Peer effects, Rewards, and Image Concerns in Energy Decision (PRICED)

We have partnered with a renewable energy provider to implement a field experimental study to test the effects of image concerns and financial incentives on the private adoption of renewable energy tariffs. To causally identify the effect of each intervention on behavior in the experiment, we will study the geographical distribution of customers and treat randomly selected defined geographical areas, ensuring that comparable areas are allocated across treatment and control groups. In the study proposed here, we will carefully select triplets comprised of three similar household clusters to each receive one of the treatments. A first treatment will allow us determine the effect of social image (as defined in Benabou and Tirole, 2006) on green energy adoption and customer retention. By comparing with a second treatment, in which financial incentives are combined with social rewards, the effect of each of these two elements can be disentangled. Financial incentives may either reinforce (crowd in) or attenuate (crowd out) the effect of social rewards. Finally, the experimental design will allow us to determine the effects of the visibility of pro-social behavior on local peer adoption. Importantly, the study will allow us to generate exogenous variation in the visibility of pro-social behavior, potentially leading to additional adoption through social contagion effects.

2019-02-27

2020-06-30

1. Intent to be part of the study as reported in the randomized consent elicitation process, as a function of the messaging received (control, treatment 1, treatment 2)
2. Self-reported use of signposts, and window clings, in treatment 1 (no financial incentive) compared to control (no intervention)
3. Difference in the self-reported use of signposts and stickers, between treatment 1 and treatment 2

1. Compliance (conditional on having consented) in treatment 1, treatment 2, for the use of signposts and window clingers, as a function of the messaging received
2. Difference between observed compliance and stated compliance and between estimated compliance and stated compliance
3. Adoption by others as attributable to our interventions (social contagion) with aggregate data analysis at the level of randomization (control groups both with consent—and therefore access to energy data—and without)
4. Adoption by others as attributable to our interventions (social contagion) using household-specific data on referrals (referral code unique to households)
5. Adoption by others as attributable to our interventions (social contagion) as a function of the characteristics of the “seeds”, which in turn may be a function of the messaging received (as well as other characteristics that will be observed in the data)
6. Potential contamination and spillover effects that our localized intervention could not prevent; Hawthorne-like effects if the design may allow them to take place in the untreated (in principle) control group that provides consent

There are two sources of exogenous variation in the study. Exogenous assignment to receive materials constitutes the first level of randomization. Households will first be assigned to either the control group (receiving no materials), treatment 1 (receiving materials with no incentive for displaying them), or treatment 2 (receiving materials with a financial incentive to display them). Within the treatment groups, half will receive a recruitment email that details the intervention to which they have been assigned, while half will receive a recruitment email that provides no information on the intervention (identical to the email sent to the consenting control group).

Households will be assigned to either control, treatment 1 (sign - no incentive), or treatment 2 (sign - incentive) using a stratified randomization. We randomly allocate 3,320 eligible postcode sectors (from the postcode sectors with at least one Good Energy customer as of February 2019) to be in one of the following 5 groups:
–
Control, Control consent email
–
Treatment 1, Control consent email
–
Treatment 2, Control consent email
–
Treatment 1, Treatment 1 consent email
–
Treatment 2, Treatment 2 consent email

Individuals in treatment groups 1 and 2 will receive either a garden sign or a window cling depending on whether they live in a detached, semi-detached house, or ground-floor flat (sign assignment) or a terraced home or apartment building (sticker assignment). Housing type will be determined according to the following procedure. Upon consent provision, participants will be asked to select their dwelling type from a list of options, representative of housing types in the United Kingdom (i.e. flat, terraced home, detached/semi-detached, cottage/bungalow, etc.). If consenting participants leave our platform without having provided information on their housing type, signs or clings will be assigned based on the characteristics, in terms of dwelling types, of the area where they live, which is identified with postcode sector level data. Information on whether determination is based on self-reported information or administrative records can be used to reduce noise in the empirical analyses. Additional information (e.g. floor number for flats) may be optionally requested in a study debrief survey to inform our analyses on the role of visibility for both behaviors driven by social rewards and social contagion.

Stratified randomization (done in office by a computer).

Postcode sector. Treatment assignment is defined at the postcode sector, hence all existing Good Energy customers (“seeds”) in a given postcode sector are assigned to the same treatment.

Yes

Households within postcode sectors. 664 postcode sectors per treatment arm.

Upper bound at the total number of GoodEnergy customers (some 60,000). Consent is necessary to participate in the experiment and so the number of observations varies depending on the fraction of consenting individuals.

664 postcode sectors per treatment arm.