The Power to Conserve: A Field Experiment on Electricity Use in Qatar

Our intervention consists of two randomized 'nudge'-style interventions to motivate reductions in electricity use by evoking both identity and agency. The first is a message quoting a passage from the Qur'an stressing the importance of conservation (Religious Message). The second is a message reminding households that the government of Qatar prioritizes energy conservation (National Message). The treatments that we use leverage injunctive norms (what people "ought" to do) rather than descriptive norms (what people actually do), and are intended to leverage individuals' identity. Additionally, both messages evoke agency, by mentioning that "You have the power to conserve!".

2019-05-01

2019-10-31

Monthly electricity use in kWh.

Since Akerlof and Kranton's seminal work, economists have theorized that identity can affect economic choices through taking actions that preserve individuals' self-image (Akerlof and Kranton, 2000). Normative prescriptions that are inherent in certain aspects of identity can increase personal utility when individuals take actions conforming to such prescriptions. Based on this idea, we use explicit primes leveraging injunctive norms of conservation inherent in two aspects of identity in Qatar.

First, we leverage the fact that religion is an important part of Qatari life, and its residents are predominantly Muslim. Therefore, priming religious values promoting conservation could lead to behavior change if people value their self-image as Muslims. In our treatment message, we use a specific verse from the Qur'an that asks its followers to 'waste not by excess'. This treatment is similar in spirit to the message used to encourage credit card debt repayment in Bursztyn et al. (2019).

Second, residents of Qatar may see themselves as playing a role in the country's stated desire to develop its economy in a more sustainable manner. Qatar instituted the National Program for Conservation and Energy Efficiency in 2012 - a campaign to encourage conservation of electricity and water. We reference this program and its patronage by the Emir of Qatar in our second treatment message to prime individuals' identity as Qatari residents and remind them that conserving energy is congruent with this identity.

Finally, in both treatment messages we remind customers that they have the agency to conserve energy. This part of the message serves to reinforce intrinsic motivation to conserve energy, which is important since extrinsic financial rewards to conserve energy are not well established in Qatar.

To select customers into the sample for randomized assignment, we proceeded in the following way. First, we keep only those meters in the database for which there exist at least one bill-month observation for the period between April 2018 to March 2019. Second, we keep only those meters that are billed as either a flat or a villa within this time period. Third, we include only those meters that are registered as belonging to "Regular Customers" or "Qatari Owners", excluding properties that are registered as being "Rented out by Qataris". Fourth, we consider only those customers who have a cellphone number registered with the utility to randomize into treated and control groups while balancing on observed average monthly electricity use over April 2018 to March 2019. Finally, we note that electricity use varies not just across months of the year, but also by type of residence (flats or villas) and ownership category (national or non-national). We anticipated heterogeneous effects among these groups and, hence, stratify the experiment on type of residence and ownership category. This results in three strata – (i) flats (n=4,803), (ii) villas owned by non-Qatari individuals (n=647), and (iii) villas owned by Qatari individuals (n=665). We do not include Qatari owners of flats as a separate stratum due to the low numbers of such premises relative to other categories in the customer database.

Overall, and within each strata, we divided customers into the two treatment groups and a control group in a 2:2:1 ratio within the sample of customers who had registered phone numbers in the utility's database.

We use a pre-intervention survey on a randomly selected sample of 247 customers to understand what characteristics of customers explain heterogeneous responses to the treatment messages.

Randomization in office by a computer.

Individual customers.

No

N/A

The planned experiment design was to include 109,728 customer-month observations - 6,096 customers over 18 months. However, due to errors in electricity meter information and reading types that we describe in the paper, we restricted the number of customers and increase the number of months of pre-intervention electricity use data. The process and exclusion criteria are detailed in the paper. Our preferred sample comprises 4,836 customers with pre-intervention use collected over January 2016 to February 2020. Our preferred sample comprises 161,254 customer-month observations.

We planned to have 2,438 customers in Religious message group, 2,438 customers in National message group, 1,220 customers in Control group in the experiment. After exclusions detailed in the paper, our preferred sample comprises 1,951 customers in the Religious message group, 1,937 customers in the National message group, and 944 customers in the Control group.