The Impacts of Informational and Emotional Messages on Attitudes Towards Migrants

We will conduct a between-subjects online experiment to estimate the causal effect of an informational and an emotional video (relative to a placebo video) on attitudes towards migrants.

2021-11-01

2021-12-31

Migrant Acceptance Index, Beliefs Index and Trust

Migrant Acceptance Index. The index is based on three questions that Gallup asked in 140 countries in 2016 and 2017 and updated again in 145 countries in 2019. The questions ask whether people think migrants living in their country, becoming their neighbors, and marrying into their families are good things or bad things. The index is a sum of the points across the three questions, with a maximum possible score of 9.0 (all three are good things) and a minimum possible score of zero (all three are bad things). The higher the score, the more accepting the population is of migrants.

Beliefs Index. The index is based on seven questions where respondents are asked about their level of agreement, on a 4-point Likert scale ranging from 1 (”totally disagree”) to 4 (”totally agree”), with statements about the impacts of migration on the respondents’ country. The index will be constructed following the methodology in Kling et al. (2007), by z-scoring the answer to each one of the seven questions and then taking a simple average, with the variables re-coded that more positive beliefs have higher scores.

Trust. This outcome corresponds to the self-reported level of trust towards migrants. It will consist of the (z-scored) level of agreement, on a 4-point Likert scale ranging from 1 (”totally disagree”) to 4 (”totally agree”), with the statement: “Migrants living in my country can be trusted just as much as you can trust [gentilic for respondent’s country]”.

Policy index and altruism

Policy Index. The index is based on three questions where respondents are asked about their level of agreement, on a 4- point Likert scale ranging from 1 (“totally disagree”) to 4 (”totally agree”), with their government implementing certain policies to integrate migrants. The index will be constructed following the methodology in Kling et al. (2007), by z-scoring the answer to each one of the seven questions and then taking a simple average.7

Altruism. It will measure the intention to donate to a migrant organization rather than an actual donation. The outcome of interest is a binary variable equal to one if the respondent answers they would donate to a migrant organization and zero otherwise.

Randomize at the individual level to three different treatment arms

The participants will first answer a baseline survey and then they will be randomly assigned to a treatment arm. The survey software will randomize the individuals into the treatment arms.

Individuals

No

Our total sample size will be n = 21,000, where the seven larger countries will each have a sample size of n = 2,400 and the two smaller countries will each have a sample size of n = 2,100

Our total sample size will be n = 21,000, where the seven larger countries will each have a sample size of n = 2,400 and the two smaller countries will each have a sample size of n = 2,100

In each of the countries, a third of participants will be assigned to the informational video treatment, a third to the emotional video treatment, and a third to the placebo video. The countries with a sample size of n = 2400 will have n = 800 individuals assigned to the placebo, n = 800 individuals assigned to the informational video and n = 800 individuals assigned to the emotional video. The countries with a sample size of n = 2100 will have n = 700 individuals assigned to the placebo, n = 700 individuals assigned to the informational video and n = 700 individuals assigned to the emotional video. In total, we will have n = 7000 individuals assigned to the placebo, n = 7000 individuals assigned to the informational video and n = 7000 individuals assigned to the emotional video.

We assume a power equal to 0.8, a level of statistical test equal to 0.05 and that the variables that we collect at baseline (i.e. before the treatment) explain 0.3 of the variance in the final result when we pool all countries (and when we pool the seven large countries) and 0.2 when we use individual countries (and also when we pool the small Caribbean countries). We also show MDEs for the whole sample (n = 21,000), for the large country group (16,800), for the small country group (4,200), and for individual countries (n = 2,400 and n = 2,100). Under these assumptions, using the formula described above, the MDE for the treatment effect of each treatment against the control (where p = 0.5) is .032σ for the whole sample, 0.036σ for the large countries sample, 0.077σ for the small countries sample, .10σ for each individual large country, and 0.11σ for each individual small country.