Beyond Observational Relationships: Evidence from a Ten-Country Experiment that Policy Disputes Cause Affective Polarization
Beyond Observational Relationships: Evidence from a Ten-Country Experiment that Policy Disputes Cause Affective Polarization (DOI: https://doi.org/10.1017/S0007123425000158) was written by Noam Gidron, James Adams, Will Horne and Thomas Tichelbaecker in 2025. It was published in the British Journal of Political Science (vol. 55).
The authors estimate the causal effect of ideological disputes on affective polarization. The experiment is a 2 wave survey experiment, wherein the first survey prompts for an open ended response. The treatments prompt respondents for a way that parties compete on economic policy ("such as taxes, economic inequality and the welfare state"), or a way that parties compete on social policy ("such as multiculturalism, immigration and national identity"). The control cohort receives an apolitical prompt. The key metric, captured in the second survey, asks: "How much of the time do you think you can trust supporters of opposing political parties to do what is right for your country?"
This survey was administered by Latana in 10 countries:
About 1000 responses were collected from each country, 11001 in total. The sample was balanced by age, gender, and urbanicity. Responses were translated using DeepL. Content coding was automated by creating codebook dictionaries, designed through review of about 15% of responses per condition.
The cohorts were balanced on all covariates except for political interest and affective out-party evaluation, so only those two are controlled for in regression analysis.
The authors model the treatment effect on the outcome measure of affective polarization. They use fixed effects for country.
They find that priming with an economic or social policy dispute significantly lowers trust of partisan opponents. This is robust to jackknife resampling and subsetting to self-identified partisans.
A large portion of the treated responses mention immigration: 37% of the social policy treatment responses were about immigration; 12% for the economic treatment. As a result, there is interest in modeling this intersction specifically.
Because mentioning immigration is not random, the authors model propensity to mention immigration to identify predictors and then use entropy balancing on these predictors and the covariates that were already unbalanced. Respondents mentioning immigration were more likely to be older, more interested in politics, and more likely to identify with conservative or radical parties.
They model the effect of mentioning immigration in the treatment; the estimate has greater magnitude for both treatment cohorts. Note however that the effect is only statistically signficant among the social treatment cohort.
From an 11-point scale of partisan self-identification, the authors classify respondents in 0-4 as liberal and 6-10 as conservative. Nearly all respondents identifying with conservative parties, and all respondents identifying with radical parties, are thus classified as conservative.
The effect of mentioning immigration in the treatment is now modeled separately for liberals and conservatives. The authors find that the effect disappears for liberals, but is significant for conservatives, suggesting that the estimated effect is driven by conservatives. Once again, this is only statistically significant among the social treatment cohort.