A test of imagined contact as a means to improve cross-partisan feelings and reduce attribution of malevolence and acceptance of political violence

ABSTRACT

In this study, we test the depolarizing potential of intrapersonal communication through imagined intergroup political contact. A randomized experiment was embedded into a quota-stratified online survey distributed to 583 U.S. adults drawn from an online panel proportioned to match the U.S. population. Structural equation modeling was used to test the direct, indirect, and conditional effects of imagined contact. We found that imagined contact with a political outgroup directly reduced negative affect toward the political outgroup regardless of the primed valence of the imagined interaction. Furthermore, we found that imagined contact indirectly reduced attribution of malevolence to the political outgroup as well as the acceptance of political violence. Implications for intergroup theory and political polarization are discussed.

KEYWORDS:

• Political polarization
• intergroup contact hypothesis
• imagined interactions
• political violence

Acknowledgements

The authors would like to thank Dr Tamara Afifi and the anonymous reviewers for their indispensable feedback. Their suggestions dramatically shaped and improved this manuscript.

Notes

1 The two phases were initially split into separate experiments but, given the contiguous timeframe of data collection and substantial similarity in findings, the data was pooled. This decision has the benefit of generating a more cohesive and parsimonious set of findings and increasing statistical power.

2 Fitting the model without parceling the items in the negative affect variable resulted in substantially worse model fit, χ²(43) = 280.36, p < .001, RMSEA = .097 (.088–.107), CFI = .905, NNFI/TLI = .879, SRMR = .054. Inspection of the modification indices revealed substantial correlated residuals for respectful/contempt and friendly/hostile as well as suspicious/trusting and cold/warm. Allowing those residuals to correlate resolved misfit, χ²(40) = 99.31, p < .001, RMSEA = .050 (.039–.062), CFI = .976, NNFI/TLI = .968, SRMR = .044. A chief advantage of parceling is the ability to reduce correlated residuals without altering latent relationships (Little et al., 2013).

Additional information

Funding

This research was supported by a Faculty Research Council award from the University of Missouri Research Board (grant number URC-15-095).