The Power of Framing Loss Aversion and Energy Consumption - an Austrian RCT

ll participants receive weekly feedback via a mobile app that compares their electricity use to a personalized baseline. This baseline is calculated as the average of the past eight weeks, adjusted for outliers (e.g., holiday weeks) and corrected for individual consumption variability using the household’s standard deviation. To avoid rewarding random fluctuations, points are only awarded when consumption reductions exceed typical variability.

Scoring System:
For every 0.5 percentage points by which participants reduce their weekly consumption below the baseline threshold, they receive 1 point—up to a maximum of 10 points per week.

Study Design:
The intervention lasts four weeks and randomly assigns participants to one of three groups:
• Gain Frame (N = 500):
Participants start with 0 points and can earn up to 10 points per week for reducing consumption below the threshold.
• Loss Frame (N = 500):
Participants start with 40 points and lose up to 10 points per week if they do not save enough. The point loss mirrors the gain group’s scoring in reverse.
• Control Group (N = 300):
Participants receive the same electricity feedback via the app but do not earn or lose any points.

Incentives & Feedback:
• Each point corresponds to a lottery ticket for a weekly prize draw. The total prize pool is up to €3,000.
• Participants receive:
• Weekly push notifications summarizing their performance,
• Midweek reminders,
• Real-time app updates showing:
• current point balance (lottery tickets),
• cumulative electricity use,
• daily and weekly baselines.

The intervention is a four-week behavioral energy conservation game implemented through a mobile application. All participants receive weekly feedback comparing their household electricity consumption to a personalized baseline. This baseline is calculated using the household’s average weekly electricity use over the eight weeks prior to the intervention, adjusted for outliers (e.g., unusually low or high usage weeks) and corrected for normal variation using the standard deviation of past consumption. This approach ensures that only meaningful savings—beyond typical fluctuations—are rewarded.

Participants are randomly assigned to one of three groups:
• Gain Frame (N = 500): Participants start with 0 points and can earn up to 10 points per week depending on how much they reduce electricity consumption below their baseline. For every 0.5 percentage point reduction, 1 point is awarded.
• Loss Frame (N = 500): Participants start with 40 points and lose up to 10 points per week depending on how little they reduce their electricity consumption. The loss function mirrors the gain group’s scoring in reverse: for every 0.5 percentage points not reduced, 1 point is lost.
• Control Group (N = 300): Participants receive identical electricity feedback through the app but do not accumulate or lose points.

Each point corresponds to a lottery ticket in a weekly prize draw, with a total prize pool of up to €3,000. Participants are informed of their point status, energy use, and baselines through weekly push notifications and can track their progress in real time within the app. Midweek reminders are sent to increase engagement. The intervention is designed to test the effect of gain vs. loss framing on electricity conservation, as well as to study dynamic belief updating and reference point adjustment under real-world conditions.

2025-07-06

2025-08-03

1. Weekly Electricity Consumption (kWh):
Total household electricity use per week, measured via smart meters during the four-week intervention period. This is the primary behavioral outcome to evaluate the effect of incentive framing (gain vs. loss) on energy conservation.
2. Points Accumulated or Lost:
Number of points earned (gain group) or retained (loss group) per week and in total, based on weekly consumption relative to the personalized baseline. One point is awarded (or not lost) for every 0.5 percentage point reduction below the household-specific baseline. This outcome captures behavioral responsiveness and effort intensity.

1. Percentage Deviation from Baseline:
Computed as the percentage change in weekly electricity consumption relative to a participant’s personalized baseline. Negative values reflect reductions. Used to assess relative performance and validate effects alongside absolute kWh savings.
2. Success Rate (Binary):
A binary indicator equal to 1 if a household’s weekly consumption falls below its baseline; 0 otherwise. Captures whether participants “succeeded” in saving energy in a given week. Used in binary models and subgroup analyses.
3. Total Points Earned or Retained (0–40):
The cumulative number of points earned (Gain Frame) or retained (Loss Frame) over the four-week intervention. Serves as a direct measure of sustained behavioral compliance under each framing condition.
4. App Engagement Metrics:
Includes number of logins, session duration, and response to push notifications (e.g., click-through rates). Used as control variables and for heterogeneity analyses to assess whether engagement moderates treatment effects.
5. Belief–Behavior Correlation:
Measures the relationship between participants’ stated expectations (points they expect to earn or retain) and actual performance (points or energy saved). Explores whether motivational beliefs predict effort and how this varies by frame.
6. Heterogeneity by Baseline Consumption and Appliances:
Pre-specified subgroup analyses based on quartiles of pre-intervention electricity consumption and household characteristics (e.g., presence of PV, EV, heat pump). Tests whether treatment effects vary by baseline energy use or technology adoption.
7. Notification Responsiveness:
Change in intra-week electricity usage following midweek push notifications. This exploratory outcome investigates whether reminders differentially trigger behavioral adjustments depending on treatment frame.

This randomized controlled trial tests how behavioral framing—specifically, gain versus loss framing—affects household electricity consumption. The study involves approximately 1,300 Austrian households, each of which receives weekly, app-based feedback on electricity use relative to a personalized baseline over a four-week period.

Participants are randomly assigned to one of three groups and stay in the group for 4 weeks (duration of the trial).
• Gain Frame (N = 500): Participants start with 0 points and can earn up to 10 points per week by reducing their electricity consumption below a personalized baseline. For every 0.5 percentage point (p.p.) reduction below the baseline, they earn 1 point.
• Loss Frame (N = 500): Participants start with 40 points and lose up to 10 points per week if they fail to reduce electricity use. For every 0.5 p.p. reduction, they avoid losing 1 point.
• Control Group (N = 300): Participants receive the same electricity feedback via the app but do not accumulate or lose points.

Each point corresponds to a lottery ticket for a weekly prize draw (total prize pool: up to €3,000). The app displays each household’s weekly electricity usage, personalized daily and weekly baselines, and current number of points. Participants receive one midweek push notification (reminder) and one end-of-week push message summarizing their performance and point status.

⸻

Recruitment is conducted by a regional utility provider, which sends postal invitations to households equipped with smart meters. Interested participants complete an online registration and baseline survey, which gathers data on socio-demographics, appliance ownership (e.g., PV systems, EVs, batteries, AC), and participation motives. Participants consent to connect their smart meter data to a mobile app. Electricity consumption is measured at 15-minute intervals.

Participants will be randomly assigned to treatment groups using block randomization to ensure balance across key pre-treatment covariates. Blocking will be based on:
• PV ownership (yes/no)
• Grid operator (Netzbetreiber: Linz Netz or Netz Oberösterreich)
• Data quality status, indicating whether both delivery and feed-in meters provide valid data or only the main meter does.

Randomization will be conducted via computer in R studio using a fixed random seed to ensure reproducibility. Assignment will take place after all participants have completed registration and consented to participate.

The unit of randomization is the individual household (app user). Each household is assigned to one of the three treatment arms (gain frame, loss frame, control) independently. There is a single level of randomization; no group- or cluster-level assignments are used.

No

0

N=1,300 individuals

Sample size (or number of clusters) by treatment arms

500 households – Gain Frame

600 households – Loss Frame

300 households – Control Group

Total planned sample: 1,300 households.

Details are in the pre analysis plan