Experimental Design
About 75000 members of a Norwegian grocery chain's loyalty program are invited by email to take a survey. We expect about 3000 people to respond and they constitute our sample.
The participants first answer a block of question on demographics, shopping habits and preferences. After that they are randomly assigned to one of three groups:
Control 1 (C1) = No questions about CO2 emissions.
Control 2 (C2) = Questions capturing participants beliefs about CO2 emissions of different meat products.
Treatment (T)= Questions capturing participants beliefs about CO2 emissions of different meat products, AND information about correct CO2 emissions for the same products.
We decided to have two control groups for three reasons. First, it might be that the additional questions asked to the treatment group (T) affects the response rate, which would make it difficult to casually interpret a comparison between T and C1. By including C2 we reduce this risk as that control group answers the same number of questions. Second, it might be that simply asking respondents about their beliefs on CO2-emissions impact purchase behavior, and this we can capture by comparing C2 to C1. Third, by including C2 we can analyze if the treatment effect differ wrt baseline beliefs on CO2 emissions.
IMPORTANT: The main outcome variables are measured using transaction data from the grocery chain covering a period of several months after the survey has been conducted, and the transaction data from the grocery chain will not be accessible to the researchers before October 1, 2025. Hence, there is a clear distinction between the intervention period and the evaluation period. To make a more informed evaluation of the treatment effects we will update this trial after analyzing the survey data BUT before getting access to the transaction data, specifying the primary treatment comparisons to use as well as more details on the primary outcome variables.