The rate of increase in grant applications in the 12 months following community based social marketing activity will be higher in the intervention area compared to the mean of the control areas, compared with the 12 months preceding. [Directional]

This study is an experimental field trial. It is a between subjects design. Geographical regions approximately matched on social deprivation index scores, historical rates of retrofitting grant applications and population are randomly assigned to treatment or control conditions. The treatment consists of a CBSM information event and associated marketing campaign happening in the region. In the control condition region, there is no CBSM.

Three geographical regions in Ireland were randomly assigned using Excel random number generator to treatment and control groups.

Individuals are blind to the treatment in that they are unaware that an outcome variable is being measured. As such, they are unaware that they are in a treatment condition. People in the control group are naturally unaware also. No individual level outcomes are measured.

The outcome variable is the number of owner-occupied homes that make retrofit grant applications as a proportion of the total owner-occupied homes in the regions. Thus, it is a group level outcome. We aggregate data across grouped neighbouring electoral divisions that are in our pre-defined intervention and control regions. We compare the proportion of owner-occupied homes in the regions pre and post intervention and compare the difference in difference between intervention and control groups a year after the intervention has finished.

Data will be collected from a database of grant applications a year after the CBSM events have taken place. There is thus no participant data collection required for the trial.

Hypothesis: The rate of grant applications in the 12 months following the event will be higher in event areas compared to control areas.
We will conduct a group level difference in difference analysis to assess the effect of the intervention.
We will report a precise p-value, without cut-off criterion, alongside the effect size.

The sample size for the randomized trial will equal the amount of owner-occupied houses in the pre-defined regions, in addition to owner-occupied homes in surrounding electoral divisions that are within 20km, at the time of the study.

Significant consideration was given to the selection of the areas:

Step 1: Using Irish electoral division (ED) level data on owner-occupied homes, proportion of homes with a building energy rating (BER) of C2 or lower and proportion of solid fuel homes, a priority index was developed.
Step 2: By setting thresholds on the priority index, deprivation index, and number of residences in the ED and neighbouring EDs, a set of priority EDs were identified. The thresholds set under tight fitting are as follows:
• EDs with priority index in the 4th quantile
• EDs with number of residences greater than 2999 and less than 4000
• EDs with deprivation index greater than zero
• EDs which are not categorised as a city.
Step 3: Using GIS mapping, these priority EDs were mapped out and clusters of EDs were identified.
Step 4: Major towns near or within these clusters of priority EDs were identified.
Step 5: Using the town as the centroid, the radius was expanded to include the priority areas. Expanding the radius was done on the basis of population density i.e., setting a threshold that decreases gradually and setting the borders at the lowest threshold.
Step 6: Based on step 5, 27 intervention areas were identified. Past grant applications and retrofits in EDs within the ‘intervention areas’ were assessed to check if there is any significant difference in socio-demographics and grant application history.
Step 7: 27 areas identified were filtered by SEAI’s marketing and communications team based on their area of expertise. 8 intervention areas were suggested considering numerous factors relevant to hosting events.
Step 8: A basic descriptive analysis of grant applications, owner-occupied homes, C2 or lower homes, solid fuel homes, and geographical suitability were considered. The availability of nearby SECs was also considered in the selection process.
Step 9: Based on steps 7 and 8, the 8 areas suggested by Marcomms, and 2 areas suggested by BEU (Portlaoise and Charleville) were selected in the final list.
Step 10: Using this list of 10 areas, a new set of data was generated. Using the towns as a centroid, EDs within a 15 Km radius were included in the intervention area. Data on grant applications, owner-occupied homes, proportion of C2 or lower homes, and proportion of solid fuel homes were assessed.
Step 11: Observing the geographical proximity of areas (to avoid spill overs) and the potential to host multiple events, 3 clusters were identified which include Cork (Mitchelstown and Charleville), Roscommon (Roscommon and Carrick-on-Shannon), and Laois/Offaly(Portlaoise and Tullamore). Data on numerous ED characteristics were assessed to ensure they were similar.
Step 12: The final 3 areas were randomised as Treatment (Laoise/Offaly) and Control (others).

We do not expect a need for data exclusion to arise.

We do not expect missing data.