How much help shall we give to refugees? An online experiment on attitudes towards temporary protection and helping behaviours in the EU -- Second-wave survey

As in the first wave, we can divide the second-wave survey into four sequential sections:

1. demographic questions for screening (these are used to reach certain fixed quotas for country representativeness),
2. (conjoint) choice experiment,
3. perspective-taking experiment, and
4. collection of socio-demographic variables

These sections of the second wave of the study will be identical to the first wave, except for the following changes:

In section (1), on top of the randomisation into blocks for the choice experiment (see previous pre-registration plan), we will also randomise the choice task into three conditions where respondents are asked to give their preference for different types of temporary protection, as for first wave. However, compared to the first wave, we replaced one treatment arm, ending up with the following:

* Free movement: In this condition, we test whether an emphasis on giving refugees the right to move freely within the EU has any effect on the individuals’ support for temporary protection. We observed in the first wave that people are comparatively less supportive of free-movement rights, thus we expect this emphasis may stimulate less inclusionary attitudes. We expect the effect to be larger for vulnerable individuals who may have a greater fear of direct migrant competition (see, e.g., discussion in Ford and Mellon 2020)

* Housing: In this condition, we test whether putting more emphasis on giving refugees access to housing has any effect on the individuals’ support for temporary protection.

* Generic protection (“Control”): This is the control condition with no emphasis on the implications of temporary protection for housing or the labor market.

In section (2), we updated the text of some questions to reflect that the Temporary Protection Directive has been extended for another year between the data collection waves.

No changes have been made to the design of section (3).

In section (4), we added a new randomised intervention (“belief correction”) to investigate the effect of information on people’s perceptions of refugee characteristics. First, we ask participants to guess the fraction of adult women among the refugees in their country. Subsequently, we ask them to revise their initial guess with one of the following four messages or belief-correction treatments that will be displayed randomly across the participations:

* [Official correct:] The statistics available for six EU countries, including [respondent’s country], indicate that women represent about 75% of adult Ukrainian refugees there.

* [Survey correct:] One group of people we surveyed in six EU countries, including [respondent’s country], believe that women represent about 75% of adult Ukrainian refugees there.

* [Survey low:] One group of people we surveyed in six EU countries, including [respondent’s country], believe that women represent about 45% of adult Ukrainian refugees there.

* [Control:] [Empty row, no information]

We have the following hypotheses concerning the belief-correction intervention in section (4):

1. We expect minor belief updates in the control treatment (probability of updating non-greater than 5%).

2. We expect the probability of updating the initial belief and the expected update to be higher, the larger the difference between the initial belief and the provided information.

3. We expect the effect of the intervention on the probability of revising the belief of the "Official correct" condition to be larger than the effect of the "Survey correct" condition due to the source of information having the approval of the statistical office and, thus, being more reliable.

4. Conditional on initial beliefs, we expect the effect of the belief-correction intervention on the probability of updating to vary depending on the level of education, self-reporting higher social trust, attitudes towards Ukrainian refugees, and views about migration.

5. We also expect an effect of the belief-correction intervention on the respondent's perceived link between refugees and crime. Compared to the control condition, we expect a more negative assessment of the effect of refugees on the crime situation in the "Survey low" condition and a more positive assessment in the "Survey correct" and "Official correct" conditions. The higher the initial guess, the higher will be the effect of displaced Ukrainians on the crime situation.

6. The effect on crime perceptions could vary by gender, age, education, and other key demographics.

2023-03-13

2023-03-19

As in the first wave, we use the following questions as outcome measures:

DCE:

- To measure individual preferences towards refugees and support for temporary protection, we ask study participants to choose between one or none hypothetical refugee situations in a choice experiment ("Please, compare groups A and B described in the table and indicate which of the two, if any, you would prefer to offer temporary protection in the EU."). We will repeat the question six times under different scenarios (see later).

Perspective-taking/getting:

- Attitudinal measure: To measure support for the protection of Ukrainian refugees in the current situation, we ask whether respondents agree or not with granting temporary work permits to Ukrainians ("Do you agree or disagree with granting displaced people from Ukraine temporary work permits?")

- Behavioural measure: We also ask whether respondents are willing to send a short text supporting Ukrainian refugees (encouraging volunteering, making donations, or raising awareness) to be shared via Twitter ("Would you agree to write a short text to persuade people in [Country] to donate? We will share your message on our Twitter account (You can also edit our template)").

- To measure support for policies already in place, we ask whether participants support the European Council's decision to grant temporary protection to the Ukrainians ("do you support or do you oppose this decision?") and whether the EU should extend it ("Assuming the situation in Ukraine does not change much, do you think the EU should extend the temporary protection for Ukrainian people beyond March 23, 2024?")

- We also measure individual support for Ukraine to join the EU ("To what extent do you agree or disagree with the statement: 'Ukraine is part of the EU family'"?).

- We ask about what rights (housing, labour market, medical, ...) they would like to include in temporary protection.

Belief-correction intervention:

- We hypothesise that providing participants with official statistics about refugees (i.e., the fraction of adult women) will "correct" possible wrong perceptions about refugees. We measure the effect of the intervention by asking respondents whether they want to revise their guesses and by how much.

- We measure the effect of belief correction on attitudes with a question about whether they perceive crime as being made better or worse by the presence of refugees from Ukraine (11-point scale).

As in the first wave, we are interested in evaluating how individuals' attitudes toward refugee protection vary with demographics and other variables and how individuals can be persuaded to adopt more inclusionary attitudes towards refugees by taking or getting the refugees' perspectives; pre-analysis plan RCT ID: AEARCTR-0009682.

Here, we report the following additional research questions/testable hypotheses for the second wave:

1. Measuring changes in attitudes over time
2. Measuring the effectiveness of perspective-getting/taking interventions over time
3. Measuring the effect of belief-correction interventions

## Measuring changes in attitudes over time

How stable are the attitudes toward views on migrants and refugees over time? Some studies have suggested that public attitudes toward immigration are stable over time [@kustov2021stability]. However, analysis of polls indicates a gradual change in Europe, with countries like Germany becoming more inclusive while others adopting more excluding attitudes [@heath2016europeans]. Moreover, in response to Russia's invasion, Europeans' support of Ukrainians who sought shelter in the EU has been overwhelmingly positive compared to pre-war attitudes [@de2022refugee]. But it is unclear whether increased contact with refugees through an extended presence in the country and daily life may result in more positive attitudes [as suggested in @de2021contact] or an adverse reaction [as in @enos2014causal].

Understanding how and whether these attitudes will change requires further research.

<!-- For example, one might expect that an initial emotional reaction to an unexpected war in Europe may decrease in intensity over time. If so, people may revert to pre-war less inclusionary attitudes towards Ukrainians (and immigrants in general). -->

By comparing results across the two waves of this study, we want to assess to what extent public attitudes toward refugees can change over time.

- We expect personal preferences towards refugees in general and willingness to pay for their protection to be stable across cross-sectional surveys. Specifically, we estimate individual tendencies toward protection for different refugee groups using a discrete choice model (as described in wave one) and anticipate only minor differences in the estimated parameters between the two cross-sections.

- At the regional level (NUTS), we expect the average difference in support of Ukrainian refugees between the first and second waves to be driven mainly by environmental changes, such as an increase in the presence of refugees.

## Effectiveness of perspective-taking/getting interventions over time

Will the effectiveness of public interventions to promote inclusionary attitudes and help the refugees diminish over time? Here, we focus on the case of perspective-taking and perspective-getting interventions (as described in the first wave). Previous work has shown that the effect of perspective-taking diminishes over time for the same individual [@adida2018perspective]. We investigate whether the effectiveness of the same intervention differs when implemented at different stages of a refugee crisis across people with comparable characteristics.

As in wave one, we will estimate the average treatment effect on attitudes and behaviours (i.e., supporting Ukrainians on social media) of stimulating participants to take the perspective of a refugee or read narratives that give a refugee's perspective. We expect the treatment to be less effective in the second wave for several reasons. First, participants had more opportunities to interact with the refugees by the time of the second wave. This may have triggered a sense of empathy making our intervention redundant and less effective. Second, people may have already helped Ukrainian refugees and may be comparatively less inclined to make further contributions, lowering the effect size of our intervention.

<!-- - For example, frequency of interactions with them, mostly because intergroup contact negatively affects inclusionary attitudes [see, e.g., @enos2014causal]. -->

<!-- https://link.springer.com/article/10.1007/s10640-015-9913-1 -->

## Belief correction intervention

How to correct wrong beliefs about refugees that could negatively affect people's attitudes toward refugee protection? In the first wave, we observed a bias in how study participants perceived key characteristics of the Ukrainian refugees in their country. They tended to underestimate the fraction of adult women among the incoming refugees, among other characteristics. And we hypothesise that providing study participants with more information (specifically, statistics about the fraction of adult women among the Ukrainian refugees in their country) will "correct" possible wrong perceptions about refugee characteristics, which is important to determine personal attitudes toward refugee protection.

Specifically, we investigate to what extent participants will revise their beliefs about refugees in the direction of the information provided: upwards when the intervention gives a higher figure than the initial belief and downwards if it is lower.

We also hypothesise:

- The effect of the belief correction on the probability of revising the initial belief of adult women among the refugees will be stronger the farther the initial belief is from the provided information.

- An effect of the source of information, with official statistics being more effective in prompting study participants to change their beliefs about the fraction of adult women than other sources.

- A direct effect of the belief-correction intervention on other aspects of how people view the refugees. Specifically, since statistics have reported consistently that men commit more criminal acts than women, we expect that correcting beliefs about the fraction of women will lead to more positive views about the impact of the Ukrainian refugees on the crime situation in the host country (which we also ask in the survey).

Computer randomisation.

Same as in the first wave (individual respondents, clustered by country)

Yes

We aim at 14,000 respondents from six countries, with demographic quotas; see first wave registration.

14,000 respondents

55 NUTS regions across 6 countries