January 23, 2013
Understanding Energy Efficiency Rebound: Interview with Steven Sorrell
Dr. Steven Sorrell trained as an electrical engineer and spent four years developing signal processing systems in industrial R&D laboratories before gaining a MSc in Science & Technology Policy in 1991. Since joining the Science Policy Research Unit (SPRU) at the University of Sussex, Dr. Sorrell has undertaken a range of academic and consultancy research in the area of energy and climate policy. He is currently deputy director of the Sussex Energy Group (SEG) at SPRU, co-manager of the Technology and Policy Assessment function of the UK Energy Research Centre (UKERC) and Honorary Senior Fellow at the Centre for Environmental Policy, Imperial College.
The new International Energy Agency report says that there is a huge amount of energy efficiency improvements that firms could easily take but don't. Why don't firms do more energy efficiency if it's so profitable? Is it because CEOs can't be bothered to focus on changing light bulbs?
The conventional answer to this is that there are a whole host of "barriers" that prevent them. But people disagree on what these are, how important they are and whether governments should do anything about them. I wrote a book on this topic about 10 years ago that drew on 50 real-world case studies of organizations in an effort to understand what was stopping them from improving energy efficiency.
So yes, it's the staff time involved, the hassle, the lack of information, the split incentives, and so on. Often the management isn't going to be bothered with light bulbs. There are transaction costs to efficiency that can certainly be reduced in some circumstances, but there are multiple overlapping reasons why organisations neglect those opportunities.
How would you sum up your view of where rebound stands in the literature?
While I don't think there is evidence that energy efficiency will always lead to greater energy consumption, to claim rebound is close to zero is to fly in the face of the evidence.
What do you think of the IEA's use of a 9 percent number for rebound?
I don't think a global across the board figure is appropriate. You should try to disaggregate the assumption as far as the evidence permits. A 9 percent average seems too low given what we know about macroeconomic rebound effects, as well as the much higher rebound effects in developing countries. Perhaps a more useful guide would be the global macroeconomic modelling study that Terry Barker led a few years ago. This explicitly modelled the rebound effects from the energy efficiency policies in a previous IEA World Energy Outlook (2006). The team simply assumed a magnitude for direct rebound effects, but then modelled various indirect effects. The net result was a total rebound effect of 52 percent - which is a lot bigger than nine percent. It's hard to argue about a number, given that the evidence is so mixed. But the best way to deal with this uncertainty is to emphasize it and to conduct sensitivity tests, rather than just picking a low number with limited justification.
How did you get started on rebound?
Our research group here is part of the UK Energy Research Centre (UKERC), which decided to do the rebound study in 2005. It met the criteria of being important and controversial and with a growing evidence base. The debate on rebound was really polarized. An early rebound researcher, Len Brooke, made some very strong arguments on the topic and efficiency advocates responded negatively to his findings. So it seemed like it was time to do a balanced review of it.
What did your research consist of?
We did what's called a "systematic review of the evidence," which is a methodology that originates from the medical field and is now increasingly used in social science as well. But for a topic as wide-ranging and uncertain as rebound, a systematic review is not easy. We thought it would take about a year but it took well over two years. The final product consisted of the main report and five technical reports, and since then we have published several journal articles. The report was received really well, gaining many citations, and a lot of good press attention in the UK.
What did you find?
When I started I was looking to understand direct rebound effects, and the household sector, which is how most people understand rebound. But it soon became obvious that it was much larger than that and that it was a macroeconomic phenomenon, which was Len Brooke's argument all along. But as soon as you recognize that it is macro you get sucked into larger questions of the relationship between energy and productivity and economic growth. This is highly complex with multiple causes, and it all starts to get rather bewildering.
What makes it so bewildering?
Trying to quantify rebound is extremely challenging. What we do is quantify those bits of the phenomena for which we have relatively good data, such as driving cars further when they become more fuel-efficient. But you immediately discover it doesn't end there — efficiency improvements lead to knock-on changes throughout the economy. And efficiency improvements rarely occur on their own — they’re normally associated with all sorts of other technical changes and economic changes.
So people miss the long-term and macro effects?
If you're looking at only short-term changes, you're missing macroeconomic changes to the economy over the long term. For example, efficiency improvements make car travel cheaper, so you might drive rather than take the train. But over the long term we respond to cheaper car travel by changing the way we live — having lower density housing, living further from where we work, changing our shopping patterns and so on. Then trying to isolate cause and effect becomes extremely difficult; the economy is a complex system with all sorts of feedbacks. Because it is so difficult to quantify long-term rebound effects — the complexities are so great — some people dismiss them altogether.
How difficult is it to measure more immediate rebound?
It's never straightforward or easy, but transport is one of the easiest areas to estimate rebound effects because you have relatively good data – either from national transport statistics or from surveys of travel behavior. If you can get data on changes in fuel prices, vehicle efficiencies and distance travelled, you can estimate the elasticity of demand for travel with respect to the price of travel – in other words, the cost per mile – and that gives you an estimate of the direct rebound effect. Usually you don't directly estimate the effect of the efficiency improvement; you use the price of travel as a proxy, which depends both on efficiency and fuel prices. You assume an improvement in efficiency is equivalent to a reduction in fuel prices. Both make travel cheaper. Fuel prices have changed more than fuel efficiency in the past, so that gives you more variation in your independent variable — which is what you need to get a good estimate.
What are the harder things to measure?
It is more difficult to estimate indirect and economy wide effects than direct effects; it is more difficult to estimate rebound effects in industry than in households; and it is more difficult to estimate long-term impacts than shorter term ones. Rebound effects are likely to get larger at macro level, but at that level it's harder to measure.
Is that because it depends on predicting the continuation or discontinuation of trends?
Yes, and because the causality is complex. And because people are often at cross-purposes about what the independent variable is and what the dependent variable is, and how you should measure them. Suppose the energy intensity of the US (energy consumption per unit economic output) has fallen. That gets equated to "energy efficiency" — but what does that mean? Energy use in the US depends on so many things. Think about how energy use has changed as you move energy-intensive manufacturing industries to China. And how do you measure US energy consumption, anyway? If you simply add up kilowatt-hours from different energy sources, you are adding apples and oranges. A shift in the energy mix between nuclear, gas, coal, and oil will change the energy intensity of the economy. The bottom line is that you have all sorts of things affecting aggregate energy intensity beyond micro-level improvements to energy efficiency.
How good are the models at estimating rebound?
Models can be an effective way to isolate the effect of individual causes. Computable General Equlibrium (CGE) models allow you to change one parameter, such as the energy intensity of a sector, and then explore the subsequent changes in things like prices, employment, and trade. You try to follow the consequences of such changes all the way through the economy. It offers an obvious way of isolating the effects of specific types of energy efficiency improvements.
I think CGE models are a good way to tackle the issue and to understand the phenomenon. But I have also been critical of these models for relying upon assumptions that are not well founded in empirical analyses.
What's an example of such a model assumption?
The ease of substitution between energy and other inputs. To what extent can you substitute manufactured capital for energy inputs? CGE models rely upon production functions. A production function is an equation that tells you the mix of capital, labor, fuel and other inputs used to obtain a particular level of output. They have to assume a form for this equation and then parameterize it — fit numbers to it. You can do this through statistical analysis of data, but often the data and studies that are available are not well matched to the models. They use different functions, different definitions, different ways of combining sectors, and so on. So frequently you just have to use your best judgment. You will find that different models make very different assumptions for the same variable. That is bound to lead to very different results.
In the case of substitution of energy and machines for other inputs, like human labor, is the problem that you have so many different cases of this (farmers being replaced by tractors, bank tellers by ATMs, factory workers by robots, etc.) that it's hard to generalize?
Yes, that's one way of putting it. Energy is a hugely complex system that behaves in complex ways.
I think we need a more systematic approach. There's an increasing amount of research being done, but it's still pretty patchy and there's not enough in the way of sensitivity testing. We need more research in other countries, especially developing countries where energy demand is growing the most, and where there is the greatest elasticity of demand, and in other sectors like manufacturing. The focus has been on OECD households in part because they are the easiest to study.
That's what makes Harry Saunders' recent work so important. He has done a very impressive job tracking rebound effects in the industrial sectors of the US economy over the last three decades. It is formidably complex — his latest paper has a 60 page annex of equations — and it needs to be repeated by others. But the message of this work is that rebound effects have generally been very large — 60% or more. This is empirical analysis, not modelling, and it needs to be taken seriously.
What are you discovering about rebound with your current research?
We have been funded by the UK government's Department of Environment Food and Rural Affairs to estimate rebound effects for households. We are trying to capture both direct and indirect effects. So, for example, if you buy a fuel-efficient car it makes driving cheaper, so you might drive further. That's a direct effect. But you also might spend the money saved on gasoline on other things, such as public transport and products made in China that also involve emissions. That's an indirect effect. Both these things offset the original energy savings and emission savings – at least at the global level. When you add it up, the erosion of savings is not trivial.
We're estimating the impacts of these choices on greenhouse gas emissions. The situation in the UK is that we plan to reduce the carbon intensity of electricity generation 80 to 90 percent by 2030, by shifting to renewables at a large scale. So by 2030 improved electricity efficiency won't save much in the way of emissions. But it will save households money that they then use to spend on other goods and services, including products made in China, which have a relatively high carbon intensity. That means you will end up with much higher rebounds or even backfire (rebound of more than 100 percent) by the late 2020s. That paper has been accepted by Energy Policy.
Why does all this matter to climate policy?
In some ways it’s a generic message of the importance of emission leakage. You squeeze a balloon and it pops out somewhere else. The point about the rebound debate is to not ignore these unexpected and unintended consequences of energy efficiency. Historically we have tended to ignore them, either because it was easier to do so or because they were inconvenient. But that doesn't make them go away.
How should efficiency be measured?
Historically there has been a bias toward modelling what an energy efficiency improvement or policy will achieve, rather than estimating what it has achieved. We have prioritised modelling over evaluation. And the models we have used usually neglect rebound effects.
How do you measure direct efficiency rebound?
You have to set up a control group and analyse the data carefully. We did it here. We were trying to figure out how households would respond to better data on their energy use. Rather than getting a bill every four months, which is the norm in the UK, you get real time data. So you select a group of households randomly and, assign some of that group smart meters with wall displays, and others with just smart meters. One gets the wall displays and the other doesn’t. Then you look at how energy consumption changed in both groups. Having the control group means that you can control for things like changes in energy prices. That allows you to isolate the effect of the wall display on energy consumption.
Did you find an effect?
In that particular study we didn't find any significant difference between the experimental and control group. For various reasons, we were unable to properly measure consumption prior to installing the smart meters and wall displays. But the general point is that we are continuing to make assumptions about energy savings based on engineering models, and continuing to neglect experimental studies and proper evaluation.
Is the UK government paying attention to this?
Some types of rebound effects are included in UK government proposals and estimates of what particular programs will deliver. They tend to be confined to things like household heating. We have a particular problem with what we call "fuel poverty" in the UK. Something like 5 million households are estimated to need to spend more than 10% of the income on energy, just to keep warm. That's partly because they are on low incomes, partly because the houses are under occupied and partly because we have the worst housing stock in Europe. If you insulate their homes, they are likely to take the bulk of the benefits in the form of more warmth, rather than reduced energy consumption. So you won't "save" much energy — they'll use a similar amount before — but their welfare will have improved. This is termed "comfort taking" by the UK government and it is allowed for in the appraisals of policies that are designed to help the fuel poor — such as subsidised insulation.
Other types of rebound effects, and rebound effects in other sectors, tend to be neglected, largely because they are more difficult to measure.
Is your experience that policymakers view the whole discussion of rebound as sort of bad news?
The argument that you can boost the economy, reduce energy consumption, reduce emissions, and boost welfare all at the same time is very seductive. If you start raising questions about whether that is really the case, it doesn't go down too well. People react defensively and don't want to hear. You end up with a highly polarized debate that generates more heat than light. It gets in the way of sensible discussion of how important these effects are and what we might do about them. In my experience, this problem is worse in the US than in Europe, but it applies everywhere.