Fear and Loathing in the American Electorate: Part 6

Partisan animosity in the 2020 election

Steve Genco


Photo credit: 1931 still from the movie Frankenstein

This is Part 6 of an eight-part exploration of the 2020 American National Election Study, focusing on the motivations and sources of information driving the American electorate, especially the Republican/conservative voters who cast their votes for Donald Trump in 2020. Here are quick links to the other seven parts.

  • Part 1: The most important question facing American democracy today
  • Part 2: Demographic predictors of 2020 vote
  • Part 3: Media consumption — Where do voters get their information and what difference does it make?
  • Part 4: Seven “deplorable” beliefs that predict 2020 Presidential vote
  • Part 5: The role of conspiracies and misinformation in the 2020 election
  • Part 7: Combining sources of behavior — Path models of the 2020 Presidential vote
  • Part 8: Republicans have crossed the Rubicon — Conclusions and implications

A major topic among academics in recent years has been the rise of what political scientist Alan Abramowitz calls negative partisanship.¹ In a nutshell, this is the increasing tendency since the turn of the century for partisan loyalty to be more a function of hating the opposing party than loving your own. As Abramowitz puts it:

The rise of negative within the American electorate implies that fear and loathing of the opposing party and its candidates, rather than affection for one’s own party and its candidates, is the most important factor in maintaining partisan loyalty.²

Feeling thermometers have been used in the ANES surveys since 1968 to measure the extent to which people feel “warmth” or “coldness” toward various targets: candidates, other political figures, political parties, other institutions. For each target, respondents choose a thermometer reading between 0 (maximum coldness) and 100 (maximum warmth), with a reading of 50 representing neutrality, no feelings either way toward the target.

The extent of partisan animosity operating in 2020 can be seen by comparing average thermometer ratings of candidates and parties by party identification (here, following Abramowitz, Dem and Repub categories include people who self-identify as independent but say they “lean” toward one party or the other). The partisan divide is even greater than during the extremely polarized 2016 Presidential election, mainly because Democrats felt warmer toward both their party and their Presidential candidate in 2020 than they had in 2016, while their average disdain for the opposing party and its now incumbent candidate increased as well.

Updated: 4/13/22

Cbias and pbias: two measures of partisan animosity

One way to capture the strength of partisan animosity in the American electorate is to create a difference measure, that is, a measure of the size of the gap between two thermometer readings. To this end, I created two variables, cbias and pbias. The variable cbias quantifies a person’s relative bias toward the Presidential candidates by subtracting their feeling score for Biden from their feeling score for Trump. This measure can range from -100 (maximum bias toward Biden) to 100 (maximum bias toward Trump). Similarly, pbias is the difference between a person’s feeling toward the Republican Party minus their feeling toward the Democratic Party. On both measures, depth of partisan polarization is unmistakable.

A person cannot achieve a cbias score near either maximum by just loving their preferred candidate. They must also despise the other candidate. In this way cbias captures the deep animosity fueling partisanship and voting in the American electorate in 2020.

Comparing histograms and means of the distribution of Biden and Trump voters’ scores on the cbias variable, it is clear that candidate bias closely tracks actual voting behavior and provides a useful measure of the distance between Biden and Trump voters in terms of relative feelings toward the two candidates. The distribution of pbias scores is similar if less extreme. Voters displayed slightly less animosity in their feelings toward the two Parties than toward the two candidates.

Like Presidential vote choice, candidate bias reflects the uneven distribution of deplorable beliefs in the American public. Dividing respondents into five equal-sized bins based on their cbias scores, we find that an increasing combination of bias toward Trump and animosity toward Biden is accompanied by increasing acceptance of all the deplorable beliefs we have reviewed. Looking at average scores on the racist, anti_outgroup, hierarchist, and anti_science scales across the five cbias bins, we see a consistent pattern: resistance to these deplorable beliefs among the 40% of the sample most biased toward Biden, then an uptick in deplorable beliefs in the middle 20%, and finally a big spike in deplorable beliefs among the 40% of the sample most biased toward Trump.

Partisan animosity by party affiliation

In the methodology of the ANES surveys, partisanship is measured in several ways. The cbias and pbias variables are indirect measures that quantify partisanship by comparing thermometer readings. Two more direct measures deployed in ANES studies for decades are a political party self-placement scale and a liberal-conservative political orientation scale. I call these measures partyscale and libcon. Each offers respondents a spectrum of choices, constructed from a series of questions and follow-up prompts, that results in placement on a 7-point scale. The partyscale results for 2020 look like this.

How should we classify independents who lean toward one party or the other? Should they be grouped together with the “pure” independents or should they be classified with members of the party toward which they lean? The easiest way to answer this question is to consider how they behave; in particular, how they vote. Do they vote more like independents or more like party members? Looking at our dataset, the answer is clear.

When it comes to how they vote, people who call themselves “independent” but “lean” toward one Party of the other, act very much like self-declared Party members. Leaners are essentially Party loyalists who don’t want to admit it.

Grouping independent leaners in with self-declared party identifiers produces these size estimates for membership in the three partisan camps in America in 2020.

In this classification, only about 12% of the voting public appears to be “truly” independent. Among the rest, Democrats enjoy a slight majority over Republicans, 46% vs 42%. But as many commentators have pointed out, this advantage does not translate into equivalent political power or electoral victories because Democrats and Republicans are not dispersed evenly between densely-populated and sparsely-populated states and Congressional districts.³

Another interesting feature of party identification in 2020 is the predominance of people classifying themselves as strong party identifiers. Self-designated “strong” partisans make up 50% of Democrats (23.1%/46%) and 57% of Republicans(21%/41.9%).

How does party affiliation relate to candidate animosity as measured by cbias? As shown below, strong identifiers in both parties display the most animosity toward the other party’s candidate and the party itself. But the results in the in-between groups show less animosity and less consistency. In both parties, “weak” partisans express somewhat less animosity toward the other party and its candidate than “independent” partisans, with the level of candidate animosity anywhere from 25% lower for weak Democrats to 60% lower for weak Republicans. Party bias results are similar, with a twist: less-than-strong Democrats express less animosity toward the Republican Party than toward its Presidential candidate, while less-than-strong Republicans express more animosity toward the Democratic Party than toward its Presidential candidate.

Like cbias, partyscale tends to strongly correlate with deplorable beliefs. For each of the four belief clusters that make up our racist, anti_outgroup, hierarchist and anti_science scales, we find a clear differentiation between members of the two parties, with Democrats at all levels expressing similar and significantly less deplorable beliefs than Republicans.

Interestingly, although “pure” independents express less animosity to candidates and parties overall, giving a slight edge to both the Democratic candidate and the Democratic Party, their attitudes on all four of these attitude scales are closer to those of Republicans than Democrats.

Partisan animosity by liberal-conservative self-placement

The second explicit measure of partisanship in the ANES surveys is a liberal-conservative political orientation scale. Like the party affiliation measure, this variable is constructed from a series of questions and follow-up prompts that produce a measure of the distribution of ideological orientations across the American voting public.

What should we do with the 13% of respondents who “haven’t thought much about this”? Should they be treated as missing data or should they be rolled into the Moderate group? Again, the best way to answer this question is to look at their voting behavior. When we do, we see that these individuals (represented in the “DK” column below) voted very similarly to members of the Moderate group. We also see that liberals of all stripes (from “slight” to “extreme”) voted nearly unanimously for Biden in 2020, while “slightly conservative” voters deviated a bit from the nearly universal Trump vote among self-identified conservatives and extreme conservatives.

Adjusting the libcon scale to combine the “haven’t thought much about this” group with the moderates produces a high-level picture of the American voting public with a large non-ideological center (34%) surrounded by a slightly larger conservative-leaning faction (36%) and a slightly smaller liberal-leaning faction (29%).

When we look at deplorable beliefs across this liberal-conservative spectrum, we see that the more conservative an American’s political self-identification, the greater their embrace of deplorable beliefs and values. As we saw with partyscale, those who place themselves in the middle the liberal-conservative continuum express a moderate bias toward the Democratic candidate and party (above), but their attitudes align more closely with those of conservatives than liberals (below).

Partisan: A summary measure of partisanship in the American public

Extremely polarized partisanship in the American public means that Party loyalty and liberal-conservative orientation have both become signifiers of the same thing: the political “team” one supports. We find fewer and fewer people who call themselves conservative Democrats or liberal Republicans. Looking at the Party affiliations of self-described liberals, moderates, and conservatives, we see that 93% of people who identify as “liberal” also identify as Democrats or Democratic leaners, while 91% of people who identify as “conservative” also identify as Republicans or Republican leaners. Only in the “moderate/don’t know” group, which does not have a clear ideological identity, do we find a mix Democrats and Republicans.

Given this tendency toward more extreme sorting in the American electorate, it makes sense to think of partisanship as a unidimensional factor that determines how people think of themselves politically. Accordingly, I’ve created another scale variable, called partisan, that combines Party ID (partyscale_f) and liberal-conservative ID (libcon_f). Although scale variables tend to be more reliable when derived from more than two source variables, partisan yields and acceptable consistency and unidimensionality (Cronbach’s alpha = .79).

Partisan is — as expected — a significant predictor of Presidential vote.

It is also an excellent differentiator of Biden vs. Trump voters, providing a stark illustration of the extreme distance separating the two partisan camps in America (mean partisan scores are .28 for Biden voters and .78 for Trump voters).

One observation that stands out from this review of partisanship and partisan animosity is that all these variables are highly correlated with each other and with voting behavior. Indeed, if we produce a weighted correlation matrix across cbias, our new partisan scale, and numeric versions of votefor (excluding third party voters), partyscale, and libcon, we see that all are capturing similar aspects of the partisan divide in the American public. Cbias has the highest bivariate correlation with Presidential vote (Pearson’s r = .90), followed by partyscale_f (r = .80) and libcon_f (r = .68).⁴ Our partisan variable blends party and ideological identities and correlates more strongly with cbias than either party or ideological affiliation alone (r = .85).

Partisan animosity: Not two sides of the same coin

Partisanship in America as captured in the 2020 ANES election surveys is not just about a radical sorting of the American public into two opposing camps with opposing beliefs on major issues — issues that just happen to have vast importance to the future of American democracy and the constitutional rule of law. It is also about Americans seeing other Americans not simply as people with different views, but as existential enemies.

The American electorate is polarized, but the polarization is not symmetrical. One side holds beliefs that are toxic to American democracy and values — racism, demonization of immigrants, opposition to equality values, denigration of science, and more. The other side finds these beliefs, and the elites who promote them with lies and conspiracies, appalling.

Those siding with the Republican Party, by and large, oppose the aspirational values that have driven the American experiment since its inception. Those siding with the Democratic Party want to protect, defend, and extend those values. The result is a deeply divided and antagonistic political system, but the motivations driving the two sides are vastly different.

The obvious next set of questions is about whether and how all these variables interact with each other and together contribute to a causal model that accurately depicts the relative impacts of major and minor drivers of American voting behavior in 2020. To answer those questions, we can deploy a multivariate statistical technique called path analysis, based on a general methodology called structural equation modeling (SEM).

Continue to Part 7: Combining sources of behavior — Path models of the 2020 Presidential vote


  1. Abramowitz, Alan I., and Steven W. Webster. “Negative partisanship: Why Americans dislike parties but behave like rabid partisans.” Political Psychology 39 (2018): 119–135.
  2. Ibid., p. 133.
  3. For an overview, see The Economist. “America’s electoral system gives the Republicans advantages over Democrats,” July 14, 2018, online here.
  4. Sample size is 5,729 for this correlation matrix.



Steve Genco

Steve is author of Intuitive Marketing (2019) & Neuromarketing for Dummies (2013). He holds a PhD in Political Science from Stanford University.