Affective Polarization During the Coronavirus Pandemic

By Matthew Gentzkow

Posted on January 22, 2021


Share

Note: this article is based on “Affective Polarization did not Increase during the Coronavirus Pandemic” by Levi Boxell, Jacob Conway, James N. Druckman, and Matthew Gentzkow (Stanford University working paper, December 2020).

 

The scientific, economic, and social challenges of responding to the coronavirus pandemic have been compounded in the US by political divisions. Studies from the early days of the pandemic show that partisan divisions were among the most significant drivers of health behaviors, concern about the virus, support for specific policies, attributions of responsibility, and even beliefs about basic facts. This echoes similar divisions among politicians and the media. It seems possible that coronavirus has been yet another force pushing toward greater polarization.

 

This is not the only possible narrative, however. A different possibility is that a major crisis like coronavirus might actually have pulled Americans together (at least temporarily)—due to a “rally around the flag” effect such as is often seen at times of war or natural disasters, or perhaps simply by giving Americans something to focus on other than politics.

 

In a recent paper, we turn to new data sources to see how polarization in fact evolved during the pandemic. We focus on affective polarization—the extent to which partisans feel more negatively toward the opposing political party than toward their own.1

 

Affective polarization in the US has been steadily increasing in recent decades, and this has generated widespread concern about its consequences including undermining democratic institutions and representation, legislative gridlock, and partisan violence. Figure 1 shows this long-term trend, using data from American National Election Study surveys going back to the late 1970s. The plot shows polarization in partisan affect, defined as the difference between respondents’ feelings toward their own party vs. the opposite party on a 1-100 scale where higher numbers indicate “warmer” feelings. The magnitude of this gap has increased from roughly 27 at the beginning of the series to roughly 45 in the most recent years. This trend has been a source of alarm to both policymakers and academics, and one recent paper calls studies of its causes and consequences “one of the most influential literatures in contemporary American politics scholarship.”2

 
Figure 1: Long-run trends in Affective Polarization

Figure 1: Long-run trends in Affective Polarization
Source: Levi Boxell, Matthew Gentzkow, and Jesse M. Shapiro, 2020,
“Cross-country trends in affective polarization,” Stanford University working paper.

 

We find no evidence that affective polarization rose during the crisis. Four of six data sources suggest that affective polarization in fact declined with the onset of the coronavirus, with the other two suggesting neither a decline nor an increase. A survey experiment adds further evidence, showing that priming respondents to think about the pandemic significantly reduces affective polarization. Taken together, our results suggest a cautiously optimistic conclusion that the coronavirus may actually have brought partisans together in the face of a common threat.

 

Details of Analysis

 

Measures of affective polarization vary in the type of attitudes elicited (e.g., feelings, trust, or behaviors) and the subject of those attitudes (e.g., voters, parties, or candidates). To avoid relying on a single measure or data source, we report trends across six different measures and data source combinations.

 

Figure 2 reports trends in affective polarization towards partisans or parties. It shows the extent to which members of each party feel more negative toward the opposite party than their own.

 

Panel A uses a large-scale survey dataset from Nationscape that covers more than 300,000 interviews between July 2019 and July 2020. This is our preferred measure due to the size, frequency, and consistent methodology of the Nationscape data. Panel A shows that, prior to the rise of the coronavirus in the US, affective polarization was relatively flat and this flat trend extends back to July 2019. However, after the first publicized coronavirus-related death in the US (on February 29th, 2020), affective polarization exhibits a significant decline. It then ticks back upward following the death of George Floyd.

 
Figure 2a: Recent trends in Affective Polarization

Figure 2: Recent trends in Affective Polarization
Panel A: Partisans
Source: Boxell, Conway, Druckman, and Gentzkow 2020.

 

Panel B shows a smaller decrease during the onset of the pandemic when measuring affective polarization using questions in the Nationscape data about feelings towards partisan members of Congress.

 
Figure 2b: Recent trends in Affective Polarization

Figure 2: Recent trends in Affective Polarization
Panel B: Members of Congress
Source: Boxell, Conway, Druckman, and Gentzkow 2020.

 

Panel C reports our measure of affective polarization towards political parties using a separate panel of respondents from a paper by Druckman et al. (2020). The estimates indicate little change in affective polarization between July 2019 and April 2020.

 
Figure 2c: Recent trends in Affective Polarization

Figure 2: Recent trends in Affective Polarization
Panel C: Political Parties
Source: Boxell, Conway, Druckman, and Gentzkow 2020.

 

Figure 3 turns from feelings toward parties to feelings toward Donald Trump, focusing on the difference in the favorability Democrats report and the favorability Republicans report.

 

Panel A reports trends in this difference in feelings using Nationscape data. While there is a slight upward trend prior to the onset of the pandemic, there is a significant decline upon the onset of the pandemic.

 
Figure 3a: Recent trends in Partisan Feelings toward Trump

Figure 3: Recent trends in Partisan Feelings toward Trump
Panel A: Feelings
Source: Boxell, Conway, Druckman, and Gentzkow 2020.

 

Panel B reports a similar decline in partisan differences in presidential approval ratings in the Nationscape data.

 
Figure 3b: Recent trends in Partisan Feelings toward Trump

Figure 3: Recent trends in Partisan Feelings toward Trump
Panel B: Approval
Source: Boxell, Conway, Druckman, and Gentzkow 2020.

 

Panel C reports estimates of affect towards Donald Trump as measured in a separate data source: the American National Election Study, using survey rounds conducted in December 2018, December 2019, and April 2020. While the difference in affect towards Trump between Republicans and Democrats is relatively constant across the 2018 and 2019 waves, the gap between the parties is significantly lower in April 2020.

 
Figure 3c: Recent trends in Partisan Feelings toward Trump

Figure 3: Recent trends in Partisan Feelings toward Trump
Panel C: Feelings
Source: Boxell, Conway, Druckman, and Gentzkow 2020.

 

Priming Strategy Analysis

 

To supplement these results, we also conduct an experimental analysis to see if priming people to think about coronavirus leads people to express more or less polarized attitudes. This priming strategy follows previous work by our coauthor Jamie Duckman, among others.

 

The experiment has three conditions. First, in the coronavirus treatment, respondents are asked to read two news article excerpts that cover the initial phases of the coronavirus pandemic and to reflect on their own experiences and faith in the United States' ability to address the pandemic at its onset. Second, in the placebo treatment, subjects are asked to perform an analogous exercise where they reflect on news articles unrelated to coronavirus—specifically, articles about Prince Harry and Meghan Markle’s separation from the UK royal family (which occurred just before the onset of the pandemic). Finally, in the control group subjects do not read news articles. We then ask subjects standard questions about their feelings toward political parties and groups.

 

The experimental results show that, relative to the control group, the coronavirus group displays significantly lower affective polarization. This is not true for the placebo group. The effect mainly reflects less negative sentiment toward the opposing party.

 

Looking at this evidence together, we conclude that the coronavirus pandemic is unlikely to have increased affective polarization and may well have decreased it.

 

References:
This article is based on: Levi Boxell, Jacob Conway, James N. Druckman, and Matthew Gentzkow, 2020, “Affective Polarization did not Increase during the Coronavirus Pandemic,” Stanford University working paper.

 

1 - See Shanto Iyengar, Gaurav Sood, and Yphtach Lelkes, 2012, “Affect, not ideology: A social identity perspective on polarization,” Public Opinion Quarterly, 76(3): 405–431.

 

2 - David E. Broockman, Joshua L. Kalla, and Sean J. Westwood, 2020, “Does affective polarization undermine democratic norms or accountability? Maybe not,” UC Berkeley working paper.


Share