Numerical Example of Rebound

From
Jesse Jenkins
Sent at 1/25/11, 11:32 AM (GMT-08:00).
Skip, Amory, Lee and all,
This email chain is quite revealing of the state of knowledge and debate on rebound effects in the United
States.
I'm surprised how quickly Amory, and even Skip and Lee (who I know have directly engaged with this
literature and taken it quite seriously in the past), move to dismiss the importance of rebound to energy
analysts and policy makers.
The three of you are among (if not the) leading energy efficiency analysts and advocates in the nation,
and collectively, your general response to Dr. Saunders' effort to clarify the terms of rebound effects and
guide the discussion to some of the literature has been to dismiss it as "hand waving" and yet another
reincarnation of a "a logical fallacy," coupled with relatively unrelated arguments that the U.S. economy
clearly has a long way to go to capture efficiency opportunities and trim waste and who knows what the
future will hold (see more on that in the post script).
What Amory classifies as mere "hand-waving" and a "logical fallacy" is actually a well developed field of
energy economics literature, a body of analysis including well defined economic theory, empirical
surveys, econometric study, and general equilibrium and integrated modeling analysis. There are dozens
if not hundreds of peer reviewed journal articles and studies, many commissioned by the UK government
in the past decade, and much of it rigorously developed since the last time American analysts seriously
engaged with rebound effects circa 2000 and before that around 1990.
Steve Sorrell (also CCed here), led an extensive review of the literature on rebound for the UK Energy
Research Commission in 2007, which you can find here: http://www.ukerc.ac.uk/support/tikiindex.php?page=0710ReboundEffects
Around the same time, scholars at several academic institutions in the UK have significantly advanced
the state of the art in modeling and analysis of macroeconomic and even global rebound effects, including
Karen Turner, Terry Barker and their colleagues (who I've added to this email exchange). Meanwhile, Dr.
Saunders has developed a potentially ground-breaking and first of its kind econometrically-dervied
analysis of direct rebound effects in the producing sectors of the U.S. economy, an area that has received
incredibly little study to date. All of this builds on prior work examining direct rebounds in end-use sectors
of developed economies, and many CGE modeling analyses of macroeconomic and indirect rebound
effects.
To call all of this extensive academic research mere "hand waving" and dismiss rebound effects out of
hand strikes me as remarkable -- and given the implications for climate mitigation strategy, is a stance
that would be a quite irresponsible to maintain.
I say that not to snipe, but rather because I know all of us on this email chain take the climate mitigation
challenge quite seriously, and are all committed to a high degree of analytical rigor. As Harry says, it's
time for us all to "roll up our collective sleeves" and finally take a deep dive into the issue.
In that vein, along with my colleagues Michael and Ted, I have led a new, comprehensive review of the
extensive rebound literature, which will be the first published in the U.S. and will build on the great work
done by UKERC to incorporate the fast evolving state of the art in the field over the past four years since
Mr. Sorrell's review was published. Breakthrough Institute will release the review in February in just a
couple short weeks.
As a preview, the overall conclusions of the energy economics and rebound literature are clear and have
strong implications for climate mitigation strategy:
• Direct rebound effects in end-use consuming sectors are typically on the scale of 10-30% before any
indirect or macroeconomic effects are considered. As Dr. Saunders work illustrates, direct
rebounds can be far higher in the producing sectors of the economy, with his study finding a
weighted average of 62% rebound across thirty producing sectors of the US economy. Rebound
effects are also likely to be far higher in the developing world, where demand for energy is largely
unmet, elasticities are far higher than in rich nations, and quality energy services are a key
constraint on economic activity. As Lee Schipper and his colleague Michael Grubb noted in a
2000 Energy Journal article, "the shadow of Jevons lurks" in developing nations. There has been
frightfully little study of rebound effects in developing nations, where conditions look far more like
the industrial revolution Western nations of Jevons' day, and where the IEA projects that more
than 90% of projected energy demand growth will occur in the next 20 years.
• Indirect rebounds erode more on top of this, although getting a precise measure is tough. It's perhaps
best estimated along with macroeconomic effects at a larger scale of analysis. Barker and
colleagues estimate that indirect effects can erode an additional 7-25% of projected energy
savings, for example, with a UK-economy-wide weighted average of 11%. Rebounds again could
be higher in developing nations that developed economies, and larger still when a global scope of
analysis is considered. With much of the energy content of our goods and services in the Western
world now imported from overseas, a thorough analysis of indirect effects must necessarily be
global and macroeconomic in scope.
• At a macroeconomic scale, numerous CGE models project that combined, economy-wide rebound
effects at the scope of a national economy (so again ignoring imported energy) typically project
rebounds greater than 50%, and in a surprising number of studies, project "backfire," or rebound
greater than 100%.
• The excellent work of Barker et al, which everyone here should familiarize themselves with (attached),
is the first attempt at a rigorous, integrative modeling analysis of rebound effects at a global,
macroeconomic scale (integrative meaning integrating detailed bottoms up energy-sector
modeling with more generalized economic techniques for the economy as whole). They find that if
the world were to pursue the full basket of "no regrets" cost effective energy efficiency efforts
outlined by the IEA and IPCC, the combined impact of rebound effects would erode more than
half (52%) of projected energy savings (and thus climate mitigation benefits) by 2030. That
finding alone is stunning, and should be a wake-up call to all of us. Yet I find there are several
reasons to believe this may actually under-estimate total rebounds. Lacking any solid estimate of
direct rebounds in the developing world or in producing sectors, the authors utilize very
conservative estimates, even assuming no direct rebound in energy intensive industrial sectors
and very low values in agriculture (which Harry's paper shows have some of the largest direct
rebounds) and the same values for direct rebound in developing economies as the lower values
found in more mature, developed economies. Considering these and other limitations, global,
economy-wide rebound may be quite larger, eroding the majority of expected energy savings,
even if not leading to backfire (100% rebound) as a norm.
• In the end, any of these analysis will also find it impossible to project in advance what I've dubbed
"frontier effects" that result when efficiency gains unlock whole new areas of the production
possibility frontier, leading to potentially vast new markets, or even whole new industries for
energy services. This is the dynamic driving the backfire in energy demand that Jevons famously
comments on, as tremendous gains in steam engine efficiency enabled enormous areas of new
and productive uses for steam engines and thus coal consumption, sparking entire new industries
and activities, from railroads to steel consumption and the core of the industrial revolution itself.
Steve Sorrel correctly notes that this dynamic is most likely for 'general purpose' technologies
from lighting to lasers to microchips and computing, to steel and other metals production, as well
as electricity generation itself, and here is when backfire is most likely. These effects are
extremely hard to predicts in advance, yet their frequency in economic history should give us
pause.
• Furthermore, if efficiency improvements lead to gains in other factors of productivity along with energy - e.g. labor or capital productivity gains -- rebound effects can be compounded and multiplied. As
Amory and Skip and other efficiency advocates often (and correctly) note, improvements in
energy efficiency often come with improvements in labor or capital productivity as well, as
efficient day-lit buildings boost labor productivity in offices, or better lighting or electric motors can
drive big boosts in productivity on the factory floor, for example. Here again, much larger
rebounds are possible.
The core conclusions for energy analysis and climate mitigation strategy is this: we can no longer afford
to assume that a simple, direct relationship between energy efficiency gains and declines in energy
consumption.
To say that we can improve U.S. energy efficiency by 30% and thus cut U.S. energy use by 30% from
some baseline is overly simplified and incorrect. The economy is far more dynamic than that, and even if
backfire is rare (and it may not be as rare as we think), the clearly significant scale of total rebound effects
renders the relationship between energy efficiency/productivity gains far from simple and direct.
Not since the days of the Luddites have we worried that improvements in labor productivity would lead to
widespread unemployment and direct reductions in demand for labor at a macroeconomic level. Rather,
economists and policy makers widely expect labor productivity (aka efficiency) gains to drive overall
employment by spurring economic growth and encouraging substitution effects that drive up employment
opportunities overall.
As Skip notes, there are ways in which rebound may be mitigated (e.g. by increasing the price of fuels to
counteract some of the decrease in implicit price of energy services driven by rebound), but to begin with,
we must clearly understand the scale and impact of rebound effects, the differences in sector and scope
of analysis, and the forms of efficiency improvement most likely to drive backfire. And energy forecasts
must explicitly model and account for rebound effects, or we should consider them suspect and
potentially misleading in our efforts to formulate effective and urgently needed climate mitigation
strategies.
The time has come to get clear on the many indirect and complex relationships between energy
productivity or efficiency and overall energy demand as well. To do that, we must grapple honestly and
clearly with rebound effects.
I look forward to further serious discussion and friendly debate.
Jesse Jenkins
Director of Energy and Climate Policy
Breakthrough Institute
P.S. I hope that we are all clear that no one is claiming that the world has run out of ways to improve the
productivity of energy or avoid waste, or to argue anything but that we should vigorously pursue such
opportunities wherever they are a net boon to the economy.
The question of interest is what net contribution those efficiency or productivity gains will have on total
global energy consumption (particularly of fossil fuels) and on global greenhouse gas emissions.
I would also note however, that defining all energy lost in conversion processes as "waste" (as in noting
that we currently "waste" 83% of all energy used) is a bit over-simplified in thinking. We routinely 'refine'
or 'upgrade' low-quality energy sources into higher quality and more valuable energy forms all the time.
We "lose" BTUs along each step of the refinement process, yet the end product is of extremely high value
compared to the initial BTUs. Mark Mills, (CCed here), argues convincingly in his book (with Peter Huber)
that while only an extremely small portion of primary energy ends up as useful energy at the tip of an
industrial laser, we would nonetheless be loath to classify all that energy used to refine diffuse primary
energy into an extremely focused and potent laser as "waste."
Yes, there are ways to improve the efficacy of each stage of the conversion process that should be
pursued, and in many cases, we can reuse some of the heat produced in the process for the lower quality
energy needs out there (e.g. direct demand for heat). But let's think twice before we consider all of these
BTUs as "waste." After all, what good is a BTU of primary energy? It only becomes useful once we
convert it to higher order and infinitely more useful purposes, from powering devices to etching silicon to
curing cataracts, a process that necessarily involves "wasted" electrons as we refine and upgrade energy
each step of the way. As the economy advances and we find even more uses for "highly refined energy"
we may even expect the portion of primary energy lost as heat to increase over time, and that would be a
good thing, not something to fear...
From
Amory B. Lovins
1/25/11, Sent at 12:01 PM (GMT-07:00)
It's great to have so many smart people on this thread, and I hope this discussion will help get us all
closer to understanding what's true. I'm indeed acquainted with the UKERC, Barker, and some other
leading recent studies, and with the arguments, Jesse, that you helpfully summarize below. There is no
dearth of evidence about correlation. What I'm trying to understand is evidence about causality. Some
concrete examples in English might help those of us who are having trouble understanding the causality
claims to understand their logic better than economic equations (which show only correlation) can reveal.
Thanks -- ABL
Amory B. Lovins
Chairman and Chief Scientist
Rocky Mountain Institute
from
Jonathan Koomey
1/25/11, Sent at 3:27 PM (GMT-08:00).
To be clear for the less technical folks in the audience, there is no consensus (or any strong empirical
evidence that I've seen) that indicates that rebound is large across the entire economy. There is also no
consensus with Roger's statement that "rebound doesn't matter for stabilization one way or the other". If
you have real efficiency gains, that means you need fewer expensive low carbon energy supplies. If you
can't count on those efficiency gains to actually reduce demand, then you need many more of the
expensive supply side resources. So whether rebound is big or small affects the cost and feasibility of
achieving climate stabilization.
There are models and correlations that sometimes indicate that rebound may be significant, but no one
has to my satisfaction explained the causal links that would lead to that result except in a small minority of
special cases (like data centers, where the delivery of computing services is often explicitly limited by a
power constraint in these facilities and power-related costs are a very large fraction of total costs). So as
Amory points out, finding that there is some high-level correlation over time between energy efficiency
improvements and GDP gains is not sufficient to prove that there is rebound. What is needed is a clear
causal explanation of how it is possible for companies and consumers for whom energy is a tiny part of
their total costs (many of whom also are subject to the principal agent problem or have devices that don't
change their energy use in response to user behavior) to have significant rebound effects. Absent such
causal links these claims are indeed all just "hand waving" (or "model waving" if you prefer).
And I still don't understand Harry's statement in the paragraph below:
Jon: Skip has this one about right. Energy efficiency gains affect the
whole space of production possibilities, including the use of other factors
of production. Even if energy is a small component, cost-wise, of
production, think of how a doubling, say, of "effective energy" (cf. note
above to Michael) might cause the "optimal scale" of production to expand
when all factors are considered. I see such scale effects routinely among my
industrial business clients. It's not that the energy cost savings are all
that's available for investing in new capacity; it's that the whole space of
production possibilities is altered.
Perhaps a specific example would be helpful here, because it still isn't clear to me how affecting a minor
part of a company's total cost can have such an effect on the production possibility frontier, unless you
are explicitly linking other benefits of a technology change to energy efficiency, but that's not an
appropriate thing to do when assessing rebound. It is only the increase in energy efficiency that is directly
attributable to efficiency (and ONLY efficiency) that is rebound. Everything else is, well, everything else.
Here's one example that comes to mind (I'm adding Evan Mills to the cc: list, because he's the real expert
on this technology). Suppose, for example, that an LED lamp is substantially more efficient, more flexible
in use, less costly, and much longer lived than conventional alternatives, and offers the possibility of
doing things we couldn't do before with conventional lighting technologies at reasonable cost. If there is
an explosion of new uses for LED lighting, some small part of that increase in use is attributable to
efficiency, but the main driver of the increased use is the VALUE of the services delivered, which is tied to
all those other attributes of these lamps. The constraint is not energy use in that case, it's the fact that
before the advent of this technology we weren't able to do the things we're able to do with LED lamps,
and people want those services. You can't then say that any increases in electricity demand for lighting
with LED lamps is due to rebound, because the causal link to efficiency ALONE is not clear (and there
are many examples, like my 6 W LED bedside lamp that replaced a 40 W incandescent lamp, where
rebound is nonexistent).
So it would help those of us who are skeptical of significant rebound effects to hear from those who are
less skeptical about how they can address the causal linkages that I describe above, most likely by giving
some specific examples. That may be a path towards resolution of what misunderstandings exist among
the group.
Thanks to all for an interesting discussion.
Jon
From
Jesse Jenkins
Sent at 1/26/11, 7:39 PM (GMT-08:00).
Dear all,
As promised, in answer to Jon's challenge to show some causal chain driving rebound, I've drafted this
simplified example. I've made up all of the values herein, but I think they aren't too far off, nor far fetched,
and should serve our illustrative purposes here. Actual values could easily be higher or lower in some
sector or another, as would the resulting rebound effects. I hope, as Jon said, this example will facilitate
'true understanding' in this discourse.
My apologies for the length of what follows -- Jon, Amory, you asked for it after all! ;-)
OK. Let's take Dow Chemical as an example, since that's one of Amory's examples in a prior email, and I
believe, one of his clients. Let's assume that a few dozen energy saving projects that Dow pursues across
it's US plants reduce the energy consumed in their processes and facilities by a total of 30%, so that if
initial energy use was 100 units (the units don't really matter, so we'll use generics), it is now expected to
be 70 units, based on engineering-level estimates that Amory has calculated for his client.
Now let's start in with the rebound effects, beginning with the direct, and moving to the indirect.
First, we'll consider the substitution component of direct rebound. Let's assume that substitution
elasticities between energy services -- say heat or electricity in the production of chemical feedstocks and
final products -- and labor or capital or other materials is about 0.5 on average. Greening et al. 2000 (in
the EP issue Lee edited) reports that typical substitution elasticities for firms often range from 0.4 to 0.8
and in some cases closer to 1.0, but we'll consider a relatively conservative value from that range of 0.5.
So as the price of energy services in Dow processes and facilities falls by 30%, plant managers over time
rely more heavily on now-cheaper energy and less on other inputs -- materials, labor, capital. I'm not
familiar with Dow's plants, so I can't give clear specifics on what these substitutions would be, but for sake
of conjecture, let's say they new insulation on chemical vats now make it cost effective for plant managers
to rely on higher temperature processes that reduce materials needs, or perform chemical reactions
faster, thus reducing labor demand. There are probably dozens of small adjustments like this over time,
which yield the aggregate substitution elasticities typically observed.
So a 30% energy efficiency improvement yields a 30% reduction in the implicit price of energy services,
so with a substation elasticity of 0.5 that means that Dow increases their consumption of energy services
by 15% following the efficiency gains, or an increase in raw energy demand of 70 energy units*15%=10.5
units, or a direct rebound due to the substitution component of the rebound effect of 10.5/30 =
35%. The firm now consumes 80.5 units of energy, not 70.
That's the substitution component of the direct rebound effect alone. As Harry has noted in his studies,
this is typically the much larger component of rebound. But let's now consider the second component of
direct rebound, the output component.
For this, we'll assume that energy makes up 15% of total costs for Dow. This is an energy intensive sector
-- chemicals -- so we'll assume something a bit higher than the 6-10% of costs that may be more typical,
but not too high (and after all, energy intensive sectors are where we often look hardest for efficiency
opportunities). The efficiency gains and the cost-effective substitution of now-cheaper energy services for
other inputs to production (which will still yield net cost savings) thus reduce Dow's bottom line by
15%*30%=4.5%. The management then decides to chose to pass this savings on in the form of lower
prices of their products, undercutting their competition and driving greater demand for their products. I
don't know what elasticities of demand are for chemical inputs, but I assume they are fairly elastic, so let's
assume it's 0.8. That means demand for Dow's products rises 0.8*4.5% = 3.6%, driving up demand for all
factors of production, including energy services, by 3.6%. So that's another 80.5*3.6% = 2.9 units, or a
rebound from the output component of direct rebound of 2.9/30 = 9.7%. (Note, we won't even get
into the energy required to make and support the other factors of production that are now in greater
demand; that's an indirect effect probably best looked at at a macroeconomic scale, as the economy
adjusts in a series of complex chain reactions to this one change in demand at the chemical plants… we'll
keep it simpler for now).
So now, total rebound from direct effects -- substitution and output effects together -- is 44.7%.
Let's round to 45% for ease. The firm now consumes 100-30+10.5+2.9 = 83.4 units of energy, for a
16.6% net reduction in energy demand, not the 30% reduction you'd expect from summing up the initial
engineering-level energy savings.
By coincidence, this is close in magnitude to the values Harry calculates for direct rebounds in the US
chemicals sector writ large in the paper he's circulated (53% over the long term). And as a back of the
envelope example illustrating the clear mechanisms at work and the simplest of microeconomic
elasticities shows how the effects can tally up, and how results like those Harry (and a dozen other
scholars of rebound effects) are not far fetched at all.
But this is just the direct effects. Let's consider the indirect rebound mechanisms as well. The simplest
could be the re-spending/re-investment effect.
Since Dow still reduces energy demand by 16.6% their total costs will fall by 16.6%*15% (energy's share
of costs) = 2.5%. That's a 2.5% increase in corporate profits, shared amongst investors who re-spend and
re-invest that money and recirculates throughout the economy (I'm not even considering how much more
the increased volume they are now selling may increase profits, just the raw energy cost savings).
If energy was 15% of total Dow costs (before the efficiency gains), then total costs, measured in the
monetary equivalent of energy units, is 100/15% = 667 units (all economic activity is calories or BTUs at
its core, so maybe this conceit is excusable, as it avoids having to think about specific monetary units).
Then that savings amounts to 667*2.5% = 16.7 units of economic value (measured in equivalent energy
units). If as a whole, energy makes up 8% of economic inputs (economy-wide), then that 11.7 units of
economic value re-circulating in the economy will increase energy demand by 16.7*8% = about 1.3 units.
Not huge, but that's another 1.3/30 = 4.33% rebound due to this indirect re-spending/re-investment
effect. Total rebound is now 49.3%.
Now let's consider another indirect mechanism, the embodied energy effect: the energy efficiency
upgrades themselves required energy to produce. Let's assume that the suite of upgrades Dow employs
have on average a 3 year payback period, so for every dollar they cost, they save 1/3rd of that amount
each year. And they have an average life time of 15 years before replacement (some things like lighting
being shorter, and others like boilers maybe a bit longer; again, I'm just making these up for illustration).
That means in their lifetime, the combined efficiency measures save 15*(1/3) = about 5 times more in
energy costs than the upfront costs for the upgrades. A nice return. But still, that means that if we invert
that ratio, for every 100 units of energy costs saved, the up-front cost (in economic value terms) was 20
units (1/5 = .2). Then if we again assume that energy makes up about 8% of the cost of these units of
economic value, then that makes 1.6 units of energy consumed in making and installing the efficiency
upgrades for every 100 units they save. Since the upgrades saved Dow 30 units of energy per year (at
the engineering-level, where these returns are calculated), then that's another 30*1.6% = 0.5 units of
indirectly increased energy consumption. In other words, another rebound due to embodied energy
effects of 0.5/30 = 1.7%. A small addition, but total rebound is now over half, at 51%.
So total rebound due to indirect and direct effects alone -- considering nothing that may be emergent at
more macroeconomic scales, and no economic "model waving" -- would erode half of the projected
savings in total energy consumption projected by the kind of engineering level analysis that firms like
Amory's would show to clients like Dow. Dow still saves a lot of money and sells a lot more of their
product, so they're happy, and the economy at large is too. But the overall climate and energy demand
reduction benefits are oversold by 2:1 if we fail to consider these rebound effects -- that is, the efficiency
savings only cut energy demand, and thus emissions, by just half as much as initially promised -- and
that's what matters from a climate mitigation perspective.
Furthermore, while this is the end of my simplified example illustrating the chain of causal mechanisms at
work in the term 'rebound effect,' that's not the end of the rebound story. To get at the full scale of
rebound effects at the scope that really matters to climate mitigation concerns -- that is the global scope -does indeed require some economic modeling.
Jon may want to reject the entire field of economic analysis, and after the financial crisis, we certainly can
all certainly be forgiven for taking economic modeling with a little grain of salt...
But the economy is a complicated beast. Like the global climate, we are forced to rely on models to try to
approximate and examine the macroeconomic and often emergent phenomena that make up our
complex, interconnected global economy.
And yes, to get these models working, economic analysts must rely on observed and measured
correlations for many factors as they develop models that use this correlative relationships to produce
outputs that as closely as possible match the real world.
Yes, correlation does not equal causality. But I've shown you some clear causal mechanisms here at a
micro scale. And the correlated factors chosen to do macroeconomic modeling will have some clearly
plausible causal connections underlying them as well -- economists won't use the number of pirate
attacks annually as an independent variable driving the dependent variable of employment in the U.S.
chemical sector, after all!
Unfortunately, if we care about things like climate change or the impacts of efficiency in a complex global
economy, these are the murky waters we must wade into if we hope to move beyond the simple and
microscale effects sketched above.
This is an epistemological and analytical challenge, sure, but it is also a set of limitations and abstractions
that we routinely accept as a means of informing policy making and as we seek greater understanding of
complex macro-scale phenomena, from the global economy to global climate patters to neural networks
to ecosystems.
Jon's rejection of dozens of peer reviewed studies published in numerous reputed journals of energy
economics and policy (including by several analysts on this email chain), thus strikes me as somewhat
akin to climate skeptics rejecting all climate modeling work out of hand, simply because it relies on
models that depend on observed correlations between atmospheric CO2 levels and global temperatures.
Some skepticism of model outputs is warranted, surely, but outright rejection? Far from it. And there's
clearly a casual relationship between the two correlated factors, even if we can only rely on observed
correlations to get a clear read on their ultimate radiative forcing (leading to continued uncertainty, but not
outright ignorance of the important effects).
That said, we'll find that at the macro-economic scale, many of the rebound effects and causal
mechanisms described at a micro scale above may compound or magnify in various ways that are hard to
sketch out using the simple math I used herein.
What will the impact of 5% cheaper chemical impacts be throughout the many sectors of the economy
that utilize such inputs as intermediate inputs to the many final products that are produced and consumed
in dozens of economic sectors? What will happen throughout the economy when suppliers respond to the
increase in demand at Dow due to their now higher output? What impact would a set of aggregate energy
productivity improvements throughout the U.S. chemical sector have on the consumption patters, overall
incomes, and economic growth trajectories of the American family, the U.S. economy, or the Chinese
economy, let alone the global economy as a whole?
And yes, some of these effects will not even be predictable based on prior observable correlations. For
example, if a 5% reduction in the price of steel has historically corresponded to an elasticity of demand of
0.8, how are we to project in advanced what amount prices must fall to be enough to enable a machine
tool manufacturer in China to produce an entire new line of products, that then go on to fuel economic
growth throughout the manufacturing sectors of Guandong and elsewhere? If the elasticity is 0.8 for a 5%
drop, will it be 1.5 for a 30% drop? Who knows?
I hope the examples herein satisfy the skepticism of Jon, Amory, or others, that rebound effects are not
just real, but can be quite substantial. This is just one set of examples. There are many others.
I also hope it will also give us all pause before rejecting out of hand dozens of studies and analyses on
the macroeconomic scale of rebound effects. This scale is of the utmost importance to climate mitigation
challenges, after all. And despite the limitations of modeling in general and the economics discipline more
specifically, it would be irresponsible to dismiss all such analysis and continue to assume that rebound
effects must be inconsequential. As I noted earlier, my colleagues and I will be publishing a
comprehensive review of the field of rebound literature and analysis shortly. I will be sure to send it to all
of you, and hope you give it a fair reading after our long but important email discourse here.
Sincerely,
Jesse Jenkins
From
James L Sweeney
1/27/11, Sent at 12:33 AM (GMT-08:00).
Jesse:
That is useful in clarifying the differences in views.
I agree that with the substitution effect, as one of the micro effects. For many industrial processes this
may be large. For many residential and commercial uses of energy it is likely to be smaller. And
transportation it is likely to be much smaller. But I bet we all agree on this concept, although we may
differ on its numerical values across the economy, including residential, transportation, and commercial.
The output component is a different issue. As you discuss it, Dow would gain market share on its
competitors who use more energy per unit of output. It is true that such a change would increase Dow's
use of energy but that would be more than compensated by the reduction in energy use by the competitor
who loses market share -- after all that competitor uses more energy per unit of output after the efficiency
improvement. My way of defining the rebound would be to include Dow's energy use increase and the
energy use reduction of its competitors. That would give zero net value of the output component (or even
a net negative value.) Your example suggests that your way of defining the rebound would be to include
Dow's energy use decrease but to exclude the energy use reduction of its competitors. This surprises
me. Your method would give a possibly large value of the output component. But if your example does
correspond to your definition of one part of the rebound, then that difference in definition of rebound could
make a large difference in our beliefs about the magnitude of rebounds.
I agree with the indirect re-spending effect. I think of it more as a change in the production of goods and
services but that corresponds to the change in the consumption of goods and services, at least on a
world-wide basis. However, as the example I cited (way below in this chain) shows, that depends on this
being a costless gain in efficiency. If the efficiency gain is costly, then the re-spending effect is reduced
correspondingly. For example, if the cost of the efficiency improvement were exactly equal to the energy
cost reduction, then there would be no increase in profits (or GDP) and there would be no re-spending
effect.
I agree also with the embodied energy effect. That also is in my example. But that depends on there
being a costly gain in efficiency. If the efficiency gain is costless, then the embodied energy effect is
zero.
But, you cannot simply add the indirect re-spending effect and the embodied energy effect without paying
close attention to how costly is the energy efficiency gain. Note that in your example you calculated the
re-spending effect as if the energy efficiency gain were costless and calculated the embodied energy
effect as if the efficiency gain were costly. Note, however, that under some homogeneity assumptions
(the ones I made in my examples), the sum of these two does not depend on how costly is the energy
efficiency gain. The sum is equal to either the re-spending effect with costless efficiency or to the
embodied energy effect with efficiency exactly as costly as the savings in energy. Your example may not
come out that way because you took a particularly energy intense Dow facility rather than one with an
average energy intensity.
Thus if you have been adding these two effects, taking inconsistent assumptions about the costliness of
the energy efficiency, then that would also help explain why you have argued for a larger rebound effect
than have I.
So, from my perspective, the issue is not whether there are rebound effects. There are. But the issue is
how they are measured and when measured correctly, how big are they? Do they overcome most of the
original efficiency gain or a relatively small share? And from my perspective, the magnitude of the
rebound effect cannot be one number. In driving automobiles, the elasticity of vehicle miles with respect
to cost-per-mile of driving is probably in the -0.2 range. Here the rebound effect is apt to be small in
total. For industrial processes involving many ways of producing the product, the substitution effect can
be quite large. Here the rebound effect is likely to be significant because the micro-effect, in particularly
the substitution effect, is likely to be large and because energy is a relatively large share of the cost of
production.
And then finally from my perspective, the issue is what policy should be undertaken if there are significant
rebound effects. I believe that we should have a carbon price equal to the estimated environmental
damages per tonne of the incremental carbon. If that were the case, then we would not be worrying
about whether there is a rebound effect, at least for well educated optimizing users of energy. The firms
who might substitute among inputs would internalize that externality and would face incremental costs
equal to incremental carbon damages from any adjustments they made. It is only the failure to add an
appropriate carbon price that makes this issue meaningful. But that same failure makes many other
consumption and production decisions non-optimal. Thus we all worry about interventions that would not
be needed were there an appropriate carbon price. And for all of them, like it or not, we are (at best) in a
second-best world. And it should be noted that some of the macroeconomic impacts of efficiency
changes are driven because the efficiency changes makes people richer. As an economist, I tend to
think that is a good thing, not a bad thing. And if we get the appropriate carbon price, that will be an
unmitigated good thing.
Jim Sweeney
From
Steven Sorrell
1/27/11, Sent at 6:44 AM (UTC).
Dear All
I hadn't checked my email for two days, so I'm left reeling by the blizzard of messages on this topic! It is
fascinating stuff though. Thanks to Harry for an excellent blog piece and for all the thought-provoking
responses. Jesse's summary of the current state of evidence is entirely consistent with what we
concluded in 2007 and shows that some real progress is being made. Rebound effects are real, variable
and frequently significant and it is a serious mistake to neglect them. They are also measurable or least
'estimatable', but doing so is remarkably difficult.
We are currently trying to make some estimates of the indirect effects (in terms of GHG emissions) of
household re-spending (<http://www3.surrey.ac.uk/resolve/Docs/WorkingPapers/RESOLVE_WP_0510.pdf>). Our initial attempts are pretty crude but serve to demonstrate that these are far from trivial. The
literature on such effects is still remarkably thin, so there is plenty of scope for more work. I've attached a
draft paper that tries to identify a suitable approach and summarise the results of existing work. Any
feedback on this is most welcome.
I saw some mention of holding a US workshop on this topic. If this gets off the ground, I'd been very keen
to attend!
Best wishes
Steve
From
Jonathan Koomey
1/27/11, Sent at 4:37 PM (GMT-08:00).
All,
This is a continuation of the Andy Revkin request thread. I've tried to reunite that thread by including
Jim's most recent reply on top of Michael's most recent reply to Ted. My more detailed technical
comments I'll send on the Jim Sweeney/Jesse technical thread.
Andy,
I strongly agree with Jim Sweeney's framing of the problem, but wanted to add a couple of points.
It is incorrect to say that rebound is zero in all cases, and it is incorrect to say that rebound is 100% (or
more) in all cases. The question is better phrased as "under what conditions is rebound a problem, and
in those cases, how big is it?".
We're still in the middle of our process, but we're definitely moving in a constructive direction, thanks to
Jesse's creation of the heuristic example. This process should allow us to define terms and move closer
to agreement on what the issues are, so in that sense I think this debate is somewhat different than the
one with the climate folks you were monitoring (up to a point). There are still major issues with the use of
macroeconomic models in this and many other contexts (see, for example, Stephen Decanio's book from
2003. Economic Models of Climate Change: A Critique. Basingstoke, UK: Palgrave-Macmillan, which is a
devastating critique from inside the economics community of the guts of the models that Ted and others
rely on so heavily for their conclusions about macro rebounds).
Putting those big issues about the macroeconomic models aside for the moment, we need to take Jesse's
heuristic example and create more like it for different situations, and we need to figure out which parts of
the economy are more or less affected by rebound, so we know what factors drive the results. As part of
that process we need to examine things like the elasticity of substitution (which Jesse pegged at 0.5 as
per the literature), not just to see whether the number we use is correct for a particular type of industry or
end-use, but to see whether real firms and consumers actually behave that way in the real world, given
the cognitive limitations, information asymmetries, split incentives, principal agent problems, increasing
returns to scale, and other issues that make real world decisions diverge from the simple models. That
means understanding the functions and other assumptions that are used to estimate elasticities as well
as the models into which those elasticities are inserted, and understanding how real firms and consumer
actually behave.
The skepticism that Amory, I, and others have expressed about these results is that real firms and
consumers almost invariably do NOT behave like the models say, so just giving us the results of
macroeconomic modeling and saying that they prove rebound always wipes out efficiency is very
unconvincing to us (and to economists, like Jim Sweeney and Steve Decanio, who have looked at how
real firms behave).
Finally, I would like to distinguish (as Jim also does) between the popular summary of rebound in (for
example) the New Yorker article, which conflates the results of increased wealth and other unrelated
trends with rebound, and the somewhat more nuanced picture that comes from review of the relevant
academic literature, which is generally, though not always, more careful to define things in a sensible
way. I would avoid deriving any conclusions at all from the popular literature, and stick closer to the
primary academic materials that Harry and Jesse have helpfully supplied.
Jon
from
Amory B. Lovins
1/27/11, Sent at 9:21 PM (GMT-07:00).
On 26 Jan 2011, at 20:39, Jesse Jenkins wrote:
Dear all,
As promised, in answer to Jon's challenge to show some causal chain driving rebound, I've
drafted this simplified example. I've made up all of the values herein, but I think they aren't too far
off, nor far fetched, and should serve our illustrative purposes here. Actual values could easily be
higher or lower in some sector or another, as would the resulting rebound effects. I hope, as Jon
said, this example will facilitate 'true understanding' in this discourse.
My apologies for the length of what follows -- Jon, Amory, you asked for it after all! ;-)
Absolutely, and I hope I speak for everyone in thanking you, Jesse, for providing this very helpful aid to
specificity in our discussion. Please forgive me if my reply makes it even longer.
OK. Let's take Dow Chemical as an example, since that's one of Amory's examples in a prior
email, and I believe, one of his clients.
Not formally, but we've done things together informally over the decades. However, I have worked
formally for other similar firms on many multi-billion-dollar facilities, new and retrofit.
Let's assume that a few dozen energy saving projects that Dow pursues across it's US plants
reduce the energy consumed in their processes and facilities by a total of 30%, so that if initial
energy use was 100 units (the units don't really matter, so we'll use generics), it is now expected
to be2 70 units, based on engineering-level estimates that Amory has calculated for his client.
In practice I'd usually expect to save more and pay less than you assume, although 30% is a reasonable
strawman for standard practice. FYI, though, in 1981, Ken Nelson at Dow's 2,400-worker Louisiana
Division set up a shop-floor-level contest for energy-saving ideas with at least a 50%/y ROI. The first
year's 27 projects averaged 173%/y ROI; the second year's 32 projects averaged 340%. Twelve years
and almost 900 implemented projects later, the workers had averaged (in the 575 projects subjected to
post hoc audit) 202%/y predicted and 204%/y audited ROI. In the later years, the returns and savings
were both getting bigger, because the engineers were (as usual) learning faster than they were
exhausting the efficiency resource. In only one year did returns dip into double digits (97%/y ROI). By
1993, the whole suite of projects was paying Dow's shareholders $110 million every year. However, when
Ken retired, his superiors -- whom he had kept largely uninformed so they wouldn't interfere by appointing
committees, writing slogans, instituting procedures, etc -- did so and reorganized the effort. The effects
included no more monitoring. One therefore can't tell what happened after that, although Ken believes
some further savings have occurred. (Perhaps monitoring has lately resumed—I need to check.) Ken also
notes that with no similar champion, and despite knowing of his results, the adjacent Texas Division did
relatively little all along: an instructive tale for skeptics about market failures.
I mention this story just as a practitioner's-eye view of what's possible even without RMI's special sauce
(integrative design), and as an indication that some of Dow's efficiency engineers are very, very good.
That, and bottom-up processes like Ken's in some parts of the company, are the main reasons their ~$1b
of efficiency investment has so far returned ~$9b.
Now let's start in with the rebound effects, beginning with the direct, and moving to the indirect.
OK, back to the subject at hand:
First, we'll consider the substitution component of direct rebound. Let's assume that substitution
elasticities between energy services -- say heat or electricity in the production of chemical
feedstocks and final products -- and labor or capital or other materials is about 0.5 on average.
Greening et al. 2000 (in the EP issue Lee edited) reports that typical substitution elasticities for
firms often range from 0.4 to 0.8 and in some cases closer to 1.0, but we'll consider a relatively
conservative value from that range of 0.5.
[Economic digression: Their 2000 paper also says on p. 390: "For consumers, the direct effect of a price
reduction may be decomposed into a substitution effect and an income effect. If the price of the energy
service drops, consumers should substitute indefinitely for a given energy service. However, this ignores
the potential for satiation for a given service, which limits potential levels of substitution, and the tradeoffs
with other expenditures that consumers make within a budget constraint. Currently, since all empirical
estimates are uncompensated estimates, direct effects cannot be decomposed into substitution and
income effects. Further, since the majority of these estimates are short-run, we cannot examine the longrun effects of changes in capital costs." They also note on p. 391 that your assumption that cheaper
energy services get substituted for other factors of production depends on an assumption of "only
diminishing returns to increased use of energy services," which may not always be true.
The elasticity values you cite appear to be for residential space heating (p. 393), among the highestrebound uses (p. 398). But for industry, p. 397 warns: "Empirical estimates of the relationship between
energy and other factor inputs...exhibit a wide range of values, which may not always indicate
substitution....[D]epending on the functional specification, the industries included, and the time period
represented, the magnitude and sign of the estimated relationship between energy and other factor
inputs...varies." They cite four examples of papers that "estimated negative elasticities of substitution
between energy and capital implying complementarity, or as the input of capital increases the input of
energy increases," and two papers that estimate substitution elasticities >1. Caveat calculator.
Further, on pp. 398 they note that Saunders (1996) "demonstrated that growth models using CobbDouglas production functions will always predict that technological progress will result in increased, rather
than decreased energy consumption. Models using CES production function, however, demonstrate that
this is not necessarily the case. The impact of energy efficiency gains depends on the elasticity of
substitution between fuel and other inputs in the production of energy services. As noted above, if the
elasticity of substitution of fuel for capital and labor (jointly) is less than 1.0, technical efficiency
improvements will reduce overall energy use. The Cobb-Douglas function is incapable of producing this
result, since the elasticity of substitution in a Cobb-Douglas production function is always unity (Varian,
1992). As we have also noted above, the evidence on this point is mixed."
Finally, on p. 397 they also conclude that what you call the output effect is small, and on p. 398 they
report one study of 0–20% rebound in [industrial] process uses and four studies of 0–2% rebound in
lighting, both short-run. End of economic digression. Back to our story:]
So as the price of energy services in Dow processes and facilities falls by 30%, plant managers
over time rely more heavily on now-cheaper energy and less on other inputs -- materials, labor,
capital. I'm not familiar with Dow's plants, so I can't give clear specifics on what these
substitutions would be, but for sake of conjecture, let's say they new insulation on chemical vats
now make it cost effective for plant managers to rely on higher temperature processes that
reduce materials needs, or perform chemical reactions faster, thus reducing labor demand. There
are probably dozens of small adjustments like this over time, which yield the aggregate
substitution elasticities typically observed.
So a 30% energy efficiency improvement yields a 30% reduction in the implicit price of energy
services, so with a substation elasticity of 0.5 that means that Dow increases their consumption of
energy services by 15% following the efficiency gains, or an increase in raw energy demand of 70
energy units*15%=10.5 units, or a direct rebound due to the substitution component of the
rebound effect of 10.5/30 = 35%. The firm now consumes 80.5 units of energy, not 70.
This is your most important term, and the one where as a practitioner I am having the most trouble
coming up with a plausible physical interpretation of this generalized literature elasticity of substitution. In
my experience, chemical process plants generally don't work in the way you posit. They are designed to
perform specific reactions on specific raw materials under the conditions required to produce a specified
slate of products within quality and safety parameters. Capacity of existing plant is increased by
systematic debottlenecking, of which energy efficiency is sometimes a byproduct but seldom a causal
driver. Efficiency is increased by pinch (thermal optimization), insulation, more efficient heat provision and
recovery, better sensors and controls, variable-speed drives, better pumps and motors, lookahead
algorithms, etc. It's probably more common for efficiency to increase as a byproduct of changes made to
increase capacity than the reverse. In the example you mention—which I appreciate was only a
conjectural illustration—if a higher-temperature or -throughput process made sense, it would probably
have been designed that way and the insulation bought to optimize it; thin insulation probably reflected
old energy prices, fast-payback rules, first-cost emphasis, or mere inattention and rules of thumb
("infectious repetitis"); and it would be very unusual for better insulation to drive the reaction kinetics much
faster (within safety constraints) or to permit a superior new process. Process equipment is costly,
insulation cheap.
Also, chemical plants, and many other heavy process plants like pulp-and-paper or cement, are already
generally pretty good on thermal and molecular efficiency but much less good on electric-auxiliary
efficiency (pumps, fans, stirrers, conveyors, grinders, etc). Much of my work in such plants in recent years
has therefore focused on order-of-magnitude savings in fluid-handling (e.g. pumping), typically specified
poorly by a junior engineer. But when we do such improvements in a carpet factory, refinery, or whatever,
they have little or no effect on capacity or output, which are limited by other aspects of the design. The
result is thus an essentially pure saving of energy and opex. I'm having trouble interpreting these
observations from scores of plants in any way consistent with significant energy rebound.
Of course, the company's saved energy dollars may be partly respent on expanding production rather
than on marketing, R&D, executive bonuses, etc, but I think that's a different issue, which you and I will
address below.
I'd also be inclined to interpret any increased output that did occur as another way to expand desired
production with less money and energy than would otherwise occur, rather than simply as an increase in
energy use without regard to the increased value and production. This question of "what's the
denominator" keeps cropping up, whether in Dow's plant or in GDP.
I also agree with Jim that the substitution effect is likely to be much larger in industrial production (but
varying very widely between segments, plants, and products -- there are well over 70,000 main products
in the chemical industry alone -- so generalizations are dangerous) than in buildings, and probably larger
in buildings (especially households) than in transportation. Thus its overall effect is much smaller when
efficiency is applied to the whole economy than when we consider industry alone. Buildings happen to
use ~71% of U.S. electricity, industry only ~29%, so the distinction is very material. (Industry is relatively
gas-hungry, but gas is cheaper and less carbon-intensive, and lacks electricity's system losses' carbon
multiplier. I do not think Harry's convention of considering electric generation a "productive" sector rather
than allocating its inputs and outputs to the sectors it serves is appropriate.) Of course, all the other
effects that also constrain microeconomic rebound—saturation, opacity about what actions saved how
much money or what implicit service prices are, principal/agent problems, etc—apply particularly in
buildings.
Finally, we need to make sure that what is being counted as an elasticity of substitution doesn't doublecount or overlap any other effect considered separately below. I'm not yet fully confident of that.
Summarizing where I come out on substitution effects: competent econometricians have crunched lots of
numbers for lots of diverse industries and found a wide range of implicit elasticities of substitution. Yet I
am finding it very hard to square that theoretical concept with what I routinely see happening in actual
industrial plants and C-suites when technical energy efficiency is raised. That is, the economic model
doesn't seem to match the observed behavior.
Jon Koomey has noted data centers as a possible counterexample where utilities constrain their electric
supply. Well, the last data center my team and I codesigned provides ~3x the previous versions'
computing per watt, with 98% savings in cooling and pumping—all at normal capex. Its builder EDS (now
HP) told me that had the client allowed adoption of all our recommendations, they'd have saved ~95% of
the energy and ~50% of the capex. The logical consequence of this sort of breakthrough efficiency
potential is that it will put the inefficient data centers out of business, probably saving much more energy
than additional electricity use from heightened demand for the new, lower-cost data centers. That's pretty
much what's starting to happen in that industry. Same for chip fabs, where our latest design is expected
to save two-thirds of the energy and half the capex, before redesigning the tools—a likely route to ~8–10x
total savings, just getting underway.
That's the substitution component of the direct rebound effect alone. As Harry has noted in his
studies, this is typically the much larger component of rebound. But let's now consider the second
component of direct rebound, the output component.
For this, we'll assume that energy makes up 15% of total costs for Dow. This is an energy
intensive sector -- chemicals -- so we'll assume something a bit higher than the 6-10% of costs
that may be more typical,
Whoa, Nellie! It's roughly 1–2% for the average U.S. business—I think the average is a hair over 1%. For
the relatively energy-intensive U.S. chemical sector, the average energy intensity measured in dollar
fraction (energy costs / value of shipments) was 3.6% in 2008: $27b energy costs / $751b revenues. A
nice little primer is at www.edf.org/documents/11206_LCMI-Chemicals-Citations.pdf and its citations. A
15%-intensity chemical plant would be quite unusual—that's like a steel mill.
but not too high (and after all, energy intensive sectors are where we often look hardest for
efficiency opportunities). The efficiency gains and the cost-effective substitution of now-cheaper
energy services for other inputs to production (which will still yield net cost savings) thus reduce
Dow's bottom line by 15%*30%=4.5%. The management then decides to chose to pass this
savings on in the form of lower prices of their products, undercutting their competition and driving
greater demand for their products.
There are many other calls on earnings—dividends, bonuses, plant investments to keep or gain market
share, marketing, R&D, etc—so devoting 100% to price-cutting is pretty unusual. But let's adopt that
extreme assumption for now; if we didn't, your output effect would be smaller and some other effect(s)
could be larger, depending on where the money goes.
I don't know what elasticities of demand are for chemical inputs, but I assume they are fairly
elastic, so let's assume it's 0.8. That means demand for Dow's products rises 0.8*4.5% = 3.6%,
driving up demand for all factors of production, including energy services, by 3.6%. So that's
another 80.5*3.6% = 2.9 units, or a rebound from the output component of direct rebound of
2.9/30 = 9.7%. (Note, we won't even get into the energy required to make and support the other
factors of production that are now in greater demand; that's an indirect effect probably best
looked at at a macroeconomic scale, as the economy adjusts in a series of complex chain
reactions to this one change in demand at the chemical plants… we'll keep it simpler for now).
Renormalizing to actual chemical-industry-average 2008 energy intensity from 15% to 3.6%, your 9.7%
drops to 2.3% (an adder of 0.7 not 2.9 units). This shows why substitution effects are potentially
significant only in a handful of extremely energy-intensive sectors like primary metals, which in 2009 were
only 1.3% of GDP; chemicals were 4.4% of GDP.
I agree with Jim Sweeney that your "output effects" are actually a zero-sum game with equal competitors;
worse, making your output-effects component negative, if Dow's competitor has a lower energy efficiency
than Dow—which it practically must under your assumption that Dow's energy efficiency is what let Dow
grab market share. Thus, if Jim and I are understanding this part of your example correctly, we're saying
it's fallacious because you counted Dow's increased output but not its losing competitor's decreased
output...putatively at even higher energy intensity. What part of this am I misunderstanding, please?
So now, total rebound from direct effects -- substitution and output effects together -- is
44.7%. Let's round to 45% for ease. The firm now consumes 100-30+10.5+2.9 = 83.4 units of
energy, for a 16.6% net reduction in energy demand, not the 30% reduction you'd expect from
summing up the initial engineering-level energy savings.
Neglecting your output-effect term, because you didn't count the competitor's lower energy from losing
market share, we're still at 2.3% rebound so far for an average 2008 chemical plant—meaning that it
saves not 30% but 27.7%. But that improves if we count, as we should, its losing competitor's lower
energy efficiency. For example, if we assume the nearby Jones Chemical plant was identical but failed to
do the 30% energy saving, then Dow's bottom line improves by 3.6% [actual average chemical-industry
energy-intensity] * 30% = 1.08%. Using your assumption that Dow puts all of this windfall into pricecutting and demand elasticity is 0.8, demand for Dow's products rises 0.864%. Meanwhile, Jones, whose
energy intensity stays fixed, loses the same sales, decreasing its energy use by 1/(1-0.3) or 43% more
than Dow saved. Net result: Dow's 30% energy saving gains 1.08% sales by knocking out the same
amount of inefficient Jones production, boosting Dow's 30% direct energy saving to 30% - 1.08% +
(1.43*1.08% - 1.08%) = 30.46%. Less 2.3% substitution effect, that's 28.2%—so far yielding a 1.8%-point
rebound effect. (You could also run Jones's substitution elasticities backwards.)
If I've done the arithmetic wrong, please forgive me—it's rather late in a long day—but I think the principle
is right. No doubt someone can better calculate Jim's correct concept.
Another thought: Unless I'm missing something, nobody has yet counted a second-order effect that may
be material: when Dow saves energy, the industries that supplied its energy also lost sales and used less
energy to provide it. Shouldn't this be subtracted from rebound? For example, if Dow saves 1 kWh, its
utility (ignoring details about Dow's cogen-rich supply system and fungibility with grid power) sells 1.0x
kWh less (the .0x being grid losses—small because Dow buys mostly at high voltage), the coal and gas
suppliers sell ~3-4 kWh(t) less, and their reduced sales are further enhanced by their (and their respective
railway and pipeline partners') not needing to use so much energy to process and deliver the fuel to the
utility. The process and delivery energy varies widely with fuel and many other details, but a 1.1x
multiplier is generally a decent first approximation, so saving 1 kWh(e) probably saves on the order of
3.3-4.4x in primary fuel. Since you're accounting in dollars, not kWh or J, we could express the avoided
processing and delivery energy as a "negative rebound," increasing the 30% direct saving by a few
percent in physical terms, albeit off Dow's premises. That takes us back up over 30% net savings even if
the processing-and-distribution energy use is only 5%, a conservative estimate. The energy suppliers
could also save capital, labor, etc, running some of your indirect-effect loops modestly backwards after
lags.
I would of course agree that whenever anyone saves energy, that also slightly depresses everyone's
price for the same sort of energy in the same or linked markets (and indirectly perhaps for other sorts of
energy too). But in practice, even when a firm the size of Dow saves, say, 30% of its direct energy use
across the board, this effect on energy prices is very small, since much of the energy involved is more or
less fungible in very large, even global, markets that normally display secular growth anyhow.
By coincidence, this is close in magnitude to the values Harry calculates for direct rebounds in the
US chemicals sector writ large in the paper he's circulated (53% over the long term). And as a
back of the envelope example illustrating the clear mechanisms at work and the simplest of
microeconomic elasticities shows how the effects can tally up, and how results like those Harry
(and a dozen other scholars of rebound effects) are not far fetched at all.
But this is just the direct effects. Let's consider the indirect rebound mechanisms as well. The
simplest could be the re-spending/re-investment effect.
Since Dow still reduces energy demand by 16.6% their total costs will fall by 16.6%*15%
(energy's share of costs) = 2.5%. That's a 2.5% increase in corporate profits, shared amongst
investors who re-spend and re-invest that money and recirculates throughout the economy (I'm
not even considering how much more the increased volume they are now selling may increase
profits, just the raw energy cost savings).
I'd make that roughly 30+% * 3.6% [chemical-industry average energy intensity] = 1.08+%. But wait a
minute...I thought you'd allocated all of the saved energy dollars to cutting costs when calculating your
output effect. Aren't you now giving the same money to shareholders—thus double-counting? Or am I
misunderstanding something obvious?
If energy was 15% of total Dow costs (before the efficiency gains), then total costs, measured in
the monetary equivalent of energy units, is 100/15% = 667 units (all economic activity is calories
or BTUs at its core, so maybe this conceit is excusable, as it avoids having to think about specific
monetary units). Then that savings amounts to 667*2.5% = 16.7 units of economic value
(measured in equivalent energy units). If as a whole, energy makes up 8% of economic inputs
(economy-wide), then that 11.7 units of economic value re-circulating in the economy will
increase energy demand by 16.7*8% = about 1.3 units. Not huge, but that's another 1.3/30 =
4.33% rebound due to this indirect re-spending/re-investment effect. Total rebound is now
49.3%.
Since you assumed 4x actual energy intensity, your 1.3-unit adder gets much smaller too, even if you
weren't double-counting.
Jim is also right that we need to subtract from Dow's energy savings the cost of achieving them. If it's as
much of a nearly-free lunch as my Dow Louisiana example above, this correction won't be large, but if
efficiency is as modest and costly a resource as some economic theorists on this thread suppose, the
correction will often be large.
Also, I realize we're using the 8% for simplicity, but in practice, whether Dow reinvests its saved energy
dollars in more efficiency or in some sort of energy profligacy depends on whether it understands what it
just did to earn those dollars. My experience with such firms suggests that successful efficiency programs
tend to win management confidence and get more money to save more energy. I say this notwithstanding
what happened after Ken Nelson retired: in fact, his success is now being more widely multiplied because
he's consulting for many other firms—a common pattern for such great practitioners.
Now let's consider another indirect mechanism, the embodied energy effect: the energy efficiency
upgrades themselves required energy to produce. Let's assume that the suite of upgrades Dow
employs have on average a 3 year payback period, so for every dollar they cost, they save 1/3rd
of that amount each year. And they have an average life time of 15 years before replacement
(some things like lighting being shorter, and others like boilers maybe a bit longer; again, I'm just
making these up for illustration). That means in their lifetime, the combined efficiency measures
save 15*(1/3) = about 5 times more in energy costs than the upfront costs for the upgrades. A
nice return. But still, that means that if we invert that ratio, for every 100 units of energy costs
saved, the up-front cost (in economic value terms) was 20 units (1/5 = .2). Then if we again
assume that energy makes up about 8% of the cost of these units of economic value, then that
makes 1.6 units of energy consumed in making and installing the efficiency upgrades for every
100 units they save. Since the upgrades saved Dow 30 units of energy per year (at the
engineering-level, where these returns are calculated), then that's another 30*1.6% = 0.5 units of
indirectly increased energy consumption. In other words, another rebound due to embodied
energy effects of 0.5/30 = 1.7%. A small addition, but total rebound is now over half, at
51%.
Having helped write the "generally accepted accounting principles" for net-energy analysis at Niels Bohr
Institutet in the 1970s, I'm pretty familiar with this concept, and would refine it both by Jim's points (the
energy saving is seldom costless, and you need to be consistent in the cost you assume between this
and the previous indirect effect). Also, I think embodied energy in energy-efficiency investments, say in
chemical plants, is generally much smaller than the economy-wide average of ~8%, for three reasons.
First, embodied energy in (say) tweaking a control algorithm is close to zero, while embodied energy in
using fat, short, straight pipes instead of skinny, long, crooked pipes is often negative (because so much
less copper etc get used, especially in the much smaller pumps, motors, and electricals). Second, the
value chains for efficiency-improving equipment work in such a way that much of the "value" is simply
supplier markups that embody very little energy. Third, after the operating life is over, typically 20+ years,
most of the embodied energy can be and is profitably recovered by normal repair, remanufacturing,
reuse, or recycling. On balance, I think the embodied-energy term is usually negligible. The equipment is
seldom made of energy-intensive stuff like enriched uranium or titanium.
So total rebound due to indirect and direct effects alone -- considering nothing that may be
emergent at more macroeconomic scales, and no economic "model waving" -- would erode half
of the projected savings in total energy consumption projected by the kind of engineering level
analysis that firms like Amory's would show to clients like Dow. Dow still saves a lot of money
and sells a lot more of their product, so they're happy, and the economy at large is too. But the
overall climate and energy demand reduction benefits are oversold by 2:1 if we fail to consider
these rebound effects -- that is, the efficiency savings only cut energy demand, and thus
emissions, by just half as much as initially promised -- and that's what matters from a climate
mitigation perspective.
I think your mode of analysis, with double-counting removed and one 4x-high number (chemical industry's
energy intensity) adjusted to actual, shows just the opposite—that, as I'd expect, the total effect of a 30%
saving in a small factor cost is small, down in the few-percent range.
Furthermore, while this is the end of my simplified example illustrating the chain of causal
mechanisms at work in the term 'rebound effect,' that's not the end of the rebound story. To get at
the full scale of rebound effects at the scope that really matters to climate mitigation concerns -that is the global scope -- does indeed require some economic modeling.
Jon may want to reject the entire field of economic analysis, and after the financial crisis, we
certainly can all certainly be forgiven for taking economic modeling with a little grain of salt...
But the economy is a complicated beast. Like the global climate, we are forced to rely on models
to try to approximate and examine the macroeconomic and often emergent phenomena that
make up our complex, interconnected global economy.
And yes, to get these models working, economic analysts must rely on observed and measured
correlations for many factors as they develop models that use this correlative relationships to
produce outputs that as closely as possible match the real world.
Yes, correlation does not equal causality. But I've shown you some clear causal mechanisms
here at a micro scale. And the correlated factors chosen to do macroeconomic modeling will have
some clearly plausible causal connections underlying them as well -- economists won't use the
number of pirate attacks annually as an independent variable driving the dependent variable of
employment in the U.S. chemical sector, after all!
Unfortunately, if we care about things like climate change or the impacts of efficiency in a
complex global economy, these are the murky waters we must wade into if we hope to move
beyond the simple and microscale effects sketched above.
This is an epistemological and analytical challenge, sure, but it is also a set of limitations and
abstractions that we routinely accept as a means of informing policy making and as we seek
greater understanding of complex macro-scale phenomena, from the global economy to global
climate patters to neural networks to ecosystems.
Jon's rejection of dozens of peer reviewed studies published in numerous reputed journals of
energy economics and policy (including by several analysts on this email chain), thus strikes me
as somewhat akin to climate skeptics rejecting all climate modeling work out of hand, simply
because it relies on models that depend on observed correlations between atmospheric CO2
levels and global temperatures. Some skepticism of model outputs is warranted, surely, but
outright rejection? Far from it. And there's clearly a casual relationship between the two
correlated factors, even if we can only rely on observed correlations to get a clear read on their
ultimate radiative forcing (leading to continued uncertainty, but not outright ignorance of the
important effects).
That said, we'll find that at the macro-economic scale, many of the rebound effects and causal
mechanisms described at a micro scale above may compound or magnify in various ways that
are hard to sketch out using the simple math I used herein.
What will the impact of 5% cheaper chemical impacts be throughout the many sectors of the
economy that utilize such inputs as intermediate inputs to the many final products that are
produced and consumed in dozens of economic sectors? What will happen throughout the
economy when suppliers respond to the increase in demand at Dow due to their now higher
output? What impact would a set of aggregate energy productivity improvements throughout the
U.S. chemical sector have on the consumption patters, overall incomes, and economic growth
trajectories of the American family, the U.S. economy, or the Chinese economy, let alone the
global economy as a whole?
That shouldn't be hard to examine, since the U.S. chemical industry has more than doubled its energy
productivity since 1973 and is in the process of doubling it again over several decades. The trouble will of
course be trying to figure out what part of general economic growth was caused specifically by energy
inefficiency in that sector (or in all sectors).
And yes, some of these effects will not even be predictable based on prior observable
correlations. For example, if a 5% reduction in the price of steel has historically corresponded to
an elasticity of demand of 0.8, how are we to project in advanced what amount prices must fall to
be enough to enable a machine tool manufacturer in China to produce an entire new line of
products, that then go on to fuel economic growth throughout the manufacturing sectors of
Guandong and elsewhere? If the elasticity is 0.8 for a 5% drop, will it be 1.5 for a 30% drop? Who
knows?
More likely, I think, is that steel will lose most of its market, over many decades, to substitutes —
advanced composites, better design, etc. — just as it has long been doing.
I hope the examples herein satisfy the skepticism of Jon, Amory, or others, that rebound effects
are not just real, but can be quite substantial. This is just one set of examples. There are many
others.
I also hope it will also give us all pause before rejecting out of hand dozens of studies and
analyses on the macroeconomic scale of rebound effects. This scale is of the utmost importance
to climate mitigation challenges, after all. And despite the limitations of modeling in general and
the economics discipline more specifically, it would be irresponsible to dismiss all such analysis
and continue to assume that rebound effects must be inconsequential. As I noted earlier, my
colleagues and I will be publishing a comprehensive review of the field of rebound literature and
analysis shortly. I will be sure to send it to all of you, and hope you give it a fair reading after our
long but important email discourse here.
I will read your and Harry's new papers when I get a chance, but for now, I agree with your source
Greening et al., which concluded at p. 391 that "Since energy is a relatively minor share of an individual
consumer's total expenditures, the secondary effects are probably insignificant." They found that while
economy-wide effects are "highly uncertain" (especially since "potential paths of technological change are
stochastic") and deserve "further study," and "transformational effects" are possible, they "could result in
more or less energy consumption." On p. 397, seeking a general-equilibrium model to illuminate
economy-wide effects, they note that when Kydes (1997) used NEMS to compare nine scenarios,
decreasing energy intensity by 6.5 percentage points extra caused a 5% drop in U.S. energy demand and
a 0.4% rise in GDP (ignoring foreign adoption of U.S. technological change). And way back in 1978, Len
Brookes said that distinguishing income effects from other effects makes attribution of specific changes in
society to increases in fuel efficiency problematic (id.). It still is.
So here's where this longwinded discussion of your straw example leaves me:
With the biggest (substitution) of your four illustrative effects open to the model-vs-reality and other
questions I've raised, the second (output) effect fallacious and wrong-signed, the third (respending) effect
double-counted among other problems, and the fourth effect (embodied energy) probably negligible, we
can better understand why different students of rebound get such different results.
Most strikingly, I don't yet see in your example any significant macroeconomic rebound effects, which I
thought were the ones you were going to show were so big! Seven-eighths of the total effect you
calculate —44.7% out of 51%—is direct effects internal to Dow and its sector. Of the remaining eighth
that you add to illustrate macro effects, 72% is from the double-counted reinvestment of the same capital
you already spent on cost-cutting, and the other 28% for embodied energy is grossly exaggerated. I'm left
grasping for macro effects that so far look pretty insubstantial.
I grant you that fundamentally new inventions like, say, the Internet can reshape a macroeconomy (while
incidentally streamlining commerce, saving space and transport, etc. in ways that save considerably more
energy than the IT uses, as Joe Romm showed some years ago). But I thought your claim was that just
garden-variety energy savings must trigger a large rebound, mainly macroeconomic, that would offset
much, all, or more than all of the direct savings. You've actually shown, I think, that the total effect, to the
extent you've described its logic, is quite small. If there are so many other and
unillustrated macroeconomic effects that collectively they're very large, I'd love to see you illustrate them
with similar concreteness.
This exercise has usefully improved my understanding, and I hope others', of how total rebound should
be calculated. Remaining is the separate question of what baseline it should be compared with ("what's
the denominator"): whether to emphasize the outcome of saving modestly less energy than expected
from a pure engineering calculation (I don't think anyone on this thread is that unsophisticated) or the
outcome of having achieved higher GDP, production, income, welfare, or whatever using much less
energy than would otherwise be the case. Let's leave that for another day.
Best – ABL
from
Harry Saunders
1/28/11, Sent at 11:53 AM (GMT-08:00).
Geez. Where to start?
I want to adhere to brevity, but let me start by directly addressing the claim implicit in Amory’s epistle that
this is a battle between “practitioners” and “theorists.” To do this, I need to make clear that I, too, am a
practitioner. In the “real” part of my life, I have done many, many projects with chemical companies and
many other companies, numbering probably as many as Amory’s, probably more.
But in my business this involves looking not just at the energy component but at large-scale decision
making typically involving huge investments in new capacity or in whole new businesses—capital, labor,
and materials as well as energy (the importance of this distinction for the present discussion will become
clearer below). I am at a disadvantage here, though, because I am bound by confidentiality agreements
that prevent me from being too explicit. (I am surprised that Amory isn’t bound by such agreements.)
Points:
·
What matters for rebound is how firms reconfigure *new* capacity. While my experience, like
Amory’s, is that debottlenecking is common in the chemical industry, it is a minor source of capacity
increases. The equation changes radically when you start from scratch. A greenfield plant has many
more opportunities to optimize the use of all factor inputs and to employ new technologies. I have seen
numerous examples of radical shifts (but I’m bound to say no more). The time horizon of relevance here
is decades. The relevant question is, “what will new capacity look like in the future given energy
efficiency technologies?” Even in the short 20-year period of my analysis, the history indicates this
reconfiguration was substantial enough to create significant sector-wide rebound in chemicals. A micro
look at individual energy-specific projects puts too small a boundary around firm decisions and misses the
larger dynamic.
·
Amory has been relying for what seems like forever on Lorna Greening’s survey to make his
arguments about income/output and substitution effects (a distinction I believe I introduced to the
academic literature.) The literature has grown substantially in the meantime and there are new analyses
available that provide new and better answers. Steve Sorrell’s UKERC survey comes to mind. The
Evans/Hunt book. The Herring/Sorrell book. The literature is becoming vast, although it largely
originates in Europe.
·
In this connection, Amory could have read my paper twice over in the time taken to write his
analysis. If he had, he would have learned that the idea of a single elasticity of substitution for energy is a
construct no longer needed. Also that we can now use highly general flexible functions (Translog) for
rebound measurement. The Cobb-Douglas paradigm he complains about no longer applies. Nor CES. If
he wants, I can send him an article showing the merits of using flexible forms for rebound measurement.
·
The zero-sum game argument is a red herring. The output effect comes along with a reduction in
output price and represents a gain in both producer and consumer welfare due to the efficiency gain.
Consumers consume more; producers produce more. Anyone wanting an undergraduate-level
explanation for producer-consumer interactions given rebound can refer to the paper [Theoretical
foundations of the rebound effect. In: Herring H, Sorrell S (Eds). Energy efficiency and Sustainable
Consumption: Dealing with the Rebound Effect. Palgrave Macmillan; 2009.]
Harry
from
Amory B. Lovins
1/28/11, Sent at 12:44 PM (GMT-07:00).
On 28 Jan 2011, at 12:53, Harry Saunders wrote:
Geez. Where to start?
I want to adhere to brevity, but let me start by directly addressing the claim implicit in Amory’s
epistle that this is a battle between “practitioners” and “theorists.” To do this, I need to make clear
that I, too, am a practitioner. In the “real” part of my life, I have done many, many projects with
chemical companies and many other companies, numbering probably as many as Amory’s,
probably more.
True and respected, Harry. I think we first met when you were chief economist of TOSCO back in the oilshale days! I'm glad you share that nuts-and-bolts experience in a way few theorists do, alas. But how
one views the logic of Jesse's example shouldn't depend on whether one comes from either sort of
experience or both.
But in my business this involves looking not just at the energy component but at large-scale
decision making typically involving huge investments in new capacity or in whole new
businesses—capital, labor, and materials as well as energy (the importance of this distinction for
the present discussion will become clearer below).
Absolutely, and part of our shared background.
I am at a disadvantage here, though, because I am bound by confidentiality agreements that
prevent me from being too explicit. (I am surprised that Amory isn’t bound by such agreements.)
I am, and have respected them as usual. There are of course no secrets in my postings. RMI's client
agreements typically include explicit provision for publishing brief agreed statements about what was
achieved, though typically not how; often the clients themselves also publish the results at this level of
generality.
Points:
·
What matters for rebound is how firms reconfigure *new* capacity. While my experience,
like Amory’s, is that debottlenecking is common in the chemical industry, it is a minor source of
capacity increases. The equation changes radically when you start from scratch. A greenfield
plant has many more opportunities to optimize the use of all factor inputs and to employ new
technologies. I have seen numerous examples of radical shifts (but I’m bound to say no more).
The time horizon of relevance here is decades. The relevant question is, “what will new capacity
look like in the future given energy efficiency technologies?” Even in the short 20-year period of
my analysis, the history indicates this reconfiguration was substantial enough to create significant
sector-wide rebound in chemicals. A micro look at individual energy-specific projects puts too
small a boundary around firm decisions and misses the larger dynamic.
In older industries like chemicals, as in the buildings sector, retrofits seem more important than greenfield
newbuild (of which we both have much experience). Both are important in some new sectors like chip
fabs and data centers, where RMI has also been active. I was discussing retrofits because that was
Jesse's example. I agree that savings are bigger and easier in new than old plants. RMI's heavy-industry
work, lately embracing >$30b of diverse projects, broadly saves ~30–60% of energy on retrofit with
paybacks of typically a few years, and saves more, up to 90-odd percent, in greenfield newbuild, typically
with reduced total capex.
If you really feel "what matters for rebound is...new capacity," I hope you, Jesse, or someone else will
construct a new-plant scenario analogous to his retrofit scenario -- but first let's explore whether his
retrofit scenario is properly analyzed.
·
Amory has been relying for what seems like forever on Lorna Greening’s survey to make
his arguments about income/output and substitution effects (a distinction I believe I introduced to
the academic literature.) The literature has grown substantially in the meantime and there are
new analyses available that provide new and better answers. Steve Sorrell’s UKERC survey
comes to mind. The Evans/Hunt book. The Herring/Sorrell book. The literature is becoming
vast, although it largely originates in Europe.
Thank you for this reminder. I quoted Lorna et al's survey not because it's current but because that's
where Jesse said he got his illustrative elasticity of substitution, so I added some caveats and conclusions
(contrary to his) from the same source.
·
In this connection, Amory could have read my paper twice over in the time taken to write
his analysis. If he had, he would have learned that the idea of a single elasticity of substitution for
energy is a construct no longer needed. Also that we can now use highly general flexible
functions (Translog) for rebound measurement. The Cobb-Douglas paradigm he complains
about no longer applies. Nor CES. If he wants, I can send him an article showing the merits of
using flexible forms for rebound measurement.
Yes, please. This is all good to know. However, I prefer for now to leave those advances to the economic
theorists on this thread, and to focus for now on scrutinizing Jesse's numerical straw example, as I did
below. I'm sorry if I distracted anyone from that task, and am eager to learn what I got wrong.
·
The zero-sum game argument is a red herring. The output effect comes along with a
reduction in output price and represents a gain in both producer and consumer welfare due to the
efficiency gain. Consumers consume more; producers produce more. Anyone wanting an
undergraduate-level explanation for producer-consumer interactions given rebound can refer to
the paper [Theoretical foundations of the rebound effect. In: Herring H, Sorrell S (Eds). Energy
efficiency and Sustainable Consumption: Dealing with the Rebound Effect. Palgrave Macmillan;
2009.]
This sounds like a macro effect, but I thought Jesse was specifically describing a micro effect. (Jim?...)
Also, Lee Schipper 01/28/11 2:06 PM asked: "[H]ow do we account for energy embodied in our huge
import surplus of goods[?] They're not necessarily energy intensive (other than raw materials imports)
but their quantity is so large as to possibly dwarf the rebound effect in manufacturing." Good question.
NAS/NRC's America's Energy Future gives this number, perhaps in the efficiency volume, though I can't
lay my hands on it right now.
Thanks – ABL
from
Jonathan Koomey
2/4/11, Sent at 3:23 PM (GMT-08:00).
All,
As the person who proposed the specific numerical example on rebound, I feel somewhat responsible for
prompting the crowd to continue the dialogue (which seems to have died out), so here goes.
I've attempted to piece together below the full thread on the specific example to facilitate discussions, with
the latest emails on top, and earlier emails following in reverse chronological order. These include (at the
bottom) my initial challenge, Jim Sweeney's explication of his simple calculation, Jesse's calculation, and
Amory's comments on Jesse's calculation. No one has yet responded to Amory's critiques of Jesse's
calculation so I'm writing now to encourage advocates for large rebound effects to address Amory's and
Jim's specific technical points so that we can at least identify the areas of agreement and disagreement.
There is no real substitute for the numerical example and I encourage those who advocate for the
existence of large rebound effects to lay their reasoning out on the table for all to see and evaluate. This
way we avoid talking past each other. If rebound is indeed "physically comprehensible", as Harry says,
then it ought to be possible to work through this simple example and discuss the various parameters that
factor into such a calculation, thus leading to a better understanding of underlying causality. That is what
Amory, Jim, I, and others have been arguing for all along.
Thanks,
Jon
from
Jesse Jenkins
2/4/11, Sent at 3:43 PM (GMT-08:00).
Hi Jonathan,
You've caught me just as I'm preparing to leave for a week's long vacation, so I'm afraid won't be able to
write a thorough reply to Amory's critique.
My brief response however is this: you asked for a simple but specific numerical example of how rebound
works. I provided one, expressly in generic terms, with numbers that were roughly appropriate for an
industrial sector, but clearly not drawn from specific analysis of the example sector I (perhaps foolishly)
decided to specify (e.g. I decided to use a chemicals plant as our example, since it could help us visualize
what I was talking about more concretely, when perhaps it should have just been left as a generic
industrial facility). Amory then proceeded to provide critique of the specific figures I utilized in this generic
example, based on very specific experiences he has from clients within the chemical sector.
I didn't reply with a detailed rebuttal, because the point of the exercise was not to provide a concrete and
rigorously accurate description of rebound at an actual chemicals plants, but rather to provide a clear
sketch of the mechanisms at work, and how they can together sum to quite significant rebound.
I provided this example because you and others had decided to completely ignore a whole set of
published academic research that does provide far more rigorous calculations and depictions of rebounds
at work in various industrial and end-use contexts.
Suffice it to say, I believe having clearly shown how rebound mechanisms operate in plain terms,
providing you with the asked-for explanation of the casual mechanisms at work, the realm of debate
should center now not on Amory's rebuttal to my straw example, but on your as of yet absent rebuttal of
any of the specific academic research on the topic. In other words, let's not ignore the dozens of papers
on the subject at hand, as we instead focus on picking nits over a straw example provided for illustrative
purposes only.
Hope the discussion continues on those grounds while I'm gone, and look forward to continuing our dialog
when I return. Cheers,
Jesse Jenkins
From
Jonathan Koomey
2/13/11, Sent at 8:37 PM (GMT-08:00).
Jesse,
Thanks for clarifying your position on the specific example.
I wrote up a summary of where this leaves us in the attached document, which can also be downloaded
at https://files.me.com/jgkoomey/0aqqfm
The key points from that memo are:
1) In an effort to avoid misunderstandings about the complex phenomenon of rebound, I proposed to an
email list of about 30 analysts and energy/environmental reporters that those supporting large rebound
effects produce a simplified example, so we could dig into the buried assumptions that always afflict such
analyses.
2) Jesse Jenkins of the Breakthrough Institute offered such an example, but failed to respond to
substantive critiques of the assumptions behind that model that reduced the calculated rebound effect by
factors of roughly 10 to 20.
3) Nevertheless, there was general agreement that the relevant research question should be “under what
conditions is rebound a problem, and in those cases, how big is it?” Once this question is accepted,
however, making blanket categorical statements like “energy efficiency never saves energy” (as blog
posts on the Breakthrough Institute’s web site routinely do) is no longer appropriate.
4) The normal burden of proof is on those advocating the existence of some unexpected and novel effect
to show the underlying causal mechanisms that lead to that result, so the assumptions can be peer
reviewed. I can’t prove that large rebounds don’t exist, just like I can’t prove that black swans don’t exist
in the absence of a perfectly accurate universal census of swan colors, but if someone brings me a black
swan, the problem is solved. And that’s what those of us skeptical about large rebound effects continue
to request: bring us a black swan!
We would love to know where Amory and Jim's critiques of your example are in error, and await your
response. If I read your email below correctly you are not inclined to respond, and of course that's your
right. But if I were in your shoes and were able to dispute the technical points raised by serious critiques
that would significantly reduce the size of the claimed rebounds, I would not hesitate to do so. And if
others also want to respond, that's fine too. Until such a response is forthcoming and withstands
technical scrutiny, however, the claim of large and pervasive rebound effects remains unproven.
For those who are time constrained, if you read the 4 points above and the last 4 paragraphs on the last
page of the memo, you can get a pretty good sense of the argument.
Please feel free to email me with any questions.
Best regards to all,
Jon
from
sender-time
Ted Nordhaus
Sent at 12:37 AM (GMT-08:00).
Jon,
Thanks for the highly selective interpretation of the 200 plus emails in this string. I would point out the
following:
1.
You and Jim didn’t ask for the same thing. Jim gave a simplified example of macro economic
rebound and asked us to show how rebound might be substantially larger than he calculated. You asked
us to take a specific efficiency measure and show how it resulted in macro economic rebound. Jim’s was
a fair request and should have served as the basis for further deliberation. Yours was, as has
subsequently become apparent, not offered in good faith nor particularly relevant to the question at hand,
which was how much energy efficiency measures, aggregated at an economy wide scale, would be
eroded by macro-economic rebound. Your request, predictably and perhaps intentionally, send the
discussion back into a debate about micro-economic rebound, the tried and true tactic that has allowed
efficiency advocates to avoid seriously dealing with the macro-economic phenomena for thirty years. The
debate about the scale of rebound at the micro-level is important in its own right, for it is rather
determinative of the scale of micro-economic energy savings that are available to aggregate at the
macro-economic level. But once we posit, as Jim did, that we have reduced energy use by X%, we are
talking about a different set of phenomena that are operating at a completely different scale. Whether
Amory is right or Jesse is right about substitution elasticities within the chemical sector is irrelevant.
Whether rebound is 5% or 50% in the specific plant or industry in question, it all adds up to what it adds
up to at the macro level. And the question analysts interested in macro-economic rebound then need to
ask is “what happens then?”
2.
The demand that you made, Jon, is in fact how efficiency advocates have short-circuited a serious
discussion of rebound at macro-economic scales for decades. Demanding that someone show you how
your compact florescent lightbulb results in macro-economic rebound or arguing about whether chemical
companies invest their energy savings on back massages and more efficiency projects as Amory
suggests or new productive capacity, as Harry suggests is really irrelevant. A discussion of how macroeconomic rebound works starts with positing some scale of initial macro-economic energy savings.
Demanding that we trace those savings back to the light bulb just turns the discussion of macro-economic
rebound into one about micro-economic rebound.
3.
Jesse made a good faith effort to honor your request and was greeted by four claims:
a.
That he hadn’t demonstrated macro-economic effects. This was a
consequence not of a flaw in Jesse’s analysis but a flaw in your demand. You can’t
show macro-economic effects from a single micro-economic case. Now if we
posited that the result of total aggregate efficiency in some sector or industry was X,
and this resulted in an economy wide reduction of energy use of Y, we could then
discuss macro-economic effects, but then you would have attacked him for not
showing directly how the micro-economic measures he was discussing resulted in
macro-economic rebound.
b.
Jim suggested that increased market share from Firm A improving its energy
productivity would come at the cost of Firm B. But this again confuses a micro
phenomena with a macro phenomena. Firm A might capture more market share as
a result, or Firm B might also implement energy saving measures, or improve other
factors of productivity, or sacrifice short term profits to cut prices and maintain
market share. But in all of the cases above, the result is price declines across the
sector, higher levels of demand, and greater production.
c.
Amory suggested with the utmost authority, in one of those, “he just made so
many claims that I can’t keep track of them” moments that are his trademark, that
his clients were reinvesting their energy savings in new efficiency programs. Indeed,
it seems rather remarkable that after 30 years of these practices Amory’s clients
use any energy at all. But even taking Amory’s claims on their face, it seems likely
that after his clients have reinvested their energy savings in every energy efficiency
project that Amory can think of, they fire Amory and hire Harry to figure out how
best to invest their newfound energy savings riches in new productive capacity.
d.
Amory further offered a bunch of anecdotal claims as a practitioner about what
his clients do. Unfortunately the plural of anecdote is not data and Amory’s claims
mostly conflict with the literature that Jesse cited when he specified his
assumptions.
4.
Our fairly exhaustive review of the peer reviewed literature will be out next week. It will document
the fairly strong consensus from that literature that:
a.
Direct rebound in end use sectors of the economy is often a good deal greater
than you and Amory have suggested.
b.
Direct rebound in end use sectors of developing economies, where the lion’s
share of future emissions and energy use will come from, appears to be
substantially greater than studies of developed economies, where most such
studies have been conducted.
c.
Direct and indirect rebound at the micro-economic level in productive sectors
of the economy, where two thirds of energy is used, also appear to be much higher
than studies in end use sectors, where most such studies have been conducted,
also appear to be substantially higher than in end use sectors.
d.
Virtually every study of macro-economic rebound predicts substantial
rebound, with most predicting rebounds of 50% or higher.
e.
That both neo-classical growth theory and alternative models developed by
ecological economists predict backfire.
5.
So to summarize, our view is consistent with the observed record of rising energy use and declining
energy intensity, neo-classical and ecological economic theory, most models of macro-economic
rebound, and pretty much every sector of the global economy excepting direct rebound in end use
sectors of developed economies, and you keep insisting that because the notion is counter-intuitive (to
you and your colleagues) the burden of proof is on us. This position appears to be supported by the
following:
a.
Self appointed claims of authority (“I speak as an empiricist and a
practitioner”)
b.
Studies of direct rebound in end use sectors of developed economies.
c.
Attacks on any and all data to the contrary. You don’t like the Jorgenson data set, or Harry’s
production functions, or the data on historic expenditure on lighting, or Greenings substitution elasticities.
In place of this we get mostly anecdotal claims about how it works in your experience. You guys have
been at this thirty years. You’ve had substantial resources at your disposal. One would think that if you
took the question of rebound at macro scales seriously you’d have invested some time and resources into
developing better data sets that could answer these questions to your satisfaction. But instead what I
mostly see is still more studies of direct rebound in end use sectors of developed economies and this
suggests to me that there hasn’t been a lot of interest in answering these questions in a manner that
might satisfy the apparent concerns that you’ve raised about the work of those who have attempted to do
so because you’re not sure you are going to like the answer.
from
sender-time
Harry Saunders
Sent at 1:19 PM (GMT-08:00).
Jon:
I suggest you re-read my email discussing the numerical example question, this time carefully. Also,
Jesse’s response reiterating the purpose of the rough explanatory example he constructed, as he (and I)
understood the request.
I find myself beginning to question your purpose in fixating on this “numerical example.” It does not
appear to be an attempt to understand the micro-level mechanics—I provided you three concrete,
physical examples illustrating this, which you neither acknowledged nor responded to. I also provided
you actual sector-wide data showing that factor substitution occurs in reality, to which you again did not
respond. And since you dismiss econometrics, we cannot discuss what actual measurements of
substitution elasticities reveal numerically. It does not appear you wish to engage on concrete examples
of the mechanics of rebound, nor on the numerical quantification of rebound effects found in the
academic literature.
I suggest the way forward is for you do the hard work of actually examining the literature that already
contains examples. Specifically:
1.
Read the Druckman et al paper (referenced in the blog), which I suggested very early in the thread
as containing a numerical example we could discuss, and which addresses your “new” suggestion to
focus on re-spending effects.
2.
Read through the Journal of Physics article (Tsao et al.) on lighting, an article subject to aggressive
peer review by scientists, which places the onus clearly on anyone claiming the future will not be like the
past 300 years, a responsibility you have yet to take upon yourself.
3.
Read through my article. Find the fatal flaw and let me know what it is.
I sense a certain zealotry creeping into this discussion. My appeal is that we all calm down and instead
make this about scholarly discourse, which to me requires openness to unfamiliar ideas and willingness
to admit error (on all sides). It would be a shame to have this degenerate into something like what we
have seen among the climatologists and their critics.
In fact, the science of climate change faces a similar challenge to economics. It is science, but not
science like particle physics is science. That is, climatologists cannot conduct controlled experiments
where they, say, run the global climate under certain conditions for ten years and then go back and re-run
it under different conditions. Just as economists can’t re-run history. In the climatology world, this opens
the door to critics who offer themselves as scientists but, given the absence of controlled experiments,
feel free to ignore the (to me convincing) growing cumulative weight of indirect evidence for climate
change.
Let’s not let that happen here.
Harry
from
sender-time
Jonathan Koomey
Sent at 7:27 PM (GMT-08:00).
Harry,
Upon reflection, I realized that I should have given you credit in the memo for making those other
constructive suggestions, so my apologies for that omission.
I worry we're talking past one another and missing the point. This has been a long conversation, and the
purpose of the examples I'm requesting is to simply and clarify, not to obfuscate. If you feel you have
already made your claims about causality clearly, then perhaps you would be willing to summarize them.
I'd also encourage you to respond to criticisms drawn from my previous comments, as well as those
from Jim and Amory. In return, we would be willing to entertain the same from you with respect to the
examples Jim gave, and I promise to dig into the references you cite (see below).
As I've tried to make clear, I'm interested in an example that explains all types of rebound effects, micro
and macro. This is consistent with Jim Sweeney's email to you on 26 January, which stated "I am waiting
to see an illustration of some of the macroeconomic pathways as quantitative examples. The macropathways I see, other than the one I showed as a simple example, are all smaller than the simple
example. But perhaps you can educate me with a basic illustration of a macro-economic rebound impact
that quantitatively is large."
Our request for a specific example is exactly the same one I make to anyone who comes to me with a
question or claim that puzzles me. I've been working with economic models of various sorts since the
1980s and there has literally never been a case when a complex phenomenon could not be explained in
heuristic terms like the ones I requested, as long as it was sufficiently well understood. So please
explain, and feel free to take whatever perspective you need to describe your claims. The point is to get
to the causal links that you believe lead to the results, so they can be peer reviewed.
I have read the Druckman paper and the analysis on which Tsao et al is based. I have not yet delved into
your paper. It will take time for me to synthesize it all, and like the rest of us on this list, I have many
pressing matters on hand at the moment. I promise a review of each of these in due time.
I'm convinced, based on my experience with dissecting complex modeling analyses for decades, that
working through a specific example is the best way to achieve increased understanding, which seems to
be the goal of participants in this discussion. So let's have at it!
Jon
From
Amory B. Lovins
2/16/11 Sent at 9:57 PM (GMT-07:00).
Dear Ted,
You seem to feel that Jim Sweeney's and my criticisms of Jesse's numerical example were fully
addressed by Ted and Harry. Your colleague Michael Shellenberger just posted that your and Harry's
posts had "responded in detail to both Amory and Jim's points." I respectfully disagree. Let me offer my
summary view of their much more limited focus and effect, and of where we stand, and invite you to
specify what in my summary you think is wrong. Frankly, I for one didn't previously respond to your or
Ted's replies because while they contained some interesting ideas, they fell so far short of being really
responsive to Jim's and my critiques of Jesse's example that I've been assuming they were preambles,
and awaiting the real rebuttals.
Jesse's strawman example of hypothetical energy savings at Dow Chemical posited two direct and two
indirect rebound effects, which I'll number #1–4. Jim and I posted critiques of each, often on multiple
grounds. Harry's 28 Jan post suggested that we discuss new greenfield plants rather than retrofits (if so,
an example would be very welcome, and I was delighted to hear Harry may also be about to post a
numerical example elucidating his macro model ideas). Harry's 29 Jan post did try to respond to one of
the four critiques of Jesse's output component, shown as point 2b below. Harry also suggested four other
ways to discuss a different example than Jesse's or not to use a numerical example at all, and said, "I
hope it is not assumed that I am running away from the numerical examples challenge; and that the
above is deemed a reasonable shot at attaining the same end." However, having just re-read his posts, I
don't think he has responded further to Jim's and my specific critiques, other than to repeat that he prefers
macroeconomic modeling. That is of course another method of examining rebound, but as we've seen
earlier in this conversation, it can easily degenerate into theological debates about how many price
elasticities can dance on the head of a pin, whether electric utilities are part of the "productive sector,"
whether big econometric models sufficiently match reality, and other unedifying matters. That is of course
precisely why Jim, Jon, and I requested, and Jesse kindly prepared, a specific numerical example. That
example remains on the table -- systematically criticized and not yet rehabilitated.
You also responded specifically to the same item 2b as Harry did, and, so far as I can see, to no other. I'll
comment under item 2b below on your and Harry's ideas about it.
State of play
So taking Jesse's four rebound effects in return, here's where I think we stand:
1. Substitution component of direct rebound: Based on my decades observing behavior in the sector
(and ~29 others), my 27 Jan critique questioned on about a half-dozen different grounds the behavioral
reality, physical interpretation, and size of Jesse's assumed elasticity of substitution. You rejected but
didn't rebut my observations. You seemed to have two main objections: first, that I offered too many facts
(but I don't think you established that they were wrong or inapplicable), and second, that my observations
conflicted with literature Jesse had cited (but I don't understand which literature you meant or how it
disproved any of my statements). More specificity would be welcome. Please note that all the subsequent
criticisms don't depend on this dispute, because they're all methodological, or for item 2a, numerical.
2. Output component of direct rebound:
a. I showed that Jesse had overstated the chemical sector's energy intensity by >4x, turning a small into
a big rebound effect. Skip then reinforced this point using other metrics; I'd simply used the same metric
as Jesse.
b. Jim and I said Jesse had counted Dow's output gain but not its competitors' equal output loss, which
would incur a net-negative rebound. You claimed on 14 Feb, in your point 3b, that this is more than
overcome by macro effects: "Jim suggested that increased market share from Firm A improving its energy
productivity would come at the cost of Firm B. But this again confuses a micro phenomena with a macro
phenomena. Firm A might capture more market share as a result, or Firm B might also implement energy
saving measures, or improve other factors of productivity, or sacrifice short term profits to cut prices and
maintain market share. But in all of the cases above, the result is price declines across the sector, higher
levels of demand, and greater production. That does indeed sound like a macro effect, which Jesse's
example omitted and you said it couldn't possibly include. It also assumes behavior by Firm B that
effectively removes Firm A's competitive market or creates offsetting negative rebounds (by Jesse's
computational logic) or both. Similarly, Harry posted on 28 Jan that "The zero-sum game argument is a
red herring. The output effect comes along with a reduction in output price and represents a gain in both
producer and consumer welfare due to the efficiency gain." I asked if that wasn't a macro effect, which
Jesse's example didn't try to show. Then on 29 Jan, Harry elaborated: "May be a definitional thing, but I
consider this a micro effect. Consider a situation in which all competitors in a sector engage energy
efficiency gains that reduce their output price. The math shows each competitor’s output goes up along
with this reduction. But this happens for all of them. Consumers consume more overall; producers
produce more overall; and economic welfare is increased. No zero-sum game here. Each producer at
the micro level produces more (and uses more energy to do it)." I may be missing something here -perhaps Jim could enlighten us -- that's a very different situation than Jesse's example of a single
producer that raises its process energy efficiency within a competitive market. Harry is simply eliminating
competitive effects by assuming that all competitors behave identically, so many firms add up to a macro
effect, unencumbered by inconvenient microeconomic competitive effects. Harry, did I misunderstand you
here? I thought the point of a microeconomic numerical example was to examine competitive-market
effects, not to assume them away. Nobody would deny the tautology that if all firms get more energyefficient, all can sell for less and earn more, but it's far from Jesse's example. If we all agree that Jesse
provided a microeconomic example, flaws in that example cannot be waved away by appealing to
macroeconomic mechanisms it didn't include.
c. I noted that Jesse had neglected the [small] negative rebound from sales lost by Dow's energy
suppliers.
d. I noted that Jesse had neglected a tiny third-order macroeconomic positive rebound from energy price
elasticity (macro effect A below).
3. Respending/reinvestment component of indirect rebound:
a. I pointed out this had been double-counted—Jesse had Dow respend its saved energy dollars in step
1, then in step 3 had Dow distribute the same money as shareholder dividends.
b. Jim noted that the efficiency gain's cost hadn't been debited either.
c. I said that even if these two apparent errors were corrected, as an empirical matter, the actual Dow
Chemical, like many smart firms, could and does reinvest in more energy efficiency. (So far the real Dow
has very systematically invested $1b in efficiency for a $9b saving.) It's therefore fallacious to assume
that respent energy dollars will necessarily incur average or greater energy intensity. The more aware a
rational firm is of where its saved energy dollars came from, the more it would feel reinforced in seeking
more such winnings.
4. Embodied energy component of indirect rebound:
a. This too neglected the efficiency gain's cost—contrary to step 2's assumption that the efficiency gain
required an investment.
b. I also suggested several practical reasons why the actual effect should be negligible.
So here's how I would summarize the effects of these ~15 distinct critiques if they continue unrebutted:
1. This key term remains in doubt, subject to very wide uncertainty in size and possibly in sign.
2. This term seems fallacious and wrong-signed.
3. This term seems doubled-counted, fallacious, and possibly wrong-signed.
4. This term depends on inconsistent assumptions and is disputed both methodologically and empirically.
The right number is not zero but is very small.
Net quantitative effect summarized from my detailed 27 Jan post: Jesse's total rebound would fall
from 51% to roughly 2–3%, in line with what Jim, Jon, I, and most analysts would expect a priori.
Presumably this reduction would at least proportionately reduce any macro mechanisms that amplify
microeconomic rebound. Of course, Jesse's example does nothing to establish the reality of large
macroeconomic effects because it posits none (and you said it couldn't because it's micro). Revealingly,
though, his two indirect effects, #3-4, accounted for just one-eighth of his posited total rebound, and
three-fourths of their contribution came from double-counting (step 3). This fits my normal intuition that
second- and higher-order effects tend to be smaller than first-order effects, so integrating over all nthorder effects should converge on a number not greatly larger than first-order effects alone.
A reduction from 51% to ~2-3% rebound (before, but driving, any macro effects) is a very material change
that would seem to call for a specific, point-by-point response. I have seen none yet -- so I think Jon's
description is correct -- and I would be grateful if you and your colleagues would kindly provide one. I'd
love to know where our criticisms erred and how. Otherwise we can't understand why Jesse's example
should be accepted as valid in logic or result. Jon, Jim, and I also hope to see added to Jesse's numerical
illustration any specific macroeconomic mechanisms he may have overlooked. If they are real--even if
they are more than the sum of all the micro parts of the economy--they should be describable in words
and numbers understandable by this expert group. Lacking any concrete illustration of how they work, we
don't see why they should be accepted as real, large, and important. I understand you to believe that
macro effects cannot be illustrated by any numerical example because they emerge only from the
behavior of large theoretical models, but none of us have ever encountered, in any kind of complex
system, at any scale, important phenomena that cannot be clearly described in words and illustrated by
numbers. Do I understand you correctly that you wish us to take on faith that the macro phenomena you
posit are of this previously unobserved kind?
Another observation, relevant to Michael's posts today addressed to Danny and to Jon, is that Jim and I
between us made at least 15 specific criticisms of Jesse's example. You and Harry between us
responded to one of those (2b) -- not frontally or convincingly, but rather by redefining the basic
conditions. In carefully reviewing your and Harry's posts, I can find no refutation of the other ~14 points
either. I therefore feel Jon was correct when he said that Jesse's example remained undefended -- which
I took to mean, as I think any reasonable reader would, unrefuted. To refute each of the 9+ critiques listed
above requires specific engagement with their facts and logic, not mere disproof by emphatic dismissal or
proof by vigorous assertion. I don't think an actual refutation, i.e. a specific argument falsifying the
critique's premises or logic, has yet been provided for any of the ~15 critiques, nor attempted for ~14 out
of the ~15. If I've missed something here, please specify it.
Please indulge me also in a simpler and broader comment about your big-rebound and "backfire"
theories.
My puzzlement about macroeconomic rebound
Your group is asking us all to believe that saving part of the energy use that totals 6–8% of US GDP
somehow triggers a tsunami of complex interactive macroeconomic effects that largely, wholly, or more
than wipes out the energy saving, drives rampant economic growth, and thus causes total energy use to
increase. Even if the complex mechanisms of economic growth were well understood and agreed upon, I
for one don't understand what kinds of plausible mechanisms could have this radically amplifying effect.
Please help me understand this better. I think your hypothesis of macroeconomic amplification rests on
the combined effect of three postulated mechanisms:
A. market price: e.g., large and widespread automotive efficiency gains would drive down world oil
prices, increasing demand for many oil-using goods and services. RMI's analyses such as Winning the
Oil Endgame (Sep 2004) and Reinventing Fire (Sep 2011) have explicitly considered such price effects
and found them to be small. WTOE considered them both at the consumer level (at p. 41 we found a net
1.5% effect and then explained a conservatism that should more than offset it) and at the global level (p.
35 noted that full global, not just U.S., adoption of our huge efficiency gains would still leave oil prices
above the cost of at least half of our efficiency potential; oil depletion meanwhile would tend to raise oil
prices and focus OPEC's market power; fast-growing developing economies could meanwhile want to
take up any demand slack; lower oil prices don't cause owners of efficient vehicles, factories, and
buildings to switch back to inefficient ones; and if oil gets too cheap, it can always be taxed). I also run
into a recursive puzzle here: it's arithmetically obvious that this macro effect can't have much macro
impact on price and hence demand, in a pretty fungible global commodity market like oil, unless total net
energy savings (especially in China, India,...) are large -- which your hypothesis says can't happen
because efficiency can't actually save much if any energy, for reasons more important than this one.
B. composition, as when the previous effect plus more efficient production processes will favor, and
increase the share of, energy-intensive sectors of the economy, increasing demand for their output.
Evidence for the scale of such effects is very limited, as I think you'd agree. I also think the forces that
have led to very durable long-term decreases in the use per $GDP of energy-intensive products like steel
and cement (as shown by Ayres, Princeton, and others) are driven far more by functionalities than by
relative prices. Similarly, the coming shift from steel to advanced polymer composites in automotive
structures is driven far more by radically simplified manufacturing and >2x investment reductions than by
the relative price of carbon fiber and steel per unit of functional performance.
C. economic growth: saving energy is on balance stimulative, helping economic growth continue to
outpace efficiency gains. The scale and function of this rebound effect is much disputed, partly because
there isn't general agreement about the basic mechanisms that drive economic growth. I agree with your
observation, Ted, that if (for example) my office retrofit improves labor productivity, that could greatly
enhance my competitive advantage and grow my business -- but then we're back to Jim's zero-sum issue
in Jesse's step 2b above. You're asking us to believe that a rising tide (even a perpetual tsunami) lifts all
boats far higher than can be explained by the well-known Newtonian mechanics of lunar gravity, and I'm
trying to understand why that should be so.
Both at a micro and at a macro level, therefore, I continue to regard your group's claims that rebound
prevents energy efficiency gains from saving much if any energy as speculative, unproven, and
implausible. (The clearest explanation I've seen for the "implausible" part is in economist James Barrett's
lucid commentary on David Owen's piece, at realtimateeconomics.org/wp/archives/647 and /654.) No
doubt you and your colleagues sincerely believe your position to be true; but wishing doesn't make it so. I
am unconvinced by anyone's theology. I'm an empiricist. Please show me the numbers, anchored in a
concrete example like Jesse's so causal links can be clearly described and there is no seemingly
ineffable gap between model and reality.
Finally, your 16 Feb post reiterates your contention "that despite all the energy efficiency we've been
doing, we continue to use increasing amounts of energy -- and that there is a growing body of peerreviewed evidence pointing to the strong connection between those two things." That is the crux of the
matter. "Connection" does not mean "explicable and demonstrable causal logic." From all the evidence
you've adduced so far, "connection" appears to mean only "correlation." Yet you are drawing conclusions
from correlation as if it were causality. Barrett's second blog mentioned above nicely disentangles this
confusion. As to your comparison between efficiency effort and rising energy demand, my New
Yorker response to David Owen's article noted that the oft-supposed inevitability of GDP growth's
outrunning energy intensity reductions has proven false in nine of the past 34 years (through 2009) in the
US. I further asserted that it could be made systematically false every year by paying attention to energy
efficiency (especially by barrier-busting) -- much as the US did with oil, whose use fell 17% during 197785 while real GDP grew 27%. Thus that US primary energy consumption rose in 23 of the past 34 years
does not mean energy efficiency isn't working, but only that we're not yet buying enough of it fast enough
-- not because we couldn't or shouldn't. Still less does it mean that doing so wouldn't save much if any
energy.
Concluding thoughts
Our shared purpose of mutual learning as a diverse virtual community of interest, together wondering in
the bewilderness, is not well served by the incivilities and imputations that have lately marred this
conversation. A third party's choice to write an independent blog containing acerbic comments offensive
to some of you is no excuse for unworthy behavior between ourselves -- let alone for blaming other
people, notably Jon, whose contributions here have in my view been uniformly fair, accurate, thoughtful,
and courteous. I also want to appreciate here Harry's integrity and graciousness in this group's pursuit of
truth, and am very much looking forward to his forthcoming example to help us better grasp the macro
effects he posits.
I hope this conversation can still serve its intended purpose by answering a very simple question, pending
now for three weeks: were Jim's and my specific methodological (and in case 2a, numerical) critiques of
Jesse's example correct, and if not, exactly why not? Please address that burning question, preferably
using my numbering system above so your rebuttals are clearly linked to their subject. Then we can all
start to understand what the arguments are and who is right. Thank you.
With best wishes – ABL
from
Ted Nordhaus
2/17/11, Sent at 7.24 PM (GMT-07:00).
All,
Jesse will be sending on a detailed reply to Amory and Jim shortly. However I will note that once you cut
through the blizzard of arguments and counter-arguments, the lion's share of the difference in rebound
magnitude at the micro-economic level comes down to your view on substitution elasticity. If you think it is
high you will find large rebound effects, if you think it is low (or in Amory's case non-existent) you will find
small effects. Jesse, responds to Amory and Jim's further criticisms as well, and notes that his numerical
example did not include macro-rebound effects because you asked for a simple example of large rebound
resulting from a specific improvement in energy efficiency. In the interests of moving the conversation
forward, I've recommended to Harry that he start by offering a more refined numerical example that only
focuses on direct substitution and output effects in order to narrow the analytical focus of the discussion.
I don't doubt that the low rebound advocates on this email will contest most or all of Harry's assumptions
regarding substitution and demand elasticities. Unfortunately, those differences are ultimately irresolvable
without using econometric analysis to estimate sector specific elasticities, a method that Jon and Amory
have both rejected out of hand. Without econometric or some other method of quantifying those
elasticities, Amory and Jon's position that substitution elasticities are low, and rebound effects therefore
small, is unfalsifiable.
Ted
From
Jesse Jenkins
2/17/11, Sent at 7:59 PM (GMT-08:00).
Dear Amory et al.,
What follows is a point-by-point survey of our discussion of our little numerical example to date, following
the helpful structure outlined by Amory in his latest email.
One point of clarification before we begin: this numerical example has remained focused on direct and
indirect micro-economic effects. It limits the scope of analysis to this scale because Jon Koomey and
Amory Lovins had previously insisted we discuss a simple, concrete example, and simple is only possible
if we restrict our analysis to the micro. Trying to figure out what all the macro-responses are to one microscale change in one Dow plant is pretty impossible. There are simply too many threads leading off into
the various corners of the economy to fully disentangle.
If we want to look at macro-scale effects, we have to look at aggregate changes across a sector or
economy: for example a 30% improvement in energy efficiency across the chemicals sector as a whole,
or across all end-use energy uses. The principles and mechanisms here are similar, except in the
aggregate, and also clearly understood (substitution and demand responses, growth effects,
compositional changes in the economy) and would also include the aggregate impact of all of the microscale direct and indirect effects discussed herein. This is appropriately the domain of macro-economic
modeling approaches (which Jon has pretty much rejected at this point). Moving to the aggregate and the
macro would force us to into far more abstract grounds, however, than the clear, easy-to-understand
example that Amory and Jon requested. I note this because Amory in his various replies both chides us
for not considering macro-rebounds, then at one point flags that we are (inappropriately) including macroscale effects (which we in fact were not).
Also, I’d like to note that while Jon Koomey’s 2/13 summary of the debate at that point in time conflates
both Jim Sweeney and Amory Lovins’ replies to my numerical example as clear critiques, Jim’s response
actually includes general agreement with all but one of the rebound mechanisms I outlined. While I deal
with Jim’s response to each point below as well, it should be noted then that the criticisms of my example
arise primarily from Amory, not Jim.
Ok, on to each point…
1. Substitution component of direct rebound.
As a reminder, for this effect, I chose a middle-of-the-road substitution elasticity value of 0.5. Using this
value, and an estimated energy savings of 30% at our example chemical plant, I calculated a substitution
rebound effect of 35%. My initial calculations:
So a 30% energy efficiency improvement yields a 30% reduction in the implicit price of energy
services, so with a substitution elasticity of 0.5 that means that Dow increases their
consumption of energy services by 15% following the efficiency gains, or an increase in raw
energy demand of 70 energy units*15%=10.5 units, or a direct rebound due to the
substitution component of the rebound effect of 10.5/30 = 35%. The firm now consumes
80.5 units of energy, not 70.
In reply, Jim Sweeney (correctly) observed that “the rebound effect cannot be one number” and is
different in various sectors and contexts (which is exactly right and fully consistent with the literature
summarized in our review). He then went on to note what appears to me to be a general corroboration of
my estimate here, writing on 1/27, “I agree that with the substitution effect, as one of the micro effects.
For many industrial processes this may be large. For many residential and commercial uses of energy it
is likely to be smaller.“ Jim offered no apparent criticism of my calculations above.
In contrast, Amory replied on 1/27 by challenging the idea that substitution of various factors of production
actually occurs in industry. He rejected all estimates of elasticity of substitution, writing “This is your most
important term, and the one where as a practitioner I am having the most trouble coming up with a
plausible physical interpretation of this generalized literature elasticity of substitution.” Against the idea
that firms will rearrange their various inputs to maximize profits to the best of their ability – represented in
the field of economics through estimates of elasticity of substitution – Amory cites his anecdotal
experience as an efficiency consultant to assert that substitution effects must be small. Then, instead of
proposing a more appropriate estimate for substitution in his view, Amory went on to entirely ignore the
substitution effect throughout the remainder of his reply, thus eliminating entirely the 35% rebound I had
calculated do to substitution.
One may question whether substitution is higher or lower than the mid-range value I assumed in my
example, and in different sectors it certainly is higher or lower than 0.5. And in the short-run, substitution
is probably smaller than in the long-run. Amory focused entirely on short-run examples while Harry
Saunders replied on 1/28 with illustrations of how longer-run substitution is far easier. Let’s also be clear
that if we are primarily concerned with efficiency’s ability to contribute to climate mitigation strategies – as
we are at the Breakthrough Institute and as our literature review clearly states is our concern – then we
must consider long-run effects over the decadal timescales relevant to climate concerns.
In short, to entirely eliminate the substitution effect, as Amory has done, is to assume that firms are not at
all capable of finding new and creative ways to arrange the various factors or production in order to
maximize profit – even as Amory makes a business out of helping firms productively rearrange factors of
production to minimize energy expenditures. This seems preposterous, and I believe most economists
and business managers alike would agree.
The size of rebound due to substitution effects depends, as I’ve always stated, on the ability of firms to
respond to changes in the price of energy services by substituting now cheaper energy services over time
for other factors of production. That ability is certainly not negligible, as Amory has assumed, and in many
cases can be quite large. If this remains difficult to grasp in concrete terms, I believe Harry will reply in
the near future with several concrete examples of how substitution can work, but suffice to say, for now,
Amory has not offered any sufficient basis to claim that rebound due to substitution effects is
effectively nil, as he did in his earlier rebuttal.
2. Output component of direct rebound:
For refresher, here is my calculation of the output component of rebound:
For this, we'll assume that energy makes up 15% of total costs for Dow. This is an energy
intensive sector -- chemicals -- so we'll assume something a bit higher than the 6-10% of costs
that may be more typical, but not too high (and after all, energy intensive sectors are where we
often look hardest for efficiency opportunities). The efficiency gains and the cost-effective
substitution of now-cheaper energy services for other inputs to production (which will still yield
net cost savings) thus reduce Dow's bottom line by 15%*30%=4.5%. The management then
decides to choose to pass this savings on in the form of lower prices of their products,
undercutting their competition and driving greater demand for their products. I don't know what
elasticities of demand are for chemical inputs, but I assume they are fairly elastic, so let's
assume it's 0.8. That means demand for Dow's products rises 0.8*4.5% = 3.6%, driving up
demand for all factors of production, including energy services, by 3.6%. So that's another
80.5*3.6% = 2.9 units, or a rebound from the output component of direct rebound of
2.9/30 = 9.7%.
This calculation was challenged on two grounds.
1. First, Jim Sweeney (in his reply of 1/27) challenged this assumption by pointing out that if Dow’s
increase in output comes at the expense of competitors’ market share, then the offset energy use
associated with production at these (presumably less efficient) competitors would more than offset any
rebound due to output increases at Dow.
Harry and Ted made two replies to this:
·
First, Harry pointed out (on that this zero-sum dynamic only applies if we assume
no change in overall market size for the chemical products in question. This assumption
appears erroneous over the long-term, as competitive pressures will drive down prices for
products throughout the sector, increasing demand for the products and increasing
overall market size. “The zero-sum game argument is a red herring,” Harry wrote, as
“consumers consume more; producers produce more.” It’s not a zero-sum game, for
even as Dow takes market share from competitors, the size of the pie they are competing
for increases.
·
Second, while there may be some countervailing effects here, if demand for the
products is not sufficiently elastic, Ted noted (on 2/14) these are macro-economic
questions of where final and intermediate demand for products settles in the longer term
after a price change begun at Dow is best captured in the kinds of CGE models or
integrative models that feature prominently in the literature (and which have been ignored
in this debate so far). Some of these effects will add to rebound and others will detract,
but our simplified example explicitly excluded full consideration of macroeconomic
effects, so as to keep it simple enough to offer easily calculable examples here. Amory
criticizes us for not showing this, even though our restricted scope is due to a good faith
effort to respond to Jon’s request for a simple example, which originates at a specific
plant.
2. Amory challenged my estimate of the share of energy expenditures in a firm’s overall expenditures.
He claimed “It’s roughly 1-2% for the average U.S. business … For the relatively energy-intensive U.S.
chemical sector, the average intensity … was 3.6% in 2008,” citing an EDF fact sheet as the source of the
chemicals sector figure (no source is provided for the average figure). He argued that I had therefore
over-estimated the output effect by a factor of roughly 4x, and he then recalculated my rebound estimate
of 9.7% from output effects given his lower estimate of energy as a share of expenditures, going to (9.7%
/ (15/3.6) = 2.3%
I was admittedly a little surprised by this. I had cited Greening et al. (2000) in initially writing that energy
typically makes up “less than 10%” of a firm’s total production costs (p. 297), while energy expenditures
make up a much larger share of overall GDP (about 6-8% I believe) and about 10% of per capita U.S.
consumer expenditures (according to Boyce and Riddle 2007). How then could industry, particularly the
energy-intensive chemicals sector, expend so much less of their overall expenditures on energy than
either the national average, or a typical household consumer? However, upon looking further, it appears
that the Jorgensen data set shows energy share of production costs for the U.S. chemicals sector to have
averaged about 8% in 1980 rising above 10% for a time before falling to about 4.5% in the latest date
reported, 2005. So Amory’s value here may not be far off, although the apparently increasingly-efficient
U.S. chemicals sector is probably still non-representative of similar industrial sectors in developing
economies, where energy is likely a larger share of firm’s costs. Furthermore, historic cases of rebound,
of the kind considered by Saunders in his latest paper, would also be higher (as energy was formerly a
larger share of production costs).
In summary, rebound due to output effects in our US chemicals sector example would thus be
reduced to about 2.4% under Amory’s assumptions. Lower or higher values are clearly possible for
different sectors and national economies and different periods of history, depending on both the share of
energy in that sectors production costs, and the elasticity of demand for their products.
[As a side note from our chemicals example, the equivalent direct effect (the income effect) for enduse energy consumers will be larger (although substitution effects likely smaller), as consumers
will respond to final prices of energy services directly, responding based on their elasticity of demand.
Hence the roughly 10-30% direct rebounds typically found for end-use consumers in rich nations (as
summarized by Greening et al. 2000 and Sorrell 2007). With greater elasticity of demand and less
saturation of energy services in developing nations, direct rebounds here will be larger still (with
studies showing a range from 40-80%, although this has received pitiably little study to date given the
importance of developing nations to future energy trends).]
3. Re-spending effects.
As I noted, since rebound is not 100% yet due to the above factors, our firm still saves money that can be
used to re-spend or re-invest. Using the same assumption that energy was 15% of the firm’s expenses
and assuming the freed up money re-circulated in an economy where 8% of expenditures eventually
flowed to energy, then total rebound from this effect would be 4.33%.
There were several responses here, which I’ll take in turn.
1. Jim first responded by saying “I agree with the indirect re-spending effect,” but noted that I had not
incorporated the cost of the efficiency upgrades themselves, which would re-duce the money available for
re-spending. This is correct, and if I correct my omission, and use the same assumption as for the
embodied energy effect calculation below, namely, that the efficiency measures save five times more
than the initial cost, then I should reduce this effect by 1/5th (the up-front costs are 1/5th of the amount of
energy saved). So the corrected value would yield an indirect re-spending effect of 4.33% * (4/5) = 3.5%.
2. Amory’s response first asks “wait a minute, I thought you’d allocated all of the saved energy dollars
to cutting costs when calculating your output effect.” I believe Amory is actually correct here, and
apologize if this was a key error. If calculated in this manner our example here, re-spending
effects would be minimized or eliminated here. I actually noted something to this effect in my literature
review writing that competitive pressures would encourage savings to be passed in on prices (triggering
output effects) but limiting re-spending effects for productive sectors of the economy.
[As an aside, one macro-scale effect here is that as the energy savings are passed along to consumers in
the form of lower product prices, and if the elasticity of demand response is not 1.0 (as in our example
here), then consumers will wind up saving money on the purchase of the firm’s products, and then respend that money elsewhere in the economy, etc. Again, this is outside of the scope of our example here,
but just wanted to raise one of the many complex threads snaking off here…]
[Also looking beyond our specific sector here for a minute, I should note that re-spending effects will be
more pronounced for consumer energy savings, where they do not have any such competitive pressures,
and will surely re-spend any net savings on other goods and services, or invest such savings in financial
instruments which will increase net investment elsewhere in the economy (being re-lent by banks, etc).
So a re-spending effect on this rough order of magnitude would be likely for an equivalent energy
savings amongst consumers (where direct fuel and electricity use makes up about 10% of
expenditures and embodied energy in other goods pushes this upwards, according to Boyce and Riddle,
2007).]
3. Amory asserts that firms will re-invest savings from efficiency in more efficiency, rather than in other
areas of production. While this is certainly possible, it is equally likely that firms will re-invest savings in
corporate dividends, or in their areas of core business, where their internal ROI is likely higher than
efficiency measures. It all depends on the precise case, but using an approximate economy-wide value
seems appropriate for this rough sketch (although we’ve already eliminated this factor in the case of
producers).
4. Embodied energy effect.
Assuming a 5:1 payback over the life of the efficiency measures, and assuming energy makes up about
8% of the share of average expenditures in the economy, I calculated that energy embedded in the
efficiency investments themselves eroded 1.7% of the net energy savings.
Jim’s reply noted, “I agree also with the embodied energy effect,” and noted he had also included this
effect in his example.
Amory replied with an anecdote-filled discussion of how the embodied energy may differ from case to
case, which is certainly plausible, but didn’t refute the assumption’s I made for our hypothetical example
here. So this effect stands unchallenged, adding 1.7% to total rebound.
Conclusion
In summary, Amory offered no justification for entirely eliminating the substitution effect, nor any
alternative estimation of a more accurate elasticity of substitution. I’ll thus continue to assume that the
35% rebound I calculated is a plausible hypothetical example for our straw man discussion here. Higher
and lower values are reported in the literature, and I chose a middle-of-the-rode value of 0.5, so in this
case, our straw example and the various studies in the academic literature seem to be consistent. The
output effect may be around 2.3% for our specific example, but can be higher depending on the
assumptions one has about share of energy in production costs and elasticity of demand to a firm’s
products. Again, I’ll defer to the literature here for various other sectors. I do concede that in this case, we
assumed competitive pressures result in lower product prices that eliminated net energy savings, and
while contributing to output effects, means that I erred in counting an additional re-spending effect for our
specific chemicals sector example. Finally, my calculations of the embodied energy effect went without
substantive challenge, so we’ll continue to assume a rebound of 1.7% due to this effect. That would still
yield a total rebound due only to direct and indirect micro-economic effects of 39.1%. That is hardly
insignificant, and is entirely consistent with the body of literature delving into the operation of rebound
effects in more precise terms than this hypothetical example.
As Ted noted (and as Harry and I have been clear about in each of our prior emails) the scale of the
micro-scale rebounds in this example are thus primarily dependent on substitution effects. He will offer
some clarifying examples of how substitution works in industrial sectors, which I’m sure we can discuss
further.
Furthermore, the macro-economic impacts of aggregate energy efficiency improvements like this, say,
across the entire chemicals sector, would include other macro-scale composition, economic growth and
market price effects as well, but that is not the focus of this simplified numerical example. These
readjustments of the economy operate under the same basic causal mechanisms – elasticity of demand
and substitution in response to the changes in final and intermediate prices for goods and services, and
the contribution of productivity and income growth due to these savings to economic growth – each
clearly understood in the economics discipline. While precise values will depend on the various methods
to estimate the final result of these many economic readjustments, this is the domain of macro-economic
modeling work, which is surveyed in our literature review. To summarize the literature again:
·
CGE models show rebounds ranging from 15% to greater than 300% (serious backfire) for
various national economies (thus excluding any effects outside the bounds of the national borders), with
most studies finding rebounds of 30-50% or higher.
·
Integrative modeling study of the global response to wide-spread efficiency improvements (e.g.
Barker et al. 2009) finds rebounds totaling 52% of expected savings.
Finally, and at the risk of introducing more topics to this discussion, I should note that while these
examples do not illustrate high potential for backfire (rebound > 100%) that is because they ignore the
two dynamics most likely to lead to backfire: multi-factor productivity gains, and frontier effects.
I described these in my initial email summarizing the state of the literature for this group, and in detail on
pages 42-48 of our literature review. Here is the summary from my email dated 1/25 again:
•
In the end, any of these analysis will also find it impossible to project in advance what
I've dubbed "frontier effects" that result when efficiency gains unlock whole new areas of
the production possibility frontier, leading to potentially vast new markets, or even whole
new industries for energy services. This is the dynamic driving the backfire in energy
demand that Jevons famously comments on, as tremendous gains in steam engine
efficiency enabled enormous areas of new and productive uses for steam engines and thus
coal consumption, sparking entire new industries and activities, from railroads to steel
consumption and the core of the industrial revolution itself. Steve Sorrell correctly notes that
this dynamic is most likely for 'general purpose' technologies from lighting to lasers to
microchips and computing, to steel and other metals production, as well as electricity
generation itself, and here is when backfire is most likely. These effects are extremely hard
to predicts in advance, yet their frequency in economic history should give us pause.
•
Furthermore, if efficiency improvements lead to gains in other factors of productivity
along with energy -- e.g. labor or capital productivity gains -- rebound effects can be
compounded and multiplied. As Amory and Skip and other efficiency advocates often (and
correctly) note, improvements in energy efficiency often come with improvements in labor or
capital productivity as well, as efficient day-lit buildings boost labor productivity in offices, or
better lighting or electric motors can drive big boosts in productivity on the factory floor, for
example. Here again, much larger rebounds are possible. [And to drive this point home
with an addition: in a 2005 paper, Amory notes that these multi-factor “side-benefits” of
efficiency upgrades can be “one or two orders of magnitude more valuable than the energy
directly saved.” This would imply that re-spending and output effects would also be one or
two orders of magnitude larger! You don’t have to add more than one or two zeroes to the
rebound figures above, before you are in serious backfire territory. The better the efficiency
upgrades are economically, the bigger these rebound effects are going to be. While that
makes the business case for efficiency all the better, and Amory’s talents all the more
desirable for firms and businesses, it makes the long-term climate benefits of such actions
all the more suspect.]
So once again, I reiterate the conclusions of our report: rebound effects are real, significant, and can no
longer be ignored in energy analysis and policymaking. They combine at economy-wide levels to erode
much and in some cases all of the energy savings expected. In other words, rebound means that for
every two steps forward we take in climate terms through below-cost efficiency measures, rebound takes
us one (or more) steps backwards.
If we don’t want to risk coming up dangerously short of climate stabilization goals – and no one here
wants that! – then we should all be very keen on launching much greater study of precisely when
rebounds are small and when they are large, while placing greater emphasis on the other major climate
lever at our disposal: decarbonizing the supply of energy itself through zero-carbon energy alternatives.
Thanks for your patience in bearing with this response,
Jesse Jenkins
_________________________
Director of Energy and Climate Policy
The Breakthrough Institute
http://thebreakthrough.org
from
Amory B. Lovins
2/17/11, Sent at 8:25 PM (GMT-07:00).
Jesse, many thanks! I can see at a quick read that this should help to reestablish the sort of point-by-point
discussion we'd sought, and that I for one will have some things to say next...but not tonight because of
severe book deadlines, so I'll get back as soon as I can. Meanwhile, I appreciate your return to this
specific example, and hope others will begin the next stage of commenting on it meanwhile. Thanks also,
Jesse, for attaching my 16 Feb summary below so people can check which points have been addressed.
My original critique was posted 27 Jan.
It would help me if you could meanwhile kindly reflect and comment on whether you disagreed with my
specific comments on the physical interpretation and behavioral reality of the substitution elasticity you
posited, and if so, where you thought they erred. I don't think my 27 Jan comments rejected the whole
concept that firms substitute between factor inputs to maximize profits. Rather, I cast doubt on how this
would actually work in an energy-efficiency retrofit in a chemical manufacturing company, and hence
whether the actual effect would be material. Though the details of those comments would change in a
different sector (I've worked in ~30 sectors), they do illustrate the practical difficulties of applying your
theoretical concept, in just the same sense as your having picked a chemical company for your strawman
example was meant to lend concreteness to our previously abstract discussion. So since we're
hypothetically in a chemical plant, and Harry also knows what that's about, I feel it would be useful to stay
there awhile until we've clarified any important differences between theory and practice: this example is
as good as another, only better because we've got it and are well into it.
Best – ABL
from
Harry Saunders
2/18/11, Sent at 10:46 AM (GMT-08:00).
Hi, gang:
Apologies for the delay. Attached are two things: one, a user-friendly interactive model (Numerical
Example 2-18-11.xlsm) for us to work with; and two, a Word document describing a particular case I
personally feel is conservative. The Word document also contains a short User Guide at the end. As you
will see, the model allows you to change assumptions to your liking and see the rebound that results.
In the Word document, I have tried to be very explicit about the mechanisms at work. I know (or at least
expect) we will have vast differences in our views of the appropriate quantifications. But I’d like to think
this model is a very honest shot at addressing Jon’s request to help us visualize things in concrete
numeric terms, as a way to advance our discussion and, I hope, our common understanding of what is
really going on here.
Harry
[Word doc pasted here]
Numerical Example of Rebound
Harry Saunders
February 17, 2011
I see my purpose here as providing a numerical example we can work through, debate
about, and see if it doesn’t advance our common understanding. I anticipate the numbers
used will be hotly debated, and that’s fine. I may have made errors, and that’s fine, too.
We’ll all learn something, I’m sure. To this end, I have created the example as a working,
interactive model (Excel workbook) designed to allow anyone to enter assumptions to
their liking. And play around with it. And try to break it. And if so, try to fix it. I have
limited the model, at this point, to direct effects, as Ted recommended.
I’m hopeful that, at very minimum, this example will illustrate the mechanics of
how such rebound works, even if we disagree on the magnitudes. I assume nobody will
dispute that this is at least part of our goal here. It’s a big part of mine. It’s what I hear Jon
asking for.
The Excel model is quite user-friendly. Neither the spreadsheet nor the (extremely
simple) Visual Basic code is protected, in the spirit of open-source collaboration.
Below I work through an example with the particular numbers I have used. At the
end of this document is a short User Guide to working with the model.
I have avoided anything like numerical substitution elasticities, since that will
without doubt get us into endless debates. The example below represents, to me, a middleof-the-road to conservative set of assumptions, but I can cook up assumptions that give
rebound magnitudes as low as 17% if I seriously stretch my credulity (or easily as high as
80% plus). I’m sure Amory will find a set of assumptions that delivers .001% (sorry for the
barb, Amory—meant in jest ).
Example
Reference (Base Case)—Old Plant
I am a manufacturer who has an old plant and is contemplating building a new one. The
input proportions for my existing plant look like the following:
[these factor proportions reflect actuals for 1996 for the US Chemicals industry, but you
can change them to anything you like in the model.]
Here’s what my plant operations look like in my existing plant:
[you’ll see this when you open the model.]
You can see the factor and output proportions in physical units (I have set factor
prices and the price of output to unity). Capital, labor, energy and materials are designated
as K , L , E , M , and real output is designated as Y . I have set energy to a value of 100 to be
consistent with Jesse’s example. The price of output (what I can sell it for) is P .
To the right (green cells) are two measures of profit. One I call “Economic Profit,”
which is how economists think of profit—revenue minus payments to all factors (i.e.,
). It considers that I must pay some return to my
PY pKKpL
L pEEpMM


shareholders. This is not profit as corporations would normally report profit—my income
statement can show positive profit even if I’m not meeting the threshold return. Also,
accountants, god love ‘em, take extreme liberties with the timing of inflows and outflows in
reporting profit. To bound this problem, the second measure I call “Accounting Profit,”
which is payments to all factors except capital and actually is more like ongoing cash flow
(excluding investment costs). In reality I should be using an NPV calculation: if someone
insists on this I can do this, but I’m hoping these current measures adequately convey the
concepts. I’d rather not deal with taxes, depreciation schedules, etc., etc., etc. …
You can see that my Economic Profit is zero. You can think of this as me being the
marginal producer in this sector. Still, I am realizing positive cash flow, so I won’t shut the
plant down. Rebound is showing as zero since we’re using this as the reference case.
New Plant with RMI “Secret Sauce”
Now I want to build a new plant. I plan to make it the same size (output capacity) as my old
one. But I want to design it a little differently. My engineers have told me about the Rocky
Mountain Institute and have suggested I bring in Amory to help me make it more energy
efficient. Amory works his magic, showing me how I can significantly and profitably reduce
my energy costs by some combination of better heat recovery, more efficient pumps,
reduced heat loss, and a whole bunch of clever things I had never even thought of
(evidently returning to his hotel room each night to consult pentagrams, sacrifice small
animals, etc.). Amory presents a plan that will allow me to reduce energy consumption by
30% compared to my old plant. [This seems to be one of the few numbers agreed upon in
the last round of numerics.]
My new design looks like the following:
All my other inputs stay the same (Amory might include a small added capital cost),
but my energy consumption is now not 100, but 70. This is decidedly cool, but I almost wet
my pants when I look over at the profit gains—30 for Economic Profit and 195 for
Accounting Profit/Cash flow.
Amory leaves my office with a big bonus for his team and my promise to spread the
word. I contemplate kissing him before he leaves, but then think better of it, fearing my
wife could catch wind of it and misinterpret.
Economists will describe this type of production change as “Leontief”—that is, one
factor input alone is reduced via efficiency gains, output is preserved, and nothing else
changes. (It is commonly viewed by these folks as a far-too-simplistic view of how the
world really works.)
New Plant with Added Automation
Now my engineering team tells me we can make things even better. They have found a new
system that can automate some of my processes. It is costly, but the cost can be offset by a
reduction in my labor force. They present me the following design:
It requires I spend 5% more capital for this plant, but it will allow me to operate
with 10% fewer workers. [Not to be too specific to Chemicals, these workers in my old plant
may have been used to load/unload materials, move inputs and products around my plant—
now done, in this design, with conveyor belts, automated assembly of electronic components,
automated operation of textile machinery, …etc. And now I have fewer administrative types
to do payroll, manage operators…. Or, perhaps I can provide my construction workers more
sophisticated tools or equipment that, while they may be more energy efficient than the old
tools some of them have or equipment they have been using, allow them to accomplish much
more in less time. With this new technology, I don’t simply hand my workers more energy
efficient tools to replace any they already have (which would imply that I use less energy to
run them, reducing energy use on a one-for-one basis with the efficiency gain, Leontief-style);
instead, I use fewer workers with more, new kinds of power tools and equipment that do more
things, allowing me to build the same structure in the same, or shorter, time frame with fewer
workers. …Or, in the financial industries sector, we have seen huge substitution of computers
for manual labor (far fewer workers needed to produce the same services—huge gains in
labor productivity), representing substantial substitution of energy (and capital) for labor.
...and so forth.].
However, this automation is electricity-intensive, and I find I’ll need to use 5% more
energy.
I’m looking at the profit consequences and am pleased, but I’m also a little pissed.
“Why the h***,” I ask, “did you not tell me about this when we built the old plant?” They
respond by explaining to me that they had indeed looked at this option at the time, but the
systems available then were both extremely expensive, capital-wise, and were notorious
energy hogs. “We couldn’t see how to do it with these old systems without actually
reducing our profits.”
I am mollified, but as a closet rebound analyst, I do take note of the fact that this new
proposal will lead to a 23% rebound in my energy use, even though my output capacity is
the same. I reflect on the fact that this rebound is the result of a substitution of energy (and
capital) for labor.
New Plant with Lower Quality Materials
I’m feeling pretty good about all this when one morning my chief engineer enters my office
to tell me about an idea one of the junior engineers has conjured up. It appears there are
low-quality materials available on the market for cheap. These are currently used in
processes that produce low-quality outputs for a “trash” market. But this engineer has
discovered that we could reconfigure our processes to extract what we need out of these
materials provided we up our energy use. [In chemicals, this may be using new reactors that
up the severity; or maybe I can use lower quality ore to produce my primary metals; or maybe
I can use discarded log trimmings to create quality wood products if I “cook” it better; or
maybe I can take a deeper “cut” into the bottom of the barrel and use heavier crudes…].
When they put together the economics for this new process configuration, my new plant
looks like the following:
My capital has increased 10% (my reactors are more expensive); labor is
unchanged; energy use goes up 5%; materials use goes down 5%. [I have chosen to reduce
the quantity of materials, but I just as easily could have reduced their price 5%.]
My profit goes up further. I feel my currency/stroke/standing with my Board of
Directors going up by the minute. I can almost smell the bonus that is coming my way. But
I do take note of the fact that rebound has increased to 36%, and reflect that this is as
substitution of energy (and capital) for materials. [Implied in this is a quantitative
elasticity of substitution, but please, let’s all agree to keep this a secret from Jon. He may
decide to demand some kind of econometric measurement to validate it….]
New Plant with Higher Yields
Thus far we have contemplated only factor substitution effects. But this is only part of the
story.
My superstar engineering team has come up with another new idea. God exalt and
preserve ‘em. They say we can, through some chemical engineering magic, extract even
more product from these low quality materials if we use a double pass-through serpentine
flow process, or some such engineering bullshit. I pretend to care what it actually is. Now,
it takes more energy, but what the heck? They show me the economic analysis:
I am ecstatic. My first inclination is to look over at the profits. Geez, Louise. I’m
already spending that bonus in my mind. Yes, I have to spend 20% more capital and burn
5% more energy, but I know I can sell that to the Board. They’ll see the 10% more output I
can realize. And the profits, well... I’ll have to explain to them that growing my market
share to accommodate this higher output will require a slight price reduction, all of which
will displease my competitors, but look at my profits.
As it happens, my nerd-encumbered self happens to observe that rebound is now at
50%. I recognize this as an output effect, but keep this to myself.
[A famous energy economist, Jim Sweeney, has noted that this kind of competition
looks a lot like a zero-sum game: Since I have taken market share from competitors, now we
will wage a share battle whose rebound dynamics will depend on output lost here and output
gained there, with the corresponding energy use result being a complex interplay among
more and less energy efficient players. In the short term, sector-wide rebound will be much
less than my 50%. This is the kind of general equilibrium dynamic that researchers like Karen
Turner and Steve Sorrell are best positioned to deal with.]
New Plant but Competition Catches Up
I have already mentally spent my bonus about four times over. My time spent most nights
with my wife (allow me to share this rather personal delight) has been very special,
perhaps not coincidental to the bonuses. As I have been promising her, the bonuses
routinely materialize, and this happy state of affairs continues for about two years.
Then one morning I reach my office and am greeted with an urgent message. About
10 minutes after I return the call, a sober, nervous team enters the conference room. The
news is not good. It seems our top competitor has announced that they will be building a
new plant that, from the sound of things, looks exactly like ours. When I (rather caustically,
I suppose I am big enough to admit) mention that we have spent significant dollars to
protect our IP via a host of patent filings, my attorneys inform me that our truly evil
competitor has evidently found a “work around” that does not appear to violate our specific
patents, but replaces these steps with ones concocted by their own spawn-of-the-devil
engineers. Holy cripes! I picture this year’s bonus fading away… I resist the temptation to
blame the messengers, and instead retreat to my office to sulk.
I know it will take my competitors at minimum two years to bring their plant on
line, so I have some time. Against this, my engineers have told me the “work around” is so
straightforward that in time all my smaller competitors will be able to match what I have
done.
My economics team (damn them) delivers me the bad news of what they think the
five-year picture will look like:
My competitors will no doubt seek to gain market share at my expense by reducing
their output prices. This looks to leave us all back exactly where we started, in profit terms.
On the good news front, by the time this all shakes out, my economics team believes the
overall market for all our products will grow by 6%, owing to economic growth. (I think
this is conservative, given the time that will have elapsed, but…) My rebound has retreated
to 39%.
But in the end, if I manage to keep my share by matching their prices, and they keep
theirs, my nerd self notes that we’ll all be back where we started, and rebound will be at
39% for all players. Even though our Economic Profit has now been competed down to
zero.
New Plant but Consumers Respond
One happy morning I am reading over my sales reports, and notice that, while my market
share has remained relatively constant, my sales have been creeping up over the last
several quarters. My chief economist explains to me that, even correcting for growth in the
overall economy, consumption of my product per capita is rising. I should have thought of
this already, but she explains that consumers can better afford my product because the
price war with my competitors has made everyone’s product cheaper.
“Consumer price elasticity,” she explains to me, with what I sense is a certain degree
of condescension. I find myself contemplating the possibly of breaking out a special bottle
of wine with my wife this evening. Maybe even picking up some flowers on the way home.
Flowers, as I have found others of my persuasion observing, have some entirely-beyondthe-rational salutary effect. I find myself unable to imagine anything that could deliver a
greater ROI. (Not to mention that they are patently transparent in their intent.) I mean,
seriously: a bunch of ordinary, cheap plants that happen to have blossoms on them? fer
godsake! Who would imagine? I finally conclude that rebound effects are simple compared
what on earth is going on there… But I digress.
Whatever, here’s the situation she lays out for me:
She says my profits are going to go up because of this consumer response.
Consumers like cheaper product. I say fine, but your analysis indicates that our output is
going to have to go up to match it. I thought, I say to her, we were already at capacity. No,
she says, remember we lost volume to our competitors. And even if volumes grow beyond
this, we can simply run the plant more hours per day. If necessary, she says, we can even
put on an overnight shift. Our workers love overtime. And the next plant we build can be
capacitized larger. I’m thinking her consumer response is extremely conservative: I mean,
a 10% decline in my output price means only a 2% increase in my sales volume (and that of
my competitors)? This does not match with my experience. Nonetheless, I am happy with
her report. I’m thinking of buying a flower shop.
My happiness is only partially compromised by my insistent inner nerd, who brings
to my attention that rebound has now increased to 44%.
Late Night Ponderings…
Sitting with a glass of wine late one night, I begin to imagine what the long-term future will
look like for my company, my sector, and, as the wine bottle mysteriously begins to empty
itself, for the global energy economy. I mean, if this has happened to me, is this happening
more globally?
And will my competitors and I continue to find ways to improve energy efficiency,
only to have it eaten away by the kind of dynamics I have witnessed? It feels a lot like the
Red Queen hypothesis those evolutionary biologists like to talk about, in reference to Lewis
Carroll’s Red Queen pronouncing, “It takes all the running you can do, to keep in the same
place.”
This disturbs me, as I remember being told that climate change economists foresee
large reductions in energy use coming from energy efficiency gains. If my ponderings are
right, this means we have less time than we think. I want to believe I am wrong. Me being
wrong would be a very good outcome for the world. But I find myself falling into a deeply
troubled sleep…
IN SUMMARY: The economy, in my view, is extremely creative and flexible. I want to
repeat something I said a while back to Amory, now in the hope that it may be better
understood. To quote:
“Amory, I think you vastly underestimate the flexibility of the economy to adjust to technology
gains so as to maximize overall profit, not just minimize energy cost.”
Model User Guide
Open the model (Numerical Example 2-18-11.xlsm).
You will see that it takes you to the tab on the worksheet entitled, “Control Panel.”
This is where you can work to change the input assumptions to your liking. If your screen
resolution is not high as mine, you may have to go to View/Zoom… to see it all at once.
The protocol I have adopted is that white cells are those you can change and then
observe the Rebound results in the red/pink cells to the right called “Rebound.” Colored
cells are calculations or empty cells. Of course you can change any of these colored cells as
you wish—open-source collaboration and all that—but be alert to inadvertently
overwriting formulas.
The “Quick Links” at the top allow you to move around this tab to the different cases
with one-click navigation.
There is a second tab, “Calculations,” which shows the underlying calculations (which
are extremely simple). There are a few explanatory cell comments you may want to look
at.
At the top of the “Control Panel” tab you will see the “Old Plant” results, which
provide the Base, or Reference Case, against which the “New Plant” configurations are
compared.
Cells H5:J5 allow you to change the factor proportions as you wish. Of course, since
proportions must sum to unity, the Materials proportion is (arbitrarily chosen as) a
calculation that balances everything out if you change the proportions of K , L , and E .
Specifically, the share of Materials, sM , is arbitrarily chosen as the proportion to be
calculated from the other three. It is shaded gray, not white, to indicate that it contains a
formula.
I’ve designed it to be fairly user-friendly. For example, if you click on the Quick Link
“Direct-Secret Sauce,” it will take you to the section of the worksheet where Amory’s
energy efficiency gain can be entered (cell F25) Clicking on the link “Direct Automation”
takes you to the area of the spreadsheet where you can enter in the white cells the
corresponding values for the percentage increase/decrease in capital (cell F38), labor (cell
F39), and energy (cell F40) you think best reflects what would be required to add a new
automation process to your new plant.
Clicking on any of the other links at the top takes you to the corresponding
reconfigurations for which you can change any of the inputs in the white cells of column F.
The rebound magnitudes calculated are cumulative. That is, each configuration change is
assumed to be in addition to what is already in place (e.g., for direct effects, the rebound
result cumulates going from left-to-right in the Quick Links in row 2.) Cell T8 at the top
shows the total cumulative rebound associated with whatever assumptions you have
entered in each section.
After you have played with it for a while, you will have forgotten where you started.
Clicking on the button labeled “Reset to Base” will restore the inputs to the values I have
used in this document. If you would rather work from a different Base Case, simply enter
your preferred values in column AT (and, for factor proportions, in cells AV5:AX5; and the
output price calculated in cell M83 can be put into cell AW83). When you click the “Reset
to Base” button, it will restore all inputs to your preferred values.
One small point: in the “competition catches up” case, you need to find the decline in
output price that drives profit to zero. Clicking on the “Find Zero-profit Output Price”
button will do this for you.
Have fun with it! And let me know of any model improvements you make, so I can
keep a central repository of the model that can be shared among all of us. Open-source
collaboration, as the geeks out there are wont to call it.