How do income-based social norms influence environmental decision-making?

Participants will be randomly assigned to one of three experimental groups based on type of information received: (i) information about actual prevalence of PEBs in their own income group; (ii) information about actual prevalence of PEBs in the other income group; and (iii) no information (control).

With an evenly split sample between high- and low- income respondents, this design yields two control conditions (High_Control and Low_Control) and four treatment conditions: two ingroup (High_High, Low_Low) and two outgroup (High_Low, Low_High) comparisons, depending on whether the income group featured in the PEB information corresponds to the respondent’s own or another income group.

The treatment information on the prevalence of PEBs in high- and low-income groups is based on statistics collected in a previous survey conducted in the UK, where we surveyed high- and low- income respondents about actions they have taken to fight global warming.

2025-08-20

2025-12-31

1. Prior beliefs about the prevalence of fighting global warming among different income groups.
2. Posterior beliefs about the prevalence of specific PEB engagement by high- and low-income groups. To test for belief revision, we compare posterior beliefs across treatment and control groups.
3. Participants’ willingness to fight global warming.
4. Participants’ attitudes towards support for redistributive climate policies.

1. Perceived descriptive norms are measured by an incentivised question in which participants estimate, for the previous survey, the proportion of the UK high- and low- income population who self-reported engaging in climate action.
2. Posterior belief is measured by an incentivised question in which participants estimate, for the previous survey, the proportion of the UK high- and low-income population who have taken a series of specific actions to fight global warming over the past year.
3. Individual willingness to fight global warming is measured by an incentivised, real-stake donation task. Specifically, participants decide how to allocate £150 between donating to the organisation Atmosfair for carbon offsetting and keeping it for themselves.
4. Individual attitudes towards climate policies are measured by (non-incentivised) stated support for 3 different types of climate policies, including carbon tax in general, carbon tax that mainly favours the high-income and carbon tax that mainly favours low-income people.

None.

None.

Recruitment and eligibility:
The online experiment will be conducted on Prolific and programmed in Qualtrics. Eligible participants are required to be non-student residents of the UK, aged 18 or older, and with a Prolific approval rating above 90%. Low-income and high-income respondents are defined as those with net equivalised household incomes below £21,817 and at least £52,121 per year, respectively. These thresholds correspond to the bottom and top quintiles of the UK net equivalised household income distribution for the 2023-2024 tax year, based on the modified OECD equivalence scale and data from the Institute for Fiscal Studies (IFS).

Procedures:
1. Personal PEBs: Participants are asked to indicate their engagement in a series of PEBs over the past year.
2. Income group information and prior beliefs: Participants are informed of their positions in the UK income distribution based on their net equivalised household income, following which their perceived descriptive norms towards low- and high-income groups are elicited.
3. Intervention: Treatment groups receive information on prevalence of PEBs for either their own income group or the other income group.
4. Willingness to fight global warming: Measured through a real-stake donation task.
5. Attitudes toward climate policies: Participants report their support for a series of redistributive climate policies with different distributional impacts.
6. Posterior belief: Participants’ estimates of the prevalence of specific PEBs among low- and high-income groups.
7. In-group closeness: Measured via the Inclusion of Other in the Self (IOS) scale (Aron et al., 1992).
8. Advantageous and disadvantageous inequality aversion: Measured by incentivised distributional choice tasks (Bartling et al., 2009).
9. Risk and future preferences: Measured through self-reported willingness on a 0-10 scale (Falk et al., 2018).

Not available

Randomization will be done using embedded features of Qualtrics.

Individual.

No

Not applicable (individual randomisation)

A total of 1,200 participants will be included in the main experiment, with 600 from the high-income group and 600 from the low-income group.

Participants will be randomly assigned to treatment arms within each income group.
High-income participants: High_High: 200; High_Low: 200; High_Control: 200.
Low-income participants: Low_High: 200; Low_Low: 200; Low_Control: 200.

Based on our planned sample size of 200 participants per arm, we are powered to detect a minimum effect size of approximately 0.28 standard deviations (Cohen’s d) between two arms, assuming α=0.05 and 80% power. This calculation is based on standard assumptions for a two-sample comparison and does not rely on pilot data.