may be more robust to individual differences in reporting behaviour. The paper proceeds as
follows: Section 2.2 outlines the data employed for this analysis and describes how
‘in fo rm a tio n ’ ch a n g e is m ea su re d in the surveys. T he econ o m e tric m eth o d s are d esc ribed in Sections 2.3. Section 2.4 presents the results and Section 2.5 c o n c lu d e s the analysis.
Figure 2.1: Research Questions
S ta g e 1 S ta g e 2 S tage 3 Feedback Im provem en ts in Inform ation ? A . D em and R eductions R esearch Q u estio n 1 R esearch Q u estio n 2
2.2
Data - The Residential Sm art M eter Trial
T h e Irish residential sm art m e te r trial w as carried ou t betw een 2009 and 2010 and involved the installation o f o v e r 5000 sm art meters into residential h o u seh o ld s (C E R (201 la ) and C E R (201 Ic)).^ T he overall objective o f the trial w a s to test the im pact and viability o f sm art m etering tec h n o lo g y in Ireland, and to ex p lo re the d em a n d reducing effects o f various feedback m e c h a n is m s and tim e -o f-u se tariffs. R ecru itm en t o f the nationally representative stratified ra n d o m sam ple involved a n u m b e r o f phase d postal invitations (five), with each ro und add in g n e w participants with the goal o f increasing the
representativeness o f the sam p le (according to location and electricity use).
® The overall project com m enced in 2007 and was overseen by the C om m ission for Energy Regulation (CER) with trials carried out by ESB N etw orks and E lectric Ireland.
A benchm ark analysis was conducted (July to Decem ber 2009) w here pre-trial dem and data (half-hourly readings) was collected and control/treatm ent groups w ere established.^ During the test period (January to D ecem ber 2010), treated households received one o f three levels o f feedback and were assigned to one o f four tariff rates. The control group, which also had sm art m eters installed, did not receive any new inform ation and also had no changes to their usual billing process. In addition to dem and data, pre and post-trial surveys were carried out (late 2009 and early 2011 respectively), in w hich a large am ount o f household inform ation was collected, including characteristics o f the dwelling (including building type and size, appliances use and heating/w ater system s), dem ographics and attitudes tow ards energy use.
Feedback stim uli were applied to households in three treatm ent groups. All treated households received a new E nergy Usage Statem ent which contained detailed information on the household’s electricity use by day o f the week, tim e-of-use and relative to previous bills and other custom ers, plus m ore generalised inform ation detailing average appliance consum ption levels and tips to lower costs, particularly during peak times.^ One group {BI- M S T henceforth) received ju st the energy statem ent on a bi-m onthly (every two m onths) basis, while a second group (M ST) was billed on a more frequent m onthly basis. The third group {IHD) received an electronic in-house display in com bination with the bi-m onthly usage statement.* This unit provided real-tim e consum ption, cost and ta riff inform ation.
Four tim e-of-use ta riff types w ere developed (Table 2.1). W hile the control group were charged their usual 14.1 cent per kilow att hour (c/kW h), the treatm ent tariffs (A through D)
^ U sing available usage data and pre-trial survey responses, an allocation algorithm ensured that each cell (feedback and tariff com bination) was approxim ately the same across a number o f behavioural, demographic and attitudinal perspectives. See CER (201 lb ) for further details.
W e acknow ledge that such generalized information is not ‘feedback’. This information, how ever, is auxiliary to consum ption feedback, in that it aids households to act upon their sp ecific feedback. H ow ever, both ‘feedback’ and ‘generalized inform ation’ w ould both improve what w e define as ‘k n ow led ge’ below . * A third stimulus - an O vera ll L o a d Reduction (O LR ) - financially rewarded households for m eeting a reduction target. This stimulus is excluded from our analysis as the data w as unavailable.
generally involved large price increases during peak day-tim e hours and reductions for night-time hours to reflect the underlying wholesale cost o f electricity at these times (all prices exclude value added tax). The time-of-use price increases/decreases were designed to keep the average household’s electricity cost unchanged (peak cost increases to balance with savings off-peak).^ The samples receiving each treatment com bination are displayed in Table 2.2.
Table 2.1: Tim e-of-use Tariffs (€/kW h) excluding VA T
Night Day Peak
Tariff A 0.120 0.140 0.200
Tariff B 0.110 0.130 0.260
Tariff C 0.100 0.130 0.320
Tariff D 0.090 0.120 0.380
Table 2.2: Treatment Matrix
Feedback Stimulus
CONTROL BI-MST MST IHD TOTAL
CONTROL 758 0 0 0 758 A 0 225 237 232 694 Tariff B 0 90 96 82 268 C 0 249 244 232 725 D 0 92 95 90 277 TOTAL 758 656 672 636 2,722
The pre and post-trial surveys explore each household’s self-reported stock o f information on how to reduce their electricity usage. The inform ational statements - to which there are five response options from strongly agree ( ‘1 ’) to strongly disagree ( ‘5 ’) - are summarised in Table 2.3 and 2.4. "KV explores self-reported information on electricity reducing actions (termed ‘general inform ation’ henceforth) and asks respondents if 7 know what I need to do in order to reduce electricity u sa g e’. This first statement is unquestionably broad, and the types o f information captured by such a question could range from operational (knowing how to adjust appliance controls to save electricity) to behavioural
^ A fifth ‘weeicend’ ta riff also involved large increases at peak tim es, but only on w eekdays. This treatm ent w as not com bined with a feedback stim ulus and is therefore not considered.