Quantity misperception in household electricity consumption

A total of 1000 households will be visited for a brief survey of household appliances, behaviors, and knowledge. Out of these, a total of 400 households will be part of the control group. The remaining 600 households will receive one of two treatments:

Treatment 1 (n=300 households) will receive the meter-reading tutorial only. The surveyor shows the respondent how to read their electric meter, typically wall or pole-mounted near the front door. Since the meter displays cumulative consumption from the date of installation, surveyor then demonstrates how to calculate running consumption using a simple online calculator. The consumer simply enters the last bill’s final reading and reading date, along with today’s reading and today’s date. The website automatically calculates total consumption for the current period, number of days passed, and consumption per day (on which the subsidy is applied). An informational pamphlet reinforcing the same information is left with the consumer.

Treatment 2 (n=300 households) will receive the meter-reading tutorial (including the informational pamphlet), along with reminder SMS messages every 2 weeks for up to 1 month. SMS will contain the following message in local language: “Hello! As mentioned during our recent survey, we’re sending you a friendly reminder to check your electric meter to help you monitor your consumption. Use the free calculator <link>. Just enter the last bill's kWh reading + date and today's reading + date to calculate usage for the current period.”

To isolate the effects of the learning and cognitive costs, we provide basic information about the price discontinuity structure to all households. The control group receives a small information sheet about the price discontinuity structure (which is also contained in the informational pamphlet provided to treated households) to serve as an information placebo. Also, phone numbers are requested from all households, even though only treatment 2 receives SMS messages.

2025-06-09

2025-07-31

Our key outcomes of interest include:
1. The proportion of households with consumption per day below 10 kWh.
2. The variance of the distribution of consumption per day.

Reducing the costs of acquiring accurate consumption information should lead to a stronger response to the price discontinuity and a convergence in the distribution toward the threshold of 10 kWh/day.

1. We hypothesize that treatment would result in a higher proportion of consumers falling below the 10 kWh/day threshold, with the effect being stronger for treatment arm 2. To test this hypothesis, we will use the following specification:
y_i=α+β_1 T_i+X_i γ+ϵ_i
for household i where T_i is a treatment indicator, X_i are household-level controls, and \beta_1 is the coefficient of interest.

2. We hypothesize that treatment would result in a lower variance in the distribution of consumption per day as consumers are better able to target the 10 kWh/day threshold. We test this hypothesis using a Levene’s test of equal variances.

We are unable to use 2025 administrative data for this study. Instead, we will collect baseline consumption information at the time of the survey by requesting consumption data shown on the latest bill. We will then conduct phone surveys 1-2 months after the in-person survey to request information about the next bill (post-treatment).

We calculate consumption per day by dividing total consumption for that billing period by the number of days in the period. The subsidy is applied for the consumption per day metric to account for differing billing period lengths across households.

We consider heterogeneous treatment effects by household demographics, such as income, education, stock of appliances, and expectations over the current cycle. We test this by interacting the treatment indicator with different demographic variables w_i collected through the survey.
y_i=α+β_1 T_i+β_2 T_i*w_i+δw_i+X_i γ+ϵ_i

Since billing dates vary across households, we may survey and treat some households toward the beginning of their cycle and others toward the end of their cycle. We consider heterogeneous treatment effects by this timing of treatment.

A total of 1000 households will be surveyed on their behaviors, appliances, beliefs, and knowledge about their electricity consumption and the subsidy design. We focus on households that were recently relatively close to the price discontinuity in summer months.

Sampling for these 1000 households was done as follows:
1. Using administrative billing data from summer 2023 for one sub-division (i.e. city), households with average summer consumption of 8-12 kWh per day were identified.

2. The localities of these households were identified and localities with >50% meters defined as industrial or commercial were dropped.

3. Surveyors will visit the identified localities and complete surveys with sampled households. Treatment assignments are randomized at the household level and uploaded to SurveyCTO. The tablet notifies the surveyor of the household’s treatment status and shows them the relevant script(s) at the end of the survey:
- Control: "Thank you for your time. Your responses will help design policies and programs that could benefit domestic consumers."
- Treatments 1 and 2: "Thank you for answering these questions. I would now like to show you how you could check your own meter readings and keep track of your consumption. For this, I would, 1. first ask you to turn off all ACs in the home. 2. We would then spend 5 minutes looking at your electric meter and talking about how you could check the readings. 3. Lastly, I will show you a simple table with which you can note down your readings over time."
Note: We request the respondent to temporarily turn off ACs to reduce the load for safety reasons.

4. Two weeks and four weeks after the survey visit, households in Treatment group 2 will receive a reminder text message via WhatsApp.

5. A phone surveyor will call all households 1-2 months after the in-person survey visit to request the latest bill information.

The survey and treatment activities are completed with an adult member of the household. Surveyor will ask to speak to someone who is “knowledgeable about household energy use”. Survey will take approximately 30 minutes. Meter-reading tutorial will take approximately 5 minutes.

Households are randomly pre-assigned to a treatment group using Stata following a non-weighted, sampling-without-replacement method.

Surveyors are informed of this treatment assignment toward the end of the survey through Survey CTO, which uses prefills based on the household ID entered by the surveyor at the beginning of the survey.

Household

No

1000 households

1000 households

Treated group 1: 300, Treated group 2: 300, Control: 400