No Negativity Bias in Experimental Tests of Retrospective Voting

• In Event: Electoral Accountability

Abstract

A sizable literature in psychology finds that information that is negatively appraised attracts more attention and weighs more heavily in evaluative reasoning than positively appraised information. This research aligns with the longstanding conventional wisdom among scholars of retrospective voting, particularly economic-voting scholars, that incumbents are more strongly punished for bad performance than they are rewarded for good performance. Observational research on this retrospective-voting expectation is, however, limited and the conclusions are mixed. At the same time, there are almost no experimental tests of negativity bias in retrospective voting.

In this paper, we investigate asymmetry in retrospective voting using an experimental approach pioneered by Huber, Hill and Lenz (2012) and developed extensively in our own prior research (Hart and Matthews, 2022, 2023). The approach uses an abstract experimental framework that requires participants to monitor the performance of a hypothetical “incumbent” over time and then to “vote” on whether to reappoint or replace the incumbent. Notably, prior work based on this framework has replicated findings familiar in previous retrospective voting research, including incumbency and recency biases and benchmarking (or reference-point) effects.

The present paper combines reanalysis of existing data – from Huber and colleagues and our own prior work – with three new experimental studies to provide an extensive experimental study of negativity bias in retrospective voting. We report multiple tests pertaining to four expectations that are fundamental to standard accounts of negativity bias in retrospective voting: that voters’ incumbent-retention decisions respond to the valence of performance; that retention decisions are more responsive to variation in negatively than positively valenced performance; that the disutility of poor performance outweighs that of good performance; and that voters are more attentive to negative than positive performance outcomes.

The results are almost uniformly inconsistent with the existence of negativity bias in retrospective voting. While we find some evidence that valence matters, there is no significant evidence of negativity bias in participants’ retention decisions or in the accuracy of their recollections of incumbent performance. We conclude by identifying implications of the findings for our understanding of the operation of retrospective voting in real-world elections, suggesting that, to the extent observational studies identify negativity bias, it is more likely to reflect asymmetries in political communication than in political psychology.

References:

Hart, A., & Matthews, J. S. (2022). Unmasking accountability: Judging performance in an interdependent world. The Journal of Politics, 84(3), 1607-1622.

Hart, A., & Matthews, J. S. (2023). Quality Control: Experiments on the Microfoundations of Retrospective Voting. Cambridge, UK: Cambridge University Press.

Huber, G. A., Hill, S. J., & Lenz, G. S. (2012). Sources of bias in retrospective decision making: Experimental evidence on voters’ limitations in controlling incumbents. American Political Science Review, 106(4), 720-741.

Authors

• Austin Hart, American University

• Scott Matthews, Memorial University