Intergroup contact and opinion polarization

Increasing political polarization is a challenge for developed and developing democracies worldwide (Boxell, Gentzkow, and Shapiro 2020). Going back to De Tocqueville (1835), a long line of research suggests that the well-being of democracies depends on intersectional social relations. The idea builds on two streams of literature. First, a literature on intergroup contact suggests that contact with outgroup members, such as direct interpersonal conversations, can reduce prejudice (Allport, Clark, and Pettigrew 1954). Overall, the literature provides mixed evidence.1 For example, recent field experiments showed that inter-group contact can be effective in increasing cohesion between different religious (Scacco and Warren 2018) economic (Rao 2019), or cast (Lowe 2021) groups. However, the research also found that inter group contact might not always lead to a change in political views (Heuser and Stötzer 2022), might not translate into contexts outside of the intervention (Mousa 2020), and might not reduce prejudice (Scacco and Warren 2018). Moreover, positive effects might fade out after a short while (Santoro and Broockman 2021). Related evidence suggests that conversations between politically like-minded individuals may lead to more extreme political views (Heuser and Stötzer 2022). A second line of research suggests that effects of political conversations depend on characteristics fo processes and can be by-products of nonpolitical social processes (Minozzi et al. 2020). Building on these ideas, recent interventions showed that political communication between participants who share non-political characteristics but differ on political attitudes seems to be an effective strategy to promote consensus on divisive political topics (Balietti et al. 2021). Yet, previous work did not study whether these effects hold across different types of topical areas. In addition, previous research designs that randomly assign participants of an experiment to a treatment of interpersonal interaction (Balietti et al. 2021) can sometimes make it difficult to recover effects is settings where we expect self-selection (Acemoglu 2010). For this, we need to understand who would engage in interpersonal interaction both as a sender and receiver, i.e. who would select into treatment (Humphreys and Scacco 2020). Lastly, experiments rarely examine the short and long-run effects of these interventions. This study addresses these issues by replicating and extending the design of Balietti et al. (2021) in two ways. First by studying treatment effects in a controlled environment (respondents get randomized into contact with in-group/out-group) and a self selection environment (with self selection into contact with in-group/out-group). Second by varying the type of topic under discussion from a traditional issue (preferences on redistribution) to a more novel issue (preferences on the mandatory vaccination). The table below illustrates.