The many attributes of residential energy efficiency improvements: How do households vary in the attributes they
value most?
Auren Clarke and Paul Thorsnes
Dept. of Economics
University of OtagoDunedin, New Zealand
IntroductionGeneral issue: slow uptake of residential energy efficiency improvements
E.g., rate of EE improvements in Europe less than half that of other types of renovations (Jakob, 2006)
Similar problem in NZ, despite subsidies/social marketing
A growing literature focuses on understanding the relative values households place on various aspects, or ‘attributes’, of EE improvements
Some report results of discrete choice survey experimentse.g., Poortinga (2003), Banfi et al. (2008), Farsi (2010), Nair et al. (2010), Achtnict (2011), Achtnicht and Madlener (2012)
Plus earlier work of our own which focus on heterogeneity across households in the relative values of attributes
In this studyWe focus on heterogeneity in the attributes themselves
Organization of the presentation
1. Describe the survey software
2. Describe the sample
3. Report results
4. Next steps
Unique decision survey software1000Minds
Web-based multiple-attribute decision software
Key feature: an efficient algorithm for presenting choices
Starts by identifing all ‘undominated pairs’ of two attributes
An undominated pair forces the respondent to make a trade-off
Then presents one such choice pair for the respondent to evaluate
Screenshot of a survey choice pair
Unique decision survey software1000Minds
Web-based multiple-attribute decision software
Key feature: an efficient algorithm for presenting choices
Identifies all ‘undominated pairs’ of two attributes
Presents one choice pair for the respondent to evaluate
Eliminates from the survey all other choices implied by transitivity
Which reduces considerably the number of choices required to rank all combinations of two attributes
Continues until all pairs are evaluated explicitly or implicitly
Relative values (or utilities) are then estimated using a linear program
The result is a complete set of relative utility values for each respondent
Screenshot of a survey choice pair
Relative values of attributes of water heating systems
Average of 30 choices to rank 80 undominated pairs
Means
Upfront cost 14.6
Running cost 16.4
Reliable supply 17.7
Confident in technology 12.4
Fits with house 12.2
Doesn't disturb neighbours 13.8
Off grid 7.3
Upgradable 5.6
Respondents/cluster 586
Size as % of sample 100%
‘Clusters’ of respondents with similar relative values
Average of 30 choices to rank 80 undominated pairs
Means Thrifty Reliable Considerate Independent
Upfront cost 14.6 22.8 15.1 12.4 12.1
Running cost 16.4 25.5 16.7 13.8 14.1
Reliable supply 17.7 11.0 26.0 19.1 12.8
Confident in technology 12.4 9.9 14.3 12.0 12.7
Fits with house 12.2 7.9 10.6 14.9 12.8
Doesn't disturb neighbours 13.8 8.4 7.9 20.0 14.0
Off grid 7.3 9.4 4.3 3.3 14.0
Upgradable 5.6 5.1 5.2 4.7 7.6
Respondents/cluster 586 94 134 203 155
Size as % of sample 100% 16.0% 22.9% 34.6% 26.5%
Next step…
Conventionally, the researcher chooses the attributes of interestTo obtain estimates of their relative values
But identifying the attributes of interest may itself be of interestThe number of attributes of EE improvements is relatively large
A review of the literature reveals more than 20
In this pilot study, we take advantage of the web-based interface to:Allow each respondent to choose from a list the 6 attributes most important to him or her
A 7th attribute – upfront cost – was imposed on everyone
Then work the respondent through a choice survey based on those 7 attributes
The choice model becomes tailored to the respondent
Respondent chooses attributes
Screenshot of a choice pair
Recruiting a sample for the pilot studyOwner-occupiers in Dunedin, New Zealand
Climate similar to Seattle’s
Recruited in three census neighborhoodsAnalogous to census tractsCombined demographics similar to NZ as a whole
Initial contact through an invitation letter in early winter 2012The letter directs the householder to the survey web siteInducement
A $10 shopping voucher upon completion, ORA 10% chance of winning a $100 voucher
450 letters sentAbout 15% response rate in the first weekRate increased to 33% after follow-up telephone calls(149 responses)
Clusters on attributesCluster One Two Three Four Five Six
Proportion of respondents 30.2% 22.1% 17.5% 14.8% 8.7% 6.7%
Value for money 0.87 0.85 0.92 0.82 0.85 0.70
As energy efficient as advertised 0.84 0.94 0.85 0.23 0.46 0.30
Works reliably 0.89 0.97 0.54 1.00 0.15 0.20
No structural alterations 0.09 0.24 0.50 0.55 0.85 0.00
Lifespan 0.71 0.30 0.08 0.73 0.38 0.30
Environmental benefits 0.07 0.73 0.31 0.32 0.54 0.40
Independence from the grid 0.20 0.58 0.31 0.05 0.23 0.90
Capitalizes into home value 0.73 0.06 0.42 0.36 0.23 0.60
Frequency of maintenance 0.62 0.36 0.15 0.14 0.77 0.30
DIY install 0.07 0.09 0.19 0.14 0.08 0.90
Time for daily operation 0.20 0.03 0.00 0.05 0.69 0.30
Well-ventilated home 0.20 0.33 1.00 0.18 0.00 0.10
Home safety 0.13 0.03 0.46 0.82 0.08 0.30
Not too fiddly 0.02 0.18 0.08 0.09 0.00 0.20
Appearance 0.11 0.12 0.04 0.14 0.31 0.20
Potential to disturb me 0.13 0.15 0.08 0.09 0.15 0.00
Potential to disturb neighbours 0.02 0.03 0.08 0.09 0.08 0.00
Large size 0.04 0.00 0.00 0.05 0.15 0.20
An information tool?Any EE improvement can be defined in terms of its attributes
Various sources assist household decision-makers by describing attributes of potential improvements
But the list of improvements can be long
The choice survey provides information about the household
This information can be used to rank-order potential improvements
Based on their attributes and the household’s preferences
That rank ordering helps reduce the information burden on households
By helping prioritise the information search
Or, the information could be useful to energy consultants
More heterogeneity…
Choice algorithm strengths and weaknessesStrengths
Each choice is as simple as possibleJust two profiles defined on just two attributes (at a time)
A relatively small number of choicesTo get respondent-specific utility weights
Ideal for investigating preference heterogeneitye.g., can cluster respondents on the basis of utility weights
Weaknesses
Imposes a simple additively separable utility functionNo interactions across the attributes as included in the model
Potentially sensitive to inaccurate choicesEach choice eliminates choices implied by transitivity
Policy implicationsPolicy targeted toward characteristics of each group:
• A relatively small cost-constrained group– Consistent with limited response to subsidies; subsidies
necessary for some but not sufficient for many
• A group willing to invest, but concerned about recovering upfront cost upon sale of house– Suggests perhaps home energy audit and certification program
• A relatively large group concerned about functional reliability– Suggests aggressive independent testing and certification
• Another group concerned about aesthetics– Suggests support for customized installations
• A surprisingly large group interested in independence from the grid– Support for solar systems?
The New Zealand contextHousehold energy use has historically been inefficient
Low prices due to abundant local energy resourcesHydro-electricity, wood, coal, natural gas w/small population
Many houses are poorly insulated and heatedNo insulation requirements until 1978Efficient heating systems are rarely installed at construction
Interest is growing in cleaner/more efficient energy useHigher prices as local energy resources become more scarceConcerns about the health impacts of cold/damp housesConcerns about negative environmental impacts
Particulate emissionsGreen-house gasesDevelopment in sensitive areas
Policy issue: slow up-takePolicy efforts to encourage domestic investment
Subsidies
Persuasive advertising
There has been some consumer response
Partial insulation retrofits
Installation of un-ducted heatpumps and efficient wood burners
Limited information?
Heterogeneous households in heterogeneous houses
Difficult to know what works in context
Good information can be difficult to obtain
Dissatisfaction with relatively expensive improvements is common
Clusters on attributesCluster Six One Four Five Two Three
Proportion of respondents 30.2% 22.1% 17.5% 14.8% 8.7% 6.7%
M. Value for money 0.87 0.85 0.92 0.82 0.85 0.70
E. As energy efficient as advertised 0.84 0.94 0.85 0.23 0.46 0.30
D. Works reliably 0.89 0.97 0.54 1.00 0.15 0.20
S. No structural alterations 0.09 0.24 0.50 0.55 0.85 0.00
L. Lifespan 0.71 0.30 0.08 0.73 0.38 0.30
O. Environmental benefits 0.07 0.73 0.31 0.32 0.54 0.40
P. Independence from the grid 0.20 0.58 0.31 0.05 0.23 0.90
N. Capitalizes into home value 0.73 0.06 0.42 0.36 0.23 0.60
F. Frequency of maintenance 0.62 0.36 0.15 0.14 0.77 0.30
G. DIY install 0.07 0.09 0.19 0.14 0.08 0.90
H. Time for daily operation 0.20 0.03 0.00 0.05 0.69 0.30
Q. Well-ventilated home 0.20 0.33 1.00 0.18 0.00 0.10
R. Home safety 0.13 0.03 0.46 0.82 0.08 0.30
J. Not too fiddly 0.02 0.18 0.08 0.09 0.00 0.20
B. Appearance 0.11 0.12 0.04 0.14 0.31 0.20
C. Potential to disturb me 0.13 0.15 0.08 0.09 0.15 0.00
I. Potential to disturb neighbours 0.02 0.03 0.08 0.09 0.08 0.00
K. Large size 0.04 0.00 0.00 0.05 0.15 0.20