We are interested in the following primary research questions:

1. Does CCR myth buster content result in more accurate CCR knowledge, more positive affect towards CCR, and higher support for carbon pricing relative to control conditions? Does all CCR myth busting content perform equally? 
2. Does CCR operational transparency content result in more accurate CCR knowledge, more positive affect toward CCR, and higher support for carbon pricing relative to control and myth busting conditions? 

We hypothesize that participants in the myth-busting conditions will have significantly more knowledge of CCR, positive affect and higher support for CCR compared to the control conditions. Relative to the active and passive controls and myth busting conditions, participants in the operational transparency condition will have significantly more accurate knowledge of CCR, positive affect and higher support for CCR.

This study will deploy a between-subjects design, with each participant randomly assigned to view a different version of the experimental stimuli. A randomized controlled trial will be conducted to compare seven experimental conditions.

To begin, all participants will complete three screening questions to ensure their eligibility for the study and three pre-intervention questions to assess their initial awareness of CCR. Participants will then be randomly assigned to 1 of 7 conditions (see the “Experimental Conditions” section of the attached Analysis Plan for more details on the conditions) :

1. Passive control
2. Active control
3. Myth-buster 1 condition: CCR eligibility
4. Myth-buster 2 condition: CCR labels on bank statements
5. Myth-buster 3 condition: Using CCR payments
6. Myth-buster 3 condition: Receive more than you pay in carbon pricing
7. Operational transparency condition

All conditions (except the passive control) will be presented with a mock tweet about CCR, which will include an animated infographic and a clickable link for more information on the rebate. Testing will be conducted using Qualtrics Survey Software, an online experimentation platform. Participants will be recruited from Qualtrics survey panel. All participants will then be asked the same set of key outcome measures followed by a series of attitudinal and demographic questions to enable richer insights and possible segmentation.

Primary Outcomes 

1. CCR knowledge (frequency of payments, payment amount, how payments are sent/received, purpose of CCR (i.e., rebate for carbon price)) 
2. Emotions (affect) about CCR
3. Perceived effectiveness and fairness of carbon price/fuel charge and CCR 
4. Policy support for the federal carbon price/fuel charge 

 Secondary Outcomes 

1. Policy support for CCR in its current form vs. alternative approaches for deploying carbon price revenues 
2. Click-through-rate to ECCC CCR website 
3. Policy support for other federal climate change, environment, and energy policies 
4. PARCA segmentation measures (climate change beliefs, social norms, affect (emotional response), perceptions of efficacy (to take climate action), and willingness to act) 
5. Self-reported climate action intentions for the next two months 
6. Trust in government 

The outcomes mentioned above are measured by asking participants to rate their levels of agreement or support on a series of statements, such as CCR knowledge (10 items) or policy support for carbon price (6 items). Wherein appropriate, we will perform Principal Component Analysis or create index variables to reduce the number of variables.

For segmentation measures, latent class analysis will be used to segment the sample based on attitude and behaviour-based groups.

The data cleaning steps taken by the Qualtrics Panels include:

1. Exclusions of partial responses (i.e., incomplete surveys), participants who reported being under 18 years of age or not being a resident of Canada.
2. A data scrub which will identify and remove poor quality responses, such as “straight-lining” responses, poor open-ended question responses, duplicate participants, speeding responses, etc.

Additionally, responses will be reviewed for any inconsistencies and exclusion will be decided on a case-by-case basis including participants who fail all three attention checks.

Wherein appropriate, we will perform PCA or calculate index variables to reduce the number of variables. This is most appropriate for questions on behavioural intent, carbon pricing, and awareness of CCR. Latent class analysis (LCA) will be used to segment the sample based on attitude and behaviour-based groups.

To ensure random assignment to conditions was successful, a series of balance tests will be conducted. Chi-square tests will be run to ensure equal distribution of participants across the seven conditions, and statistical tests will be performed to ensure conditions are balanced on key demographic variables. Descriptive and frequency analyses will be conducted to get a sense of the responses of the sample, such as their demographic information and sample average responses on key measures.

To examine the primary research questions, a series of one-way ANOVA tests will be run to study the effect of experimental conditions on the key outcomes (i.e., CCR knowledge, affect towards CCR, and support for carbon pricing). Post-hoc examination will be conducted to investigate the differences between treatment and control conditions using Tukey's Honest Significant Difference (HSD) test when sample sizes are approximately equal across groups. Alternatively, Scheffe's test will be utilized when sample sizes are unequal across groups (Family-wise Type I error rate: .05). These analyses are dependent on the fact that continuous dependent variables are normally distributed; if variables are not normally distributed (i.e., skewed), they will be transformed accordingly. In addition, the analyses may be changed to a non-parametric test. (Note- see the ‘Analyses’ section in the attached Analysis plan for a more detailed overview of the primary confirmatory questions and exploratory questions.)

Participants will consist of a sample of approximately 3000 Canadians in provinces currently eligible to receive CCR payments. The sample will consist of 50% women and 50% men for each age group (18 to 34; 35 to 54; 55+), and for each region:  
Atlantic Canada (Newfoundland, Prince Edward Island, Nova Scotia, New Brunswick);  
Ontario;  
Manitoba/Saskatchewan;  
Alberta.  
  
We will apply stratified randomization to ensure the balance of the treatment/control groups with respect to age and gender

The data cleaning steps taken by Qualtrics will exclude:

1. Removing data from partial responses (i.e., incomplete surveys), potential participants who reported being under 18 years of age or not being a resident of a CCR-eligible Canadian provinces.
2. Removing poor quality responses, such as “straight-lining” responses, poor open-ended question responses, duplicate participants, speeding responses, etc.

Additionally, responses will be reviewed for any inconsistencies and exclusion will be decided on a case-by-case basis including participants who fail all three attention checks.