does differential off-peak electricity pricing affect usage ?
DESCRIPTION
Does Differential Off-Peak Electricity Pricing Affect Usage ?. John Williams, Rob Lawson and Paul Thorsnes School of Business. Synopsis of Project. Mercury Energy contacted Otago University for help with a pricing experiment Rob and Paul responded and set up study, John joined later - PowerPoint PPT PresentationTRANSCRIPT
Does Differential Off-Peak Electricity Pricing Affect Usage?
John Williams, Rob Lawson and Paul ThorsnesSchool of Business
Synopsis of Project
• Mercury Energy contacted Otago University for help with a pricing experiment
• Rob and Paul responded and set up study, John joined later
• Question: Does pricing household electricity differently at peak and off-peak times induce “load shifting”?
• Peaks strain the physical infrastructure and have negative financial impacts on retailers
Study Design: Experimental Groups• Five experimental groups (four treatment
groups + one control group)Name Information Price Difference On-Peak Off-Peak
High Yes 20¢ 30.79 10.79
Med Yes 10¢ 24.52 14.52
Low Yes 4¢ 20.29 16.29
Info Yes None18.29
Control None None
• “Off-peak” is from 7PM to 7AM weekdays; weekends & public holidays
Study Design: Sample
• Approximately 400 households in Auckland (Pakuranga)
• Recruited by Mercury Energy • Allocated by Mercury to experimental groups• All participants got:– A monthly report of usage, including daily and
monthly peak and off-peak usage– Access to usage info via the Web– A list of energy-saving tips
Study Design: Data
• Study ran from 1 August 2008 to 31 July 2009• Mercury supplied us with daily readings for
both peak and off-peak periods (i.e. two readings per day for each household)
• Also supplied data for corresponding period one year before the experiment began
• Technical problems with data: only December 2007 onwards is usable
Energy Usage: Seasonality
Proportion of Off-Peak Use
Christmas
ANZAC
Waitangi Easter
Group EffectStart of Experiment
Panic!• Identified systematic variations across
experimental groups which confound results• Significant amount of unusable data• Solution: compare within households– Examine the differences in energy use in a period
(week, month, year) during the experiment and compare with the corresponding period before the experiment
– Scale: proportional change from baseline (+ve values indicate increase in study period)• (Before – During) / Before
Total Usage Change
Proportional Usage Change
Prop. Off-Peak Usage Change
Differences by Year: Total (%)
Differences by Year: Prop Off-Peak(%)
Summary
• Systematic differences between experimental groups complicates analysis enormously– Not possible to directly detect influence of pricing
• Comparison to previous period is suspect– Don’t know if change was part of a pre-existing trend
• Solution: comparison to baseline, expressed as a proportion, puts all groups on common metric and allows comparison between groups
• Result: possibly a conservation effect (“significant” but R2 tiny); no evidence of a switching effect
Where to from here?• Caveats: data is difficult to deal with, i.e. Missing values
and outliers — have not fully investigated impacts of this yet
• May need to take other non-random differences into account (characteristics of households)
• Not 100% (or even 95%) confident of results yet• Mercury ran a post-survey, but we haven’t had time to
search it for clues yet ...• Some households did use less energy, and some used
more off-peak: what makes them different from those who didn’t?
Tentative Conclusions
• Absolute magnitude of financial incentives may have been too low — but note the large price difference is outside the margins that a retailer could realistically offer
• Attitudes and values may have bigger impact than $$$, also could be interactions (further analysis)