Experimental Design Details
The first part of the survey asks the respondents a series of questions about themselves. This includes both a series of standard demographic questions and a set of questions to tests the respondents' knowledge in areas relevant to the survey. The questions on the respondents' demographics include: gender, age, level of education, religious observance, and, as a measure of political ideology, which party they supported in Turkey's most recent parliamentary elections. The surveyors also record the region and size of community where the respondent lives. The survey also asks the respondents' personal exposure to refugees.
The second part of the survey presents the experimental vignette. The respondents are told about refugees in Turkey, specifically, the following (translated into Turkish): "You know people are migrating to other countries because of the civil war in Syria. Turkey has already accepted more than 3 million refugees. But people still continue to migrate to Turkey and other countries. This issue has been discussed very much in Turkey. Suppose that the Turkish government decided to not accept the new-coming refugees anymore."
I had several objectives in crafting the vignette. The first was to keep the treatment short, as the survey was conducted in face-to-face interviews, and one concern is that people do not keep focus on long statements that are read out to them. The survey firm KONDA emphasized this point, based on their extensive experience in Turkey. Second, I decided not to give them the arguments for and against accepting refugees. The reason is that these arguments are complicated and many people are familiar with them anyway. Also, giving these arguments would add so much additional language to the vignette that it would be hard for people to process, increasing the risk that the treatments would get lost in that information.
The survey experiment comprises five variations (four treatments and one control) and differ in just one respect: to what extent international law or norms promote Turkey's continuing to accept refugees. The outcome is the degree to which the respondent supports "the government’s proposed new policy to start turning away additional refugees from Syria."
The first version, which provides no information on international norms, is intended as a control.The second version, that international law requires Turkey to accept refugees, is intended to test the effect of binding international law on citizens' foreign policy preferences. This treatment will be effective in promoting willingness to accept refugees if and only if respondents view international law as a constraining force-- either because they expect that non-compliance will trigger sanctions (through international shaming, ostracizing, economic, or military), because violating it would be immoral, or because its widespread acceptance signals that complying would also be desirable for Turkey. The third version, that Turkey is one of many countries that has promised other countries to accept all arriving refugees, is intended to test the effect of people's valuing upholding international commitments. This treatment is similar to the international law treatment above, but it emphasizes the effect of voluntary, "bottom-up" commitment, as opposed to the "top-down" imposition of rules on a country and its citizens. The fourth version, that many other countries have accepted refugees and plan to continue doing so, is intended to test the effect of acculturation, or people taking some action because others are taking it. The fifth version, that Turkey is not obligated under international law to accept all arriving refugees, is -- like the second version -- intended to test the effect of the effect of binding international law on citizens' foreign policy preferences, but in the negative. This treatment will be effective in reducing willingness to accept refugees if and only respondents view international law as a constraining force, and knowing that it does not constrain shapes their view on the policy for one reasons above (sanctions, morality, or signaling).