H1: Support for sufficiency-oriented policies will be different between framing manipulation groups to whom they are presented as a) energy demand reduction policies (the energy demand condition); b) quality of life (variously health, housing, cost of living) policies (the co-benefits condition); c) both energy and quality of life policies (the combination condition), such that:
a) Support for policies will be higher amongst those who see the co-benefits (participants in the co-benefits and combination conditions) compared to participants in the energy demand condition. [Directional hypothesis]
b) Support among participants in the combination condition will be different to participants in the co-benefits condition. [Non-directional hypothesis]
c) The framing manipulation will interact with climate attitudes, such that co-benefits will have a stronger positive effect on support levels among people for whom climate is a lower priority. [Directional hypothesis]

H2: Support for sufficiency-oriented policies will be greater when it is stated that the policies are already implemented elsewhere (the exists manipulation). [Directional hypothesis]
H3: Perceived support for policies will be lower than actual support for policies (which we will refer to as a perception gap). [Directional hypothesis]
H3b: The perception gap will be affected by the framing manipulation, such that it will be larger in the energy demand condition compared to the two other conditions. [Directional hypothesis]
H4: Willingness to express support or opposition will depend on the degree to which the policy is supported or opposed. [Non-directional hypothesis]

This study will use a mixed experimental design.
Each participant will respond to 6 policies that are randomly drawn from a pool of 16. 3 will drawn from a pool of 8 “enabling” policies, and 3 will be drawn from a pool of 8 “restrictive policies”, i.e. the numbers of enabling and restrictive policies are counterbalanced. We will refer to this counterbalanced variable as policy restrictiveness.
Participants will be randomly and evenly assigned to one of three groups for the framing manipulation (a between group manipulation with three conditions). For one of the groups, the policies will be framed as energy demand reduction measures (the energy demand condition); for a second group, they will be framed as quality-of-life measures (the co-benefits condition); the third group, both energy demand and quality-of-life benefits will be combined (the combination condition).
For every participant, the second three policies they see will include a statement about where they are already implemented or supported (a within-subject manipulation, which we will refer to as the exists manipulation).
For each of the 6 policies in turn, participants will 1) rate their support, 2) estimate national support, 3) indicate whether they would express their opinion to others, 4) rate its fairness, and 5) rate its effectiveness.
Afterwards, each participant will be presented with one of the same 6 policies again, and asked to indicate how it could be changed in such a way that would increase the extent to which they would support it.
Participants will then rate how well they understand climate issues/how to save energy, and how important climate, health, housing, and cost of living are to them on 7-point scales.
Participants will then complete two tasks that measure knowledge about climate issues.
Participants will then indicate their support for political parties and political ideology.
Finally, participants will complete personal and socio-demographic information.

Balanced block randomisation will be used to randomly split the sample into three groups as described in the previous section.
Participants will be presented with 6 policies to rate in the study. Three of these will be “enabling”, and three “restrictive”. Each set of three policies will be randomly selected form a longer list of 8 and 8. The order of presentation will be randomised. And one policy will be randomly selected from the 6 to inquire further about.

--------------------------------
Manipulated variables:
Framing manipulation: participants will be randomly and evenly assigned to one of three groups for the framing manipulation (a between group manipulation with three conditions). For one of the groups, the policies will be framed as energy demand reduction measures (the energy demand condition); for a second group, they will be framed as quality-of-life measures (the co-benefits condition); the third group, both energy demand and quality-of-life benefits will be combined (the combination condition).
Policy-existence: for every participant, the second three policies they see will include a statement about where they are already implemented or supported (a within-subject manipulation, which we will refer to as the exists manipulation, which has two conditions).

Policy support – response on a 7-point scale from “fully oppose” to “fully support” to the question “would you oppose or support having this policy in Ireland?”
Effectiveness of policy – response on a 7-point scale from “not at all effective” to “very effective” to the question “how effective do you think the policy would be?”
Fairness of policy – response on a 7-point scale from “very unfair” to “very fair” to the question “how fair do you think this policy is?”
Perceived policy support of others – categorical response [options: 1) more people would oppose than support the policy; 2) most people would not be sure or care either way; 3) more people would support than oppose the policy] to the question “do you think other people in Ireland would support this policy?”
Willingness to express support – categorical response [options: 1) no, I would keep it to myself; 2) yes, but only with others I think share my view; 3) yes, with anyone, regardless of their views] to the question “would you feel comfortable sharing your opinion on this policy with other people?”

---------

Other measured variables
Reaction times – time spent on each policy.
Time on introduction page– time spent on the introduction page.
Factors that increase support – open text response that will be coded exploratively.
Self-rated knowledge – using the following question from the “Climate Change in the Irish Mind” survey:
How much do you know about climate change? Would you say you… • Have never heard of it • Know a little about it • Know a moderate amount about it • Know a lot about it
Personal importance of
a) Climate change
b) Health
c) Housing
d) Cost of living
- Each will be measured on 7-points scales from “not at all” to “very important”.
Climate attitudes - using the following question from the “Climate Change in the Irish Mind” survey:
Do you think climate change should be a very high, high, medium, or low priority for the Government of Ireland?
Knowledge – (sub-)sectoral final energy consumption – participants will be asked to rank 6 domains of energy use from most to least consuming. Participants will score a point for each of the sectors that they rank within one place of its correct rank.
Knowledge – effective climate actions – participants will see five pairs of climate actions and be asked to select which action in the pair is most effective in mitigating climate change. We will create another score variable, based on one point per correct response, a range of 0-5.
Political party support – participants select up to three political parties from a list that best align to their views.
Political ideology – 1-7 scale from “extreme left” to “extreme right”.

Age
Gender
Education
Locality (region, county and city/town/village/rural setting)
Household composition (sharing, number of people, children, people with chronic health issues or a disability)
Socio-economic status – social grade categorisation based on occupation of chief income earner of the household
Employment status
Household income (annual gross)
House tenure
House type, size
Typical approximate frequency employed people work from home
Mobility (car use, bike/similar use, number flights taken per year, personal and business)

H1 (a, b, c), H2: to test whether support is influenced by the framing manipulation and/or the exists manipulation, we will run an ordinal logistic regression model in which support is the dependent variable (pooling all policies together, i.e. six observations per participant), and the manipulation variables are independent variables. We will use standard errors clustered at the individual participant level and random intercepts by participant. We will run a model with an interaction term for the framing manipulation and climate attitudes variables (hypothesis 1c).
We will add relevant sociodemographic and household variables (age, gender, education, social grade, household composition, income) as covariates. We will control for order effects of the presentation order of the policies.
The models will include our measures of knowledge as well as a categorical variable for policy restrictiveness. In secondary models, we will add fairness, and perceived effectiveness, to check if these variables are mediators.
Note that depending on the distribution of the outcome variable, we might instead categorise the outcome variable to suit. We will run linear regression models as robustness checks.

H3: to test if perceived support for policies is lower than actual support (i.e. whether there is a perception gap), we will compare a) the proportion of support ratings that were 5 or above across all polices to b) the proportion of responses indicating a belief that most other people in Ireland would support the policy. Primarily, we will use descriptive statistics to compare the responses, and break these out by policy sector and type. We will use a McNemar’s chi-square test to assess whether differences are statistically significant. Note that the variables used will be binary.
For variable a) in the comparison above, 0 will denote a support rating below 5; 1 will denote a support rating of 5 or higher. For variable b) 0 will denote that participants either indicated they thought that most people would be unsure or not care, or that more people would oppose it than support it; 1 will denote that they thought more people would support it than oppose the policy.
H3b: to test whether the perception gap is affected by the framing manipulation (such that it is larger in the energy demand condition compared to the others), we will first compute the gap by subtracting the two variables described in the previous paragraph. We will then conduct a Pearson's chi-squared test to first determine if there is an overall significant difference in the distributions across the framing manipulation groups, and then we will use post-hoc pairwise tests to identify where any differences are.

H4: to test whether willingness to express support or opposition is influenced by the degree of support or opposition, we will run an ordinal regression model in which willingness to express support or opposition is the dependent variable (pooling all policies together, i.e. six observations per participant), and the policy support is the main independent variable. We will use standard errors clustered at the individual participant level and random intercepts by participant.
We will add relevant sociodemographic and household variables (age, gender, education, social grade, household composition, income) as covariates.
The models will include our measures of knowledge as well as a categorical variable for policy restrictiveness. In secondary models, we will add fairness, and perceived effectiveness.
Note we will check if there is a linear relationship between the log-odds of the cumulative probabilities of the willingness to support and support variables. In the case that the relationship is non-linear, we will use the most suitable alternative option (likely to a partial proportional odds model or a multinomial logistic model (which does not assume proportionality).

----------------------
Exploratory analyses:
We will exploratively code and describe responses to the question that asks participants what would increase their support for one random policy.
We will check for effects on 1) support and 2) willingness to express support of political ideology and relative perceived importance economic growth compared to climate change by adding these variables to the primary statistical models of those two variables described above.
We will check if different types of co-benefits (health-, housing-, or cost-of-living-related) are more or less associated with policy support ratings, by running a version of the primary support model described above in which the ¬co-benefits condition of the framing manipulation variable is replaced by the more granular variable of co-benefit type. We will run this model first including everyone but we will also check if results are different when excluding the energy demand condition group (to whom co-benefits were not explicitly mentioned).
We will check if the model results line up with importance given to the issues (health, housing, cost of living, climate) on the 7-point rating scales. This will be an informal comparison without use of statistical tests.
We will also model 1) perceived fairness and 2) perceived effectiveness of policies to test whether they are influenced by a) the experimental manipulations, b) climate knowledge, c) policy restrictiveness, and d) political variables. We will add relevant sociodemographic and household variables (age, gender, education, social grade, household composition, income) as covariates.
Note we will also report any perceived fairness and effectiveness effects on support, and willingness to express support – we included these variables in the models described in the statistical models section above that test primary hypotheses.
We will check the extent to which various other variables impact support for relevant policies, for example:
- mobility variables (car use, bike use, flights taken per year) on transport and aviation policies
- geographical variable on policies tied to specific places (e.g. urban traffic, proximity to coast for sail & rail)
- household variables on housing policies.
If we find certain socio-demographic characteristics predict policy support, we will check for interactions between them and the experimental manipulations.

------------------------------
Other analysis:
In a survey report separate to the experiment write-up to which the rest of this document pertains, we will describe individual policy results and model different groups of policies separately (by sector and co-benefits) to further describe sociodemographic, household, political associations with policy support.
These separate models for policies grouped by sector and outcome will include different independent variables that are most likely to be relevant to the different types of policies.

We will collect data from 2,000 people. Each participant will respond to 6 policies from a pool of 16 policies. This will provide 750 responses per policy. Because we also seek to use the study as a survey (which is not the subject of this pre-registration but is relevant to the sample sizing), this is important. 750 responses will allow us to get a representative view of Irish people for each of the policies.
In terms of the experiment, the 2,000 participants will each be randomly assigned to one of three groups. Here responses to individual policies are the primary interest. We will pool policy responses for modelling and use participant clustered standard errors. Our sample size will have 1,333 cases for the interaction with the most combination, which will be enough to detect effect sizes of interest (in this case of sector (transport, electricity, built environment)/policy outcome (housing, health, cost of living) x framing condition on support).