. When they saw these data, they got me involved to look at the economics of it. In the end, our results show 100 percent rebound over three centuries, five technologies and six continents.
This raises a huge question and places the burden of proof on those who argue that solid state lighting is a prescription for reduced energy use for lighting.
Have other studies been done that back up Tsao's research?
A couple of articles recently have appeared in the same vein: one by Roger Fouquet, and one by Fouquet and Peter Pearson. These researchers looked at transportation in the UK over 150 years, and lighting over 200 years, and they showed 60 – 70 percent rebound.
One troubling concern this raises is that many researchers may be undertaking rebound analyses that comprehend time horizons far too short to reveal the full power of the rebound dynamic.
What are you working on now?
I'm working for the Asian Development Bank, a regional organization similar to the World Bank, on an energy report. One thing we are looking at is rebound and backfire propensity in the developing regions of Asia. In the energy research realm I’m working on an empirical analysis of indirect rebound effects on the productive side of the US economy.
You said you first looked at efficiency in 1989. What did you do?
I created a model extending Solow's neoclassical growth model to include energy and energy efficiency explicitly. I thought it would be interesting to see what neoclassical growth theory would tell us about what kind of energy reductions could be expected for different levels of energy efficiency gain. But when I put in any level of energy efficiency gain I saw that energy use went up! I thought I'd screwed something up in coding the model but I double-checked it and it everything checked out. So I sat down and went through the math and, sure enough, it proved that an energy efficiency gain would always increase energy use.
When I later introduced a more general functional form, I discovered that rebound magnitude depends on the ease with which energy can be substituted for other inputs to production (factors of production). I also discovered that efficiency gains for other inputs always led to an increase in energy use – a kind of backfire associated with efficiency gains more generally.
I thought I had been the first to discover that standard economic theory either predicted backfire or that energy efficiency gains would not reduce energy much at all.
Was it a major discovery?
My elation at this discovery was short-lived. Someone told me Len Brookes and Dan Khazzoom had been talking about something like this, and I went back for a closer look at the literature and discovered that Len and Dan had scooped me by 10 years! Len had argued qualitatively that energy efficiency would increase energy use; and Dan had shown analytically that mandated standards for household appliances would result in far lower reductions in energy use than were being claimed.
Both Len and Dan credited a 19th century economist, W. Stanley Jevons, with the original insight. It’s one thing to be scooped by 10 years; but exceptionally painful to be scooped by 150 years!
So Brookes was early backfire and Khazzoom was early rebound?
Yes. Len Brookes has been a defender of the idea that energy efficiency gains will always produce increased energy use. He even hates the term “rebound,” as it gives credence to the idea that energy efficiency gains could reduce energy use below where it otherwise would be. Both Brookes and Khazzoom use basic classical economics arguments, but neither touched growth theory.
What happened after you published your 1992 paper?
In various conferences in the 1990s I kept making this case and finally caught the attention of Lee Schipper, a well-known, super-dynamic, and entertaining energy efficiency guy. He was highly skeptical and we argued a lot back and forth in these conferences, but I could tell it troubled him. Finally, in 2000, Lee got a group of people together to do a special issue of Energy Policy devoted to rebound. That’s when it came alive again in the literature.
Did you write for that?
I had two papers in that volume.
What was in that issue?
Lee assembled all the usual suspects in the energy efficiency field at the time. Len contributed. Peter Berkhout, my fellow Canadian Mark Jaccard, and Skip Laitner (whose names appear everywhere in the efficiency literature even today) also contributed. Rich Howarth and Fatih Birol (current chief economist of the IEA) were also part of this effort. Joyashree Roy provided us the first hints that rebound effects might be significantly larger in developing countries than industrialized ones.
Probably the most widely-cited article there, other than the issue itself, was by Lorna Greening and David Greene (and Carmen Difiglio, whom I don’t know). They conducted a review of what was a very sparse literature at the time, and concluded that “the range of estimates for the size of the rebound effect is very low to moderate.” But they did point to a number of methodological issues. And they provided what was really the first attempt at a rebound taxonomy (including the identification of “indirect” rebound effects), since improved by Steve Sorrell and Karen Turner. Amory leapt on this paper and has continued to cite it as the definitive word on rebound being a trivial effect even though it is long since dated.
How did Greening, Greene, and Difiglio come up with such small rebound?
Greening, Greene, and Difiglio were seriously hampered because the data were sparse and there were few empirical studies to draw from. Also, the studies they relied on tended to focus on end use consumption, which represents only 1/3 of the global energy economy, while 2/3 of energy is consumed in the production and movement of goods and services.
But nobody had done a good analysis of the productive side of the economy. Their analysis was necessarily fairly shallow. Another problem was that they ignored the huge problem that rebound mechanisms in developing countries are likely to be far different and are likely to produce very large rebounds, as Joyashree pointed out. But Joyashree’s analysis was largely ignored, and continued to be ignored in the decade after this.
How could they leave out those two huge areas?
Poor data and analyses available at the time. Also, a prejudice in favor of looking only at end-use consumption in industrialized countries. Additionally, we had the emergence of the somewhat tribal notion in policy circles that energy efficiency was a solution for all of our energy woes – scarcity, pollution, climate change and the like.
Did anything happen between that issue and Steve Sorrell's 2007 UKERC study?
Not much. One day out of the blue I got a call from the UK Prime Minister's office. Apparently some very bright Lord in the House of Lords started looking at the promises made by the government about how energy efficiency would significantly reduce energy demand in the UK and noticed it wasn't happening. This Lord raised hell and challenged the government to explain.
After the UKERC report came out rebound exploded in the literature. Every month there are articles coming out.
What does your new study do and find?
My new study looks at 30 sectors and uses an extremely meticulous and well-known data set from Dale Jorgensen at Harvard. This is arguably the best efficiency rebound data set available in the United States. It uses a highly flexible functional form, consistent with what I have shown to be necessary. It empirically measures the parameters of this function and all the associated elasticities. It measures the technology gains for all of the factors – capital, labor, energy and materials. And then using those it measures rebound over the period of 1980 – 2000 by sector. The results show varying rebound magnitudes across sectors, but when aggregated together the total shows 50 – 60 percent rebound for the US. And this only includes so-called “direct” rebound.
What about indirect and long-term rebound?
When we look over long time frames that consider both direct and indirect effects, we get very large rebounds – in the case of lighting it's 100 percent rebound and close to backfire. Equally as worrisome, there are indications that backfire phenomena are far from uncommon in the developing countries of Asia. While the magnitudes are still under study, the mechanism is understood: when energy efficiency gains are sufficiently large to make energy affordable to those previously experiencing energy poverty, vast new realms of welfare-creating energy use are opened up.
In another example, Chinese scholars recently measured rebound in passenger transportation from 1992 to 2010 and found it was 107 percent. All this adds further evidence to Joyashree’s argument that rebound can be very high in developing countries.
What do you conclude from all this?
To the extent that the energy use projections of the IPCC, the IEA and others understate future energy consumption because they ignore (or generously, inadequately account for) rebound effects, the world has less time than we think to devise climate change solutions.