122 4 Methods Between March and August 2020, I fielded online survey experiments with 3,056 citizens of France, Germany, and the Netherlands, administered by survey agency Kantar Public. This survey agency makes use of large-scale panels, representative of the population in each country. The panelists receive e-mail invitations to be part of a survey every few weeks. Response rates fluctuate around 54 percent. Kantar Public has many checks in place to ensure respondents are representative of the population. After completing a survey, respondents receive points with which they can purchase online gift cards. The points respondents receive for a fifteen minute survey are worth approximately two euros. See appendix for the research report as delivered by Kantar Public. The survey consisted of attitude questions and candidate/conjoint experiments. Researchers have been using candidate experiments to mimic low-information voting settings for decades (e.g. McDermott, 1998), with renewed interest since the introduction of conjoint experiments (Hainmueller et al., 2014). But despite countless experiments (e.g. Aguilar et al., 2015; Kevins, 2019; Lemi and Brown, 2019; Martin and Blinder, 2020; Teele et al., 2018), we still do not know whether certain voters prefer politicians with specific demographic characteristics out of simple preference for that group (Carey and Lizotte, 2019) or because of the policies they expect such politicians to advance (Arnesen et al., 2019). To answer this open question (Webster and Pierce, 2019: 636) and to unearth the dynamics of stereotyping and projection, I asked respondents what policy on same-sex adoption they expect from politicians. I presented respondents 6,112 short profiles of fictitious politicians, varying the politician’s religion, migration background, and gender (see the appendix). I matched a common name to their migration background and signaled gender through first name and the pronouns she/he. An example: “Sebnem Yılmaz has a Turkish background and she practices Islam.” After each short profile, I asked respondents what concrete policy measure they expected from this politician. They could choose between A. Homosexual couples are not allowed to adopt children. B. Homosexual couples are allowed to adopt children. C. I don’t know. I dropped respondents who answered “I don’t know” from the analysis in this article (outcomes with the “don’t know” category are found in the appendix). I recoded the answers by coding those who answered A. as 0 and B. as 1. I measured attitudes towards homosexuality by asking respondents whether they agree/disagree with the statement “homosexual couples are allowed to adopt children”22 on a scale of 0 to 10. I ask the respondents what policy they expect politicians to stand for and, for the purpose of consistency, I formulated this statement 22 VAA (2017) Voting Advice Application, French Elections 2017. Available at: http://vaa-research.net/.
RkJQdWJsaXNoZXIy MjY0ODMw