Can We Reduce Affective Polarization in the Mass Public?

Download data and study materials from OSF

Principal investigator:

Matthew Levendusky

University of Pennsylvania



Sample size: 2821

Field period: 01/14/2016-04/28/2016


There is growing evidence that Americans are affectively polarized--ordinary Democrats and Republicans increasingly dislike one another. Such affective polarization is normatively troubling because it lowers political trust and contributes to governmental gridlock and dysfunction.

This study looks at 3 different interventions that might overcome this affective discord: priming American national identity, increasing partisan ambivalence, and self-affirmation techniques. The results suggest that it is difficult to overcome partisan animus.


Hypothesis/Research Questions:

The main hypothesis is that each of the three interventions (priming American identity, self-affirmation, and partisan ambivalence) will decrease affective polarization.

Experimental Manipulations

Subjects are randomized into one of 4 experimental conditions:

1. Control (subjects read a brief apolitical news story and answer some apolitical primes)

2. American identity prime: Subjects select why people are proud to be American from a list, and then write about why they are proud to be American in a brief essay

3. Self-Affirmation: subjects select an apolitical value/trait that is important to them, and write an essay reflecting on a time when they embodied that value/trait.

4. Ambivalence: subjects list two things they dislike about their own party, and two things they like about the opposing party

Outcome Variables:

Measures of affective polarization: feeling thermometer rating of the opposing party, belief that the opposing party is a threat to the nation, and comfort being close friends with those from the opposing party.

The study also includes a downstream consequence of affective polarization: willingness to engage in heterogenous political discussion.

Summary of Results

The American identity prime increases the likelihood that subjects rate the opposing party above 50 degrees, and that subjects are willing to engage in heterogeneous political discussion. The overall treatment effects of both the ambivalence treatment and the self-affirmation treatment are null for all variables.

I did not pre-register a moderation hypothesis, but I find that among ideological moderates (where ideology is measured using the GfK panel ideology variable), both the American treatment and the ambivalence treatment decrease polarization. There is no moderating effect of ideology for the self-affirmation treatment.

Additional Information

The ambivalence treatment itself contains evidence of affective polarization: the majority of subjects (nearly 60%) are unwilling to list 2 things they like about the opposing party.


Working papers are currently in preparation; please contact the author for a copy.

Sample size: 2821
Field period: 01/14/2016-04/28/2016