Recent scholarship has challenged the long-held assumption in the social sciences that Conservatives are more biased than Liberals, yet litt
By: Bo M. Winegard, Cory Clark, Connor R. Hasty, Roy F. Baumeister
Published: Aug 23, 2023
Abstract
Recent scholarship has challenged the long-held assumption in the social sciences that Conservatives are more biased than Liberals, yet little work deliberately explores domains of liberal bias.
Here, we demonstrate that Liberals (some might call them Progressives) are particularly prone to bias about victims’ groups (e.g. women, Black people) and identify a set of beliefs that consistently predict this bias, termed Equalitarianism.
Equalitarianism, we believe, stems from an aversion to inequality and a desire to protect relatively low status groups, and includes three interrelated beliefs: (1) demographic groups do not differ biologically; (2) prejudice is ubiquitous and explains existing group disparities; (3) society can, and should, make all groups equal in society.
This leads to bias against information that portrays a perceived privileged group more favorably than a perceived victims’ group. Eight studies and twelve mini meta-analyses (n=3,274) support this theory. Liberalism was associated with perceiving certain groups as victims (Studies 1a-1b).
In Studies 2-7 and meta-analyses, Liberals evaluated the same study as less credible when the results portrayed a privileged group (men and White people) more favorably than a victims’ group (women and Black people) than vice versa. Ruling out alternative explanations of normative reasoning, significant order effects in within-subjects designs in Study 6 and Study 7 (preregistered) suggest that Liberals believe they should not evaluate identical information differently depending on which group is portrayed more favorably, yet do so.
In all studies, higher equalitarianism mediated the relationship between liberalism and lower credibility ratings when privileged groups were portrayed more favorably. Although not predicted a priori, meta-analyses also revealed Moderates to be the most balanced in their judgments. These findings do not indicate whether this bias is morally justifiable, only that it exists.
A recent meta-analysis found that both Liberals and Conservatives were roughly equally biased when evaluating information with conclusions that were more or less congenial with their preferred beliefs (Ditto et al., 2019a, 2019b). Although many scholars continue to dispute the possibility of roughly symmetrical bias among Liberals and Conservatives, even the most skeptical scholars agree that Liberals likely are biased on some issues (e.g., Baron & Jost, 2019; Van Bavel et al., 2020).
Indeed, it has been argued that bias is a natural human tendency that evolved at least partially to facilitate group cooperation and status attainment within social groups (e.g., Clark et al., 2019; Clark & Winegard, 2020; Winegard & Clark, 2020), and thus Liberals, as humans, likely are biased in at least some domains.
Despite widespread agreement that Liberals are susceptible to biases, little work has explored domains in which Liberals display biases. Here we explore one such domain: low status groups. We contend that Liberals are biased in their evaluations of information that portray low status groups unfavorably (relative to high status groups). We also find evidence that this bias is at least partially explained by a set of interrelated beliefs about low status groups that are endorsed more strongly by modern Liberals than modern Conservatives.
[...]
General Discussion
Taken together, the data from these studies strongly support the equalitarian account of liberal bias.
First, Liberals appeared committed to intrinsic group equality. They were biased such that they found vignettes that stated that two demographic groups were equal more (although not statistically significantly relative to Women Higher in Study 5) credible than vignettes that stated that one group outperformed another.
Second, they were consistently biased against results that favored a privileged group over a victims’ group. In every single study, they rated the Privileged Group Higher vignette as less credible than the Victims’ Group Higher.
And third, scores on our equalitarian measure mediated our results in every study such that higher scores on the equalitarian measure predicted more bias among Liberals. Consistent with our hypotheses, meta-analyses revealed a significant interaction effect overall as well as within studies that manipulated sex and those that manipulated race, such that increased iberalism predicted a larger bias against information that portrays high status groups more favorably than low status groups relative to the reverse.
Also consistent with hypotheses, meta-analyses revealed the hypothesized effects for Liberals overall and within both types of studies. The meta-analyses also revealed a relatively smaller but still significant effect of the sex manipulation for Conservatives in the same direction as for Liberals, and a significant effect of the race manipulation for Conservatives in the opposite direction as for Liberals. Among Moderates, there were no significant effects of the conditions.
Our theory builds from previous work, but goes beyond it, providing a framework for understanding a powerful and largely empiricallyunexplored—but not undiscussed—source of bias. Many scholars have noted—some lamenting and some championing—that many Liberals have protective concerns for victims’ groups (e.g., Bawer, 2012; Haidt, 2012; Mac Donald, 2018; Pinker, 2003) and that those concerns can lead to powerful biases about victims’ groups. Therefore, at minimum, our theory is a prioriplausible.
But it is also largely congruent with many previous analyses of Liberals and provides a potential explanation for previous findings. For example, Liberals have a stronger pro-black bias than Conservatives (Axt et al., 2016), Liberals but not Conservatives are less willing to sacrifice the life of a Black man than a White man to save 100 others (Uhlmann et al., 2009), one of the largest discrepancies among liberal and conservative bias in the Ditto et al. (2019a) meta-analysis came from a study involving victims’ groups (Crawford et al., 2013), Liberals are more inclined to impute motives to scientists who propose biological explanations for life outcomes than those who propose more extrinsic explanations (Hannikainen, 2018), and Liberals are particularly opposed to research on male-favoring sex-differences (Stewart-Williams et al., 2021).
And this theory makes novel predictions, one of which was supported in this paper. The consistency of our results across studies and with established empirical data and with recent controversies in the academic community increases our confidence in our theory and persuades us that it might be a powerful framework for understanding certain political and even scientific biases.
Furthermore, our theory contributes to a burgeoning area of research on liberal bias that has challenged prior assumptions about the relation between political ideology and bias. For a long time, many scholars contended that Conservatives were more prone to bias than Liberals (e.g., Jost et al., 2003). However, recent evidence and arguments have challenged this asymmetry argument, asserting that bias is likely equal across political ideologies (Ditto et al., 2019a; Guay & Johnston, 2022).
Although some scholars have been troubled by this (e.g., Baron & Jost, 2019), our results also challenge the asymmetry argument and illustrate the importance of exploring many different areas of bias. Because most social psychologists are liberal, they may take liberal biases for granted; that is, they simply assume that liberal biases are correct and are not biases at all. Indeed, even when people are aware bias exists, they seem unable to identify bias in themselves (Pronin, Lin, & Ross, 2002).
Furthermore, liberal social psychologists might not be as motivated to discover and shed light on liberal bias as they are for conservative bias, because conservative thought seems more peculiar and foreign to them. When scholars have looked in the right places, though, they have found more equivalent levels of bias between ideological groups (e.g., Crawford, 2012; 2014; Brandt et al., 2014).
Of course, our results cannot settle this important debate, but they do add plausibility to the symmetrical bias thesis, or at least a ‘not as asymmetrical as previously thought’ hypothesis. And they forward a novel domain and direction of bias among Liberals.
==
As noted in the beginning, liberalism in this sense refers to American progressivism (vs conservatism) rather than Classical Liberalism, which is bipartisan.
Anecdotally, I can confirm this phenomenon.
You can point out the violence statistics. Such as 77 black or brown children killed by neighborhood gun violence in one year, versus about 12 unarmed - but not necessarily not-dangerous - black men killed by police each year. One gets national coverage, one does not. The response? Well, it must be incomplete (it's not), it must be from a far-right source (Washington Post, really?), it... it doesn't reflect the "lived reality" of black people in America today (that one might be true, only because you've fear-mongered a narrative that's in defiance of the statistics).
You can point out studies that show a police interaction is less (~27%) likely to end in a police shooting death if the citizen is black than if white. Well, you can't say that, that's racist. You want to deny systemic racism.
You can point out the dometic violence statistics which show women are more often the aggressor, men experience sexual assault, including rape, almost as often as women, lesbian relationships are more violent than straight ones, which are more violent than gay male ones, and that men are more often the fatalities in domestic violence cases when all deaths are accounted for. That's some misogynistic claims from bogus far-right stats (the CDC is not far-right, and it's been reproduced across hundreds of studies).
The narrative has to be protected at all costs. Especially when people have attached their identities to these narratives.
