A field experiment on energy conservation with manipulation

As a treatment measure, the study introduces a rebate (monetary incentive) based on electricity savings during peak hours (5:00 PM to 9:00 PM). The baseline for calculating the rebate is each household’s average electricity consumption during the same month of the previous year (August and December, respectively). The event periods are set to seven days in both summer (August 18–24) and winter (December 15–21).

If a household’s electricity consumption during the event period falls below the baseline, a reward of 200 yen per kilowatt-hour saved will be provided, up to a maximum of 3,500 yen. If the average consumption during the event period exceeds the baseline, the household will be treated as having achieved zero savings and will not receive any reward.

2025-08-18

2025-12-21

electricity consumption

A total of 2,500 households, randomly selected from across Japan, will be divided into the following 10 groups to participate in a two-period (summer and winter) treatment experiment.

[1] Groups with Fixed Treatment Assignments (Groups 1–4)
Treatment status in each period is predetermined and does not depend on household behavior:

Group Period 1 (Summer) Period 2 (Winter)
UU No treatment No treatment
UT No treatment Treatment
TU Treatment No treatment
TT Treatment Treatment

[2] Groups with Behavior-Contingent Treatment Assignments (Groups 5–10)
Households receive treatment in the second period only if their electricity-saving rate in the first period exceeds a specified target (10%, 30%, or 50%):

Group Period 1 Treatment Savings Target Period 2 Treatment
U(C=10%) No 10% Conditional
U(C=30%) No 30% Conditional
U(C=50%) No 50% Conditional
T(C=10%) Yes 10% Conditional
T(C=30%) Yes 30% Conditional
T(C=50%) Yes 50% Conditional

Information about treatment assignment and baseline electricity usage will be communicated to each household via email or other means at the beginning of each event month and again during the event period.

Not available

Stratified randomization done in office by a computer (Households will be clustered by dividing their past summer and winter electricity consumption based on the median, classifying them into groups with consumption above or below the median for each season)

households

No

N/A

2,500 households (expected)

250 households (expected)

We determined that 10 groups, as described above, were necessary to achieve the objectives of this experiment. We then considered the number of households we could realistically recruit in the field and decided to assign 250 households to each group. Our power calculation indicated that, with 250 households per group, detecting a statistically significant treatment effect at the 5% level would require at least a 7.5% reduction in electricity consumption. In our previous study (Ida et al., 2023), we offered a rebate of 100 Japanese Yen per kWh and observed an average reduction in electricity usage of 5%. Since a 100 Yen rebate was unlikely to produce a reduction of 7.5% or more, we decided to increase the rebate to 200 Yen per kWh in order to reach the target effect size. Ida, T., Ishihara, T., Ito, K., Kido, D., Kitagawa, T., Sakaguchi, S., & Sasaki, S. (2024). Choosing Who Chooses: Selection-Driven Targeting in Energy Rebate Programs. (No. w30469). National Bureau of Economic Research.

The Pre-Analysis Plan for “A field experiment on energy conservation with manipulation”