29 September 2010

Why Energy Efficiency Does not Decrease Energy Consumption

[NOTE: This is a guest post by Harry Saunders, cross-posted from The Breakthrough blog.]

Why Energy Efficiency Does not Decrease Energy Consumption

By Harry Saunders

I recently co-authored an article for the Journal of Physics ("Solid-state lighting: an energy-economics perspective" by Jeff Tsao, Harry Saunders, Randy Creighton, Mike Coltrin, Jerry Simmon, August 19, 2010) analyzing the increase in energy consumption that will likely result from new (and more efficient) solid-state lighting (SSL) technologies. The article triggered a round of commentaries and responses that have confused the debate over energy efficiency. What follows is my attempt to clarify the issue, and does not necessarily represent the views of my co-authors.

More Efficient Lighting Will Increase, Not Decrease, Energy Consumption

Our Journal of Physics article drew on 300 years of evidence to shows that, as lighting becomes more energy efficient, and thus cheaper, we use ever-more of it. The result, we note, is that "over the last three centuries, and even now, the world spends about 0.72% of its GDP on light. This was the case in the UK in 1700 (UK 1700), is the case in the undeveloped world not on grid electricity in modern times, and is the case for the developed world in modern times using the most advanced lighting technologies."

The implications of this research are important for those who care about global warming. In recent years, more efficient light bulbs have been widely viewed as an important step to reducing energy consumption and thus greenhouse gas (GHG) emissions. Moreover, the Intergovernmental Panel on Climate Change (IPCC) of the United Nations and the International Energy Agency (IEA) have produced analyses that assume energy efficiency technologies will provide a substantial part of the remedy for climate change by reducing global energy consumption approximately 30 percent -- a reduction nearly sufficient to offset projected economic growth-driven energy consumption increases.

Many have come to believe that new, highly-efficient, solid-state lighting -- generally LED technology, like that used on the displays of stereo consoles, microwaves, and digital clocks -- will result in reduced energy consumption. We find the opposite is true, concluding "that there is a massive potential for growth in the consumption of light if new lighting technologies are developed with higher luminous efficacies and lower cost of light."
The good news is that increased light consumption has historically been tied to higher productivity and quality of life. The bad news is that energy efficient lighting should not be relied upon as means of reducing aggregate energy consumption, and therefore emissions. We thus write: "These conclusions suggest a subtle but important shift in how one views the baseline consequence of the increased energy efficiency associated with SSL. The consequence is not a simple 'engineering' decrease in energy consumption with consumption of light fixed, but rather an increase in human productivity and quality of life due to an increase in consumption of light." This phenomenon has come to be known as the energy "rebound" effect.

The Empirical Evidence for Rebound

The findings of our SSL research inspired The Economist magazine to write a commentary about the study that was mostly correct but made a couple of errors, which we responded to in a letter. In our response, we clarified that energy prices would need to increase 12 percent, not three-fold, in order to reduce the consumption of electricity for lighting, which, to its credit, The Economist posted on its web site and published in its letters section.

Evans Mills of the Lawrence Berkeley National Laboratory wrote on the Climate Progress blog that The Economist had "inverted" our findings. However, The Economist did not "invert" our findings, it had simply overstated an implication of them.

Efficiency advocates sometimes dismiss rebound by only looking at "direct" energy consumption -- that is, consumption by households and for private transportation. Examples of rebound in this part of the energy economy would be driving your Prius more because gasoline costs you very little, or turning up the thermostat in your efficient home. But these "direct-use" rebounds are small in comparison to "indirect-use" rebounds in energy consumption. Globally, some two-thirds of all energy is consumed indirectly-- in the energy used to produce goods and services. A residential washing machine may be energy efficient in terms of function, but in terms of production, the metal body alone requires energy to mine, smelt, stamp, coat, assemble and transport it to a dealer showroom and eventually a residential home. The energy embedded in your washing machine, or just about any product or service you consume, is very large. And remember that any money you save on your energy bills through efficient appliances or the like is re-spent on other goods and services, which each take energy to produce, all while more productive use of our money (e.g. in spending, savings and production) spurs a more robust economy, demanding even more energy.

As our recent SSL research suggests, there is strong empirical evidence that even in the "direct" part of the economy, the rebound effect can sometimes be so substantial as to eliminate essentially all energy reduction gains. But in my new research (which relies on a detailed, theoretically rigorous econometric analysis of real data), the rebound effect found in the larger "indirect" part of the economy is even more significant -- and more worrisome.
Varying degrees of rebound occur because the phenomenon works in several ways. Increasingly efficient technologies effectively lower the cost of energy, as well as the products and services in which it is embedded. This results in firms consuming more energy relative to other production inputs and producing more output profitably. Firms and individuals benefit from cheaper and more abundant products and services, causing them to find many more uses for these (and the energy they contain). A more efficient steel plant, for example, produces cheaper steel that, in turn, allows firms and individuals to afford to find more uses for the same material.

While some find the notion that increased energy efficiency increases energy consumption to be counter-intuitive, the economic theory is remarkably commonsensical. Mills claims that the idea that the rebound effect "has been postulated in theory but never shown empirically to be significant" is not the case. After many years, rebound theory has advanced to the point that it is now a reliable foundation for empirical study and the empirical evidence firmly suggests rebound exists. And remember that the "rebound effect" for other factors of production is expected, even welcomed; economists have long expected labor productivity improvements to drive even greater economic activity, for example, thus increasing demand for labor and creating new employment opportunities in the economy as a whole, even as efficient production may eliminate a handful of jobs at one factory.

The Implications of Rebound

There are significant potential implications of high levels of rebound. One is that greater energy efficiency may be a net positive in increasing economic productivity and growth but should not be relied upon as a way to reduce energy consumption and thus greenhouse gas emissions. Particularly in a world where many billions lack sufficient access to modern energy services, efficient technologies such as solid-state lighting may be central to uplifting human dignity and improving quality of life through much of the world. One might even argue that energy efficiency is still important from a climate perspective, because when efficiency leads to greater economic growth, societies will be better able and more willing to invest in more expensive but cleaner energy sources. But in this way energy efficiency is no different from other strategies for increasing economic growth. What should be reconsidered is the assumption that energy efficiency results in a direct, net decrease in aggregate energy consumption when there is a growing body of research suggesting the opposite.

Dr. Harry Saunders has a B.S. in Physics from the University of Alberta, an M.S. in Resources Planning from the University of Calgary, and a Ph.D. in Engineering-Economic Systems from Stanford University. Saunders coined the "Khazzoom-Brookes Postulate" in 1992 to describe macro-economic theories of energy rebound, and has published widely on energy economics, evolutionary biology, and legal theory. He can be reached at: hsaunders@decisionprocessesinc.com.


  1. Points to why we need efficiency AND conservation - e.g., use CFLs AND turn them off when you are not in the room, drive a hybrid, but take transit or walk or bike when you can, etc.

  2. Bill, even if one takes those steps, they are likely to use the money saved from their conservation to purchase things that require energy. Basically, the only option to reduce energy usage is to make it more expensive. My concern is if making energy more expensive is more damaging long term than the emissions being released due to energy usage.

  3. "A residential washing machine may be energy efficient in terms of function, but in terms of production, the metal body alone requires energy to mine, smelt, stamp, coat, assemble and transport it to a dealer showroom and eventually a residential home. The energy embedded in your washing machine, or just about any product or service you consume, is very large."

    "Very large" is not really a meaningful term. (As the old joke goes, "How's your wife?" "Compared to what?")

    Here is a website that discusses the emboddied energy in the steel in a washing machine:


    Per that website:

    1) A washing machine uses about 99 kWh per year,

    2) A washing machine contains about 33 pounds of steel, and

    3) A pound of steel requires about 5 kWh to produce (note that he initially started with a value of 25 kWh per pound, but decided that was wrong).

    Therefore, per that website, the emboddied energy of the steel in a washing machine is equal to 33 pounds of steel times 5 kWh per pound = 165 kWh. If the machine consumes 99 kWh per year, that means the emboddied energy of the steel is equal to 1.7 years of energy consumption.

    Is that "very large"? Well, "compared to what?"

    P.S. None of this in any way detracts from the main point of your post. I just like to nitpick.

    P.P.S. Probably a more definitive site on the emboddied energy in steel is this:


    It has the emboddied energy in recycled and primary steel as 8.9 and 32 MJ/kg, respectively. That would be 2.5 and 8.9 kWh per kg, or 1.1 and 4.0 kWh per pound, respectively. Using the average of those two would be about 2.6 kWh per pound. So then the steel in a washing machine would have an emboddied energy of 33 x 2.6 = 86 kWh...a little less than a year's energy use.

  4. It's telling that there is no mention of automobiles here, where IIRC the production to consumption energy ratio over the lifetime of the vehicle is on the order of 1:10, and where long-term rebound effects are nowhere near as signicant as Dr. Saunders seems to suggest using other examples.

    Another point that seems to be glossed over and which is worth mentioning is that there is a diminishing returns element here that typically limits rebound effects. for example, it doesn't follow that as dishwashers become twice as efficient that you'll run them twice as often (the limiting factor here being the number of dirty dishes you produce).

    On the other hand, if it now costs less to run your diswahsher and drive your car, then maybe you'll be more likely to buy a beer fridge or an extra large plasma TV. IOW, money=energy, and money/energy saved in one place will simply be used somewhere else.

    all of which of course seems to support some form of carbon price to complement energy efficiency standards....

  5. FYI

    A good discussion of the rebound effect can be found here:


  6. David Leonhardt of the NY Times periodically has charts that show how what percentage of our earnings we spend on certain things such as housiing, clothing, energy, education, healthcare, etc. What is remarkable is how much some of the proportions have changes with things like clothing dropping from 20% of income more than 30 years ago to less than 5% now. The cost of Housing has stayed relatively constant at 20-25% (but that may reflect averaging in a lot of older home owners with low mortgages with young home owners with high expenses). One of the areas of the home budget consuming an ever larger portion of earnings is health care. It was ~5% in the early 70's and exceeds 20% now. Considering that incomes are falling in real terms yet the cost of health care continues to rise at an anual clip of 10-15% per year, it doesn't take a rocket scientist to realize that any money/energy saved is going to be allocated to the health care industry. So the government with the assistance of a grossly innefficient health delivery system, obviates the need to charge us more for carbon based energy because few can afford its discretionary consumption.

  7. This points up how increasing fuel efficiency standards for autos never seem to deliver actual reductions in fuel consumption.

    Why move 20 miles closer to that new job when the fuel cost is so little? Why give up the flexibility and time savings of solo commuting, for the car pool when the cost savings of pooling is so low? Plus you don't have to sit next to Mr. Stinky.

    The bottom line is increased efficiency leads to increased utilization.

  8. Sean,

    I think your point about increasing expenditure on health care is a good one and it's something that would seem to minimize the indirect rebound effect as it's not a particularly energy-intensive service...

  9. This has hugely important (to use a terribly non-descript term) implications on the economics of climate change. So, essentially what this article is saying is that all of the positive things that people generally cite in favor of "greening" the economy will, in fact, cut against the overall objective of reducing impacts of GHGs on the climate. Wow.

    I once heard mention of a theory of low-growth economics. Is it possible to conceive of a first world economy that is sustainable at, say, less than 1% GDP growth rate.

    Put another way, vis-a-vis Sean's comment, if the economy shifted to the point that our expenses were largely concentrated on things like services, experiences, or low-impact consumption (rather than durable and disposable goods), then the economy might continue to grow without the adverse effects of energy consumption.

  10. My point about healthcare is two fold. Certainly the obvious one is that the amount of discresionary income for the average working class American has gone down and is getting smaller as income stagnates and expenses rise. Health insurance costs for hourly workers in our company is a quarter of total compensation and rising. Compensation increases mostly end up going to health insurance providers and little goes to take home wages. To think its important to increase the cost of carbon to reduce fossile energy consumption does not recognize the economic pressure most working class people are under these days. Who are we to try and tell these people what consumption is appropriate and what consumption is not? Harsh realities of this economy is doing that already.

  11. This is hardly new. Jevons propounded this theory in 1865 (http://en.wikipedia.org/wiki/Jevons_paradox).

    There does come a point, however, where (as other commenters have pointed out above) a wealthy society starts to choose consumption patterns that are less energy-intensive. For example, rich people buy hand crafted wood furniture rather than factory mass-produced stuff, or they choose to employ people to perform services for them (ie invest in their own leisure time) rather than buying more stuff.

  12. "...it doesn't follow that as dishwashers become twice as efficient that you'll run them twice as often (the limiting factor here being the number of dirty dishes you produce)."

    Not always by any means. If its cheap to run, running it daily rather than waiting three or four days for a full load becomes a more attractive option for some.

  13. The rebound idea is far from new. In economics it is known as the Jevons' paradox and can be dated back to 1865. See http://en.wikipedia.org/wiki/Jevons'_paradox

    Crucial quite is "1865, the English economist William Stanley Jevons observed that technological improvements that increased the efficiency of coal use led to the increased consumption of coal in a wide range of industries. He argued that, contrary to common intuition, technological improvements could not be relied upon to reduce fuel consumption"