Seminar talk from Steven Sorrell (University of Sussex) at Centre for Climate Change Economics and Policy (CCCEP): http://www.cccep.ac.uk

1. Chasing our tails? - Rebound effects
from improved energy efficiency
Seminar at the School of Earth and Environment,
University of Leeds, 16th Nov 2011
Steve Sorrell
Sussex Energy Group, University of Sussex, UK
s.r.sorrell@sussex.ac.uk

2. Topics
• Defining and measuring
rebound effects
• Rebound effects for
consumers
• Rebound effects for
producers
• Jevons Paradox
• Summary and implications

3. Defining and measuring rebound effects

4. Rebound effects - consumers
Indirect
Lower Holiday
gasoline in Spain
Embodied
energy bills
More energy
Less energy
Lower More energy
Driver further
running or more often
costs
Direct

5. Reinforcing rebound effects!

6. Rebound effects - producers
Indirect
Lower More
Embodied cost cars car
energy travel
More energy
Less energy
Lower More energy
More steel
cost
production
steel
Direct

7. Economy-wide rebound effect
Actual energy
savings
Estimate of
energy
savings Indirect
rebound
effects Economy-wide
rebound effect
Direct rebound
effect

8. A polarised debate
“The concept of a nontrivial rebound “With fixed real energy prices, energy
effect…..is without basis in either theory or efficiency gains will increase energy
experience. It is, I believe, now widely consumption above what it would be
accepted to be a fallacy whose tedious without these gains)” (Saunders,
repetition ill serves rational discourse and 1992)
sound public policy” (Lovins, 1988)
Quantification hampered by inadequate data, uncertain causal
relationships, endogenous variables and complex, long-term dynamics
Existing evidence base only captures a subset of effects in a limited
number of countries/sectors over limited periods of time

9. Defining an energy efficiency improvement
• Energy efficiency may be measured in a
variety of ways for a variety of system
boundaries
• Sources and consequences of changes
and hence rebound effects will vary with
the measure chosen
• Example: If energy efficiency policy
reduced the US energy/GDP ratio by
30%, GDP would need to grow by 55%
to offset the energy savings. But
rebound effects may prevent the energy
intensity reduction

10. Isolating an energy efficiency
improvement
• Energy efficiency improvements rarely occur
in isolation – new technologies provide
multiple benefits
• Rebound effects need not be small just
because the share of energy in total costs is
small – and win-win improvements will have
bigger rebound effects
• Example: Mean payback for 52 energy
efficiency projects fell from 4.2 to 1.9 years
when non-energy benefits taken into account
• Example: green commercial buildings
improved labour productivity by 16% (labour
costs ~25*energy costs)

11. Measuring aggregate energy efficiency
improvements
• Aggregate measures of represent 450
400
the net result of multiple variables
350 GDP
• Measuring energy consumption on
300
In d e x (1 9 4 7 =1 0 0 )
the basis of thermal content 250
Primary
Energy Use
amounts to summing apples and 200 Quality
Adjusted
oranges – and will be misleading 150 Final
Energy Use
• Economic phenomena require 100
50
economic measures
0
• Estimated rebound effects will
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
Year
differ if we use carbon or GHGs

12. Rebound effects for consumers

13. Evidence for direct rebound effects for
households
• Quasi experimental studies – before & after
measurements of energy or energy service consumption following
efficiency improvements
• Econometric analysis of secondary data - estimates of
efficiency or price elasticities for household energy service demand
or energy demand
• Scope limited by data availability: largely confined to personal
transport, space heating and space cooling in OECD households
• Convergence in results, despite varying regions, time periods, data
types, econometric specifications and measures
• Many studies rely upon estimates of price elasticities and could be
upwardly biased
• Own price elasticity of energy demand provides an upper bound for
the direct rebound effect – and is usually less than unity

14. Estimates of direct rebound effects for
households
Source: Sorrell (2007)
Service Range of Best guess No. of Confidence
estimates studies
Car 3 - 87% 10-30% 17 High
transport
Space 1 - 60% 10 – 30% 9 Medium
heating
Space 1 – 26% 1 – 26% 2 Low
cooling
Other 0 – 39% <20% 3 Low
May decline in future as demand saturates and incomes increase
Higher in developing countries
Marginal consumers and key variables ignored
Long term transformational effects not captured

15. Evidence for direct + indirect rebound
effects for households
• Environmentally extended I-O models - energy/carbon/GHG
intensities of commodity groups
• Systems of consumer demand equations - expenditure, own-
price and cross-price elasticities for those commodity groups
• Cheaper energy services may lead to increased or reduced demand for
multiple commodities with widely varying intensities
• Key methodological issues:
• Level of commodity disaggregation
• National versus multiregional I-O model
• Cross-sectional versus time-series data on household expenditure
(income effects versus income + substitution effects)
• Commodities versus energy services

16. Estimates of direct + indirect rebound
effects for households
Author No. Effects Efficiency Area of Rebound
categories captured or consumption effects
sufficiency
Lenzen & Day ? Income Efficiency & Food; heating 45-123%
(2002) Sufficiency (GHGs)
Alfreddson 300 Income Sufficiency Food; travel; 7-300%
(2004) (18) utilities (carbon)
Brannlund 13 Income & Efficiency Transport; utilities 120-175%
(2007) substitution (carbon)
Mizobuchi 13 Income & Efficiency Transport; utilities 12-38%
(2008) substitution (energy)
Kratena (2010) 6 Income & Efficiency Transport; heating; 37-86%
substitution electricity (energy)
Chitnis et al 16 Income Sufficiency Transport, heating, 7-51%
(2011) food (GHGs)
Thomas (2011) 74 Income Efficiency Transport, 7-25%
electricity (GHGs)
Murray (2011) 36 Income Efficiency & Transport, lighting 5 – 40%
sufficiency (GHGs)

17. The importance of indirect energy use and
emissions
Direct carbon emissions
from UK households steady,
but indirect and total
emissions rising
Indirect emissions due to imported
goods rising faster
Source: Druckman & Jackson (2009)
17

18. The importance of indirect energy use and
emissions
Energy Consumption per Household Most countries exhibit a strong correlation
by Income, 2002 between total energy use/carbon emissions and
1000
household income – with indirect use of
Energy Use per Household
900 Embedded Consumption
800
700 Direct Consumption greater importance for higher income groups
Million BTUs
600
500
400
and over time
300
200
100
0
Less $5000 $10,000 $15,000 $20,000 $30,000 $40,000 $50,000
than to to to to to to and
$5,000 $9,999 $14,999 $19,999 $29,999 $39,999 $49,999 over
Source: Saunders (2010)
Wide variation in total energy use/emissions
between socio-economic groups, but little
evidence for significance of values and
awareness, once socio-economic situation is
allowed for
Source: Vringer et al (2007)

19. Simple ‘sufficiency actions’ –
GHG savings for average household
Household heating: reduce
thermostat by 1oC (10%)
Food: reduce food waste by
one third (33%)
Transport: replace car
journeys <2miles by
walking/cycling (23%).

20. Estimated rebound effects
2.50
2.00
Expected GHG
reductions (ΔH)
GHG emissions (tCO2e)
1.50 (tCO2e)
1.00
GHGs due to use of
34%
avoided
51% expenditure (ΔG)
0.50 (tCO2e)
25%
7%
0.00
Household Food Travel All actions
thermostat

21. Varying the re-spend from the combined
actions
14
12
Rebound
10 515%
Expected GHG
GHG emissions (tCO2e)
reduction (ΔH )
8 (tCO2e)
6
GHGs due to
use of avoided
expenditure
4
(ΔG) (tCO2e)
2 Rebound Rebound
Rebound
34% 26%
12%
0
Behaviour Worst case Best case 100%
as usual investment

22. Simple efficiency actions
– GHG savings for average1 UK household
1. Cavity wall insulation 5.9%
2. Top-up loft insulation 1.5%
3. Condensing boiler 5.9%
4. Tank insulation 1.7%
5. CFL lighting 1.2%
6. LED lighting 1.3%
7. 1-5 combined 15.6%
1 GHG savings for eligible households divided by
total number of UK households

23. Baseline – no capital costs, no embodied
energy, 5-year average
35
30
25
Rebound effect (%)
20
Baseline
15
10
5
0
Cavity wall Top-up loft Condensing Tank insulation CFL lighting LED lighting 1-5 combined
insulation insulation boiler

24. Baseline + embodied energy
35
30
25
Reb ound effect (%)
20
Embodied energy
15 Baseline
10
5
0
Cavity wall Top-up loft Condensing Tank insulation CFL lighting LED lighting 1-5 combined
insulation insulation boiler

25. Baseline + embodied energy + full capital
costs
40
30
20
10
Capital costs
Embodied energy
Baseline
0
Cavity wall Top-up loft Condensing Tank insulation CFL lighting LED lighting 1-5 combined
insulation insulation boiler
-10
-20
-30

26. Baseline + embodied energy + subsidised
capital costs
35
30
25
20
15
Capital costs
10 Embodied energy
Baseline
5
0
Cavity wall Top-up loft Condensing Tank insulation CFL lighting LED lighting 1-5 combined
insulation insulation boiler
-5
-10
-15

27. 10 years versus 5 years
30
25
20
15
10
10 years
5 years
5
0
Cavity wall Top-up loft Condensing Tank insulation CFL lighting LED lighting 1-5 combined
insulation insulation boiler
-5
-10
-15

28. Planned work
Variation of rebound effects
by income group
• Estimate Engel curves for up to
40 commodities using data from
UK Living Costs and Food
Survey
Capture of income and
substitution effects
• Estimate Linear Almost Ideal
Demand system using ONS
expenditure data for UK
households

29. Conclusions on rebound effects for
consumers
• Evidence base reasonable for direct effects
but sparse for combined effects
• Direct effects <30% for transport and
heating, smaller for most other energy
services but larger for low income groups
and developing countries
• Combined effects variable with complex
determinants - estimates range from 5 to
200%, but lower values more plausible
• Not capturing economy-wide, long-term
and transformational effects

30. Rebound effects for producers

31. Rebound effects for producers
• Substitution effect: energy services
become cheaper relative to other inputs
• Output effect: lowers production costs and
therefore price of output encouraging
increased demand
• Productivity effect: productivity
improvements lead to increased incomes,
consumption and economic growth
• Composition effect: economy shifts towards
activities where impact of efficiency
improvement is greatest

32. 0%
100%
120%
140%
20%
40%
60%
80%
Electric Utilities
Transportation
Services
Chemicals
Construction
Primary Metal
Agriculture
Financial Industries
Government Enterprises
Food & Kindred…
Paper & Allied Products
Stone, Glass, Clay
effects for US producers
Machinery, non-…
Fabricated Metal
Electrical Machinery
Lumber and Wood
Rubber &…
Textile Mill Products
stock turnover
Motor Vehicles
Non-metallic mining
Weighted Mean = 62%
Long-term Rebound
Communications
Transportation…
Printing, Publishing &…
Instruments
Apparel
Metal Mining
Furniture & Fixtures
Econometric estimates of direct rebound
Misc. Manufacturing
Leather
Tobacco
Estimation of KLEM translog cost functions for 30 US industry sectors over
period 1960-2000, including technology gain parameters and allowing for capital
Source: Saunders, 2011

33. Macro-economic modelling estimates of
economy-wide rebound effects
Only a handful of studies, focusing upon different
regions and different types and sizes of energy
efficiency improvement
• 8 CGE studies of economy-wide effects. Smallest result ~37%
and four >100%
• Macro-econometric model of global economy to 2030 found
~52% for from ‘no-regrets’ policies summarised in IPCC AR4
Results sensitive to uncertain assumptions
• e.g. varying elasticity of substitution between energy and other
inputs from 0.1 to 0.7 increased rebound effect from 7% to 60%
Results sensitive to multiple variables
• Energy intensity of sector, openness to trade, capital and labour
market adjustments etc.
• Negative multiplier effects in energy sector

34. Conclusions on rebound effects for
producers
• Effect varies widely with the nature
and location of the energy efficiency
improvement
• Problems with methodologies,
sensitivity to uncertain assumptions
and little systematic investigation
• Estimates consistently indicate
large effects (> 50%) despite
focusing upon ‘pure’ efficiency
improvements that neglect
associated benefits

35. Jevons Paradox

36. Steam engines in the 19th century
Improved efficiency of steam engines
Lower cost steam
Greater use of steam engines
“….It is wholly a confusion of ideas to suppose that the
economical use of fuel is equivalent to a diminished
consumption. The very contrary is the truth…. Every
Coal-mining Steel-making
improvement of the engine when effected will only accelerate
anew the consumption of coal…” cost steel
Lower
(Jevons, 1865)
Lower cost rail transport
Lower cost coal

37. General purpose technologies
• Potential for substantial
improvement
• Pervasive across a range of uses,
sectors, products and processes
• Complementarities with existing or
potential new technologies
• Diffusion characterised by
increasing returns, widespread
stimulus to innovation and long-
term productivity growth

38. Variability of rebound effects
0%
Energy intensive sectors and
Rebound energy producers Non-energy intensive sectors and
effect (%) households
Core process technologies
50% ‘Win-win’ technologies Non-core technologies
‘General-purpose’ technologies ‘Dedicated’ energy efficient
technologies
Developing countries
Developed countries
100%
Type and location of energy
efficiency improvement

39. Long term rebound - indices of key lighting
variables in the UK 1300-2000
10000
Light Use/cap, 6567
Efficiency, 1000
1000
Source: Fouquet & Pearson (2006)
100 85
GDP/cap, 15
14.5
10
Light price 2.9
1 1
0.26
Fuel price, 0.18
0.1 Light use per capita
0.042
0.01
0.001
Light Price, 0.0003
0.0001
Page 39 1300 1700 © Imperial College London 1850
1750 1800 1900 1950 2000
’

40. The world’s appetite for light
Consumption
0.72% = US$440B / US$60T
•Global analysis over three centuries 6.5% = 1 TWc / 16 TWc
•Per capita consumption of light increased 105.4
104
by 1.01% for every 1% reduction in the JP+KR 2005
per capita consumption of light
AU+NZ 2005
OECD-
OECD-EU 2005
cost of light
FSU 2000 US 2001
UK 2000
•Energy ‘’savings’ from improved lighting 101
WRLD-
WRLD-DEV 2005
CN 2006
UK 1950 CN 2005
efficiency almost exactly cancelled out by CN 1993
increased light consumption
φ [Mlmh/(per-yr)]
UK 1900
WRLD-
WRLD-UNDEV 1999
•Little sign of saturation in lighting 10-2
UK 1800 UK 1850
consumption in developed countries UK 1700 UK 1750
•Considerable potential for growth in Tsao and Waide, “The World’s Appetite for Light: Empirical Data and Trends Spanning
10-5 Three Centuries and Six Continents,” to be submitted to LEUKOS
consumption of both lighting and energy 10-5 10-2 101 104
for lighting β·gdp/CoL [Mlmh/(per-yr)]
0.0072 $/Mlmh
$/(per-yr)
Source: Tsao et al (2010)

41. Underlying debates
Orthodox perspective Ecological perspective
Source of Technical change and Increasing availability of high-
productivity improvements in labour quality energy, both directly and
improvements quality embodied in capital equipment
and technology
Marginal Proportional to share of Greater than the share of energy
productivity of energy in the value of in the value of output
energy inputs output
Input Considerable scope for Limited scope for substituting
substitution in substituting capital for capital for energy
production energy
Decoupling of Has already occurred with Strong link exists and will
energy considerable scope for continue to exist between quality
consumption further decoupling adjusted energy use and
from GDP
economic output
Implied Small Large
rebound effects

42. Len Brookes and ecological
economics
• “..it is energy that drives modern
economic systems rather than such
systems creating a demand for energy”
(Brookes, 1984)
• Improvements in energy productivity are
typically associated with proportionately
greater improvements in total factor
productivity. Latter raises output and
increases demand for all inputs.
• The resulting increase in demand for energy
inputs more than offsets the reduction in
energy use per unit of output – so energy
efficiency improves while energy
consumption increases

43. Conclusions on Jevons Paradox
• Theoretical arguments are stylised and
often rely upon questionable assumptions
• Cited empirical evidence is indirect
selective and frequently ambiguous
• Backfire not demonstrated to be a universal
outcome of energy efficiency improvements
• Arguments and evidence should
nevertheless be taken seriously
• Key issue is the contribution of increasing
energy services to productivity
improvements and economic growth

44. Conclusions and policy implications

45. Conclusions
• Rebound effects can be quantified
and appear to be substantial
• Estimated size of rebound effects
seems to increase with the scope
and boundary of the analysis
• Contribution of efficiency
improvements to carbon mitigation
continues to be overestimated
• Uncomfortable implications for
policy – can energy/carbon pricing
mitigate this effect?

46. Suggested policy responses
• Allow for rebound in policy
appraisals
• Introduce progressive efficiency
standards
• Impose increasingly stringent
carbon taxes or emission caps
• Seek comprehensive, global
climate agreement to prevent all
forms of carbon leakage...
46