Experimental Design
To examine the extent to which perspective-taking and social categorization messages impact Mexican’s attitudes towards migrants, a randomized online experiment will be conducted. The experiment will take place on Prolific, a platform specialized in recruiting and managing participants for online research. It is to be noted that recent research indicates that Prolific stands out among similar online platforms for providing participants who consistently offer higher quality responses, adhere to instructions, work at a normal pace, and can verify their unique IDs, thus ensuring they are not bots (Douglas et al., 2023).
The population of interest is Mexican adults registered in Prolific, surpassing 2,800 individuals. To overcome one of the limitations encountered by previous experiments - which is that most of the research done has targeted college-age individuals (Paluck et al. 2021) the sample will consist of the adult population in general rather than only college-aged individuals. The experiment will involve a survey created using Qualtrics. Initially, participants will receive a brief explanation of the study, followed by a consent form. Upon agreeing to participate, they will be first asked a series of demographic questions. After this, participants will then be assigned to one of the videos or to the control group. Following the intervention, additional questions will be posed to evaluate the potential changes in attitudes regarding migrants both in regard to their impact in public safety perception as well as in regards to their inclusion in the labour market.
To determine whether the treatments impact the outcomes of interests, two statistical strategies will be employed: mean group differences and linear in means models.
Further, to ensure the comparability of the control and treatment groups, mean balance tests will be conducted to assess the baseline characteristics across the three groups. This process aims to discover whether any significant imbalances exist between the groups before the implementation of the treatment. If the balance tests reveal substantial imbalances between the treated and control participants across variables, a robustness check will be conducted using a linear regression that includes the treatment variable as well as the control variables that were found not to be balanced. This analysis will assess whether the treatment effect remains consistent when controlling for other variables that may influence the outcome. The mathematical expression for this robustness check is as follows:
Y=β+ ŤDi+X′B+ϵi
In the above equation, Y is the dependent variable, β is the intercept, ŤDi is the treatment variable, X′B is the variable or set of variables that were not balanced and ϵi corresponds to the error term.
After completing the robustness checks, and with the aim to answer Hypothesis 1, which suggests that receiving the treatment will increase positive attitudes toward migrants, the effect of the treatments will be analized with a saturated model, that given the three-group randomized design is expressed as:
Y=α+β1T1+β2T2 +ϵ
After this, heterogeneity analyses will be conducted to assess whether there are differences in effect across subgroups of interest. In this case, to answer hypothesis 3, which suggests that the effect will be lower in less educated individuals as well as in older people, a linear regression with interaction terms between treatment and both age and education respectively will be performed as follows:
yi = β+ ŤDi + agei + (ŤDi x agei) + Xi′B + ϵi
yi = β+ ŤDi +education + (ŤDi x education ) + Xi′B + ϵi