September 28, 2010
Six Misconceptions About Rebound and Backfire
David Owen's recent New Yorker article on energy efficiency and rebound phenomena has sparked a lively debate in a number of blog posts but it has also resulted in some confusion. The purpose of this post is to lend clarity to the rebound debate by dispelling some misconceptions about rebound.
Rebound comes from several sources, requiring important distinctions. Typically, rebound analysts distinguish consumer-side effects from producer-side effects. A second distinction frequently called upon is between so-called direct and indirect rebound. On top of these rebound classifications, some analysts identify a so-called macroeconomic effect.
The direct rebound effect on the consumer side arises because an energy efficiency gain reduces the effective price of energy, causing consumers to use more of it. An example is the installation of more efficient household heating or cooling that causes the household to heat or cool more rooms.
The direct rebound effect on the producer side is subtle but measurable. By reducing the effective price of energy, producers are potentially able to increase their output. More importantly, energy efficiency gains, by again reducing the effective price of energy, allow firms to reconfigure the inputs of new capacity to use more energy profitably. This effect is dependent on the flexibility of the new productive capacity to reconfigure. Roughly speaking, the greater the flexibility, the greater the rebound from the direct effect.
Magnitude-wise, rebound from the direct effect on the producer side appears to be substantial. My latest research [Reference #1], covering thirty US sectors, shows direct rebound magnitudes to be large. In this analysis, historical rebound magnitudes over the period 1990-2000 show as being 62 percent for all thirty sectors combined.
The indirect effect on the consumer side arises from consumers taking the money saved from, say, driving a more efficient car, and spending it to purchase other goods and services that require energy to provide.
On the producer side, the indirect effect is the close analog of this. Energy efficiency gains reduce the output price of products, some more than others, causing consumers to shift toward them, carrying with this shift the "embedded" energy required to meet these demands. This is a very difficult effect to measure at this point in methodology development, although general equilibrium rebound analysts are making good progress.
The macroeconomic effects are the result of the aggregate impacts of widespread energy efficiency improvements, which combine to drive several macroeconomic mechanisms that also contribute to rebound. For example, decreases in the effective energy price due to efficiency gains can reduce the price of intermediate goods and services leading to complex economy-wide adjustments in energy use. Again, these effects are very hard to directly measure, but general equilibrium models and integrated modeling analysis have been used to study these macroeconomic dynamics. One recent study out of the UK [Barker et al. 2009, Reference #2] found that at this macroeconomic scale, rebound effects would erode more than 50 percent of the projected energy savings due to a series of worldwide, 'no regrets' energy efficiency measures recommended by the International Energy Agency and IPCC. And let's keep in mind, when it comes to mitigating climate change, the total, economy-wide, impact of rebound effects is of primary interest.
With these definitions in mind, let's now turn our attention to some of the common misconceptions surrounding energy and backfire.
Misconception #1: End use energy consumption is all that matters
An evident source of confusion in much that is written has to do with a failure to distinguish between two types of energy use.
Many of us, when contemplating the potential for reduced energy use, quite naturally reference our thoughts around those opportunities we see around us in our personal realm--energy used in the household or for private transportation. But this "end use" energy consumption represents only a relatively small fraction of the energy we actually consume. Globally, some two-thirds of all energy is consumed in the energy used to produce the goods and services we consume. Your washing machine may be very efficient in its use of energy, but think of the metal body alone and the energy required to mine, smelt, stamp, coat, assemble, and transport it to the dealer showroom where you bought the appliance. The energy "embedded" in your washing machine is substantial. The same is true for any product you purchase or service you consume.
Rebound effects may differ significantly between end use energy consumption and the energy used in production. Unfortunately, rebound effects in the much larger productive part of the energy economy are substantially more difficult, technically, to measure, and more difficult to discern. This critical distinction between the two types of energy use is nearly always lost in the political realm, and quite frequently lost even among energy analysts.
Misconception #2: Rebound must be small because we can't re-spend all energy efficiency savings on energy
A further confusion arises from analysts equating rebound effects to "re-spending" effects.
Michael Levi of the Council on Foreign Relations protests that the money saved from introducing energy efficiency cannot possibly create backfire, or even significant rebound, because the money re-spent will not be re-spent entirely on energy. While Levi might be correct that re-spending may not produce significant rebound, he unfortunately neglects the larger rebound picture. As discussed earlier, rebound occurs directly and indirectly. By focusing on re-spending, Levi is ignoring the effects of both direct rebound and further macroeconomic impacts of widespread efficiency improvements.
But Levi also may be slightly underestimating the effects of indirect rebound. As Druckman, et al. [Reference #3 Druckman, et al. forthcoming: working version available from the authors] have shown, what matters to climate change is not the energy intensity of the goods and services on which saved money is re-spent, but their carbon intensity. They show from UK data that "carbon rebound" from the re-spending component can be very large, potentially even leading to backfire from the re-spending source alone.
Clearly, all manifestations of rebound must be considered for a complete and proper analysis of the problem.
Misconception #3: The Income/output component of rebound must be small
A further categorization: The direct rebound effect is usefully decomposed into an "income/output" component and a "substitution" component. The substitution component of direct rebound on the producer side is fundamentally a measure of the flexibility producers have to substitute among inputs when they invest in new capacity. Analytically, this flexibility is formally characterized using the "substitution elasticities" of microeconomics. (The analog on the direct consumer side is capital or labor substituting for energy--e.g., insulation for heating, bicycling for driving--or the reverse.)
But it is the output component (whose analog on the consumer side is the "income" component) that seems to engender the greatest confusion.
James Barrett's recent post on rebound argues that the output effect is insubstantial. But Barrett may underestimate the extent of the output effect. The output effect comes not from energy cost reductions increasing profits, and "plow[ing] increased profits back into making product," to quote Barrett, but from the fact that energy efficiency increases are, in microeconomic terms, technology gains that expand the space of profitable production possibilities for investors. Energy efficiency gains make higher output levels profitable, irrespective of the source of capital funding.
To be fair to Barrett, rebound economists have long believed the output component to be relatively small. And when measured in the right frame, my analysis cited above [Reference #1] shows it to indeed be small in the short term--on average about 4 percent of the combined output/substitution effect. However, this component is dynamic--output gains are self-reinforcing. Over a twenty-year period, it grows to contribute as much as 20 percent to rebound. When considering climate change remedies, the time horizon needs to be decades. It would be imprudent to treat the output effect as something safely ignored.
Furthermore, while income/output effects may be relatively small, the substitution component of direct rebounds can be far larger. In my analysis of direct rebounds in thirty U.S. producing sectors, substitution effects account for more than half of long-term rebound in all but two sectors, and accounts for more than 70 percent of rebound in all but four. Clearly, a wider view of rebound effects is again warranted.
Misconception #4: Energy is being "decoupled" from the economy
This misconception is related to the "re-spending" misconception.
David Owen quotes Lee Schipper as saying, "In the end, the impact of rebound is small, in my view, for one very key reason: energy is a small share of the economy." The argument goes like this: since energy is such a small share of the economy, energy efficiency gains cannot save enough money to be released on significant magnitudes of other energy-consuming economic activity. As energy efficiency improves still further, the link between energy and economic activity diminishes.
But as we have seen, this re-spending effect is only one component of rebound, and possibly a small contributor in relative terms.
Furthermore, the link between energy and economic activity is actually very tight. Virtually any economic activity imaginable requires some level of physical energy as an irreplaceable input. And in fact, the lower the share of energy, the tighter the link becomes. Theoretical considerations indicate that the lower the energy share, the more crucial it becomes in the production of economic output (the higher is its "marginal productivity," in microeconomics-speak). As Owen explained, the outsized importance of energy to the economy is quite clear "if you imagine eliminating primary energy from the world. If you do that, you don't end up losing 'between six and eight per cent' of current economy activity ... you lose almost everything we think of as modern life."
To be fair to Schipper and the "decoupling" school, theoretical considerations also indicate that rebound magnitudes will decline with declining energy share. But this is almost entirely due to the direct output effect on the producer side, which is the smaller component of the direct effect. The substitution effect is virtually independent of energy share. Flexibility of production means producers can significantly alter their energy consumption to more profitably produce goods and services. From a rebound perspective, then, energy is anything but decoupled from economic activity.
Misconception #5: Macro rebound effects are discernible by considering micro effects
This may be the subtlest misconception. Barrett makes the entirely sensible assertion that macro effects can be nothing other than "the sum of all the micro parts." Slam dunk assertion. But the difficulty here is in the summation process.
The subtlety of the macroeconomic component is perhaps best appreciated when thinking of it as an "emergent phenomenon" - a characterization coined in a forthcoming survey of the rebound literature from the Breakthrough Institute's Jesse Jenkins, Ted Nordhaus, and Michael Shellenberger, to be published in February [Editors' note: stay tuned, Breakthrough blog readers!]. Emergent phenomena are higher order effects resulting from the complex interaction of multifold individual components that combine through non-linear and reinforcing effects. In these cases, the effects emergent at scale are often quite difficult to discern simply by looking at the various constituent micro-scale components.
Rebound as an "emergent phenomenon" can be illustrated by reference to a recent article, of which I was a co-author [Tsao et al. Reference #4], on the historical record for lighting. This analysis showed that new applications in efficient lighting have, since the 1700s, offset the energy efficiency gains from new lighting technologies almost exactly, leaving the share of global GDP spent on lighting unchanged over hundreds of years and independent of major gains in "luminous efficacy." New lighting applications have continually arisen that offset energy consumption reductions due to energy efficiency gains, for more than 300 years, across three continents and across six technologies.
Such "emergent phenomena" can be thought of this way: energy efficiency gains can provide the basis for as yet unforeseen new energy-using applications, products, enterprises or even whole new industries. The "macro" effects of energy efficiency gains, while being the sum of all the "micro" effects, nonetheless must comprehend an exceedingly complex summation. In the lighting case, think of the new possibilities that have been opened by new lighting technologies that have enabled the production of other new energy-using products and services, and production of these in non-daylight hours. "Emergent phenomena" cannot be discerned simply by looking at micro effects in isolation.
Misconception #6: Efficiency gains happen only in energy
Efficiency gains are not limited to energy alone. Other factors of production (capital, labor, materials...) also experience efficiency gains. The problem is, efficiency gains for other factors also increase energy use. In my forthcoming analysis cited above, technology gains for all factors considered together lead to significant energy backfire in nearly every sector.
An example of this is the Primary Metal sector of the US economy. My analysis in the working paper cited above measures the average annual efficiency gain over the 1980-2000 period to be 2.46 percent for capital, 3.30 percent for labor, 2.90 percent for energy, and 0.53 percent for materials. This period was characterized by the aggressive introduction of electric arc furnaces for steel production that were highly efficient in the use of both capital and labor, in addition to energy. This is likely a major contributor to a calculated "all factor" energy rebound for the Primary Metal sector of 172 percent over this time period. By this analysis, the increased efficiency of other factors not only increased energy consumption in this sector, but created significant backfire - a rebound in energy demand greater than 100 percent.
The consequence of this phenomenon is rather profound. The problem is not so much that efficiency gains targeted at energy often also improve the efficiency of other factors (a feature of energy efficiency that analysts such as Amory Lovins cite as a key ancillary benefit); the real problem is that technology gains, considered together, increase energy consumption. Without these gains, energy consumption would be lower.
Analytically, this makes "teasing out" energy-specific rebound effects extra challenging. But the larger problem is that from a climate change perspective technology gains generally are a culprit in increasing energy use.
Rebound effects are subtle and complex. And they create some thorny issues for policy makers that will need to be addressed. Most importantly, energy analysts must cease utilizing a far-too-simple assumption that efficiency gains yield direct and linear reductions in energy use. The complex and varied economic phenomena known collectively as "rebound effects" mean that we cannot expect that improving the energy efficiency of steel production by 30 percent, for example, will yield a simple and direct 30 percent reduction in the energy consumed by the steel sector, let alone the economy as whole. Just as economists expect that gains in labor productivity will contribute to greater employment overall, not less, gains in energy productivity (aka energy efficiency) are not likely to be taken up simply as direct reductions in energy demand overall. Before these key issues can be sensibly addressed, it is important that we simply understand the many ways rebound operates. And, with that, that we continue work on developing tools to measure rebound, in its multiple manifestations.