Energy-saving effect of automatic home energy report utilizing home energy management system data in Japan

Energy 125 (2017) 382e392

Contents lists available at ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

Energy-saving effect of automatic home energy report utilizing home energy management system data in Japan Yumiko Iwafune a, b, *, Yuko Mori a, Toshiaki Kawai a, Yoshie Yagita a a b

Research Center for Energy Engineering, Institute of Industrial Science, University of Tokyo, Tokyo, Japan JST, CREST, Japan

a r t i c l e i n f o

a b s t r a c t

Article history: Received 4 August 2016 Received in revised form 26 January 2017 Accepted 24 February 2017 Available online 3 March 2017

This study assesses the effects of sending home energy reports utilizing the Home Energy Management System (HEMS) data to more than 1600 households in Japan. The treatment effect was verified using a panel data regression random effects model comparing the electricity consumption of a treatment group to which the report was sent with that of a control group that was not sent. The report was effective in winter and led to a 3.4% reduction in electricity consumption compared to the previous year in the average household. A further reduction of 5.4% for the households with higher electricity consumption for whom a significant reduction of 11.4% in the use of space heating was also observed. Although the treatment effect was not significant in summer for the average household, larger households reduced consumption by an overall average of 2%, with reductions of 6.8% and 7% in terms of space cooling and hot water use, respectively, from the previous month. In contrast, smaller households increased their space cooling consumption by more than 10% on average, which might be considered an undesirable boomerang effect. The accumulative treatment effect in a detached house group was also confirmed. Additionally, an accumulative two-year winter consumption reduction of 7.5% demonstrated the effectiveness of continual intervention. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Home energy report Home energy management system Energy saving

1. Introduction Global electricity consumption continues to rise at a high pace, with residential electricity use representing 31% of electricity consumption in OECD countries in 2012 [1]. Households consumed 29% of the total electricity used in Japan in 2013, and this household electricity consumption has increased by 56% from 1990 to 2013 [2]. In June 2015, the Japanese government announced a new energy conservation target to be reached by all sectors by 2030, including a requirement that the residential sector reduce energy and electricity use by 27% and 19%, respectively, from 2013 levels [3]. Achieving this aggressive target is expected to be accomplished through the improvement of insulation levels in dwellings, the diffusion of efficient appliances and lighting, introduction of home energy management systems (HEMS), and the implementation of public awareness/educational activities. The introduction of HEMS, which involves the monitoring and control of residential energy use, to all 50 million households by 2030 is expected to be a target

* Corresponding author. 153-8505 Komaba 4-6-1, Meguro-ku, Tokyo, Japan. E-mail address: [email protected] (Y. Iwafune). http://dx.doi.org/10.1016/j.energy.2017.02.136 0360-5442/© 2017 Elsevier Ltd. All rights reserved.

of the program. As of 2015, it is estimated that 200,000 HEMS have been introduced to individual homes in Japan, a trend that has been accelerated through government support in the form of subsidies issued following the 2011 earthquake. HEMS can provide highresolution data on electricity consumption broken down by variables, including overall space heating, space cooling, and water heating, and other factors, compared to a general in-home display system. However, the utilization of HEMS data is still very limited. The goal of this study was to develop a method for tailored feedback based on the household electricity consumption reports utilizing HEMS data and quantitatively evaluate the effectiveness of such reports.

2. Literature review Several intervention programs have been instituted with the goal of encouraging households in order to reduce their energy demand [4e10]. Abrahamse et al. reviewed 38 studies conducted between 1977 and 2004 and categorized them as involving either antecedent or consequence strategies to promote household energy conservation [4]. The former strategy involves the use of

Y. Iwafune et al. / Energy 125 (2017) 382e392

Nomenclature yit

a x1 x2t

b1 b2 Wit

git 3 it

Average electricity consumption for household i for month t [kWh/day] Constants Dummy variable indicating treatment households (households sent report ¼ 1, otherwise ¼ 0) Dummy variable indicating the month after the report was sent (after treatment ¼ 1, otherwise ¼ 0) Regression coefficient for dummy variable for treatment household Treatment effect (Regression coefficient for sending the report) Control variables matrix Regression coefficient for control variables matrix Error terms (ci: Random effects error term, uit: Other error terms)

factors, such as commitment, goal setting, information, and home auditing, whereas the latter involves feedback and rewards. Of these, information feedback and energy audit programs are addressed in this study. Feedback programs involve providing information to households on their energy demand or energy savings in order to affect behavioral change. In the reviewed paper [5], requirements of successful feedback programs were identified, including: frequent and long feedback based on actual consumption; interaction with households; appliance-specific breakdown of usage, including historical or normative comparisons; and understandable and appealing report design. The contents, frequency, and delivery method of feedback affected the outcome of introduction of new devices such as smart meters or internet services. The post-2010 research has involved larger sample sizes and has increased the credibility of the validation of effect of intervention feedback programs, as pointed out in Ref. [10]. For example, in 2010, Schleich et al. sent feedback to 1500 households in Austria through a web portal and by mail and confirmed a 4.5% electricity reduction over 11 months [11]. In their program, information on previous-day electricity consumption patterns, electricity cost, and practical measures to save electricity were provided on a web portal that utilized smart meter data. In addition, written two-page feedback forms were mailed to participants once a month to increase their motivation. Houde et al. assessed real-time feedback sent to 1743 Google employees in 2012 and confirmed a 5.7% reduction of electricity use that was sustained over four weeks [12]. The effectiveness of comparative feedback for similar households was examined using very large household samples in a series of studies by OPower [13]. The energy report sent by OPower reduced energy use by 2% over a sample of 600,000 households, and follow-up research validated the results [14,15]. The use of tailored feedback in the residential sector is generally valid but involves high costs [4,16]. The Japanese environmental ministry is supporting a visiting audit program to reduce residential energy demand [17]; however, larger-scale deployment is expected to be difficult because of barriers including expense, difficulties with dispatching consultants, and low acceptance of intrusive audit measures. By contrast, the above-mentioned methods used by OPower obtained a 2% reduction in energy use [13] and have the versatility to expand to very large numbers of households. Effective feedback systems utilizing HEMS or smart meters will be needed to promote further energy conservation in the residential sector. Rapid introduction of electricity smart

383

meters in households and businesses is currently being promoted in Japan and is expected to be complete by the early 2020s [18]. Although the utilization of smart meters installed in all households for energy audits or demand response will ultimately need to be considered, our study targets automatic energy audits and the production of customized tailored messages using HEMS, which has a higher resolution than smart meters. Energy reporting using HEMS can achieve a balance between general versatility and targeted specificity. Commercial HEMS systems in Japan have been installed in circuit panels in nearly all new houses, allowing the collection of electricity consumption as cloud data for 8e32 circuits at intervals of one-half to 1 h. HEMS service providers including house builders and appliance manufacturers have developed web sites that customers can use to check their electricity consumption, make simple comparisons between customers, and obtain general energy tips; however, their utilization has remained low. Our goal is achievement of a cost-effective, diffusible energy reporting system and we investigate the development of energy reports utilizing HEMS data in this paper. We develop an algorithm for automatically generating tailored feedback and confirm the effectiveness of this reporting system. The novelty of this paper is that (i) the home energy report was automatically created using HEMS data which has higher resolution than before, (ii) the effect of the report was quantitatively evaluated by randomized experiment. The remainder of this paper is organized as follows. Section 3 presents experimental method including the description of participants, contents of home energy report, and evaluation method of the effect. Section 4 provides experimental result and Section 6 concludes on energy conservation methods including use of home energy report. 3. Method 3.1. Participants We selected households for the treatment and control groups randomly from the households registered in the HEMS database [19]. The database possesses HEMS data obtained from homebuilders, condominium developers, and HEMS data service providers at various resolutions for approximately 2000 households. In addition to electricity consumption data, the database contains survey responses regarding household resident attribute information, including building type, household data, appliance use, electricity contracts, and behavior, which were collected through questionnaires. The database is linked to weather and appliance specification databases to enable analysis of these factors in connection with HEMS data. We selected treatment and control groups from households who have installed HEMS and cooperate with the experiment. Based on the experiment schedule, households previously accepted were set as the treatment group, and households agreed later were treated as the control group. Differences between the attributes of both groups were controlled by panel data regression random effects model analysis. Households in three sites were selected for an experiment in which home energy reports were sent to a treatment group. Table 1 shows the major attributes of the sample households. Sites A and B consisted of detached houses, whereas Site C consisted of apartment buildings. Common characteristics among the three sites included relatively new houses, younger heads of household, and locations in primarily warm climate regions. Compared to average Japanese households, the sample households had larger floor areas and more members and were more likely to be centrally air conditioned. All households in Site A lived in an all-

384

Y. Iwafune et al. / Energy 125 (2017) 382e392

Table 1 Attribute data of sample households. Site

Dwelling type

Built year

Average household size

Average floor area [m2]

Ratio of electrified housea

Ratio of central AC system

Ratio of colder areab

Average electricity consumption [kWh/day] (Aug2015/Feb2016)

SiteA SiteB SiteC

Detached Detached Apartment

2010e2012 2012e2015 2010e2015

3.68 3.62 2.65

129.0 129.2 79.7

100% 72% 63%

61% 21% 0%

8% 3% 4%

21.7/34.4 17.5/25.3 12.3/17.1

a b

The term electrified house indicates houses that consume only electricity for the main air-conditioning, hot water, and other appliances. Colder area indicates the area where HDD(18-18) is more than 2500.

electrified house, indicating that these houses consume electricity only for main air-conditioning, hot water, and powering appliances, and all had a roof-top photovoltaic generation system installed. 61% of households in Site B lived in all-electrified detached houses, whereas the remainder used gas for main air-conditioning or hot water. The households in Site C lived in apartments with smaller floor areas (less than 80 m2) and smaller numbers of household members (2e3); by contrast, in Sites A and B, most households had more than 100 m2 of floor area and 3e4 members. In Site C, half of the sample were all-electrified houses, and nearly 90% of these were located in the Kanto area. 3.2. Contents of home energy report We sent home energy reports to the treatment households three times: the winters of 2015 and 2016 and the summer of 2016, as shown in Table 2. The number of participants increased gradually over this experimental period, which made it difficult to evaluate the continued effect of the report. We divided households in the treatment group into three groups based on how they were diagnosed. The “larger” group included households determined to be larger households during the target period in the report. The “smaller” group included smaller households, and the rest were counted as “medium” households. Larger households were defined as those consuming more than 1.1 times the average seasonal consumption among similar households, whereas smaller households consumed less than 0.9 times the average among similar households. The remainder of this report is as follows: - Comparison of annual electricity consumption among similar households by use (air conditioning, hot water supply, and other appliances) - Hourly and monthly power consumption pattern by circuit - Average power consumption and usage hours by air conditioning devices in winter and summer - Weekly operation pattern by main air conditioning device and outdoor temperature - Tailored comments generated in response to individual household characteristics, such as whether usage time of air conditioning is longer than average, whether refrigeration consumption is larger than average, whether inefficient electric heating appliances may be used in winter; whether base load appliance should be checked; whether the heat pump water heater operates during expensive fee periods, etc.) - General energy saving tips Note that only electricity consumption was assessed in this report and that the results were gradually improved to better reflect the opinions of the participants. We developed a home energy analysis system in HEMS that uses electricity data measured by circuits and household attribute data. Hourly and seasonal electricity consumption are aggregated by usage, which consists of space heating and cooling, hot water, and

Table 2 Sample sizes of control and treatment groups. Intervention

Site

Winter 2015

Site A 290 Site B e Site C 319

Households in treatment groupa

ALL

Control Treatment Larger

Medium

Smaller

362 e 128

e e e

e e e

e e e

Total Summer 2015 Site A Site B Site C

609 290 142 291

490 321 124 134

e 83 31 37

e 188 79 77

e 50 14 20

Total Site A Site B Site C

723 290 141 386

579 313 121 383

151 92 30 110

344 181 69 172

84 40 22 101

Total

817

817

232

422

163

Winter 2016

a

“Larger” or “Smaller” indicates that the household was determined to be a larger or smaller household during the target season of the report. Larger households consumed more than 1.1 times the average seasonal consumption, and smaller households consumed less than 0.9 times the average. The winter 2015 report did not include determination of seasonal consumption level.

other appliances, and are compared with the corresponding characteristics of similar households. Tailored advice was also produced for each household based on this data analysis. Comparison with households with similar attributes is believed to be effective to evoke social norms [4,13], and we developed a selection method of a comparable household that uses a regression analysis between electricity consumption and attribute data, a method that is effective even if the sample number is small. The procedure is as follows: (1) First, regression models by season and usage of air conditioning, hot water and other appliances are constructed, and standardized coefficients are calculated. (2) Explanatory variables are sorted in descending order using the standardized coefficients. (3) Continuous explanatory variables are classified into discrete clusters. For example, floor area is classified into groups by 50 m2 increments. (4) Sample households are clustered by the explanatory variables with the largest standardized coefficients. (5) The target household is then set. If the number of clustered households in the same cluster as the target household is larger than the reference number, Clnum, further clustering is executed using the explanatory variables with the next larger standardized coefficients. Here clustered households in the same cluster indicate “similar households” to the target household. (6) Procedure (5) is repeated until the number of similar households is less than Clnum. Finally, the smallest cluster larger than Clnum becomes the cluster of similar households, and the average consumption in the cluster indicates the electricity consumption in similar households.

Y. Iwafune et al. / Energy 125 (2017) 382e392

The reports consisted of 2e3 sheets of A4-size documents and were sent by mail. Fig. 1 provides a sample report.

3.3. Verification of the effects of the energy audit report In this study, the treatment effect was verified through comparison of electricity consumption between the treatment group, to which the report was sent, and a control group that was not sent the report. A panel data regression random effects model was adopted to verify the treatment effect [20,21]. Panel data analysis is used as a regression analysis method for handling a combination of time-series and cross section data and has been used in several experimental studies of energy saving or demand response [12e15,22e26]. The general formulation of the panel data regression random effects model is as follows:

ln yit ¼ a þ b1 x1 þ b2 x1 $ x2t þ Sgit Wit þ 3 it 3 it

(1)

¼ ci þ uit

The natural log of consumption for the dependent variables indicates the treatment effect on rate of change. The treatment effects for dummy variables can be obtained as exp(b)  1. We can also estimate the model in the case where the object variables are yit in order to quantify the treatment effect on the amount of electricity use, as shown in (2):

yit ¼ a þ b1 x1 þ b2 x1 $ x2t þ Sgit Wit þ 3 it

385

(2)

The control variable matrix Wit is added to control the household attributes and the influence of the time effects other than the treatment effect and the floor area, the number of people, the type of main air conditioning device, and the monthly average outdoor air temperature were chosen as control variables. The treatment effects were evaluated as shown in Table 3. Comparison of adjacent months can significantly reduce the risk of including the effect of changes in household attributes or replacement of appliances; however, it is possible that the day that the report is confirmed by a participant will deviate. Comparison with the previous year has the advantage of allowing treatment of the same month while removing the influence of variable factors dependent on the month, such as New Year holidays and summer vacation days; however, this comparison may have to account for large changes in household attributes or appliances. In this study, structural changes in household attributes and replacement appliances were not reflected, even for comparison with previous years, because such information could not be obtained for the control group. Finally, the credibility of the obtained effect can be verified by both types of comparison, i.e., monthly and annual. The repetition effect of the treatment could be confirmed by a comparison between February 2014 and February 2016, as we were successful in performing this assessment for the households in Site A for two years. Households for which a data reading was missing in compared

Fig. 1. Sample of winter home energy report.

386

Y. Iwafune et al. / Energy 125 (2017) 382e392

Table 3 Report delivery timing and comparison months for electricity consumption. Intervention

Report sending timing

Before treatment

After treatment

Winter 2015

End of January 2015

Summer 2015

End of July 2015

Winter 2016

Winter 2016 End of January 2016

January 2015 February 2014 June 2015 August 2014 January 2016 February 2015 February 2014

February 2015 February 2015 June 2015 August 2015 February 2016 February 2016 February 2016

Table 4 Descriptive statistics of sample households in summer 2015. Variable

Sample size Floor area [m2] Household size [Number] Monthly average air temperature in July [ C] Monthly average air temperature in August [ C] With heating system other than electricity [0/1 Dummy] With floor heating system [0/1 Dummy] With central heating system [0/1 Dummy] Apartment [0/1 Dummy] Electricity water heater [0/1 Dummy] Electricity consumption in July [kWh/day] Electricity consumption in August [kWh/day]

Control

Treatment

Mean (S.D)

Mean (S.D)

637 113.0 (29.2) 3.13 (1.22) 25.1 (0.7) 26.0 (0.8) 0.05 (0.22)

524 115.1 (33.3) 3.42 (1.16) 25.0 (0.7) 26.0 (0.9) 0.10 (0.29)

0.01 0.28 0.43 0.77 16.2 17.0

0.03 0.44 0.23 0.85 17.4 18.2

(0.09) (0.45) (0.50) (0.42) (7.6) (8.4)

(0.17) (0.50) (0.42) (0.35) (7.2) (7.8)

Table 5 Descriptive statistics of sample households in winter 2016. Variable

Control

Sample size Floor area [m2] Household size [Number] Monthly average air temperature in January [ C] Monthly average air temperature in February [ C] With heating system other than electricity [0/1 Dummy] With floor heating system [0/1 Dummy] With central heating system [0/1 Dummy] Apartment [0/1 Dummy] Electricity water heater [0/1 Dummy] Electricity consumption in January [kWh/day] Electricity consumption in February [kWh/day]

Treatment

Mean (S.D)

Mean (S.D)

667 99.2 (35.1) 3.13 (1.22) 5.4 (1.7) 6.2 (1.6) 0.05 (0.21)

678 106.4 (31.0) 3.12 (1.18) 4.9 (1.5) 5.9 (1.4) 0.09 (0.28)

0.01 0.18 0.55 0.78 24.2 23.6

0.03 0.25 0.53 0.75 24.3 23.5

(0.09) (0.39) (0.50) (0.42) (13.9) (13.3)

(0.17) (0.44) (0.50) (0.43) (12.4) (11.8)

months or less than 10 Wh per day, or for which the rate of change of consumption was larger than 50%, were eliminated from the sample. Tables 4 and 5 provide descriptive statistics of the variables used in our model. 4. Results Tables 6e9 show the results of the effect of the energy report on all households in the three sites for summer 2015 and winter 2016. Floor area, number of people, and outdoor temperature were normalized, and the standard partial regression coefficients were calculated. The treatment effect by amount of consumption is shown in Table 10 and in Figs. 2 and 3. Generally speaking, the electricity saving effect of the energy report can be observed in winter only. The treatment effect from the previous month on total consumption for summer 2015 (Table 6) was not significant. While hot water consumption was reduced by 6.5%, space cooling consumption increased by 3.7%. The effect from the previous year in summer 2015 (Table 7) had a similar trend, e.g., an increase of space cooling consumption and a decrease in hot water consumption, with a total consumption increase of 1.1% at a 10% significance level. Despite the fact that we prepared energy saving tips for cooling system usage in the summer report, there was no effect on participant behavior. This may have been caused by a rapid air temperature increase just after arrival of the report at the end of July, 2015, as shown in Fig. 4. Allcott et al. pointed out that energy saving efforts tend to peak at about 10 days after receiving a report; thus, a rapid temperature rise during the tested period discouraged participants in the experimental group from taking energy saving actions [14]. The summer treatment effect by consumption level shown in Table 10 and Fig. 2 provides further information. Although space cooling consumption in all households increased by 3.7% between

Table 6 Treatment effect of energy report in summer 2015 (July 2015/Aug 2015). Total

Space cooling

Hot water

Treatment effect (Treatment group  After intervention)

0.6% (0.6%)

3.7%** (1.8%)

6.5%*** (1.0%)

0.3% (0.5%)

Treatment group Control variables

0.8% (2.0%) 9.9%*** (1.3%) 13.3%*** (1.1%) 4.2%*** (0.3%) 0.5% (3.2%) 9.6% (7.1%) 29.7%*** (2.4%) 4.4% (2.9%) 27.0%*** (2.9%)

17.2%*** (4.8%) 5.3%* (2.9%) 11.2%*** (2.4%) 21.9%*** (0.9%) 9.7% (7.1%) 7.6% (16.4%) 109.9%*** (5.5%) 1.0% (6.5%) 4.4% (6.4%)

4.2% (3.0%) 8.6%*** (1.8%) 10.5%*** (1.6%) 14.5%*** (0.6%) 13.6%** (7.0%) 0.3% (10.6%) 12.4%*** (3.5%) 40.1%*** (4.5%)

4.7%** (2.1%) 11.3%*** (1.4%) 13.8%*** (1.1%) 0.5%* (0.3%) 0.4% (3.5%) 16.9%** (7.7%) 11.5%*** (2.5%) 10.4%*** (3.2%) 17.8%*** (3.2%)

Floor Area Household size Monthly average air temperature With heating system other than electricity With floor heating system With central heating system Apartment dummy Electricity water heater

Other appliances

Sample size

1161

965

924

1160

Adjusted R-square

0.37

0.39

0.44

0.28

Robust standard deviation in parentheses. Statistical significance * p < 0.10, **p < 0.05, and ***p < 0.01.

Y. Iwafune et al. / Energy 125 (2017) 382e392

387

Table 7 Treatment effect of energy report in summer 2015 (Aug 2014/Aug 2015). Total

Space cooling

Hot water

Other appliances

Treatment effect (Treatment group  After intervention)

1.1%* (0.7%)

7.7%*** (1.8%)

4.7%*** (1.1%)

0.5% (0.7%)

Treatment group Control variables

0.7% (2.1%) 9.0%*** (1.3%) 12.6%*** (1.1%) 6.3%*** (0.7%) 0.3% (3.5%) 5.5% (7.5%) 31.5%*** (2.5%) 5.6%* (3.1%) 26.0%*** (3.2%)

16.6%*** (4.7%) 6.0%** (2.8%) 11.2%*** (2.4%) 21.2%*** (1.6%) 8.4% (7.4%) 1.4% (16.3%) 129.7%*** (5.4%) 7.9% (6.6%) 1.9% (6.7%)

5.5%* (3.1%) 8.6%*** (1.8%) 9.4%*** (1.6%) 4.6%*** (1.0%) 11.0%* (7.0%) 6.0% (10.8%) 13.8%*** (3.5%) 41.8%*** (4.6%)

5.1%** (2.2%) 10.2%*** (1.4%) 12.8%*** (1.2%) 0.2% (0.7%) 0.4% (3.7%) 13.2% (7.9%) 11.8%*** (2.6%) 11.9%*** (3.3%) 17.9%*** (3.3%)

Floor Area Household size Monthly average air temperature With heating system other than electricity With floor heating system With central heating system Apartment dummy Electricity water heater

Sample size

1074

947

887

1086

Adjusted R-square

0.33

0.25

0.08

0.26

Space heating

Hot water

Other appliances

Robust standard deviation in parentheses. Statistical significance * p < 0.10, **p < 0.05, and ***p < 0.01.

Table 8 Treatment effect of energy report in winter 2016 (Jan 2016/Feb 2016). Total Treatment effect (Treatment group  After intervention)

1.0%** (0.5%)

3.4%*** (1.2%)

0.7% (0.6%)

0.3% (0.5%)

Treatment group Control variables

2.3% (1.9%) 10.4%*** (1.5%) 11.2%*** (1.1%) 4.3%*** (0.5%) 16.3%*** (3.3%) 5.2% (7.1%) 33.9%*** (2.8%) 8.3%*** (3.3%) 60.4%*** (3.0%)

6.8% (4.6%) 11.9%*** (3.4%) 4.6%* (2.5%) 11.2%*** (1.4%) 25.4%*** (8.0%) 24.9% (19.5%) 152.1%*** (6.6%) 0.6% (7.7%) 7.0% (7.2%)

7.6%*** (2.6%) 6.9%*** (1.8%) 14.7%*** (1.4%) 4.4%*** (0.8%) 6.9% (6.6%) 1.3% (9.0%) 5.3% (3.6%) 7.1% (4.2%)

3.8% (2.4%) 14.6%*** (1.8%) 11.1%*** (1.3%) 2.0%*** (0.6%) 8.1%** (4.0%) 2.6% (8.6%) 9.9%*** (3.4%) 14.4%*** (4.0%) 15.1%*** (3.7%)

Sample size

1345

1209

1012

1344

Adjusted R-square

0.46

0.25

0.11

0.25

Floor Area Household size Monthly average temperature With heating system other than electricity With floor heating system With central heating system Apartment dummy Electricity water heater

Robust standard deviation in parentheses. Statistical significance * p < 0.10, **p < 0.05, and ***p < 0.01.

Table 9 Treatment effect of energy report in winter 2016 (Feb 2015/Feb 2016). Total

Space heating

Hot water

Other appliances

Treatment effect (Treatment group  After intervention)

3.4%*** (0.7%)

10.1%*** (1.9%)

2.3%*** (0.8%)

0.5% (0.8%)

Treatment group Control variables

1.0% (2.5%) 9.6%*** (1.4%) 9.6%*** (1.2%) 4.4%*** (0.9%) 15.9%*** (3.5%) 3.8% (7.7%) 33.8%*** (2.8%) 8.6%** (3.5%) 60.0%*** (3.3%)

5.0% (6.1%) 7.5%** (3.6%) 9.7%*** (3.0%) 6.7%*** (2.2%) 32.7%*** (9.5%) 50.0%*** (22.7%) 159.5%*** (6.9%) 11.8% (8.9%) 12.4% (8.3%)

2.0% (3.1%) 7.3%*** (1.7%) 10.0%*** (1.5%) 7.1%*** (1.1%) 6.3% (6.4%) 0.9% (10.2%) 3.6% (3.3%) 2.1% (4.4%)

5.4%* (3.0%) 13.4%*** (1.7%) 9.5%*** (1.4%) 2.8%*** (1.0%) 6.1% (4.2%) 19.9%* (9.3%) 9.3%*** (3.3%) 14.1%*** (4.2%) 16.9%*** (3.8%)

Floor Area Household size Monthly average air temperature With heating system other than electricity With floor heating system With central heating system Apartment dummy Electricity water heater

Sample size

991

836

746

962

Adjusted R-square

0.48

0.25

0.15

0.24

Robust standard deviation in parentheses. Statistical significance * p < 0.10, **p < 0.05, and ***p < 0.01.

July and August 2015, in larger households it was reduced by 6.8%, whereas in the smaller households it increased by 10%, as shown in Table 10 and Fig. 2. This suggests that the report was more effective for households determined to be larger consuming households, as those determined to be smaller consuming households might have been using their air conditioners with less trepidation during the rapid temperature rise. This finding is consistent with previous work by Schultz et al. that revealed an undesirable boomerang effect of increasing use in households shown by an energy audit to

have low electricity consumption [27]. The same study suggested that adding an injunctive message indicating approval of the desired behavior might reduce the boomerang effect, and incorporating this tactic into our summer report may prove beneficial. The effect on amount of consumption in summer 2015 is shown in Table 10, from which it can be seen that total consumption increased 0.28 kWh per day from the previous year at a 5% significance level as a result of space cooling consumption. The treatment effects on total consumption from the previous

388

Y. Iwafune et al. / Energy 125 (2017) 382e392

Table 10 Treatment effect of energy report by consumption level. Season

Compari-son months

Use

Rate of change [%]

Amount of change [kWh/day]

Upper: Treatment G/Lower: Control G

Winter 2015

Jan/Feb 2015

Feb 2014/Jan 2015

Summer 2015

July/Aug 2015

Aug 2014/Aug 2015

Winter 2016

Jan/Feb 2016

Feb 2015/Jan 2016

All

Larger

Smaller

All

Larger

Smaller

All

All

All

All

All

All

0.0% 10.0%** 4.8%** 1.6% 1.5% 14.2%*** 8.1%*** 1.3% 0.6% 1.5% 0.3% 0.8% 1.9% 8.2%** 1.5% 0.2%

0.86*** 0.72*** 0.05 0.24*** 1.32*** 1.02*** 0.07 0.52*** 0.09 0.10* 0.12*** 0.03 0.28** 0.23*** 0.07*** 0.07 0.22* 0.22*** 0.08* 0.06 0.73*** 0.65*** 0.12* 0.01

0.00 0.27** 0.27*** 0.03 0.28 0.13 0.06* 0.17 0.67*** 0.50*** 0.03 0.15 1.85*** 1.20*** 0.35*** 0.25

0.29 0.11 0.04 0.12 0.09 0.11 0.09** 0.12 0.47** 0.35*** 0.12 0.02 0.34 0.18 0.13 0.08

2.0%*** 5.5%*** 0.7% 1.2%** 4.0%*** 9.1%*** 0.9% 2.9%*** 0.6% 3.7%** 6.5%*** 0.3% 1.1%* 7.7%*** 4.7%*** 0.5% 1.0%** 3.4%*** 0.7% 0.3% 3.4%*** 10.1%*** 2.3%*** 0.5%

Total AC Hot water Other app. Total AC Hot water Other app. Total AC Hot water Other app. Total AC Hot water Other app. Total AC Hot water Other app. Total AC Hot water Other app.

2.0%* 6.8%** 7.0%*** 0.3% 0.3% 2.0% 2.8% 0.6% 1.1% 3.5%** 0.6% 0.2% 5.4%*** 11.4%*** 3.8%*** 2.1%

Statistical significance * p < 0.10, **p < 0.05, and ***p < 0.01.

and Fig. 3 include a significant finding that larger households tended to save electricity by curtailing space heating and hot water use. Larger households reduced total electricity usage by 5.4% relative to the previous year, which is larger than the overall 3.4% household average reduction. This total quantitative reduction corresponds to 56 kWh per month as compared to the household average of 22 kWh per month. The reduction rates in AC (space heating) and hot water use were also large at 11.4% and 3.8%, respectively, at a 1% significance level. Tables 11e13 show the treatment effect of the energy report by site. There is a general reduction of sample size caused by an

month and previous year in winter 2016 are shown in Tables 8 and 9, respectively, resulting in respective reductions of 1% and 3.4% in total consumption. Quantitative reduction effects per household of 7 and 22 kWh per month were also obtained from the previous month and year, respectively. Large reductions in space heating of 3.4% from the previous month and 10.1% from the previous year were observed. Adaptation of winter energy saving measures, such as maintaining a reasonable room temperature and reducing the use of inefficient electric heating devices, appears to have been easier for customers than corresponding measures for other seasons. The results in winter 2016 by consumption level in Table 10

15

14.2*** 10.0**

Rate of change [%]

10 5

7.7*** 3.7** 0.0

0 -0.3

-0.6 -5

-6.5***

1.6

1.1*

-6.8**

1.5

-4.8**

-4.7***

-7.0***

-8.1***

-10 All

Larger

Smaller

All

July 2015/Aug 2015 Total

1.3

-0.3 -2.0 -2.8

-0.3

-2.0*

0.6

0.5

Space cooling

Larger Aug 2014/Aug 2015

Hot water

Other appliances

Shaded areas show insignificant effecs. Statistical significance * p<0.10, ** p<0.05, *** p<0.01 Fig. 2. Treatment effect in summer 2015 by consumption level.

Smaller

Y. Iwafune et al. / Energy 125 (2017) 382e392

389

5

Rate of change [%]

0.7

0.6

0 -0.3

-1.0** -5

-0.2

-1.1

-3.4***

-0.6

-0.3-0.8 -1.5

-3.5**

-0.5 -2.3***

-3.4***

-2.1

-1.9

-3.8***

-0.2 -1.5

-5.4*** -8.2**

-10

-10.1***

-11.4***

-15 All

Larger

Smaller

All

Larger

Jan 2016/Feb 2016 Total

Smaller

Feb 2015/Feb 2016

Space heating

Hot water

Other appliances

Shaded areas show insignificant effecs. Statistical significance * p<0.10, ** p<0.05, *** p<0.01

40 35 30 25 20

Report was arrived.

Osaka average

Osaka highest

Tokyo average

8/30

8/26

8/28

8/24

8/20

8/22

8/18

8/14

8/16

8/12

8/10

8/8

8/6

8/2

8/4

7/31

7/29

7/25

7/27

7/23

7/21

7/19

7/17

7/13

7/15

7/11

7/7

7/9

7/5

7/3

15 7/1

Highest and average Temperature [C]

Fig. 3. Treatment effect in winter 2016 by consumption level.

Tokyo highest

Fig. 4. Daily average and highest temperatures in major cities in summer, 2015.

increase in insignificant results; however, it is still possible to observe differences by site characteristics. Site A, which had the largest consumption and longest involvement period of the three sites, had the largest treatment effect (Table 11). The treatment effect in winter 2015 was a 3%e4.4% reduction in total consumption, which corresponds to 33 to 48 kWh per month. A 1.7% reduction from the previous month was also seen in summer 2015. Furthermore, an accumulative treatment effect over two years can be confirmed in site A, as shown in Fig. 5. The reduction effect from the previous year in winter 2015 was 4.4% in terms of total consumption, 9.1% in space heating use, 1.7% in hot water use, and 3.2% in other appliance use. The effect from the previous year in winter 2016 was smaller than of 2015, with reductions down 3.7% in terms of total consumption and 9.8% in space heating use, with no significant effect in terms of hot water or other appliance use. A month-by-month comparison for 2016 does not reveal significant changes; by contrast, a 3% month-by-month reduction in total consumption was obtained in 2015. These results indicate decreasing margins of energy saving as the diagnostic process was repeated. Finally, an accumulative effect in terms of reduction in total consumption in winter 2016 of 3.7% from the previous year and 7.5% from two years prior was confirmed. The

decline in space heating was significant in site A, possibly because this factor could be reduced relatively easily, owing to the higher possession ratio of central heating and heating levels in site A. Site B (Table 12) has fewer significant results as a result of its smaller sample size, as shown in Table 2. Space cooling consumption increased by 14.4%, which corresponded to an increase in total consumption of 3.2% in summer 2015. A treatment effect from the previous year in winter 2016 is seen as a 2.8% reduction in all households and a 6.1% reduction in larger households. The effect in which larger households incurred larger reductions than average in winter was also confirmed in Site B. The smallest effect was observed in site C, which consisted of apartment buildings (Table 13). Note that no year-to-year comparison in winter 2016 was calculated for site C because of the limited amount of consumption data extracted from the control group households. Hot water use reduction by 7%e8% from the previous month was observed in summer 2015, and space heating use reduction by 4.3% from the previous month was observed in larger households in winter 2016. In general, the inhabitants in site C did not choose the HEMS function proactively; instead, their apartment building managements had installed the system. In contrast, the detached house inhabitants were more likely to

390

Y. Iwafune et al. / Energy 125 (2017) 382e392

Table 11 Treatment effect of energy report in Site A. Comparison months

Use

Rate of change [%]

Amount of change [kWh/day]

Upper:Treatment G/Lower: Control G

Jan/Feb 2015

Feb 2014/Jan 2015

July/Aug 2015

Aug 2014/Aug 2015

Jan/Feb 2016

Feb 2015/Jan 2016

Feb 2014/Jan 2016

Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances

All

Larger

Smaller

All

Larger

Smaller

All

All

All

All

All

All

1.7% 8.1% 4.2% 0.3% 1.5% 9.7%* 8.1%** 0.8% 3.1%* 11.0%** 3.1% 2.2% 0.2% 10.8% 2.0% 0.6% 9.5%*** 16.3%** 4.2% 5.1%

1.10*** 0.81*** 0.04 0.34*** 1.58*** 1.18*** 0.13* 0.56*** 0.37** 0.05 0.11*** 0.17 0.13 0.18* 0.11*** 0.02 0.22 0.24 0.11* 0.12 1.15*** 1.04*** 0.09 0.01 2.51*** 1.53*** 0.14 0.59**

0.05 0.38** 0.26*** 0.00 0.01 0.12 0.15*** 0.10 0.99** 0.58** 0.05 0.47* 2.28*** 1.82*** 0.20 0.11 3.86*** 2.54*** 0.11 1.11**

0.78** 0.34 0.07 0.08 0.05 0.01 0.10 0.16 1.54** 0.85* 0.33** 0.28 0.64 0.01 0.33 0.06 1.39 0.05 0.24 0.57

3.0%*** 6.6%*** 0.6% 1.9%*** 4.4%*** 9.1%*** 1.7%** 3.2%*** 1.7%** 4.7%* 7.0%*** 1.4%* 0.2% 5.5%** 6.6%*** 0.7% 1.1% 4.2%*** 1.0% 0.5% 3.7%*** 9.8%*** 1.7%* 0.1% 7.5%*** 12.6%*** 2.6%* 3.6%**

1.9% 1.7% 8.0%*** 0.6% 1.9% 3.8% 6.2%** 0.6% 2.1%** 3.7%* 0.3% 1.7% 5.1%*** 11.1%*** 2.0% 0.2% 8.3%*** 14.7%*** 1.7% 4.5%*

Table 12 Treatment effect of energy report in Site B. Comparison months

Use

Rate of change [%]

Amount of change [kWh/day]

Upper:Treatment G/Lower: Control G

July/Aug 2015

Aug 2014/Aug 2015

Jan/Feb 2016

Feb 2015/Jan 2016

Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances

All

Larger

Smaller

All

Larger

Smaller

All

All

All

All

All

All

1.4% 3.1% 3.5% 1.7% 3.2%** 14.4%*** 0.1% 1.4% 1.3% 1.5% 0.0% 1.0% 2.8%* 5.3% 2.9% 2.6%

0.7% 9.7%* 3.7% 1.6% 2.0% 6.7% 3.9% 1.6% 0.1% 0.3% 2.3% 0.9% 6.1%** 8.2% 7.6%** 5.3%*

0.2% 13.4% 6.4% 1.9% 0.3% 15.3% 9.2%* 0.6% 1.6% 3.3% 2.3% 1.2% 2.5% 4.1% 1.2% 2.7%

0.12 0.14 0.08** 0.16 0.56* 0.36** 0.00 0.19 0.24 0.16 0.01 0.03 0.68 0.28 0.20 0.31

0.20 0.13 0.21*** 0.24 0.83 0.31 0.07 0.31 0.34 0.41* 0.18 0.25 2.27*** 1.03** 0.78** 0.81

0.39 0.07 0.01 0.20 0.38 0.08 0.07 0.09 0.51 0.30 0.11 0.15 0.75 0.13 0.14 0.26

choose HEMS proactively, as part of photovoltaic system purchase, in most cases, and the reports were studied more closely overall than in the apartment case. Therefore, it appears that the interest in energy savings is smaller in apartment houses than in detached houses, a trend that is coupled with smaller consumption and limited saving options. 5. Discussion It is generally difficult to compare the results of the effects of our

reports with those of similar existing experiments because of the differences in the sample size, experiment duration, feedback content, feedback method, and the original energy demand levels by different countries. However, if we attempt to compare, the 3.4% electricity-saving effect in winter, reported in our study, can be considered sufficiently higher than the results of similar existing experiments. For example, the feedback programs by Opower team [13e15], which feed monthly data back to customers, showed 2% energy savings throughout the year in the United States, which is almost the same as our results in the winter. However, there is a

Y. Iwafune et al. / Energy 125 (2017) 382e392

391

Table 13 Treatment effect of energy report in Site C. Comparison months

Use

Rate of change [%]

Amount of change [kWh/day]

Upper:Treatment G/Lower: Control G

Jan/Feb 2015

Feb 2014/Jan 2015

July/Aug 2015

Aug 2014/Aug 2015

Jan/Feb 2016

Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances Total AC Hot water Other appliances

All

Larger

Smaller

All

Larger

Smaller

All

All

All

All

All

All

1.2% 13.2% 7.7%* 2.9% 0.8% 18.8%* 6.3% 2.1% 2.2%** 1.6% 2.2% 1.9%

0.17 0.07 0.20** 0.11 0.57 0.25 0.18 0.56** 0.15 0.01 0.18*** 0.01 0.10 0.06 0.01 0.11 0.04 0.13* 0.10 0.03

0.57* 0.13 0.33*** 0.18 0.03 0.22 0.06 0.10 0.22 0.33*** 0.11 0.07

0.10 0.11 0.03 0.31 0.09 0.25 0.05 0.14 0.12 0.00 0.03 0.11

1.3% 0.1% 2.7%** 0.9% 4.1%* 9.8%* 2.5% 3.3% 1.4% 3.6% 7.9%*** 0.4% 0.8% 5.2% 0.6% 1.4% 0.3% 3.2% 1.0% 0.3%

4.8%** 13.2%** 8.7%*** 2.0% 0.1% 7.9% 1.5% 0.7% 0.0% 4.3%* 2.2% 0.9%

5

Rate of change [%]

0 -1.7** -5 -10

-4.4***

-3.2***

-1.7*

-0.1 -2.6*

-3.7***

-3.6**

-7.5***

-9.1***

-9.8*** -12.6***

-15 -20 Feb 2014/Feb 2015 Total

Feb 2015/Feb 2016 Space heating

Hot water

Feb 2014/Feb 2016 Other appliances

Shaded areas show insignificant effecs. Statistical significance * p<0.10, ** p<0.05, *** p<0.01 Fig. 5. Accumulative treatment effect in site A.

smaller room for reduction in electricity consumption in Japanese households because it is about half of that of the United State [28]. Accordingly, if we compare the reduction rate of 5.4% for households classified as larger households in our experiment and the results reported by Opower, it can be said that a greater effect was obtained in our experiment. The Opower report was also demonstrated in Japan in winter 2016 and an energy-saving effect of 0.9e1.2% across 20,000 household samples was obtained [29]. Since we were able to provide high resolution information using HEMS, we obtained a higher effect than that. The effect of our report was lower than the 4.5% reduction achieved by Schleich et al. who fed back data to 1500 households through a web portal and monthly mail based on smart meter data in Austria in 2010 [11], which might be caused by fewer feedback frequency, such as once in the season.

6. Conclusion We developed a home energy report utilizing electricity data measured on a by-circuit basis by HEMS combined with household attribute data and verified its treatment effect in a study involving more than 1600 Japanese households. We sent HEMS reports to the participants in two winter seasons and one summer season. Hourly and seasonal electricity consumption were aggregated by usage in terms of space heating and cooling, hot water use, and other appliance use and compared with figures for similar households. Tailored advice and comparison with households with similar attributes were also provided. We developed an efficient selection method of comparable households using a regression analysis between electricity consumption and attribute data. The treatment effect was verified by applying a panel data

392

Y. Iwafune et al. / Energy 125 (2017) 382e392

regression random effects model in the comparison of electricity consumption between a treatment group, to whom the report was sent, and a control group not receiving the report. Our report was effective in winter and led to a reduction in overall use of 3.4% from the previous year for average households and a further reduction of 5.4% for households classified as larger households. A significant reduction in space heating use of 11.4% was observed for larger households. The treatment effect in summer was not significant for the average household; however, larger households reduced usage by 2% overall, and reductions from the previous month of 6.8% in terms of space cooling and 7% in terms of hot water use were observed for this group. By contrast, smaller households increased their space cooling consumption by more than 10% by month, which may be considered to be an undesirable boomerang effect. Furthermore, an accumulative treatment effect over two years was confirmed in site A. The reduction in total consumption in the first winter (2015) was 4.4%; for the winter of 2016 it was smaller (3.7%). These results indicate declining margins of energy saving with repetition of diagnosis. Finally the two year accumulative effect of reduction in total consumption in winter was 7.5%, confirming the effectiveness of continual intervention. Since our samples were biased toward relatively new households that have installed HEMS, the results are not necessarily representative of the effect on the current average of Japan; however, as the use of HEMS gradually becomes more popular, the energy savings reported in our energy report will surely become achievable. There is no intervention method that works for every household; in fact, as mentioned above, in some cases intervention with households consuming less produced a reverse (boomerang) effect. Thus, it is not necessarily effective to provide feedback information to all people under all circumstances; in fact, conservation measures undertaken by larger households resulted in larger reductions to electricity use. Historically, Japanese average household energy demand has been quite smaller than in European or American countries, especially in terms of space heating demand, owing to more restricted space heating practices, such as heating only occupied rooms for occupied time only using room heating devices [29]. In order to carry out cost-effective energy audits of Japanese households, we will need to extract larger households with higher potential energy saving margins than similar households and determine a suitable methodology for engaging them. In future work, we will carefully analyze these and further experimental results and clarify what types of measure are effective for different household types based on their particular attributes. Furthermore, we will use e-mail reports in place of paper reports in order to increase the frequency of report sending to reduce the audit cost and enhance the effect. Acknowledgement This research was supported by JST, CREST and the Ministry of Environment in Japan. References [1] IEA. IEA statistics electricity information. 2014. [2] The Institute of Energy Economics, Japan. EDMC handbook of energy & economic statistics in Japan. 2015. [3] Ministry of Economy, Trade and Industry in Japan. Long term prospect of electricity demand in Japan in 2030. 2015.

[4] Abrahamse W, Steg L, Vlek C, Rothengatter T. A review of intervention studies aimed at household energy conservation. J Environ Psychol 2005;25(3): 273e91. http://dx.doi.org/10.1016/j.jenvp.2005.08.002. [5] Darby S. The effectiveness of feedback on energy consumption, A review for DEFRA of the Literature of metering, billing and direct displays. Oxford: Environmental Change Institute, University of Oxford; 2006. [6] Fischer C. Feedback on household electricity consumption: a tool for saving energy? Energy Effic 2008:1e79. http://dx.doi.org/10.1007/s12053-008-90097. [7] Faruqui A, Sergici S, Sharif A. The impact of informational feedback on energy consumptionda survey of the experimental evidence. Energy 2010;35(4): 1598e608. http://dx.doi.org/10.1016/j.energy.2009.07.042. [8] Delmas MA, Fischlein M, Asensio OI. Information strategies and energy conservation behavior: a meta-analysis of experimental studies from 1975 to 2012. Energy Policy 2013;61:729e39. http://dx.doi.org/10.1016/ j.enpol.2013.05.109. [9] Ea Energy Analyses. Impact of feedback about energy consumption. 2015. [10] Ramos A, Gago A, Labandeira X, Linares P. The role of information for energy efficiency in the residential sector. Energy Econ 2015;52(Supplement 1): S17e29. http://dx.doi.org/10.1016/j.eneco.2015.08.022. € lz S, Brunner M. Effects of feedback on residential [11] Schleich J, Klobasa M, Go electricity demanddfindings from a field trial in Austria. Energy Policy 2013;61:1097e106. http://dx.doi.org/10.1016/j.enpol.2013.05.012. [12] Houde S, Todd A, Sudarshan A, Flora JA, Armel KC. Real-time feedback and electricity consumption: a field experiment assessing the potential for savings and persistence. Q J IAEE’s Energy Econ Educ Found 2013;34(1):87e102. [13] Allcott H. Social norms and energy conservation. J Public Econ 2011;95(9e10):1082e95. http://dx.doi.org/10.1016/j.jpubeco.2011.03.003. [14] Allcott H, Rogers T. The short-run and long-run effects of behavioral interventions: experimental evidence from energy conservation. Am Econ Rev 2014;104(10):3003e37. http://dx.doi.org/10.1257/aer.104.10.3003. [15] Navigant Consulting I. Home energy report pilot program evaluation FINAL REPORT. 2015. [16] Abrahamse W, Steg L, Vlek C, Rothengatter T. The effect of tailored information, goal setting, and tailored feedback on household energy use, energyrelated behaviors, and behavioral antecedents. J Environ Psychol 2007;27(4):265e76. http://dx.doi.org/10.1016/j.jenvp.2007.08.002. [17] Ministry of Environment in Japan. Japanese eco-home diagnosis (in Japanese). 2016. 2016/11/22. [18] Minister of Economy, Trade and Industry in Japan. A report from the smart meter system planning conference security review working group. 2015. 2016/11/22. [19] Iwafune Y, Yagita Y. High-resolution determinant analysis of Japanese residential electricity consumption using home energy management system data. Energy Build 2016;116:274e84. http://dx.doi.org/10.1016/ j.enbuild.2016.01.017. [20] Electric Power Research Institute, Inc. Guidelines for designing effective energy information feedback pilots: research protocols. 2010. [21] Lawrence Berkeley National Laboratory. Evaluation, measurement, and verification (EM&V) of residential behavior-based energy efficiency programs: issues and recommendations. Lawrence Berkley National Laboratory Report. 2012. [22] Lynham J, Nitta K, Saijo T, Tarui N. Why does real-time information reduce energy consumption? Energy Econ 2016;54:173e81. http://dx.doi.org/ 10.1016/j.eneco.2015.11.007. [23] Gleerup M, Larsen A, Leth-Petersen S, Togeby M. The effect of feedback by text message (SMS) and email on household electricity consumption: experimental evidence. Energy J 2010;31(3):113e32. [24] Costa DL, Kahn ME. Energy conservation "Nudges" and environmentalist ideology: evidence from a randomized residential electricity field experiment. J Eur Econ Assoc 2013;13. http://dx.doi.org/10.1111/jeea.12011. [25] Gans W, Alberini A, Longo A. Smart meter devices and the effect of feedback on residential electricity consumption: evidence from a natural experiment in Northern Ireland. Energy Econ 2013;36:729e43. http://dx.doi.org/10.1016/ j.eneco.2012.11.022. [26] Ida T, Ito K, Tanaka M. Using dynamic electricity pricing to address energy crises: evidence from randomized field experiments. Cambridge, MA, USA: 36th Annual NBER Summer Institute; 2013. [27] Schultz WP, Nolan JM, Cialdini RB, Goldstein NJ, Griskevicius V. The constructive, destructive, and reconstructive power of social norms. Psychol Sci 2007;5(18):429e34. [28] Iwafune Y. Energy conservation and energy management in Japanese residential sector. In: The world engineering conference and convention (WECC2015), Kyoto; 2015. [29] Hirayama S, Nakagami H, Tsurusaki T, Haig K. Japan's first large-scale home energy report pilot study: impact on Japanese consumers' awareness, motivations, and electricity consumption. In: 4th european conference on behaviour and energy efficiency; 2016.